blob: 1d238d54c4840da5bdc0e5aa0ec960959b578f9a [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;
John Fastabend814abfa2017-07-17 09:27:07 -070069struct xdp_buff;
Linus Torvalds1da177e2005-04-16 15:20:36 -070070
Joe Perchesf629d202013-09-26 14:48:15 -070071void netdev_set_default_ethtool_ops(struct net_device *dev,
72 const struct ethtool_ops *ops);
Stanislaw Gruszkad07d7502013-01-10 23:19:10 +000073
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000074/* Backlog congestion levels */
75#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
76#define NET_RX_DROP 1 /* packet dropped */
77
Patrick McHardy572a9d72009-11-10 06:14:14 +000078/*
79 * Transmit return codes: transmit return codes originate from three different
80 * namespaces:
81 *
82 * - qdisc return codes
83 * - driver transmit return codes
84 * - errno values
85 *
86 * Drivers are allowed to return any one of those in their hard_start_xmit()
87 * function. Real network devices commonly used with qdiscs should only return
88 * the driver transmit return codes though - when qdiscs are used, the actual
89 * transmission happens asynchronously, so the value is not propagated to
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -050090 * higher layers. Virtual network devices transmit synchronously; in this case
91 * the driver transmit return codes are consumed by dev_queue_xmit(), and all
Patrick McHardy572a9d72009-11-10 06:14:14 +000092 * others are propagated to higher layers.
93 */
94
95/* qdisc ->enqueue() return codes. */
96#define NET_XMIT_SUCCESS 0x00
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000097#define NET_XMIT_DROP 0x01 /* skb dropped */
98#define NET_XMIT_CN 0x02 /* congestion notification */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000099#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100
Gerrit Renkerb9df3cb2006-11-14 11:21:36 -0200101/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
102 * indicates that the device will soon be dropping packets, or already drops
103 * some packets of the same priority; prompting us to send less aggressively. */
Patrick McHardy572a9d72009-11-10 06:14:14 +0000104#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
106
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000107/* Driver transmit return codes */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000108#define NETDEV_TX_MASK 0xf0
Patrick McHardy572a9d72009-11-10 06:14:14 +0000109
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000110enum netdev_tx {
Patrick McHardy572a9d72009-11-10 06:14:14 +0000111 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000112 NETDEV_TX_OK = 0x00, /* driver took care of packet */
113 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000114};
115typedef enum netdev_tx netdev_tx_t;
116
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000117/*
118 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
119 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
120 */
121static inline bool dev_xmit_complete(int rc)
122{
123 /*
124 * Positive cases with an skb consumed by a driver:
125 * - successful transmission (rc == NETDEV_TX_OK)
126 * - error while transmitting (rc < 0)
127 * - error while queueing to a different device (rc & NET_XMIT_MASK)
128 */
129 if (likely(rc < NET_XMIT_MASK))
130 return true;
131
132 return false;
133}
134
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500136 * Compute the worst-case header length according to the protocols
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137 * used.
138 */
Graf Yangfe2918b2009-02-05 21:26:19 -0800139
KY Srinivasanc0eb4542015-12-01 16:43:10 -0800140#if defined(CONFIG_HYPERV_NET)
141# define LL_MAX_HEADER 128
142#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
David S. Miller8388e3d2008-05-12 20:17:33 -0700143# if defined(CONFIG_MAC80211_MESH)
144# define LL_MAX_HEADER 128
145# else
146# define LL_MAX_HEADER 96
147# endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700148#else
David S. Miller8388e3d2008-05-12 20:17:33 -0700149# define LL_MAX_HEADER 32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700150#endif
151
Ben Hutchingsd11ead72011-11-25 14:40:26 +0000152#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
153 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700154#define MAX_HEADER LL_MAX_HEADER
155#else
156#define MAX_HEADER (LL_MAX_HEADER + 48)
157#endif
158
159/*
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000160 * Old network device statistics. Fields are native words
161 * (unsigned long) so they can be read and written atomically.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162 */
Graf Yangfe2918b2009-02-05 21:26:19 -0800163
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800164struct net_device_stats {
Ben Hutchings3cfde792010-07-09 09:11:52 +0000165 unsigned long rx_packets;
166 unsigned long tx_packets;
167 unsigned long rx_bytes;
168 unsigned long tx_bytes;
169 unsigned long rx_errors;
170 unsigned long tx_errors;
171 unsigned long rx_dropped;
172 unsigned long tx_dropped;
173 unsigned long multicast;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700174 unsigned long collisions;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700175 unsigned long rx_length_errors;
Ben Hutchings3cfde792010-07-09 09:11:52 +0000176 unsigned long rx_over_errors;
177 unsigned long rx_crc_errors;
178 unsigned long rx_frame_errors;
179 unsigned long rx_fifo_errors;
180 unsigned long rx_missed_errors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700181 unsigned long tx_aborted_errors;
182 unsigned long tx_carrier_errors;
183 unsigned long tx_fifo_errors;
184 unsigned long tx_heartbeat_errors;
185 unsigned long tx_window_errors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186 unsigned long rx_compressed;
187 unsigned long tx_compressed;
188};
189
Linus Torvalds1da177e2005-04-16 15:20:36 -0700190
191#include <linux/cache.h>
192#include <linux/skbuff.h>
193
Eric Dumazetadc93002011-11-17 03:13:26 +0000194#ifdef CONFIG_RPS
Ingo Molnarc5905af2012-02-24 08:31:31 +0100195#include <linux/static_key.h>
196extern struct static_key rps_needed;
Eric Dumazet13bfff22016-12-07 08:29:10 -0800197extern struct static_key rfs_needed;
Eric Dumazetadc93002011-11-17 03:13:26 +0000198#endif
199
Linus Torvalds1da177e2005-04-16 15:20:36 -0700200struct neighbour;
201struct neigh_parms;
202struct sk_buff;
203
Jiri Pirkof001fde2009-05-05 02:48:28 +0000204struct netdev_hw_addr {
205 struct list_head list;
206 unsigned char addr[MAX_ADDR_LEN];
207 unsigned char type;
Jiri Pirkoccffad252009-05-22 23:22:17 +0000208#define NETDEV_HW_ADDR_T_LAN 1
209#define NETDEV_HW_ADDR_T_SAN 2
210#define NETDEV_HW_ADDR_T_SLAVE 3
211#define NETDEV_HW_ADDR_T_UNICAST 4
Jiri Pirko22bedad32010-04-01 21:22:57 +0000212#define NETDEV_HW_ADDR_T_MULTICAST 5
Jiri Pirko22bedad32010-04-01 21:22:57 +0000213 bool global_use;
Vlad Yasevich4cd729b02013-04-15 09:54:25 +0000214 int sync_cnt;
Eric Dumazet8f8f1032010-09-19 11:24:02 -0700215 int refcount;
Vlad Yasevich4543fbe2013-04-02 17:10:07 -0400216 int synced;
Jiri Pirkof001fde2009-05-05 02:48:28 +0000217 struct rcu_head rcu_head;
218};
219
Jiri Pirko31278e72009-06-17 01:12:19 +0000220struct netdev_hw_addr_list {
221 struct list_head list;
222 int count;
223};
224
Jiri Pirko22bedad32010-04-01 21:22:57 +0000225#define netdev_hw_addr_list_count(l) ((l)->count)
226#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
227#define netdev_hw_addr_list_for_each(ha, l) \
228 list_for_each_entry(ha, &(l)->list, list)
229
230#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
231#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
Jiri Pirko32e7bfc2010-01-25 13:36:10 -0800232#define netdev_for_each_uc_addr(ha, dev) \
Jiri Pirko22bedad32010-04-01 21:22:57 +0000233 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
Jiri Pirko32e7bfc2010-01-25 13:36:10 -0800234
Jiri Pirko22bedad32010-04-01 21:22:57 +0000235#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
236#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
Pavel Roskin18e225f2010-04-07 16:40:09 -0700237#define netdev_for_each_mc_addr(ha, dev) \
Jiri Pirko22bedad32010-04-01 21:22:57 +0000238 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
Jiri Pirko6683ece2010-02-04 10:22:25 -0800239
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800240struct hh_cache {
Alexey Dobriyan5b3dc2f2017-04-10 11:11:17 +0300241 unsigned int hh_len;
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
Jiri Pirko2572ac52017-08-07 10:15:17 +0200774enum tc_setup_type {
John Fastabend16e5cc62016-02-16 21:16:43 -0800775 TC_SETUP_MQPRIO,
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800776 TC_SETUP_CLSU32,
Amir Vadai5b33f482016-03-08 12:42:29 +0200777 TC_SETUP_CLSFLOWER,
Jiri Pirkoade9b652017-08-07 10:15:18 +0200778 TC_SETUP_CLSMATCHALL,
Jakub Kicinski332ae8e2016-09-21 11:43:53 +0100779 TC_SETUP_CLSBPF,
John Fastabend16e5cc62016-02-16 21:16:43 -0800780};
781
Brenden Blancoa7862b42016-07-19 12:16:48 -0700782/* These structures hold the attributes of xdp state that are being passed
783 * to the netdevice through the xdp op.
784 */
785enum xdp_netdev_command {
786 /* Set or clear a bpf program used in the earliest stages of packet
787 * rx. The prog will have been loaded as BPF_PROG_TYPE_XDP. The callee
788 * is responsible for calling bpf_prog_put on any old progs that are
789 * stored. In case of error, the callee need not release the new prog
790 * reference, but on success it takes ownership and must bpf_prog_put
791 * when it is no longer used.
792 */
793 XDP_SETUP_PROG,
Jakub Kicinskiee5d0322017-06-21 18:25:04 -0700794 XDP_SETUP_PROG_HW,
Brenden Blancoa7862b42016-07-19 12:16:48 -0700795 /* Check if a bpf program is set on the device. The callee should
Jakub Kicinskice158e52017-06-21 18:25:09 -0700796 * set @prog_attached to one of XDP_ATTACHED_* values, note that "true"
797 * is equivalent to XDP_ATTACHED_DRV.
Brenden Blancoa7862b42016-07-19 12:16:48 -0700798 */
799 XDP_QUERY_PROG,
800};
801
Jakub Kicinskiddf9f972017-04-30 21:46:46 -0700802struct netlink_ext_ack;
803
Brenden Blancoa7862b42016-07-19 12:16:48 -0700804struct netdev_xdp {
805 enum xdp_netdev_command command;
806 union {
807 /* XDP_SETUP_PROG */
Jakub Kicinskiddf9f972017-04-30 21:46:46 -0700808 struct {
Jakub Kicinski32d60272017-06-21 18:25:03 -0700809 u32 flags;
Jakub Kicinskiddf9f972017-04-30 21:46:46 -0700810 struct bpf_prog *prog;
811 struct netlink_ext_ack *extack;
812 };
Brenden Blancoa7862b42016-07-19 12:16:48 -0700813 /* XDP_QUERY_PROG */
Martin KaFai Lau58038692017-06-15 17:29:09 -0700814 struct {
Jakub Kicinskice158e52017-06-21 18:25:09 -0700815 u8 prog_attached;
Martin KaFai Lau58038692017-06-15 17:29:09 -0700816 u32 prog_id;
817 };
Brenden Blancoa7862b42016-07-19 12:16:48 -0700818 };
819};
John Fastabend16e5cc62016-02-16 21:16:43 -0800820
Steffen Klassertd77e38e2017-04-14 10:06:10 +0200821#ifdef CONFIG_XFRM_OFFLOAD
822struct xfrmdev_ops {
823 int (*xdo_dev_state_add) (struct xfrm_state *x);
824 void (*xdo_dev_state_delete) (struct xfrm_state *x);
825 void (*xdo_dev_state_free) (struct xfrm_state *x);
826 bool (*xdo_dev_offload_ok) (struct sk_buff *skb,
827 struct xfrm_state *x);
828};
829#endif
830
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800831/*
832 * This structure defines the management hooks for network devices.
Stephen Hemminger00829822008-11-20 20:14:53 -0800833 * The following hooks can be defined; unless noted otherwise, they are
834 * optional and can be filled with a null pointer.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800835 *
836 * int (*ndo_init)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500837 * This function is called once when a network device is registered.
838 * The network device can use this for any late stage initialization
839 * or semantic validation. It can fail with an error code which will
840 * be propagated back to register_netdev.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800841 *
842 * void (*ndo_uninit)(struct net_device *dev);
843 * This function is called when device is unregistered or when registration
844 * fails. It is not called if init fails.
845 *
846 * int (*ndo_open)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500847 * This function is called when a network device transitions to the up
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800848 * state.
849 *
850 * int (*ndo_stop)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500851 * This function is called when a network device transitions to the down
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800852 * state.
853 *
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000854 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
855 * struct net_device *dev);
Stephen Hemminger00829822008-11-20 20:14:53 -0800856 * Called when a packet needs to be transmitted.
Rusty Russelle79d8422015-04-03 22:17:17 +1030857 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
858 * the queue before that can happen; it's for obsolete devices and weird
859 * corner cases, but the stack really does a non-trivial amount
860 * of useless work if you return NETDEV_TX_BUSY.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500861 * Required; cannot be NULL.
Stephen Hemminger00829822008-11-20 20:14:53 -0800862 *
Dimitris Michailidis1a2a1442017-01-31 16:03:13 -0800863 * netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
864 * struct net_device *dev
865 * netdev_features_t features);
866 * Called by core transmit path to determine if device is capable of
867 * performing offload operations on a given packet. This is to give
868 * the device an opportunity to implement any restrictions that cannot
869 * be otherwise expressed by feature flags. The check is called with
870 * the set of features that the stack has calculated and it returns
871 * those the driver believes to be appropriate.
Eric Dumazetcdba7562016-01-06 06:53:50 -0800872 *
Jason Wangf663dd92014-01-10 16:18:26 +0800873 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +0100874 * void *accel_priv, select_queue_fallback_t fallback);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500875 * Called to decide which queue to use when device supports multiple
Stephen Hemminger00829822008-11-20 20:14:53 -0800876 * transmit queues.
877 *
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800878 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
879 * This function is called to allow device receiver to make
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500880 * changes to configuration when multicast or promiscuous is enabled.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800881 *
882 * void (*ndo_set_rx_mode)(struct net_device *dev);
883 * This function is called device changes address list filtering.
Jiri Pirko01789342011-08-16 06:29:00 +0000884 * If driver handles unicast address filtering, it should set
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500885 * IFF_UNICAST_FLT in its priv_flags.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800886 *
887 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
888 * This function is called when the Media Access Control address
Mike Rapoport37b607c2009-04-27 05:45:54 -0700889 * needs to be changed. If this interface is not defined, the
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500890 * MAC address can not be changed.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800891 *
892 * int (*ndo_validate_addr)(struct net_device *dev);
893 * Test if Media Access Control address is valid for the device.
894 *
895 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500896 * Called when a user requests an ioctl which can't be handled by
897 * the generic interface code. If not defined ioctls return
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800898 * not supported error code.
899 *
900 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
901 * Used to set network devices bus interface parameters. This interface
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500902 * is retained for legacy reasons; new devices should use the bus
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800903 * interface (PCI) for low level management.
904 *
905 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
906 * Called when a user wants to change the Maximum Transfer Unit
Magnus Dammdb46a0e2017-06-14 16:15:24 +0900907 * of a device.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800908 *
Stephen Hemminger00829822008-11-20 20:14:53 -0800909 * void (*ndo_tx_timeout)(struct net_device *dev);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500910 * Callback used when the transmitter has not made any progress
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800911 * for dev->watchdog ticks.
912 *
stephen hemmingerbc1f4472017-01-06 19:12:52 -0800913 * void (*ndo_get_stats64)(struct net_device *dev,
914 * struct rtnl_link_stats64 *storage);
Wolfram Sangd308e382009-10-07 13:53:11 -0700915 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800916 * Called when a user wants to get the network device usage
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000917 * statistics. Drivers must do one of the following:
Ben Hutchings3cfde792010-07-09 09:11:52 +0000918 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
919 * rtnl_link_stats64 structure passed by the caller.
Ben Hutchings82695d92010-06-15 15:08:48 -0700920 * 2. Define @ndo_get_stats to update a net_device_stats structure
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000921 * (which should normally be dev->stats) and return a pointer to
922 * it. The structure may be changed asynchronously only if each
923 * field is written atomically.
924 * 3. Update dev->stats asynchronously and atomically, and define
925 * neither operation.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800926 *
Or Gerlitz3df5b3c2016-11-22 23:09:54 +0200927 * bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id)
Nogah Frankel2c9d85d2016-09-16 15:05:36 +0200928 * Return true if this device supports offload stats of this attr_id.
929 *
930 * int (*ndo_get_offload_stats)(int attr_id, const struct net_device *dev,
931 * void *attr_data)
932 * Get statistics for offload operations by attr_id. Write it into the
933 * attr_data pointer.
934 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530935 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500936 * If device supports VLAN filtering this function is called when a
Patrick McHardy80d5c362013-04-19 02:04:28 +0000937 * VLAN id is registered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800938 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530939 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500940 * If device supports VLAN filtering this function is called when a
Patrick McHardy80d5c362013-04-19 02:04:28 +0000941 * VLAN id is unregistered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800942 *
943 * void (*ndo_poll_controller)(struct net_device *dev);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000944 *
945 * SR-IOV management functions.
946 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
Moshe Shemesh79aab092016-09-22 12:11:15 +0300947 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan,
948 * u8 qos, __be16 proto);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -0400949 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
950 * int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +0000951 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
Hiroshi Shimamotodd461d62015-08-28 06:57:55 +0000952 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000953 * int (*ndo_get_vf_config)(struct net_device *dev,
954 * int vf, struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +0300955 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
Scott Feldman57b61082010-05-17 22:49:55 -0700956 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
957 * struct nlattr *port[]);
Vlad Zolotarov01a3d792015-03-30 21:35:23 +0300958 *
959 * Enable or disable the VF ability to query its RSS Redirection Table and
960 * Hash Key. This is needed since on some devices VF share this information
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -0500961 * with PF and querying it may introduce a theoretical security risk.
Vlad Zolotarov01a3d792015-03-30 21:35:23 +0300962 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
Scott Feldman57b61082010-05-17 22:49:55 -0700963 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
Jiri Pirko2572ac52017-08-07 10:15:17 +0200964 * int (*ndo_setup_tc)(struct net_device *dev, enum tc_setup_type type,
Jiri Pirkode4784c2017-08-07 10:15:32 +0200965 * void *type_data);
Florian Fainelli6a4bc2b2017-01-26 14:44:17 -0800966 * Called to setup any 'tc' scheduler, classifier or action on @dev.
967 * This is always called from the stack with the rtnl lock held and netif
968 * tx queues stopped. This allows the netdevice to perform queue
969 * management safely.
Ben Hutchingsc4454772011-01-19 11:03:53 +0000970 *
Yi Zoue9bce842011-03-09 08:48:03 +0000971 * Fiber Channel over Ethernet (FCoE) offload functions.
972 * int (*ndo_fcoe_enable)(struct net_device *dev);
973 * Called when the FCoE protocol stack wants to start using LLD for FCoE
974 * so the underlying device can perform whatever needed configuration or
975 * initialization to support acceleration of FCoE traffic.
976 *
977 * int (*ndo_fcoe_disable)(struct net_device *dev);
978 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
979 * so the underlying device can perform whatever needed clean-ups to
980 * stop supporting acceleration of FCoE traffic.
981 *
982 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
983 * struct scatterlist *sgl, unsigned int sgc);
984 * Called when the FCoE Initiator wants to initialize an I/O that
985 * is a possible candidate for Direct Data Placement (DDP). The LLD can
986 * perform necessary setup and returns 1 to indicate the device is set up
987 * successfully to perform DDP on this I/O, otherwise this returns 0.
988 *
989 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
990 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
991 * indicated by the FC exchange id 'xid', so the underlying device can
992 * clean up and reuse resources for later DDP requests.
993 *
994 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
995 * struct scatterlist *sgl, unsigned int sgc);
996 * Called when the FCoE Target wants to initialize an I/O that
997 * is a possible candidate for Direct Data Placement (DDP). The LLD can
998 * perform necessary setup and returns 1 to indicate the device is set up
999 * successfully to perform DDP on this I/O, otherwise this returns 0.
1000 *
Neerav Parikh68bad942012-01-04 20:23:39 +00001001 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1002 * struct netdev_fcoe_hbainfo *hbainfo);
1003 * Called when the FCoE Protocol stack wants information on the underlying
1004 * device. This information is utilized by the FCoE protocol stack to
1005 * register attributes with Fiber Channel management service as per the
1006 * FC-GS Fabric Device Management Information(FDMI) specification.
1007 *
Yi Zoue9bce842011-03-09 08:48:03 +00001008 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
1009 * Called when the underlying device wants to override default World Wide
1010 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
1011 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
1012 * protocol stack to use.
1013 *
Ben Hutchingsc4454772011-01-19 11:03:53 +00001014 * RFS acceleration.
1015 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
1016 * u16 rxq_index, u32 flow_id);
1017 * Set hardware filter for RFS. rxq_index is the target queue index;
1018 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
1019 * Return the filter ID on success, or a negative error code.
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001020 *
Jiri Pirko8b98a702013-01-03 22:49:02 +00001021 * Slave management functions (for bridge, bonding, etc).
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001022 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
1023 * Called to make another netdev an underling.
1024 *
1025 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
1026 * Called to release previously enslaved netdev.
Michał Mirosław5455c692011-02-15 16:59:17 +00001027 *
1028 * Feature/offload setting functions.
Dimitris Michailidis1a2a1442017-01-31 16:03:13 -08001029 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1030 * netdev_features_t features);
1031 * Adjusts the requested feature flags according to device-specific
1032 * constraints, and returns the resulting flags. Must not modify
1033 * the device state.
1034 *
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001035 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +00001036 * Called to update device configuration to new features. Passed
1037 * feature set might be less than what was returned by ndo_fix_features()).
1038 * Must return >0 or -errno if it changed dev->features itself.
1039 *
stephen hemmingeredc7d572012-10-01 12:32:33 +00001040 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
1041 * struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001042 * const unsigned char *addr, u16 vid, u16 flags)
John Fastabend77162022012-04-15 06:43:56 +00001043 * Adds an FDB entry to dev for addr.
Vlad Yasevich1690be62013-02-13 12:00:18 +00001044 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1045 * struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001046 * const unsigned char *addr, u16 vid)
John Fastabend77162022012-04-15 06:43:56 +00001047 * Deletes the FDB entry from dev coresponding to addr.
1048 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -04001049 * struct net_device *dev, struct net_device *filter_dev,
Roopa Prabhud2976532016-08-30 21:56:45 -07001050 * int *idx)
John Fastabend77162022012-04-15 06:43:56 +00001051 * Used to add FDB entries to dump requests. Implementers should add
1052 * entries to skb and update idx with the number of entries.
John Fastabende5a55a82012-10-24 08:12:57 +00001053 *
Nicolas Dichtelad41faa2015-03-17 11:16:00 +01001054 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1055 * u16 flags)
John Fastabende5a55a82012-10-24 08:12:57 +00001056 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
Nicolas Dichtel46c264d2015-04-28 18:33:49 +02001057 * struct net_device *dev, u32 filter_mask,
1058 * int nlflags)
Nicolas Dichtelad41faa2015-03-17 11:16:00 +01001059 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1060 * u16 flags);
Jiri Pirko4bf84c32012-12-27 23:49:37 +00001061 *
1062 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1063 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1064 * which do not represent real hardware may define this to allow their
1065 * userspace components to manage their virtual carrier state. Devices
1066 * that determine carrier state from physical hardware properties (eg
1067 * network cables) or protocol-dependent mechanisms (eg
1068 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
Jiri Pirko66b52b02013-07-29 18:16:49 +02001069 *
1070 * int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01001071 * struct netdev_phys_item_id *ppid);
Jiri Pirko66b52b02013-07-29 18:16:49 +02001072 * Called to get ID of physical port of this device. If driver does
1073 * not implement this, it is assumed that the hw is not able to have
1074 * multiple net devices on single physical port.
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07001075 *
Alexander Duyck7c46a642016-06-16 12:21:00 -07001076 * void (*ndo_udp_tunnel_add)(struct net_device *dev,
1077 * struct udp_tunnel_info *ti);
1078 * Called by UDP tunnel to notify a driver about the UDP port and socket
1079 * address family that a UDP tunnel is listnening to. It is called only
1080 * when a new port starts listening. The operation is protected by the
1081 * RTNL.
1082 *
1083 * void (*ndo_udp_tunnel_del)(struct net_device *dev,
1084 * struct udp_tunnel_info *ti);
1085 * Called by UDP tunnel to notify the driver about a UDP port and socket
1086 * address family that the UDP tunnel is not listening to anymore. The
1087 * operation is protected by the RTNL.
1088 *
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001089 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1090 * struct net_device *dev)
1091 * Called by upper layer devices to accelerate switching or other
1092 * station functionality into hardware. 'pdev is the lowerdev
1093 * to use for the offload and 'dev' is the net device that will
1094 * back the offload. Returns a pointer to the private structure
1095 * the upper layer will maintain.
1096 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1097 * Called by upper layer device to delete the station created
1098 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1099 * the station and priv is the structure returned by the add
1100 * operation.
John Fastabend822b3b22015-03-18 14:57:33 +02001101 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1102 * int queue_index, u32 maxrate);
1103 * Called when a user wants to set a max-rate limitation of specific
1104 * TX queue.
Nicolas Dichtela54acb32015-04-02 17:07:00 +02001105 * int (*ndo_get_iflink)(const struct net_device *dev);
1106 * Called to get the iflink value of this device.
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001107 * void (*ndo_change_proto_down)(struct net_device *dev,
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001108 * bool proto_down);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001109 * This function is used to pass protocol port error state information
1110 * to the switch driver. The switch driver can react to the proto_down
1111 * by doing a phys down on the associated switch port.
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07001112 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1113 * This function is used to get egress tunnel information for given skb.
1114 * This is useful for retrieving outer tunnel header parameters while
1115 * sampling packet.
Paolo Abeni871b6422016-02-26 10:45:37 +01001116 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1117 * This function is used to specify the headroom that the skb must
1118 * consider when allocation skb during packet reception. Setting
1119 * appropriate rx headroom value allows avoiding skb head copy on
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001120 * forward. Setting a negative value resets the rx headroom to the
Paolo Abeni871b6422016-02-26 10:45:37 +01001121 * default value.
Brenden Blancoa7862b42016-07-19 12:16:48 -07001122 * int (*ndo_xdp)(struct net_device *dev, struct netdev_xdp *xdp);
1123 * This function is used to set or query state related to XDP on the
1124 * netdevice. See definition of enum xdp_netdev_command for details.
John Fastabend814abfa2017-07-17 09:27:07 -07001125 * int (*ndo_xdp_xmit)(struct net_device *dev, struct xdp_buff *xdp);
1126 * This function is used to submit a XDP packet for transmit on a
1127 * netdevice.
John Fastabend11393cc2017-07-17 09:29:40 -07001128 * void (*ndo_xdp_flush)(struct net_device *dev);
1129 * This function is used to inform the driver to flush a paticular
1130 * xpd tx queue. Must be called on same CPU as xdp_xmit.
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001131 */
1132struct net_device_ops {
1133 int (*ndo_init)(struct net_device *dev);
1134 void (*ndo_uninit)(struct net_device *dev);
1135 int (*ndo_open)(struct net_device *dev);
1136 int (*ndo_stop)(struct net_device *dev);
Eric Dumazetcdba7562016-01-06 06:53:50 -08001137 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1138 struct net_device *dev);
1139 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1140 struct net_device *dev,
1141 netdev_features_t features);
Stephen Hemminger00829822008-11-20 20:14:53 -08001142 u16 (*ndo_select_queue)(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08001143 struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +01001144 void *accel_priv,
1145 select_queue_fallback_t fallback);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001146 void (*ndo_change_rx_flags)(struct net_device *dev,
1147 int flags);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001148 void (*ndo_set_rx_mode)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001149 int (*ndo_set_mac_address)(struct net_device *dev,
1150 void *addr);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001151 int (*ndo_validate_addr)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001152 int (*ndo_do_ioctl)(struct net_device *dev,
1153 struct ifreq *ifr, int cmd);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001154 int (*ndo_set_config)(struct net_device *dev,
1155 struct ifmap *map);
Stephen Hemminger00829822008-11-20 20:14:53 -08001156 int (*ndo_change_mtu)(struct net_device *dev,
1157 int new_mtu);
1158 int (*ndo_neigh_setup)(struct net_device *dev,
1159 struct neigh_parms *);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001160 void (*ndo_tx_timeout) (struct net_device *dev);
1161
stephen hemmingerbc1f4472017-01-06 19:12:52 -08001162 void (*ndo_get_stats64)(struct net_device *dev,
1163 struct rtnl_link_stats64 *storage);
Or Gerlitz3df5b3c2016-11-22 23:09:54 +02001164 bool (*ndo_has_offload_stats)(const struct net_device *dev, int attr_id);
Nogah Frankel2c9d85d2016-09-16 15:05:36 +02001165 int (*ndo_get_offload_stats)(int attr_id,
1166 const struct net_device *dev,
1167 void *attr_data);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001168 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1169
Jiri Pirko8e586132011-12-08 19:52:37 -05001170 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001171 __be16 proto, u16 vid);
Jiri Pirko8e586132011-12-08 19:52:37 -05001172 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001173 __be16 proto, u16 vid);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001174#ifdef CONFIG_NET_POLL_CONTROLLER
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001175 void (*ndo_poll_controller)(struct net_device *dev);
Herbert Xu4247e162010-06-10 16:12:47 +00001176 int (*ndo_netpoll_setup)(struct net_device *dev,
Eric W. Biedermana8779ec2014-03-27 15:36:38 -07001177 struct netpoll_info *info);
WANG Cong0e34e932010-05-06 00:47:21 -07001178 void (*ndo_netpoll_cleanup)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001179#endif
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001180 int (*ndo_set_vf_mac)(struct net_device *dev,
1181 int queue, u8 *mac);
1182 int (*ndo_set_vf_vlan)(struct net_device *dev,
Moshe Shemesh79aab092016-09-22 12:11:15 +03001183 int queue, u16 vlan,
1184 u8 qos, __be16 proto);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -04001185 int (*ndo_set_vf_rate)(struct net_device *dev,
1186 int vf, int min_tx_rate,
1187 int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +00001188 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1189 int vf, bool setting);
Hiroshi Shimamotodd461d62015-08-28 06:57:55 +00001190 int (*ndo_set_vf_trust)(struct net_device *dev,
1191 int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001192 int (*ndo_get_vf_config)(struct net_device *dev,
1193 int vf,
1194 struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +03001195 int (*ndo_set_vf_link_state)(struct net_device *dev,
1196 int vf, int link_state);
Eran Ben Elisha3b766cd2015-06-15 17:59:07 +03001197 int (*ndo_get_vf_stats)(struct net_device *dev,
1198 int vf,
1199 struct ifla_vf_stats
1200 *vf_stats);
Scott Feldman57b61082010-05-17 22:49:55 -07001201 int (*ndo_set_vf_port)(struct net_device *dev,
1202 int vf,
1203 struct nlattr *port[]);
1204 int (*ndo_get_vf_port)(struct net_device *dev,
1205 int vf, struct sk_buff *skb);
Eli Cohencc8e27c2016-03-11 22:58:34 +02001206 int (*ndo_set_vf_guid)(struct net_device *dev,
1207 int vf, u64 guid,
1208 int guid_type);
Vlad Zolotarov01a3d792015-03-30 21:35:23 +03001209 int (*ndo_set_vf_rss_query_en)(
1210 struct net_device *dev,
1211 int vf, bool setting);
John Fastabend16e5cc62016-02-16 21:16:43 -08001212 int (*ndo_setup_tc)(struct net_device *dev,
Jiri Pirko2572ac52017-08-07 10:15:17 +02001213 enum tc_setup_type type,
Jiri Pirkode4784c2017-08-07 10:15:32 +02001214 void *type_data);
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001215#if IS_ENABLED(CONFIG_FCOE)
Yi Zoucb454392009-08-31 12:31:36 +00001216 int (*ndo_fcoe_enable)(struct net_device *dev);
1217 int (*ndo_fcoe_disable)(struct net_device *dev);
Yi Zou4d288d52009-02-27 14:06:59 -08001218 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1219 u16 xid,
1220 struct scatterlist *sgl,
1221 unsigned int sgc);
1222 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1223 u16 xid);
Yi Zou6247e082011-02-01 07:22:06 +00001224 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1225 u16 xid,
1226 struct scatterlist *sgl,
1227 unsigned int sgc);
Neerav Parikh68bad942012-01-04 20:23:39 +00001228 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1229 struct netdev_fcoe_hbainfo *hbainfo);
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001230#endif
1231
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001232#if IS_ENABLED(CONFIG_LIBFCOE)
Yi Zoudf5c7942009-10-28 18:24:35 +00001233#define NETDEV_FCOE_WWNN 0
1234#define NETDEV_FCOE_WWPN 1
1235 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1236 u64 *wwn, int type);
Yi Zou4d288d52009-02-27 14:06:59 -08001237#endif
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001238
Ben Hutchingsc4454772011-01-19 11:03:53 +00001239#ifdef CONFIG_RFS_ACCEL
1240 int (*ndo_rx_flow_steer)(struct net_device *dev,
1241 const struct sk_buff *skb,
1242 u16 rxq_index,
1243 u32 flow_id);
1244#endif
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001245 int (*ndo_add_slave)(struct net_device *dev,
1246 struct net_device *slave_dev);
1247 int (*ndo_del_slave)(struct net_device *dev,
1248 struct net_device *slave_dev);
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001249 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1250 netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +00001251 int (*ndo_set_features)(struct net_device *dev,
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001252 netdev_features_t features);
Jiri Pirko503eebc2016-07-05 11:27:37 +02001253 int (*ndo_neigh_construct)(struct net_device *dev,
1254 struct neighbour *n);
1255 void (*ndo_neigh_destroy)(struct net_device *dev,
1256 struct neighbour *n);
John Fastabend77162022012-04-15 06:43:56 +00001257
1258 int (*ndo_fdb_add)(struct ndmsg *ndm,
stephen hemmingeredc7d572012-10-01 12:32:33 +00001259 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001260 struct net_device *dev,
stephen hemminger6b6e2722012-09-17 10:03:26 +00001261 const unsigned char *addr,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001262 u16 vid,
John Fastabend77162022012-04-15 06:43:56 +00001263 u16 flags);
1264 int (*ndo_fdb_del)(struct ndmsg *ndm,
Vlad Yasevich1690be62013-02-13 12:00:18 +00001265 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001266 struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001267 const unsigned char *addr,
1268 u16 vid);
John Fastabend77162022012-04-15 06:43:56 +00001269 int (*ndo_fdb_dump)(struct sk_buff *skb,
1270 struct netlink_callback *cb,
1271 struct net_device *dev,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -04001272 struct net_device *filter_dev,
Roopa Prabhud2976532016-08-30 21:56:45 -07001273 int *idx);
John Fastabende5a55a82012-10-24 08:12:57 +00001274
1275 int (*ndo_bridge_setlink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001276 struct nlmsghdr *nlh,
1277 u16 flags);
John Fastabende5a55a82012-10-24 08:12:57 +00001278 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1279 u32 pid, u32 seq,
Vlad Yasevich6cbdcee2013-02-13 12:00:13 +00001280 struct net_device *dev,
Nicolas Dichtel46c264d2015-04-28 18:33:49 +02001281 u32 filter_mask,
1282 int nlflags);
Vlad Yasevich407af322013-02-13 12:00:12 +00001283 int (*ndo_bridge_dellink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001284 struct nlmsghdr *nlh,
1285 u16 flags);
Jiri Pirko4bf84c32012-12-27 23:49:37 +00001286 int (*ndo_change_carrier)(struct net_device *dev,
1287 bool new_carrier);
Jiri Pirko66b52b02013-07-29 18:16:49 +02001288 int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01001289 struct netdev_phys_item_id *ppid);
David Aherndb24a902015-03-17 20:23:15 -06001290 int (*ndo_get_phys_port_name)(struct net_device *dev,
1291 char *name, size_t len);
Alexander Duyck7c46a642016-06-16 12:21:00 -07001292 void (*ndo_udp_tunnel_add)(struct net_device *dev,
1293 struct udp_tunnel_info *ti);
1294 void (*ndo_udp_tunnel_del)(struct net_device *dev,
1295 struct udp_tunnel_info *ti);
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001296 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1297 struct net_device *dev);
1298 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1299 void *priv);
1300
Vlad Yasevich25175ba2014-05-16 17:04:54 -04001301 int (*ndo_get_lock_subclass)(struct net_device *dev);
John Fastabend822b3b22015-03-18 14:57:33 +02001302 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1303 int queue_index,
1304 u32 maxrate);
Nicolas Dichtela54acb32015-04-02 17:07:00 +02001305 int (*ndo_get_iflink)(const struct net_device *dev);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001306 int (*ndo_change_proto_down)(struct net_device *dev,
1307 bool proto_down);
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07001308 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1309 struct sk_buff *skb);
Paolo Abeni871b6422016-02-26 10:45:37 +01001310 void (*ndo_set_rx_headroom)(struct net_device *dev,
1311 int needed_headroom);
Brenden Blancoa7862b42016-07-19 12:16:48 -07001312 int (*ndo_xdp)(struct net_device *dev,
1313 struct netdev_xdp *xdp);
John Fastabend814abfa2017-07-17 09:27:07 -07001314 int (*ndo_xdp_xmit)(struct net_device *dev,
1315 struct xdp_buff *xdp);
John Fastabend11393cc2017-07-17 09:29:40 -07001316 void (*ndo_xdp_flush)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001317};
1318
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001319/**
1320 * enum net_device_priv_flags - &struct net_device priv_flags
1321 *
1322 * These are the &struct net_device, they are only set internally
1323 * by drivers and used in the kernel. These flags are invisible to
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001324 * userspace; this means that the order of these flags can change
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001325 * during any kernel release.
1326 *
1327 * You should have a pretty good reason to be extending these flags.
1328 *
1329 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1330 * @IFF_EBRIDGE: Ethernet bridging device
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001331 * @IFF_BONDING: bonding master or slave
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001332 * @IFF_ISATAP: ISATAP interface (RFC4214)
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001333 * @IFF_WAN_HDLC: WAN HDLC device
1334 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1335 * release skb->dst
1336 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1337 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1338 * @IFF_MACVLAN_PORT: device used as macvlan port
1339 * @IFF_BRIDGE_PORT: device used as bridge port
1340 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1341 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1342 * @IFF_UNICAST_FLT: Supports unicast filtering
1343 * @IFF_TEAM_PORT: device used as team port
1344 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1345 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1346 * change when it's running
1347 * @IFF_MACVLAN: Macvlan device
Luis de Bethencourt6d0e24c2016-03-21 20:58:28 +00001348 * @IFF_XMIT_DST_RELEASE_PERM: IFF_XMIT_DST_RELEASE not taking into account
1349 * underlying stacked devices
1350 * @IFF_IPVLAN_MASTER: IPvlan master device
1351 * @IFF_IPVLAN_SLAVE: IPvlan slave device
David Ahern007979e2015-09-29 20:07:10 -07001352 * @IFF_L3MDEV_MASTER: device is an L3 master device
Phil Sutterfa8187c2015-08-13 19:01:06 +02001353 * @IFF_NO_QUEUE: device can run without qdisc attached
Jiri Pirko35d4e172015-08-27 09:31:20 +02001354 * @IFF_OPENVSWITCH: device is a Open vSwitch master
David Ahernfee6d4c2015-10-05 08:51:24 -07001355 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
Jiri Pirkoc981e422015-12-03 12:12:06 +01001356 * @IFF_TEAM: device is a team device
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001357 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
Paolo Abeni871b6422016-02-26 10:45:37 +01001358 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1359 * entity (i.e. the master device for bridged veth)
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001360 * @IFF_MACSEC: device is a MACsec device
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001361 */
1362enum netdev_priv_flags {
1363 IFF_802_1Q_VLAN = 1<<0,
1364 IFF_EBRIDGE = 1<<1,
Jiri Pirko0dc15492015-08-27 09:31:21 +02001365 IFF_BONDING = 1<<2,
1366 IFF_ISATAP = 1<<3,
1367 IFF_WAN_HDLC = 1<<4,
1368 IFF_XMIT_DST_RELEASE = 1<<5,
1369 IFF_DONT_BRIDGE = 1<<6,
1370 IFF_DISABLE_NETPOLL = 1<<7,
1371 IFF_MACVLAN_PORT = 1<<8,
1372 IFF_BRIDGE_PORT = 1<<9,
1373 IFF_OVS_DATAPATH = 1<<10,
1374 IFF_TX_SKB_SHARING = 1<<11,
1375 IFF_UNICAST_FLT = 1<<12,
1376 IFF_TEAM_PORT = 1<<13,
1377 IFF_SUPP_NOFCS = 1<<14,
1378 IFF_LIVE_ADDR_CHANGE = 1<<15,
1379 IFF_MACVLAN = 1<<16,
1380 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1381 IFF_IPVLAN_MASTER = 1<<18,
1382 IFF_IPVLAN_SLAVE = 1<<19,
David Ahern007979e2015-09-29 20:07:10 -07001383 IFF_L3MDEV_MASTER = 1<<20,
Jiri Pirko0dc15492015-08-27 09:31:21 +02001384 IFF_NO_QUEUE = 1<<21,
1385 IFF_OPENVSWITCH = 1<<22,
David Ahernfee6d4c2015-10-05 08:51:24 -07001386 IFF_L3MDEV_SLAVE = 1<<23,
Jiri Pirkoc981e422015-12-03 12:12:06 +01001387 IFF_TEAM = 1<<24,
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001388 IFF_RXFH_CONFIGURED = 1<<25,
Paolo Abeni871b6422016-02-26 10:45:37 +01001389 IFF_PHONY_HEADROOM = 1<<26,
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001390 IFF_MACSEC = 1<<27,
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001391};
1392
1393#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1394#define IFF_EBRIDGE IFF_EBRIDGE
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001395#define IFF_BONDING IFF_BONDING
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001396#define IFF_ISATAP IFF_ISATAP
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001397#define IFF_WAN_HDLC IFF_WAN_HDLC
1398#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1399#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1400#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1401#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1402#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1403#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1404#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1405#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1406#define IFF_TEAM_PORT IFF_TEAM_PORT
1407#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1408#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1409#define IFF_MACVLAN IFF_MACVLAN
Eric Dumazet02875872014-10-05 18:38:35 -07001410#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -08001411#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1412#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
David Ahern007979e2015-09-29 20:07:10 -07001413#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
Phil Sutterfa8187c2015-08-13 19:01:06 +02001414#define IFF_NO_QUEUE IFF_NO_QUEUE
Jiri Pirko35d4e172015-08-27 09:31:20 +02001415#define IFF_OPENVSWITCH IFF_OPENVSWITCH
Jiri Pirko8f253482015-11-04 14:59:06 +01001416#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
Jiri Pirkoc981e422015-12-03 12:12:06 +01001417#define IFF_TEAM IFF_TEAM
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001418#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001419#define IFF_MACSEC IFF_MACSEC
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001420
Karoly Kemeny536721b2014-07-30 20:27:36 +02001421/**
1422 * struct net_device - The DEVICE structure.
Mauro Carvalho Chehabd6519832017-05-12 09:35:46 -03001423 *
1424 * Actually, this whole structure is a big mistake. It mixes I/O
1425 * data with strictly "high-level" data, and it has to know about
1426 * almost every data structure used in the INET module.
Karoly Kemeny536721b2014-07-30 20:27:36 +02001427 *
1428 * @name: This is the first field of the "visible" part of this structure
1429 * (i.e. as seen by users in the "Space.c" file). It is the name
Mauro Carvalho Chehabd6519832017-05-12 09:35:46 -03001430 * of the interface.
Karoly Kemeny536721b2014-07-30 20:27:36 +02001431 *
1432 * @name_hlist: Device name hash chain, please keep it close to name[]
1433 * @ifalias: SNMP alias
1434 * @mem_end: Shared memory end
1435 * @mem_start: Shared memory start
1436 * @base_addr: Device I/O address
1437 * @irq: Device IRQ number
1438 *
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001439 * @carrier_changes: Stats to monitor carrier on<->off transitions
1440 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001441 * @state: Generic network queuing layer state, see netdev_state_t
1442 * @dev_list: The global list of network devices
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001443 * @napi_list: List entry used for polling NAPI devices
1444 * @unreg_list: List entry when we are unregistering the
1445 * device; see the function unregister_netdev
1446 * @close_list: List entry used when we are closing the device
Benjamin Poirier62d885fe2016-03-21 14:08:28 -07001447 * @ptype_all: Device-specific packet handlers for all protocols
1448 * @ptype_specific: Device-specific, protocol-specific packet handlers
Karoly Kemeny536721b2014-07-30 20:27:36 +02001449 *
1450 * @adj_list: Directly linked devices, like slaves for bonding
Karoly Kemeny536721b2014-07-30 20:27:36 +02001451 * @features: Currently active device features
1452 * @hw_features: User-changeable features
1453 *
1454 * @wanted_features: User-requested features
1455 * @vlan_features: Mask of features inheritable by VLAN devices
1456 *
1457 * @hw_enc_features: Mask of features inherited by encapsulating devices
1458 * This field indicates what encapsulation
1459 * offloads the hardware is capable of doing,
1460 * and drivers will need to set them appropriately.
1461 *
1462 * @mpls_features: Mask of features inheritable by MPLS
1463 *
1464 * @ifindex: interface index
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001465 * @group: The group the device belongs to
Karoly Kemeny536721b2014-07-30 20:27:36 +02001466 *
1467 * @stats: Statistics struct, which was left as a legacy, use
1468 * rtnl_link_stats64 instead
1469 *
1470 * @rx_dropped: Dropped packets by core network,
1471 * do not use this in drivers
1472 * @tx_dropped: Dropped packets by core network,
1473 * do not use this in drivers
Jarod Wilson6e7333d2016-02-01 18:51:05 -05001474 * @rx_nohandler: nohandler dropped packets by core network on
1475 * inactive devices, do not use this in drivers
Karoly Kemeny536721b2014-07-30 20:27:36 +02001476 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001477 * @wireless_handlers: List of functions to handle Wireless Extensions,
1478 * instead of ioctl,
1479 * see <net/iw_handler.h> for details.
1480 * @wireless_data: Instance data managed by the core of wireless extensions
1481 *
1482 * @netdev_ops: Includes several pointers to callbacks,
1483 * if one wants to override the ndo_*() functions
1484 * @ethtool_ops: Management operations
Alexander Aringf997c552016-06-15 21:20:23 +02001485 * @ndisc_ops: Includes callbacks for different IPv6 neighbour
1486 * discovery handling. Necessary for e.g. 6LoWPAN.
Eric W. Biedermand4760592015-03-02 00:11:09 -06001487 * @header_ops: Includes callbacks for creating,parsing,caching,etc
Karoly Kemeny536721b2014-07-30 20:27:36 +02001488 * of Layer 2 headers.
1489 *
1490 * @flags: Interface flags (a la BSD)
1491 * @priv_flags: Like 'flags' but invisible to userspace,
1492 * see if.h for the definitions
1493 * @gflags: Global flags ( kept as legacy )
1494 * @padded: How much padding added by alloc_netdev()
1495 * @operstate: RFC2863 operstate
1496 * @link_mode: Mapping policy to operstate
1497 * @if_port: Selectable AUI, TP, ...
1498 * @dma: DMA channel
1499 * @mtu: Interface MTU value
Jarod Wilson61e84622016-10-07 22:04:33 -04001500 * @min_mtu: Interface Minimum MTU value
1501 * @max_mtu: Interface Maximum MTU value
Karoly Kemeny536721b2014-07-30 20:27:36 +02001502 * @type: Interface hardware type
Willem de Bruijn2793a232016-03-09 21:58:32 -05001503 * @hard_header_len: Maximum hardware header length.
Willem de Bruijn217e6fa2017-02-07 15:57:20 -05001504 * @min_header_len: Minimum hardware header length
Karoly Kemeny536721b2014-07-30 20:27:36 +02001505 *
1506 * @needed_headroom: Extra headroom the hardware may need, but not in all
1507 * cases can this be guaranteed
1508 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1509 * cases can this be guaranteed. Some cases also use
1510 * LL_MAX_HEADER instead to allocate the skb
1511 *
1512 * interface address info:
1513 *
1514 * @perm_addr: Permanent hw address
1515 * @addr_assign_type: Hw address assignment type
1516 * @addr_len: Hardware address length
Alexander Aring8626a0c2016-06-15 21:20:16 +02001517 * @neigh_priv_len: Used in neigh_alloc()
Karoly Kemeny536721b2014-07-30 20:27:36 +02001518 * @dev_id: Used to differentiate devices that share
1519 * the same link layer address
1520 * @dev_port: Used to differentiate devices that share
1521 * the same function
1522 * @addr_list_lock: XXX: need comments on this one
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001523 * @uc_promisc: Counter that indicates promiscuous mode
Karoly Kemeny536721b2014-07-30 20:27:36 +02001524 * has been enabled due to the need to listen to
1525 * additional unicast addresses in a device that
1526 * does not implement ndo_set_rx_mode()
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001527 * @uc: unicast mac addresses
1528 * @mc: multicast mac addresses
1529 * @dev_addrs: list of device hw addresses
1530 * @queues_kset: Group of all Kobjects in the Tx and RX queues
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001531 * @promiscuity: Number of times the NIC is told to work in
1532 * promiscuous mode; if it becomes 0 the NIC will
1533 * exit promiscuous mode
Karoly Kemeny536721b2014-07-30 20:27:36 +02001534 * @allmulti: Counter, enables or disables allmulticast mode
1535 *
1536 * @vlan_info: VLAN info
1537 * @dsa_ptr: dsa specific data
1538 * @tipc_ptr: TIPC specific data
1539 * @atalk_ptr: AppleTalk link
1540 * @ip_ptr: IPv4 specific data
1541 * @dn_ptr: DECnet specific data
1542 * @ip6_ptr: IPv6 specific data
1543 * @ax25_ptr: AX.25 specific data
1544 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1545 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001546 * @dev_addr: Hw address (before bcast,
1547 * because most packets are unicast)
1548 *
1549 * @_rx: Array of RX queues
1550 * @num_rx_queues: Number of RX queues
1551 * allocated at register_netdev() time
1552 * @real_num_rx_queues: Number of RX queues currently active in device
1553 *
1554 * @rx_handler: handler for received packets
1555 * @rx_handler_data: XXX: need comments on this one
1556 * @ingress_queue: XXX: need comments on this one
1557 * @broadcast: hw bcast address
1558 *
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001559 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1560 * indexed by RX queue number. Assigned by driver.
1561 * This must only be set if the ndo_rx_flow_steer
1562 * operation is defined
1563 * @index_hlist: Device index hash chain
1564 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001565 * @_tx: Array of TX queues
1566 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1567 * @real_num_tx_queues: Number of TX queues currently active in device
1568 * @qdisc: Root qdisc from userspace point of view
1569 * @tx_queue_len: Max frames per queue allowed
1570 * @tx_global_lock: XXX: need comments on this one
1571 *
1572 * @xps_maps: XXX: need comments on this one
1573 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001574 * @watchdog_timeo: Represents the timeout that is used by
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001575 * the watchdog (see dev_watchdog())
Karoly Kemeny536721b2014-07-30 20:27:36 +02001576 * @watchdog_timer: List of timers
1577 *
1578 * @pcpu_refcnt: Number of references to this device
1579 * @todo_list: Delayed register/unregister
Karoly Kemeny536721b2014-07-30 20:27:36 +02001580 * @link_watch_list: XXX: need comments on this one
1581 *
1582 * @reg_state: Register/unregister state machine
1583 * @dismantle: Device is going to be freed
1584 * @rtnl_link_state: This enum represents the phases of creating
1585 * a new link
1586 *
David S. Millercf124db2017-05-08 12:52:56 -04001587 * @needs_free_netdev: Should unregister perform free_netdev?
1588 * @priv_destructor: Called from unregister
Karoly Kemeny536721b2014-07-30 20:27:36 +02001589 * @npinfo: XXX: need comments on this one
1590 * @nd_net: Network namespace this network device is inside
1591 *
1592 * @ml_priv: Mid-layer private
1593 * @lstats: Loopback statistics
1594 * @tstats: Tunnel statistics
1595 * @dstats: Dummy statistics
1596 * @vstats: Virtual ethernet statistics
1597 *
1598 * @garp_port: GARP
1599 * @mrp_port: MRP
1600 *
1601 * @dev: Class/net/name entry
1602 * @sysfs_groups: Space for optional device, statistics and wireless
1603 * sysfs groups
1604 *
1605 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1606 * @rtnl_link_ops: Rtnl_link_ops
1607 *
1608 * @gso_max_size: Maximum size of generic segmentation offload
1609 * @gso_max_segs: Maximum number of segments that can be passed to the
1610 * NIC for GSO
1611 *
1612 * @dcbnl_ops: Data Center Bridging netlink ops
1613 * @num_tc: Number of traffic classes in the net device
1614 * @tc_to_txq: XXX: need comments on this one
Randy Dunlap920c1cd2016-11-21 18:28:36 -08001615 * @prio_tc_map: XXX: need comments on this one
Karoly Kemeny536721b2014-07-30 20:27:36 +02001616 *
1617 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1618 *
1619 * @priomap: XXX: need comments on this one
1620 * @phydev: Physical device may attach itself
1621 * for hardware timestamping
1622 *
Eric Dumazet123b3652016-06-08 07:22:49 -07001623 * @qdisc_tx_busylock: lockdep class annotating Qdisc->busylock spinlock
1624 * @qdisc_running_key: lockdep class annotating Qdisc->running seqcount
Karoly Kemeny536721b2014-07-30 20:27:36 +02001625 *
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001626 * @proto_down: protocol port state information can be sent to the
1627 * switch driver and used to set the phys state of the
1628 * switch port.
1629 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630 * FIXME: cleanup struct net_device such that network protocol info
1631 * moves out.
1632 */
1633
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08001634struct net_device {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001635 char name[IFNAMSIZ];
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001636 struct hlist_node name_hlist;
Stephen Hemminger0b815a12008-09-22 21:28:11 -07001637 char *ifalias;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001638 /*
1639 * I/O specific fields
1640 * FIXME: Merge these and struct ifmap into one
1641 */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001642 unsigned long mem_end;
1643 unsigned long mem_start;
1644 unsigned long base_addr;
1645 int irq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001646
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001647 atomic_t carrier_changes;
1648
Linus Torvalds1da177e2005-04-16 15:20:36 -07001649 /*
Karoly Kemeny536721b2014-07-30 20:27:36 +02001650 * Some hardware also needs these fields (state,dev_list,
1651 * napi_list,unreg_list,close_list) but they are not
Linus Torvalds1da177e2005-04-16 15:20:36 -07001652 * part of the usual set specified in Space.c.
1653 */
1654
Linus Torvalds1da177e2005-04-16 15:20:36 -07001655 unsigned long state;
1656
Pavel Emelianov7562f872007-05-03 15:13:45 -07001657 struct list_head dev_list;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07001658 struct list_head napi_list;
Eric Dumazet44a08732009-10-27 07:03:04 +00001659 struct list_head unreg_list;
Eric W. Biederman5cde2822013-10-05 19:26:05 -07001660 struct list_head close_list;
Salam Noureddine7866a622015-01-27 11:35:48 -08001661 struct list_head ptype_all;
1662 struct list_head ptype_specific;
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001663
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001664 struct {
1665 struct list_head upper;
1666 struct list_head lower;
1667 } adj_list;
1668
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001669 netdev_features_t features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001670 netdev_features_t hw_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001671 netdev_features_t wanted_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001672 netdev_features_t vlan_features;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001673 netdev_features_t hw_enc_features;
Simon Horman0d89d202013-05-23 21:02:52 +00001674 netdev_features_t mpls_features;
Alexander Duyck802ab552016-04-10 21:45:03 -04001675 netdev_features_t gso_partial_features;
Michał Mirosław04ed3e72011-01-24 15:32:47 -08001676
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677 int ifindex;
Nicolas Dichtel7a66bbc2015-04-02 17:07:09 +02001678 int group;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001679
Rusty Russellc45d2862007-03-28 14:29:08 -07001680 struct net_device_stats stats;
Eric Dumazet015f0682014-03-27 08:45:56 -07001681
Eric Dumazet015f0682014-03-27 08:45:56 -07001682 atomic_long_t rx_dropped;
1683 atomic_long_t tx_dropped;
Jarod Wilson6e7333d2016-02-01 18:51:05 -05001684 atomic_long_t rx_nohandler;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001685
Johannes Bergb86e0282007-04-26 20:48:23 -07001686#ifdef CONFIG_WIRELESS_EXT
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001687 const struct iw_handler_def *wireless_handlers;
1688 struct iw_public_data *wireless_data;
Johannes Bergb86e0282007-04-26 20:48:23 -07001689#endif
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001690 const struct net_device_ops *netdev_ops;
Stephen Hemminger76fd8592006-09-08 11:16:13 -07001691 const struct ethtool_ops *ethtool_ops;
Scott Feldman41706042015-03-15 21:07:14 -07001692#ifdef CONFIG_NET_SWITCHDEV
Jiri Pirko9d47c0a2015-05-10 09:47:47 -07001693 const struct switchdev_ops *switchdev_ops;
Scott Feldman41706042015-03-15 21:07:14 -07001694#endif
David Ahern1b69c6d2015-09-29 20:07:11 -07001695#ifdef CONFIG_NET_L3_MASTER_DEV
1696 const struct l3mdev_ops *l3mdev_ops;
1697#endif
Alexander Aringf997c552016-06-15 21:20:23 +02001698#if IS_ENABLED(CONFIG_IPV6)
1699 const struct ndisc_ops *ndisc_ops;
1700#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001701
Steffen Klassertd77e38e2017-04-14 10:06:10 +02001702#ifdef CONFIG_XFRM
1703 const struct xfrmdev_ops *xfrmdev_ops;
1704#endif
1705
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07001706 const struct header_ops *header_ops;
1707
Karoly Kemeny536721b2014-07-30 20:27:36 +02001708 unsigned int flags;
1709 unsigned int priv_flags;
1710
Linus Torvalds1da177e2005-04-16 15:20:36 -07001711 unsigned short gflags;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001712 unsigned short padded;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001713
Karoly Kemeny536721b2014-07-30 20:27:36 +02001714 unsigned char operstate;
1715 unsigned char link_mode;
Stefan Rompfb00055a2006-03-20 17:09:11 -08001716
Karoly Kemeny536721b2014-07-30 20:27:36 +02001717 unsigned char if_port;
1718 unsigned char dma;
Joe Perchesbdc220d2011-05-09 17:42:46 +00001719
Karoly Kemeny536721b2014-07-30 20:27:36 +02001720 unsigned int mtu;
Jarod Wilson61e84622016-10-07 22:04:33 -04001721 unsigned int min_mtu;
1722 unsigned int max_mtu;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001723 unsigned short type;
1724 unsigned short hard_header_len;
Alexey Dobriyand92be7a2017-04-10 11:25:26 +03001725 unsigned char min_header_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001726
Johannes Bergf5184d22008-05-12 20:48:31 -07001727 unsigned short needed_headroom;
1728 unsigned short needed_tailroom;
1729
Linus Torvalds1da177e2005-04-16 15:20:36 -07001730 /* Interface address info. */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001731 unsigned char perm_addr[MAX_ADDR_LEN];
1732 unsigned char addr_assign_type;
1733 unsigned char addr_len;
Sebastian Siewiora0a96632013-12-12 10:15:59 +01001734 unsigned short neigh_priv_len;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001735 unsigned short dev_id;
1736 unsigned short dev_port;
Jiri Pirkoccffad252009-05-22 23:22:17 +00001737 spinlock_t addr_list_lock;
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001738 unsigned char name_assign_type;
1739 bool uc_promisc;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001740 struct netdev_hw_addr_list uc;
1741 struct netdev_hw_addr_list mc;
1742 struct netdev_hw_addr_list dev_addrs;
1743
Eric Dumazet4c3d5e72013-03-30 06:31:03 +00001744#ifdef CONFIG_SYSFS
1745 struct kset *queues_kset;
1746#endif
Wang Chen9d45abe2008-06-17 21:12:48 -07001747 unsigned int promiscuity;
1748 unsigned int allmulti;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001749
Linus Torvalds1da177e2005-04-16 15:20:36 -07001750
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001751 /* Protocol-specific pointers */
Jesse Gross65ac6a52010-10-20 13:56:05 +00001752
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001753#if IS_ENABLED(CONFIG_VLAN_8021Q)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001754 struct vlan_info __rcu *vlan_info;
Jesse Gross65ac6a52010-10-20 13:56:05 +00001755#endif
Ben Hutchings34a430d2011-11-25 14:38:38 +00001756#if IS_ENABLED(CONFIG_NET_DSA)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001757 struct dsa_switch_tree *dsa_ptr;
Lennert Buytenhek91da11f2008-10-07 13:44:02 +00001758#endif
Ying Xue37cb0622013-12-10 20:45:41 -08001759#if IS_ENABLED(CONFIG_TIPC)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001760 struct tipc_bearer __rcu *tipc_ptr;
Ying Xue37cb0622013-12-10 20:45:41 -08001761#endif
Karoly Kemeny536721b2014-07-30 20:27:36 +02001762 void *atalk_ptr;
1763 struct in_device __rcu *ip_ptr;
1764 struct dn_dev __rcu *dn_ptr;
1765 struct inet6_dev __rcu *ip6_ptr;
1766 void *ax25_ptr;
1767 struct wireless_dev *ieee80211_ptr;
Alexander Aring98a18b62014-11-02 06:44:54 +01001768 struct wpan_dev *ieee802154_ptr;
Robert Shearman03c57742015-04-22 11:14:37 +01001769#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1770 struct mpls_dev __rcu *mpls_ptr;
1771#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001772
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001773/*
Eric Dumazetcd135392010-09-16 02:58:13 +00001774 * Cache lines mostly used on receive path (including eth_type_trans())
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001775 */
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001776 /* Interface address info used in eth_type_trans() */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001777 unsigned char *dev_addr;
Jiri Pirkof001fde2009-05-05 02:48:28 +00001778
Michael Daltona953be52014-01-16 22:23:28 -08001779#ifdef CONFIG_SYSFS
Tom Herbert0a9627f2010-03-16 08:03:29 +00001780 struct netdev_rx_queue *_rx;
1781
Tom Herbert0a9627f2010-03-16 08:03:29 +00001782 unsigned int num_rx_queues;
Ben Hutchings62fe0b42010-09-27 08:24:33 +00001783 unsigned int real_num_rx_queues;
Eric Dumazetdf334542010-03-24 19:13:54 +00001784#endif
Tom Herbert0a9627f2010-03-16 08:03:29 +00001785
Eric Dumazet7acedaf2017-04-25 11:36:52 -07001786 struct bpf_prog __rcu *xdp_prog;
Eric Dumazet3b47d302014-11-06 21:09:44 -08001787 unsigned long gro_flush_timeout;
stephen hemminger61391cd2010-11-15 06:38:12 +00001788 rx_handler_func_t __rcu *rx_handler;
1789 void __rcu *rx_handler_data;
David S. Millere8a04642008-07-17 00:34:19 -07001790
Daniel Borkmann4cda01e2015-05-11 19:28:49 +02001791#ifdef CONFIG_NET_CLS_ACT
Daniel Borkmannd2788d32015-05-09 22:51:32 +02001792 struct tcf_proto __rcu *ingress_cl_list;
1793#endif
Eric Dumazet24824a02010-10-02 06:11:55 +00001794 struct netdev_queue __rcu *ingress_queue;
Pablo Neirae687ad62015-05-13 18:19:38 +02001795#ifdef CONFIG_NETFILTER_INGRESS
Aaron Conolee3b37f12016-09-21 11:35:07 -04001796 struct nf_hook_entry __rcu *nf_hooks_ingress;
Pablo Neirae687ad62015-05-13 18:19:38 +02001797#endif
Daniel Borkmannd2788d32015-05-09 22:51:32 +02001798
Karoly Kemeny536721b2014-07-30 20:27:36 +02001799 unsigned char broadcast[MAX_ADDR_LEN];
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001800#ifdef CONFIG_RFS_ACCEL
1801 struct cpu_rmap *rx_cpu_rmap;
1802#endif
1803 struct hlist_node index_hlist;
Eric Dumazetcd135392010-09-16 02:58:13 +00001804
1805/*
1806 * Cache lines mostly used on transmit path
1807 */
David S. Millere8a04642008-07-17 00:34:19 -07001808 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1809 unsigned int num_tx_queues;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07001810 unsigned int real_num_tx_queues;
Patrick McHardyaf356af2009-09-04 06:41:18 +00001811 struct Qdisc *qdisc;
Jiri Kosina59cc1f62016-08-10 11:05:15 +02001812#ifdef CONFIG_NET_SCHED
1813 DECLARE_HASHTABLE (qdisc_hash, 4);
1814#endif
Alexey Dobriyan0cd29502017-05-17 13:30:44 +03001815 unsigned int tx_queue_len;
David S. Millerc3f26a22008-07-31 16:58:50 -07001816 spinlock_t tx_global_lock;
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001817 int watchdog_timeo;
Eric Dumazetcd135392010-09-16 02:58:13 +00001818
Tom Herbertbf264142010-11-26 08:36:09 +00001819#ifdef CONFIG_XPS
Eric Dumazeta4177862010-11-28 21:43:02 +00001820 struct xps_dev_maps __rcu *xps_maps;
Tom Herbertbf264142010-11-26 08:36:09 +00001821#endif
Daniel Borkmann1f211a12016-01-07 22:29:47 +01001822#ifdef CONFIG_NET_CLS_ACT
1823 struct tcf_proto __rcu *egress_cl_list;
1824#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -07001825
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001826 /* These may be needed for future network-power-down code. */
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001827 struct timer_list watchdog_timer;
1828
Eric Dumazet29b44332010-10-11 10:22:12 +00001829 int __percpu *pcpu_refcnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001830 struct list_head todo_list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001831
Eric Dumazete014deb2009-11-17 05:59:21 +00001832 struct list_head link_watch_list;
Herbert Xu572a1032007-05-08 18:34:17 -07001833
Linus Torvalds1da177e2005-04-16 15:20:36 -07001834 enum { NETREG_UNINITIALIZED=0,
Stephen Hemmingerb17a7c12006-05-10 13:21:17 -07001835 NETREG_REGISTERED, /* completed register_netdevice */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001836 NETREG_UNREGISTERING, /* called unregister_netdevice */
1837 NETREG_UNREGISTERED, /* completed unregister todo */
1838 NETREG_RELEASED, /* called free_netdev */
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08001839 NETREG_DUMMY, /* dummy device for NAPI poll */
Eric Dumazet449f4542011-05-19 12:24:16 +00001840 } reg_state:8;
1841
Karoly Kemeny536721b2014-07-30 20:27:36 +02001842 bool dismantle;
Patrick McHardya2835762010-02-26 06:34:51 +00001843
1844 enum {
1845 RTNL_LINK_INITIALIZED,
1846 RTNL_LINK_INITIALIZING,
1847 } rtnl_link_state:16;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848
David S. Millercf124db2017-05-08 12:52:56 -04001849 bool needs_free_netdev;
1850 void (*priv_destructor)(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001851
Linus Torvalds1da177e2005-04-16 15:20:36 -07001852#ifdef CONFIG_NETPOLL
Cong Wang5fbee842013-01-22 21:29:39 +00001853 struct netpoll_info __rcu *npinfo;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001854#endif
David S. Millereae792b2008-07-15 03:03:33 -07001855
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -05001856 possible_net_t nd_net;
Eric W. Biederman4a1c5372007-09-12 11:56:32 +02001857
David S. Miller49517042008-05-12 03:29:11 -07001858 /* mid-layer private */
Eric Dumazeta7855c72010-09-23 23:51:51 +00001859 union {
Karoly Kemeny536721b2014-07-30 20:27:36 +02001860 void *ml_priv;
1861 struct pcpu_lstats __percpu *lstats;
Li RongQing8f849852014-01-04 13:57:59 +08001862 struct pcpu_sw_netstats __percpu *tstats;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001863 struct pcpu_dstats __percpu *dstats;
1864 struct pcpu_vstats __percpu *vstats;
Eric Dumazeta7855c72010-09-23 23:51:51 +00001865 };
Karoly Kemeny536721b2014-07-30 20:27:36 +02001866
Tobias Klauserfb585b42017-02-10 16:43:50 +01001867#if IS_ENABLED(CONFIG_GARP)
Eric Dumazet3cc77ec2010-10-24 21:32:36 +00001868 struct garp_port __rcu *garp_port;
Tobias Klauserfb585b42017-02-10 16:43:50 +01001869#endif
1870#if IS_ENABLED(CONFIG_MRP)
David Wardfebf0182013-02-08 17:17:06 +00001871 struct mrp_port __rcu *mrp_port;
Tobias Klauserfb585b42017-02-10 16:43:50 +01001872#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001873
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001874 struct device dev;
Eric W. Biederman0c509a62009-10-29 14:18:21 +00001875 const struct attribute_group *sysfs_groups[4];
Michael Daltona953be52014-01-16 22:23:28 -08001876 const struct attribute_group *sysfs_rx_queue_group;
Patrick McHardy38f7b872007-06-13 12:03:51 -07001877
Patrick McHardy38f7b872007-06-13 12:03:51 -07001878 const struct rtnl_link_ops *rtnl_link_ops;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001879
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07001880 /* for setting kernel sock attribute on TCP connection setup */
1881#define GSO_MAX_SIZE 65536
1882 unsigned int gso_max_size;
Ben Hutchings30b678d2012-07-30 15:57:00 +00001883#define GSO_MAX_SEGS 65535
1884 u16 gso_max_segs;
Eric Dumazet743b03a2016-04-09 11:29:58 -07001885
Jeff Kirsher7a6b6f52008-11-25 01:02:08 -08001886#ifdef CONFIG_DCB
Stephen Hemminger32953542009-10-05 06:01:03 +00001887 const struct dcbnl_rtnl_ops *dcbnl_ops;
Alexander Duyck2f90b862008-11-20 20:52:10 -08001888#endif
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001889 u8 num_tc;
1890 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1891 u8 prio_tc_map[TC_BITMASK + 1];
Alexander Duyck2f90b862008-11-20 20:52:10 -08001892
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001893#if IS_ENABLED(CONFIG_FCOE)
Yi Zou4d288d52009-02-27 14:06:59 -08001894 unsigned int fcoe_ddp_xid;
1895#endif
Daniel Borkmann86f85152013-12-29 17:27:11 +01001896#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
Neil Horman5bc14212011-11-22 05:10:51 +00001897 struct netprio_map __rcu *priomap;
1898#endif
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001899 struct phy_device *phydev;
1900 struct lock_class_key *qdisc_tx_busylock;
Eric Dumazetf9eb8ae2016-06-06 09:37:15 -07001901 struct lock_class_key *qdisc_running_key;
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05001902 bool proto_down;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001903};
Greg Kroah-Hartman43cb76d2002-04-09 12:14:34 -07001904#define to_net_dev(d) container_of(d, struct net_device, dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001905
David S. Millerb5cdae32017-04-18 15:36:58 -04001906static inline bool netif_elide_gro(const struct net_device *dev)
1907{
1908 if (!(dev->features & NETIF_F_GRO) || dev->xdp_prog)
1909 return true;
1910 return false;
1911}
1912
Linus Torvalds1da177e2005-04-16 15:20:36 -07001913#define NETDEV_ALIGN 32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001914
David S. Millere8a04642008-07-17 00:34:19 -07001915static inline
John Fastabend4f57c082011-01-17 08:06:04 +00001916int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1917{
1918 return dev->prio_tc_map[prio & TC_BITMASK];
1919}
1920
1921static inline
1922int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1923{
1924 if (tc >= dev->num_tc)
1925 return -EINVAL;
1926
1927 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1928 return 0;
1929}
1930
Alexander Duyck8d059b02016-10-28 11:43:49 -04001931int netdev_txq_to_tc(struct net_device *dev, unsigned int txq);
Alexander Duyck9cf1f6a2016-10-28 11:43:20 -04001932void netdev_reset_tc(struct net_device *dev);
1933int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset);
1934int netdev_set_num_tc(struct net_device *dev, u8 num_tc);
John Fastabend4f57c082011-01-17 08:06:04 +00001935
1936static inline
1937int netdev_get_num_tc(struct net_device *dev)
1938{
1939 return dev->num_tc;
1940}
1941
1942static inline
David S. Millere8a04642008-07-17 00:34:19 -07001943struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1944 unsigned int index)
1945{
1946 return &dev->_tx[index];
1947}
1948
Daniel Borkmann10c51b56232014-08-27 11:11:27 +02001949static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1950 const struct sk_buff *skb)
1951{
1952 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1953}
1954
David S. Millere8a04642008-07-17 00:34:19 -07001955static inline void netdev_for_each_tx_queue(struct net_device *dev,
1956 void (*f)(struct net_device *,
1957 struct netdev_queue *,
1958 void *),
1959 void *arg)
1960{
1961 unsigned int i;
1962
1963 for (i = 0; i < dev->num_tx_queues; i++)
1964 f(dev, &dev->_tx[i], arg);
1965}
1966
Eric Dumazetd3fff6c2016-06-09 07:45:12 -07001967#define netdev_lockdep_set_classes(dev) \
1968{ \
1969 static struct lock_class_key qdisc_tx_busylock_key; \
1970 static struct lock_class_key qdisc_running_key; \
1971 static struct lock_class_key qdisc_xmit_lock_key; \
1972 static struct lock_class_key dev_addr_list_lock_key; \
1973 unsigned int i; \
1974 \
1975 (dev)->qdisc_tx_busylock = &qdisc_tx_busylock_key; \
1976 (dev)->qdisc_running_key = &qdisc_running_key; \
1977 lockdep_set_class(&(dev)->addr_list_lock, \
1978 &dev_addr_list_lock_key); \
1979 for (i = 0; i < (dev)->num_tx_queues; i++) \
1980 lockdep_set_class(&(dev)->_tx[i]._xmit_lock, \
1981 &qdisc_xmit_lock_key); \
1982}
1983
Joe Perchesf629d202013-09-26 14:48:15 -07001984struct netdev_queue *netdev_pick_tx(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08001985 struct sk_buff *skb,
1986 void *accel_priv);
Amerigo Wang8c4c49d2012-09-17 20:16:31 +00001987
Paolo Abeni871b6422016-02-26 10:45:37 +01001988/* returns the headroom that the master device needs to take in account
1989 * when forwarding to this dev
1990 */
1991static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1992{
1993 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1994}
1995
1996static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1997{
1998 if (dev->netdev_ops->ndo_set_rx_headroom)
1999 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
2000}
2001
2002/* set the device rx headroom to the dev's default */
2003static inline void netdev_reset_rx_headroom(struct net_device *dev)
2004{
2005 netdev_set_rx_headroom(dev, -1);
2006}
2007
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002008/*
2009 * Net namespace inlines
2010 */
2011static inline
2012struct net *dev_net(const struct net_device *dev)
2013{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00002014 return read_pnet(&dev->nd_net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002015}
2016
2017static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07002018void dev_net_set(struct net_device *dev, struct net *net)
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002019{
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -05002020 write_pnet(&dev->nd_net, net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002021}
2022
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002023/**
2024 * netdev_priv - access network device private data
2025 * @dev: network device
2026 *
2027 * Get network device private data
2028 */
Patrick McHardy6472ce62007-06-13 12:03:21 -07002029static inline void *netdev_priv(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002030{
Eric Dumazet1ce8e7b2009-05-27 04:42:37 +00002031 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002032}
2033
Linus Torvalds1da177e2005-04-16 15:20:36 -07002034/* Set the sysfs physical device reference for the network logical device
2035 * if set prior to registration will cause a symlink during initialization.
2036 */
Greg Kroah-Hartman43cb76d2002-04-09 12:14:34 -07002037#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002038
Marcel Holtmann384912e2009-08-31 21:08:19 +00002039/* Set the sysfs device type for the network logical device to allow
Maxime Jayat3f794102013-10-12 01:29:46 +02002040 * fine-grained identification of different network device types. For
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002041 * example Ethernet, Wireless LAN, Bluetooth, WiMAX etc.
Marcel Holtmann384912e2009-08-31 21:08:19 +00002042 */
2043#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2044
Eric Dumazet82dc3c63c2013-03-05 15:57:22 +00002045/* Default NAPI poll() weight
2046 * Device drivers are strongly advised to not use bigger value
2047 */
2048#define NAPI_POLL_WEIGHT 64
2049
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002050/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002051 * netif_napi_add - initialize a NAPI context
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002052 * @dev: network device
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002053 * @napi: NAPI context
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002054 * @poll: polling function
2055 * @weight: default weight
2056 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002057 * netif_napi_add() must be used to initialize a NAPI context prior to calling
2058 * *any* of the other NAPI-related functions.
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002059 */
Herbert Xud565b0a2008-12-15 23:38:52 -08002060void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2061 int (*poll)(struct napi_struct *, int), int weight);
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002062
Alexander Duyckd8156532008-07-08 15:13:05 -07002063/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002064 * netif_tx_napi_add - initialize a NAPI context
Eric Dumazetd64b5e82015-11-18 06:31:00 -08002065 * @dev: network device
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002066 * @napi: NAPI context
Eric Dumazetd64b5e82015-11-18 06:31:00 -08002067 * @poll: polling function
2068 * @weight: default weight
2069 *
2070 * This variant of netif_napi_add() should be used from drivers using NAPI
2071 * to exclusively poll a TX queue.
2072 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2073 */
2074static inline void netif_tx_napi_add(struct net_device *dev,
2075 struct napi_struct *napi,
2076 int (*poll)(struct napi_struct *, int),
2077 int weight)
2078{
2079 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2080 netif_napi_add(dev, napi, poll, weight);
2081}
2082
2083/**
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002084 * netif_napi_del - remove a NAPI context
2085 * @napi: NAPI context
Alexander Duyckd8156532008-07-08 15:13:05 -07002086 *
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002087 * netif_napi_del() removes a NAPI context from the network device NAPI list
Alexander Duyckd8156532008-07-08 15:13:05 -07002088 */
Herbert Xud565b0a2008-12-15 23:38:52 -08002089void netif_napi_del(struct napi_struct *napi);
2090
2091struct napi_gro_cb {
Herbert Xu78a478d2009-05-26 18:50:21 +00002092 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002093 void *frag0;
Herbert Xu78a478d2009-05-26 18:50:21 +00002094
Herbert Xu74895942009-05-26 18:50:27 +00002095 /* Length of frag0. */
2096 unsigned int frag0_len;
2097
Herbert Xu86911732009-01-29 14:19:50 +00002098 /* This indicates where we are processing relative to skb->data. */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002099 int data_offset;
Herbert Xu86911732009-01-29 14:19:50 +00002100
Herbert Xud565b0a2008-12-15 23:38:52 -08002101 /* This is non-zero if the packet cannot be merged with the new skb. */
Jerry Chubf5a7552014-01-07 10:23:19 -08002102 u16 flush;
2103
2104 /* Save the IP ID here and check when we get to the transport layer */
2105 u16 flush_id;
Herbert Xud565b0a2008-12-15 23:38:52 -08002106
2107 /* Number of segments aggregated. */
Eric Dumazet2e71a6f2012-10-06 08:08:49 +00002108 u16 count;
2109
Tom Herbert15e23962015-02-10 16:30:31 -08002110 /* Start offset for remote checksum offload */
2111 u16 gro_remcsum_start;
2112
Eric Dumazet2e71a6f2012-10-06 08:08:49 +00002113 /* jiffies when first packet was created/queued */
2114 unsigned long age;
Eric Dumazet86347242012-10-08 21:38:50 +02002115
Tom Herbertafe93322014-09-17 12:25:57 -07002116 /* Used in ipv6_gro_receive() and foo-over-udp */
Or Gerlitzb582ef02014-01-20 13:59:19 +02002117 u16 proto;
2118
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002119 /* This is non-zero if the packet may be of the same flow. */
2120 u8 same_flow:1;
2121
Jesse Grossfac8e0f2016-03-19 09:32:01 -07002122 /* Used in tunnel GRO receive */
2123 u8 encap_mark:1;
Tom Herbert573e8fc2014-08-22 13:33:47 -07002124
2125 /* GRO checksum is valid */
2126 u8 csum_valid:1;
2127
Tom Herbert662880f2014-08-27 21:26:56 -07002128 /* Number of checksums via CHECKSUM_UNNECESSARY */
2129 u8 csum_cnt:3;
Eric Dumazetc3c7c252012-12-06 13:54:59 +00002130
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002131 /* Free the skb? */
2132 u8 free:2;
2133#define NAPI_GRO_FREE 1
2134#define NAPI_GRO_FREE_STOLEN_HEAD 2
2135
Tom Herbertefc98d02014-10-03 15:48:08 -07002136 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2137 u8 is_ipv6:1;
2138
Alexander Duycka0ca1532016-04-05 09:13:39 -07002139 /* Used in GRE, set in fou/gue_gro_receive */
2140 u8 is_fou:1;
2141
Alexander Duyck15305452016-04-10 21:44:57 -04002142 /* Used to determine if flush_id can be ignored */
2143 u8 is_atomic:1;
2144
Sabrina Dubrocafcd91dd2016-10-20 15:58:02 +02002145 /* Number of gro_receive callbacks this packet already went through */
2146 u8 recursion_counter:4;
2147
2148 /* 1 bit hole */
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002149
Jerry Chubf5a7552014-01-07 10:23:19 -08002150 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2151 __wsum csum;
2152
Eric Dumazetc3c7c252012-12-06 13:54:59 +00002153 /* used in skb_gro_receive() slow path */
2154 struct sk_buff *last;
Herbert Xud565b0a2008-12-15 23:38:52 -08002155};
2156
2157#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
Alexander Duyckd8156532008-07-08 15:13:05 -07002158
Sabrina Dubrocafcd91dd2016-10-20 15:58:02 +02002159#define GRO_RECURSION_LIMIT 15
2160static inline int gro_recursion_inc_test(struct sk_buff *skb)
2161{
2162 return ++NAPI_GRO_CB(skb)->recursion_counter == GRO_RECURSION_LIMIT;
2163}
2164
2165typedef struct sk_buff **(*gro_receive_t)(struct sk_buff **, struct sk_buff *);
2166static inline struct sk_buff **call_gro_receive(gro_receive_t cb,
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(head, skb);
2176}
2177
2178typedef struct sk_buff **(*gro_receive_sk_t)(struct sock *, struct sk_buff **,
2179 struct sk_buff *);
2180static inline struct sk_buff **call_gro_receive_sk(gro_receive_sk_t cb,
2181 struct sock *sk,
2182 struct sk_buff **head,
2183 struct sk_buff *skb)
2184{
2185 if (unlikely(gro_recursion_inc_test(skb))) {
2186 NAPI_GRO_CB(skb)->flush |= 1;
2187 return NULL;
2188 }
2189
2190 return cb(sk, head, skb);
2191}
2192
Linus Torvalds1da177e2005-04-16 15:20:36 -07002193struct packet_type {
David S. Millerf2ccd8f2005-08-09 19:34:12 -07002194 __be16 type; /* This is really htons(ether_type). */
2195 struct net_device *dev; /* NULL is wildcarded here */
2196 int (*func) (struct sk_buff *,
2197 struct net_device *,
2198 struct packet_type *,
2199 struct net_device *);
Eric Leblondc0de08d2012-08-16 22:02:58 +00002200 bool (*id_match)(struct packet_type *ptype,
2201 struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002202 void *af_packet_priv;
2203 struct list_head list;
2204};
2205
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002206struct offload_callbacks {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002207 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2208 netdev_features_t features);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002209 struct sk_buff **(*gro_receive)(struct sk_buff **head,
Tom Herberta2b12f32015-01-12 17:00:37 -08002210 struct sk_buff *skb);
Jerry Chu299603e82013-12-11 20:53:45 -08002211 int (*gro_complete)(struct sk_buff *skb, int nhoff);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002212};
2213
2214struct packet_offload {
2215 __be16 type; /* This is really htons(ether_type). */
David S. Millerbdef7de2015-06-01 14:56:09 -07002216 u16 priority;
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002217 struct offload_callbacks callbacks;
2218 struct list_head list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002219};
2220
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002221/* often modified stats are per-CPU, other are shared (netdev->stats) */
Li RongQing8f849852014-01-04 13:57:59 +08002222struct pcpu_sw_netstats {
2223 u64 rx_packets;
2224 u64 rx_bytes;
2225 u64 tx_packets;
2226 u64 tx_bytes;
2227 struct u64_stats_sync syncp;
2228};
2229
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002230#define __netdev_alloc_pcpu_stats(type, gfp) \
2231({ \
2232 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2233 if (pcpu_stats) { \
2234 int __cpu; \
2235 for_each_possible_cpu(__cpu) { \
2236 typeof(type) *stat; \
2237 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2238 u64_stats_init(&stat->syncp); \
2239 } \
2240 } \
2241 pcpu_stats; \
WANG Cong1c213bd2014-02-13 11:46:28 -08002242})
2243
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002244#define netdev_alloc_pcpu_stats(type) \
Felix Fietkau326fcfa2015-12-05 13:58:11 +01002245 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002246
Jiri Pirko764f5e52015-12-03 12:12:12 +01002247enum netdev_lag_tx_type {
2248 NETDEV_LAG_TX_TYPE_UNKNOWN,
2249 NETDEV_LAG_TX_TYPE_RANDOM,
2250 NETDEV_LAG_TX_TYPE_BROADCAST,
2251 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2252 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2253 NETDEV_LAG_TX_TYPE_HASH,
2254};
2255
2256struct netdev_lag_upper_info {
2257 enum netdev_lag_tx_type tx_type;
2258};
2259
Jiri Pirkofb1b2e32015-12-03 12:12:16 +01002260struct netdev_lag_lower_state_info {
2261 u8 link_up : 1,
2262 tx_enabled : 1;
2263};
2264
Linus Torvalds1da177e2005-04-16 15:20:36 -07002265#include <linux/notifier.h>
2266
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002267/* netdevice notifier chain. Please remember to update the rtnetlink
2268 * notification exclusion list in rtnetlink_event() when adding new
2269 * types.
2270 */
2271#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2272#define NETDEV_DOWN 0x0002
2273#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2274 detected a hardware crash and restarted
2275 - we can use this eg to kick tcp sessions
2276 once done */
2277#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2278#define NETDEV_REGISTER 0x0005
2279#define NETDEV_UNREGISTER 0x0006
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002280#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002281#define NETDEV_CHANGEADDR 0x0008
2282#define NETDEV_GOING_DOWN 0x0009
2283#define NETDEV_CHANGENAME 0x000A
2284#define NETDEV_FEAT_CHANGE 0x000B
2285#define NETDEV_BONDING_FAILOVER 0x000C
2286#define NETDEV_PRE_UP 0x000D
2287#define NETDEV_PRE_TYPE_CHANGE 0x000E
2288#define NETDEV_POST_TYPE_CHANGE 0x000F
2289#define NETDEV_POST_INIT 0x0010
Eric Dumazet0115e8e2012-08-22 17:19:46 +00002290#define NETDEV_UNREGISTER_FINAL 0x0011
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002291#define NETDEV_RELEASE 0x0012
2292#define NETDEV_NOTIFY_PEERS 0x0013
2293#define NETDEV_JOIN 0x0014
Jiri Pirko42e52bf2013-05-25 04:12:10 +00002294#define NETDEV_CHANGEUPPER 0x0015
Jiri Pirko4aa5dee2013-07-20 12:13:53 +02002295#define NETDEV_RESEND_IGMP 0x0016
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002296#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
Jiri Pirkod4261e52014-08-19 16:02:12 +02002297#define NETDEV_CHANGEINFODATA 0x0018
Moni Shoua61bd3852015-02-03 16:48:29 +02002298#define NETDEV_BONDING_INFO 0x0019
Jiri Pirko573c7ba2015-10-16 14:01:22 +02002299#define NETDEV_PRECHANGEUPPER 0x001A
Jiri Pirko04d48262015-12-03 12:12:15 +01002300#define NETDEV_CHANGELOWERSTATE 0x001B
Alexander Duyck7c46a642016-06-16 12:21:00 -07002301#define NETDEV_UDP_TUNNEL_PUSH_INFO 0x001C
Sabrina Dubroca296d8ee2017-07-21 12:49:30 +02002302#define NETDEV_UDP_TUNNEL_DROP_INFO 0x001D
Jason Wang08294a22016-06-30 14:45:35 +08002303#define NETDEV_CHANGE_TX_QUEUE_LEN 0x001E
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002304
Joe Perchesf629d202013-09-26 14:48:15 -07002305int register_netdevice_notifier(struct notifier_block *nb);
2306int unregister_netdevice_notifier(struct notifier_block *nb);
Jiri Pirko351638e2013-05-28 01:30:21 +00002307
2308struct netdev_notifier_info {
2309 struct net_device *dev;
2310};
2311
Jiri Pirkobe9efd32013-05-28 01:30:22 +00002312struct netdev_notifier_change_info {
2313 struct netdev_notifier_info info; /* must be first */
2314 unsigned int flags_changed;
2315};
2316
Jiri Pirko0e4ead92015-08-27 09:31:18 +02002317struct netdev_notifier_changeupper_info {
2318 struct netdev_notifier_info info; /* must be first */
2319 struct net_device *upper_dev; /* new upper dev */
2320 bool master; /* is upper dev master */
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002321 bool linking; /* is the notification for link or unlink */
Jiri Pirko29bf24a2015-12-03 12:12:11 +01002322 void *upper_info; /* upper dev info */
Jiri Pirko0e4ead92015-08-27 09:31:18 +02002323};
2324
Jiri Pirko04d48262015-12-03 12:12:15 +01002325struct netdev_notifier_changelowerstate_info {
2326 struct netdev_notifier_info info; /* must be first */
2327 void *lower_state_info; /* is lower dev state */
2328};
2329
Cong Wang75538c22013-05-29 11:30:50 +08002330static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2331 struct net_device *dev)
2332{
2333 info->dev = dev;
2334}
2335
Jiri Pirko351638e2013-05-28 01:30:21 +00002336static inline struct net_device *
2337netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2338{
2339 return info->dev;
2340}
2341
Joe Perchesf629d202013-09-26 14:48:15 -07002342int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002343
2344
Linus Torvalds1da177e2005-04-16 15:20:36 -07002345extern rwlock_t dev_base_lock; /* Device list lock */
2346
Eric W. Biederman881d9662007-09-17 11:56:21 -07002347#define for_each_netdev(net, d) \
2348 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
Eric W. Biedermandcbccbd42009-11-29 22:25:26 +00002349#define for_each_netdev_reverse(net, d) \
2350 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
Eric Dumazetc6d14c82009-11-04 05:43:23 -08002351#define for_each_netdev_rcu(net, d) \
2352 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
Eric W. Biederman881d9662007-09-17 11:56:21 -07002353#define for_each_netdev_safe(net, d, n) \
2354 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2355#define for_each_netdev_continue(net, d) \
2356 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
stephen hemminger254245d2009-11-10 07:54:47 +00002357#define for_each_netdev_continue_rcu(net, d) \
2358 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00002359#define for_each_netdev_in_bond_rcu(bond, slave) \
2360 for_each_netdev_rcu(&init_net, slave) \
Benjamin Poirier4ccce022015-01-14 16:52:35 +09002361 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
Pavel Emelianov7562f872007-05-03 15:13:45 -07002362#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2363
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002364static inline struct net_device *next_net_device(struct net_device *dev)
2365{
2366 struct list_head *lh;
2367 struct net *net;
Pavel Emelianov7562f872007-05-03 15:13:45 -07002368
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002369 net = dev_net(dev);
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002370 lh = dev->dev_list.next;
2371 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2372}
2373
Eric Dumazetce81b762009-11-11 17:34:30 +00002374static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2375{
2376 struct list_head *lh;
2377 struct net *net;
2378
2379 net = dev_net(dev);
Eric Dumazetccf43432011-01-26 18:08:02 +00002380 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
Eric Dumazetce81b762009-11-11 17:34:30 +00002381 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2382}
2383
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002384static inline struct net_device *first_net_device(struct net *net)
2385{
2386 return list_empty(&net->dev_base_head) ? NULL :
2387 net_device_entry(net->dev_base_head.next);
2388}
Pavel Emelianov7562f872007-05-03 15:13:45 -07002389
Eric Dumazetccf43432011-01-26 18:08:02 +00002390static inline struct net_device *first_net_device_rcu(struct net *net)
2391{
2392 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2393
2394 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2395}
2396
Joe Perchesf629d202013-09-26 14:48:15 -07002397int netdev_boot_setup_check(struct net_device *dev);
2398unsigned long netdev_boot_base(const char *prefix, int unit);
2399struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2400 const char *hwaddr);
2401struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2402struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2403void dev_add_pack(struct packet_type *pt);
2404void dev_remove_pack(struct packet_type *pt);
2405void __dev_remove_pack(struct packet_type *pt);
2406void dev_add_offload(struct packet_offload *po);
2407void dev_remove_offload(struct packet_offload *po);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002408
Nicolas Dichtela54acb32015-04-02 17:07:00 +02002409int dev_get_iflink(const struct net_device *dev);
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07002410int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
WANG Cong6c555492014-09-11 15:35:09 -07002411struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2412 unsigned short mask);
Joe Perchesf629d202013-09-26 14:48:15 -07002413struct net_device *dev_get_by_name(struct net *net, const char *name);
2414struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2415struct net_device *__dev_get_by_name(struct net *net, const char *name);
2416int dev_alloc_name(struct net_device *dev, const char *name);
2417int dev_open(struct net_device *dev);
stephen hemminger7051b882017-07-18 15:59:27 -07002418void dev_close(struct net_device *dev);
2419void dev_close_many(struct list_head *head, bool unlink);
Joe Perchesf629d202013-09-26 14:48:15 -07002420void dev_disable_lro(struct net_device *dev);
Eric W. Biederman0c4b51f2015-09-15 20:04:18 -05002421int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
Eric W. Biederman2b4aa3c2015-09-15 20:04:07 -05002422int dev_queue_xmit(struct sk_buff *skb);
Jason Wangf663dd92014-01-10 16:18:26 +08002423int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
Joe Perchesf629d202013-09-26 14:48:15 -07002424int register_netdevice(struct net_device *dev);
2425void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2426void unregister_netdevice_many(struct list_head *head);
Eric Dumazet44a08732009-10-27 07:03:04 +00002427static inline void unregister_netdevice(struct net_device *dev)
2428{
2429 unregister_netdevice_queue(dev, NULL);
2430}
2431
Joe Perchesf629d202013-09-26 14:48:15 -07002432int netdev_refcnt_read(const struct net_device *dev);
2433void free_netdev(struct net_device *dev);
Eric Dumazet74d332c2013-10-30 13:10:44 -07002434void netdev_freemem(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002435void synchronize_net(void);
2436int init_dummy_netdev(struct net_device *dev);
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08002437
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +02002438DECLARE_PER_CPU(int, xmit_recursion);
Daniel Borkmanna70b5062016-06-10 21:19:06 +02002439#define XMIT_RECURSION_LIMIT 10
2440
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +02002441static inline int dev_recursion_level(void)
2442{
2443 return this_cpu_read(xmit_recursion);
2444}
2445
Joe Perchesf629d202013-09-26 14:48:15 -07002446struct net_device *dev_get_by_index(struct net *net, int ifindex);
2447struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2448struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
Miroslav Lichvar90b602f2017-05-19 17:52:37 +02002449struct net_device *dev_get_by_napi_id(unsigned int napi_id);
Joe Perchesf629d202013-09-26 14:48:15 -07002450int netdev_get_name(struct net *net, char *name, int ifindex);
2451int dev_restart(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002452int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
Herbert Xu86911732009-01-29 14:19:50 +00002453
2454static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2455{
2456 return NAPI_GRO_CB(skb)->data_offset;
2457}
2458
2459static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2460{
2461 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2462}
2463
2464static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2465{
2466 NAPI_GRO_CB(skb)->data_offset += len;
2467}
2468
Herbert Xua5b1cf22009-05-26 18:50:28 +00002469static inline void *skb_gro_header_fast(struct sk_buff *skb,
2470 unsigned int offset)
Herbert Xu86911732009-01-29 14:19:50 +00002471{
Herbert Xu78a478d2009-05-26 18:50:21 +00002472 return NAPI_GRO_CB(skb)->frag0 + offset;
Herbert Xu86911732009-01-29 14:19:50 +00002473}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002474
Herbert Xua5b1cf22009-05-26 18:50:28 +00002475static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2476{
2477 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2478}
2479
Herbert Xu57ea52a2017-01-10 12:24:15 -08002480static inline void skb_gro_frag0_invalidate(struct sk_buff *skb)
2481{
2482 NAPI_GRO_CB(skb)->frag0 = NULL;
2483 NAPI_GRO_CB(skb)->frag0_len = 0;
2484}
2485
Herbert Xua5b1cf22009-05-26 18:50:28 +00002486static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2487 unsigned int offset)
2488{
Herbert Xu17dd7592011-07-27 06:16:28 -07002489 if (!pskb_may_pull(skb, hlen))
2490 return NULL;
2491
Herbert Xu57ea52a2017-01-10 12:24:15 -08002492 skb_gro_frag0_invalidate(skb);
Herbert Xu17dd7592011-07-27 06:16:28 -07002493 return skb->data + offset;
Herbert Xua5b1cf22009-05-26 18:50:28 +00002494}
2495
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002496static inline void *skb_gro_network_header(struct sk_buff *skb)
2497{
Herbert Xu78d3fd02009-05-26 18:50:23 +00002498 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2499 skb_network_offset(skb);
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002500}
2501
Jerry Chubf5a7552014-01-07 10:23:19 -08002502static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2503 const void *start, unsigned int len)
2504{
Tom Herbert573e8fc2014-08-22 13:33:47 -07002505 if (NAPI_GRO_CB(skb)->csum_valid)
Jerry Chubf5a7552014-01-07 10:23:19 -08002506 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2507 csum_partial(start, len, 0));
2508}
2509
Tom Herbert573e8fc2014-08-22 13:33:47 -07002510/* GRO checksum functions. These are logical equivalents of the normal
2511 * checksum functions (in skbuff.h) except that they operate on the GRO
2512 * offsets and fields in sk_buff.
2513 */
2514
2515__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2516
Tom Herbert15e23962015-02-10 16:30:31 -08002517static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2518{
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002519 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
Tom Herbert15e23962015-02-10 16:30:31 -08002520}
2521
Tom Herbert573e8fc2014-08-22 13:33:47 -07002522static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2523 bool zero_okay,
2524 __sum16 check)
2525{
Tom Herbert6edec0e2015-02-10 16:30:28 -08002526 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2527 skb_checksum_start_offset(skb) <
2528 skb_gro_offset(skb)) &&
Tom Herbert15e23962015-02-10 16:30:31 -08002529 !skb_at_gro_remcsum_start(skb) &&
Tom Herbert662880f2014-08-27 21:26:56 -07002530 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
Tom Herbert573e8fc2014-08-22 13:33:47 -07002531 (!zero_okay || check));
2532}
2533
2534static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2535 __wsum psum)
2536{
2537 if (NAPI_GRO_CB(skb)->csum_valid &&
2538 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2539 return 0;
2540
2541 NAPI_GRO_CB(skb)->csum = psum;
2542
2543 return __skb_gro_checksum_complete(skb);
2544}
2545
Tom Herbert573e8fc2014-08-22 13:33:47 -07002546static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2547{
Tom Herbert662880f2014-08-27 21:26:56 -07002548 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2549 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2550 NAPI_GRO_CB(skb)->csum_cnt--;
2551 } else {
2552 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2553 * verified a new top level checksum or an encapsulated one
2554 * during GRO. This saves work if we fallback to normal path.
2555 */
2556 __skb_incr_checksum_unnecessary(skb);
Tom Herbert573e8fc2014-08-22 13:33:47 -07002557 }
2558}
2559
2560#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2561 compute_pseudo) \
2562({ \
2563 __sum16 __ret = 0; \
2564 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2565 __ret = __skb_gro_checksum_validate_complete(skb, \
2566 compute_pseudo(skb, proto)); \
Davide Caratti219f1d792017-05-18 15:44:39 +02002567 if (!__ret) \
Tom Herbert573e8fc2014-08-22 13:33:47 -07002568 skb_gro_incr_csum_unnecessary(skb); \
2569 __ret; \
2570})
2571
2572#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2573 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2574
2575#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2576 compute_pseudo) \
2577 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2578
2579#define skb_gro_checksum_simple_validate(skb) \
2580 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2581
Tom Herbertd96535a2014-08-31 15:12:42 -07002582static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2583{
2584 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2585 !NAPI_GRO_CB(skb)->csum_valid);
2586}
2587
2588static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2589 __sum16 check, __wsum pseudo)
2590{
2591 NAPI_GRO_CB(skb)->csum = ~pseudo;
2592 NAPI_GRO_CB(skb)->csum_valid = 1;
2593}
2594
2595#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2596do { \
2597 if (__skb_gro_checksum_convert_check(skb)) \
2598 __skb_gro_checksum_convert(skb, check, \
2599 compute_pseudo(skb, proto)); \
2600} while (0)
2601
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002602struct gro_remcsum {
2603 int offset;
2604 __wsum delta;
2605};
2606
2607static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2608{
Geert Uytterhoeven846cd662015-02-18 11:38:06 +01002609 grc->offset = 0;
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002610 grc->delta = 0;
2611}
2612
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002613static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2614 unsigned int off, size_t hdrlen,
2615 int start, int offset,
2616 struct gro_remcsum *grc,
2617 bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08002618{
2619 __wsum delta;
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002620 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
Tom Herbertdcdc8992015-02-02 16:07:34 -08002621
2622 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2623
Tom Herbert15e23962015-02-10 16:30:31 -08002624 if (!nopartial) {
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002625 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2626 return ptr;
Tom Herbert15e23962015-02-10 16:30:31 -08002627 }
2628
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002629 ptr = skb_gro_header_fast(skb, off);
2630 if (skb_gro_header_hard(skb, off + plen)) {
2631 ptr = skb_gro_header_slow(skb, off + plen, off);
2632 if (!ptr)
2633 return NULL;
2634 }
2635
2636 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2637 start, offset);
Tom Herbertdcdc8992015-02-02 16:07:34 -08002638
2639 /* Adjust skb->csum since we changed the packet */
Tom Herbertdcdc8992015-02-02 16:07:34 -08002640 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002641
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002642 grc->offset = off + hdrlen + offset;
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002643 grc->delta = delta;
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002644
2645 return ptr;
Tom Herbertdcdc8992015-02-02 16:07:34 -08002646}
2647
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002648static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2649 struct gro_remcsum *grc)
2650{
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002651 void *ptr;
2652 size_t plen = grc->offset + sizeof(u16);
2653
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002654 if (!grc->delta)
2655 return;
2656
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002657 ptr = skb_gro_header_fast(skb, grc->offset);
2658 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2659 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2660 if (!ptr)
2661 return;
2662 }
2663
2664 remcsum_unadjust((__sum16 *)ptr, grc->delta);
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002665}
Tom Herbertdcdc8992015-02-02 16:07:34 -08002666
Steffen Klassert25393d32017-02-15 09:39:44 +01002667#ifdef CONFIG_XFRM_OFFLOAD
2668static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2669{
2670 if (PTR_ERR(pp) != -EINPROGRESS)
2671 NAPI_GRO_CB(skb)->flush |= flush;
2672}
2673#else
Steffen Klassert5f114162017-02-15 09:39:39 +01002674static inline void skb_gro_flush_final(struct sk_buff *skb, struct sk_buff **pp, int flush)
2675{
2676 NAPI_GRO_CB(skb)->flush |= flush;
2677}
Steffen Klassert25393d32017-02-15 09:39:44 +01002678#endif
Steffen Klassert5f114162017-02-15 09:39:39 +01002679
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002680static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2681 unsigned short type,
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002682 const void *daddr, const void *saddr,
Eric Dumazet95c96172012-04-15 05:58:06 +00002683 unsigned int len)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002684{
Ursula Braunf1ecfd52007-10-22 16:16:14 +02002685 if (!dev->header_ops || !dev->header_ops->create)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002686 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002687
2688 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002689}
2690
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002691static inline int dev_parse_header(const struct sk_buff *skb,
2692 unsigned char *haddr)
2693{
2694 const struct net_device *dev = skb->dev;
2695
Patrick McHardy1b833362007-10-18 05:09:28 -07002696 if (!dev->header_ops || !dev->header_ops->parse)
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002697 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002698 return dev->header_ops->parse(skb, haddr);
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002699}
2700
Willem de Bruijn2793a232016-03-09 21:58:32 -05002701/* ll_header must have at least hard_header_len allocated */
2702static inline bool dev_validate_header(const struct net_device *dev,
2703 char *ll_header, int len)
2704{
2705 if (likely(len >= dev->hard_header_len))
2706 return true;
Willem de Bruijn217e6fa2017-02-07 15:57:20 -05002707 if (len < dev->min_header_len)
2708 return false;
Willem de Bruijn2793a232016-03-09 21:58:32 -05002709
2710 if (capable(CAP_SYS_RAWIO)) {
2711 memset(ll_header + len, 0, dev->hard_header_len - len);
2712 return true;
2713 }
2714
2715 if (dev->header_ops && dev->header_ops->validate)
2716 return dev->header_ops->validate(ll_header, len);
2717
2718 return false;
2719}
2720
Linus Torvalds1da177e2005-04-16 15:20:36 -07002721typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
Joe Perchesf629d202013-09-26 14:48:15 -07002722int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002723static inline int unregister_gifconf(unsigned int family)
2724{
2725 return register_gifconf(family, NULL);
2726}
2727
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002728#ifdef CONFIG_NET_FLOW_LIMIT
Willem de Bruijn5f121b92013-06-13 15:29:38 -04002729#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002730struct sd_flow_limit {
2731 u64 count;
2732 unsigned int num_buckets;
2733 unsigned int history_head;
2734 u16 history[FLOW_LIMIT_HISTORY];
2735 u8 buckets[];
2736};
2737
2738extern int netdev_flow_limit_table_len;
2739#endif /* CONFIG_NET_FLOW_LIMIT */
2740
Linus Torvalds1da177e2005-04-16 15:20:36 -07002741/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002742 * Incoming packets are placed on per-CPU queues
Linus Torvalds1da177e2005-04-16 15:20:36 -07002743 */
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08002744struct softnet_data {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002745 struct list_head poll_list;
Changli Gao6e7676c2010-04-27 15:07:33 -07002746 struct sk_buff_head process_queue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002747
Changli Gaodee42872010-05-02 05:42:16 +00002748 /* stats */
David S. Millercd7b5392010-05-02 22:27:59 -07002749 unsigned int processed;
2750 unsigned int time_squeeze;
David S. Millercd7b5392010-05-02 22:27:59 -07002751 unsigned int received_rps;
Changli Gaofd793d82010-04-15 00:16:59 -07002752#ifdef CONFIG_RPS
Eric Dumazet88751272010-04-19 05:07:33 +00002753 struct softnet_data *rps_ipi_list;
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002754#endif
2755#ifdef CONFIG_NET_FLOW_LIMIT
2756 struct sd_flow_limit __rcu *flow_limit;
2757#endif
2758 struct Qdisc *output_queue;
2759 struct Qdisc **output_queue_tailp;
2760 struct sk_buff *completion_queue;
Eric Dumazet88751272010-04-19 05:07:33 +00002761
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002762#ifdef CONFIG_RPS
Eric Dumazet501e7ef2016-04-26 15:30:07 -07002763 /* input_queue_head should be written by cpu owning this struct,
2764 * and only read by other cpus. Worth using a cache line.
2765 */
2766 unsigned int input_queue_head ____cacheline_aligned_in_smp;
2767
2768 /* Elements below can be accessed between CPUs for RPS/RFS */
Tom Herbert0a9627f2010-03-16 08:03:29 +00002769 struct call_single_data csd ____cacheline_aligned_in_smp;
Eric Dumazet88751272010-04-19 05:07:33 +00002770 struct softnet_data *rps_ipi_next;
2771 unsigned int cpu;
Tom Herbert76cc8b12010-05-20 18:37:59 +00002772 unsigned int input_queue_tail;
Tom Herbert1e94d722010-03-18 17:45:44 -07002773#endif
Eric Dumazet95c96172012-04-15 05:58:06 +00002774 unsigned int dropped;
Tom Herbert0a9627f2010-03-16 08:03:29 +00002775 struct sk_buff_head input_pkt_queue;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002776 struct napi_struct backlog;
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002777
Linus Torvalds1da177e2005-04-16 15:20:36 -07002778};
2779
Tom Herbert76cc8b12010-05-20 18:37:59 +00002780static inline void input_queue_head_incr(struct softnet_data *sd)
Tom Herbertfec5e652010-04-16 16:01:27 -07002781{
2782#ifdef CONFIG_RPS
Tom Herbert76cc8b12010-05-20 18:37:59 +00002783 sd->input_queue_head++;
2784#endif
2785}
2786
2787static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2788 unsigned int *qtail)
2789{
2790#ifdef CONFIG_RPS
2791 *qtail = ++sd->input_queue_tail;
Tom Herbertfec5e652010-04-16 16:01:27 -07002792#endif
2793}
2794
Tom Herbert0a9627f2010-03-16 08:03:29 +00002795DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002796
Joe Perchesf629d202013-09-26 14:48:15 -07002797void __netif_schedule(struct Qdisc *q);
John Fastabend46e5da40a2014-09-12 20:04:52 -07002798void netif_schedule_queue(struct netdev_queue *txq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002799
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002800static inline void netif_tx_schedule_all(struct net_device *dev)
2801{
2802 unsigned int i;
2803
2804 for (i = 0; i < dev->num_tx_queues; i++)
2805 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2806}
2807
Denys Vlasenkof9a7cbb2016-04-08 17:51:54 +02002808static __always_inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002809{
Tom Herbert734664982011-11-28 16:32:44 +00002810 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002811}
2812
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002813/**
2814 * netif_start_queue - allow transmit
2815 * @dev: network device
2816 *
2817 * Allow upper layers to call the device hard_start_xmit routine.
2818 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002819static inline void netif_start_queue(struct net_device *dev)
2820{
David S. Millere8a04642008-07-17 00:34:19 -07002821 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002822}
2823
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002824static inline void netif_tx_start_all_queues(struct net_device *dev)
2825{
2826 unsigned int i;
2827
2828 for (i = 0; i < dev->num_tx_queues; i++) {
2829 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2830 netif_tx_start_queue(txq);
2831 }
2832}
2833
John Fastabend46e5da40a2014-09-12 20:04:52 -07002834void netif_tx_wake_queue(struct netdev_queue *dev_queue);
David S. Miller79d16382008-07-08 23:14:46 -07002835
Dave Jonesd29f7492008-07-22 14:09:06 -07002836/**
2837 * netif_wake_queue - restart transmit
2838 * @dev: network device
2839 *
2840 * Allow upper layers to call the device hard_start_xmit routine.
2841 * Used for flow control when transmit resources are available.
2842 */
David S. Miller79d16382008-07-08 23:14:46 -07002843static inline void netif_wake_queue(struct net_device *dev)
2844{
David S. Millere8a04642008-07-17 00:34:19 -07002845 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002846}
2847
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002848static inline void netif_tx_wake_all_queues(struct net_device *dev)
2849{
2850 unsigned int i;
2851
2852 for (i = 0; i < dev->num_tx_queues; i++) {
2853 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2854 netif_tx_wake_queue(txq);
2855 }
2856}
2857
Denys Vlasenkof9a7cbb2016-04-08 17:51:54 +02002858static __always_inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002859{
Tom Herbert734664982011-11-28 16:32:44 +00002860 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002861}
2862
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002863/**
2864 * netif_stop_queue - stop transmitted packets
2865 * @dev: network device
2866 *
2867 * Stop upper layers calling the device hard_start_xmit routine.
2868 * Used for flow control when transmit resources are unavailable.
2869 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002870static inline void netif_stop_queue(struct net_device *dev)
2871{
David S. Millere8a04642008-07-17 00:34:19 -07002872 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002873}
2874
Denys Vlasenkoa2029242015-05-11 21:17:53 +02002875void netif_tx_stop_all_queues(struct net_device *dev);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002876
David S. Miller4d295152012-03-07 21:02:35 -05002877static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002878{
Tom Herbert734664982011-11-28 16:32:44 +00002879 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002880}
2881
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002882/**
2883 * netif_queue_stopped - test if transmit queue is flowblocked
2884 * @dev: network device
2885 *
2886 * Test if transmit queue on device is currently unable to send.
2887 */
David S. Miller4d295152012-03-07 21:02:35 -05002888static inline bool netif_queue_stopped(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002889{
David S. Millere8a04642008-07-17 00:34:19 -07002890 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002891}
2892
David S. Miller4d295152012-03-07 21:02:35 -05002893static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
David S. Millerc3f26a22008-07-31 16:58:50 -07002894{
Tom Herbert734664982011-11-28 16:32:44 +00002895 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2896}
2897
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002898static inline bool
2899netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
Tom Herbert734664982011-11-28 16:32:44 +00002900{
2901 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2902}
2903
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002904static inline bool
2905netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2906{
2907 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2908}
2909
Eric Dumazet53511452014-10-08 08:19:27 -07002910/**
2911 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2912 * @dev_queue: pointer to transmit queue
2913 *
2914 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002915 * to give appropriate hint to the CPU.
Eric Dumazet53511452014-10-08 08:19:27 -07002916 */
2917static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2918{
2919#ifdef CONFIG_BQL
2920 prefetchw(&dev_queue->dql.num_queued);
2921#endif
2922}
2923
2924/**
2925 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2926 * @dev_queue: pointer to transmit queue
2927 *
2928 * BQL enabled drivers might use this helper in their TX completion path,
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05002929 * to give appropriate hint to the CPU.
Eric Dumazet53511452014-10-08 08:19:27 -07002930 */
2931static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2932{
2933#ifdef CONFIG_BQL
2934 prefetchw(&dev_queue->dql.limit);
2935#endif
2936}
2937
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002938static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2939 unsigned int bytes)
2940{
Tom Herbert114cf582011-11-28 16:33:09 +00002941#ifdef CONFIG_BQL
2942 dql_queued(&dev_queue->dql, bytes);
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00002943
2944 if (likely(dql_avail(&dev_queue->dql) >= 0))
2945 return;
2946
2947 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2948
2949 /*
2950 * The XOFF flag must be set before checking the dql_avail below,
2951 * because in netdev_tx_completed_queue we update the dql_completed
2952 * before checking the XOFF flag.
2953 */
2954 smp_mb();
2955
2956 /* check again in case another CPU has just made room avail */
2957 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2958 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
Tom Herbert114cf582011-11-28 16:33:09 +00002959#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002960}
2961
Florian Fainelli0042d0c2013-09-06 16:58:00 +01002962/**
2963 * netdev_sent_queue - report the number of bytes queued to hardware
2964 * @dev: network device
2965 * @bytes: number of bytes queued to the hardware device queue
2966 *
2967 * Report the number of bytes queued for sending/completion to the network
2968 * device hardware queue. @bytes should be a good approximation and should
2969 * exactly match netdev_completed_queue() @bytes
2970 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002971static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2972{
2973 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2974}
2975
2976static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
Eric Dumazet95c96172012-04-15 05:58:06 +00002977 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002978{
Tom Herbert114cf582011-11-28 16:33:09 +00002979#ifdef CONFIG_BQL
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00002980 if (unlikely(!bytes))
2981 return;
2982
2983 dql_completed(&dev_queue->dql, bytes);
2984
2985 /*
2986 * Without the memory barrier there is a small possiblity that
2987 * netdev_tx_sent_queue will miss the update and cause the queue to
2988 * be stopped forever
2989 */
2990 smp_mb();
2991
2992 if (dql_avail(&dev_queue->dql) < 0)
2993 return;
2994
2995 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2996 netif_schedule_queue(dev_queue);
Tom Herbert114cf582011-11-28 16:33:09 +00002997#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002998}
2999
Florian Fainelli0042d0c2013-09-06 16:58:00 +01003000/**
3001 * netdev_completed_queue - report bytes and packets completed by device
3002 * @dev: network device
3003 * @pkts: actual number of packets sent over the medium
3004 * @bytes: actual number of bytes sent over the medium
3005 *
3006 * Report the number of bytes and packets transmitted by the network device
3007 * hardware queue over the physical medium, @bytes must exactly match the
3008 * @bytes amount passed to netdev_sent_queue()
3009 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003010static inline void netdev_completed_queue(struct net_device *dev,
Eric Dumazet95c96172012-04-15 05:58:06 +00003011 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003012{
3013 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3014}
3015
3016static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3017{
Tom Herbert114cf582011-11-28 16:33:09 +00003018#ifdef CONFIG_BQL
Alexander Duyck5c490352012-02-07 02:29:01 +00003019 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
Tom Herbert114cf582011-11-28 16:33:09 +00003020 dql_reset(&q->dql);
3021#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003022}
3023
Florian Fainelli0042d0c2013-09-06 16:58:00 +01003024/**
3025 * netdev_reset_queue - reset the packets and bytes count of a network device
3026 * @dev_queue: network device
3027 *
3028 * Reset the bytes and packet count of a network device and clear the
3029 * software flow control OFF bit for this network device
3030 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003031static inline void netdev_reset_queue(struct net_device *dev_queue)
3032{
3033 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
David S. Millerc3f26a22008-07-31 16:58:50 -07003034}
3035
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003036/**
Daniel Borkmannb9507bd2014-02-16 15:55:21 +01003037 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3038 * @dev: network device
3039 * @queue_index: given tx queue index
3040 *
3041 * Returns 0 if given tx queue index >= number of device tx queues,
3042 * otherwise returns the originally passed tx queue index.
3043 */
3044static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3045{
3046 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3047 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3048 dev->name, queue_index,
3049 dev->real_num_tx_queues);
3050 return 0;
3051 }
3052
3053 return queue_index;
3054}
3055
3056/**
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003057 * netif_running - test if up
3058 * @dev: network device
3059 *
3060 * Test if the device has been brought up.
3061 */
David S. Miller4d295152012-03-07 21:02:35 -05003062static inline bool netif_running(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003063{
3064 return test_bit(__LINK_STATE_START, &dev->state);
3065}
3066
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003067/*
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003068 * Routines to manage the subqueues on a device. We only need start,
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003069 * stop, and a check if it's stopped. All other device management is
3070 * done at the overall netdevice level.
3071 * Also test the device if we're multiqueue.
3072 */
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003073
3074/**
3075 * netif_start_subqueue - allow sending packets on subqueue
3076 * @dev: network device
3077 * @queue_index: sub queue index
3078 *
3079 * Start individual transmit queue of a device with multiple transmit queues.
3080 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003081static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3082{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003083 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003084
3085 netif_tx_start_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003086}
3087
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003088/**
3089 * netif_stop_subqueue - stop sending packets on subqueue
3090 * @dev: network device
3091 * @queue_index: sub queue index
3092 *
3093 * Stop individual transmit queue of a device with multiple transmit queues.
3094 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003095static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3096{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003097 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003098 netif_tx_stop_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003099}
3100
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003101/**
3102 * netif_subqueue_stopped - test status of subqueue
3103 * @dev: network device
3104 * @queue_index: sub queue index
3105 *
3106 * Check individual transmit queue of a device with multiple transmit queues.
3107 */
David S. Miller4d295152012-03-07 21:02:35 -05003108static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3109 u16 queue_index)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003110{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003111 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003112
3113 return netif_tx_queue_stopped(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003114}
3115
David S. Miller4d295152012-03-07 21:02:35 -05003116static inline bool netif_subqueue_stopped(const struct net_device *dev,
3117 struct sk_buff *skb)
Pavel Emelyanov668f8952007-10-21 17:01:56 -07003118{
3119 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3120}
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003121
Florian Fainelli738b35c2017-01-11 21:13:02 -08003122/**
3123 * netif_wake_subqueue - allow sending packets on subqueue
3124 * @dev: network device
3125 * @queue_index: sub queue index
3126 *
3127 * Resume individual transmit queue of a device with multiple transmit queues.
3128 */
3129static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
3130{
3131 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
3132
3133 netif_tx_wake_queue(txq);
3134}
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003135
Alexander Duyck537c00d2013-01-10 08:57:02 +00003136#ifdef CONFIG_XPS
David S. Miller53af53a2013-10-08 23:07:53 -04003137int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
Joe Perchesf629d202013-09-26 14:48:15 -07003138 u16 index);
Alexander Duyck537c00d2013-01-10 08:57:02 +00003139#else
3140static inline int netif_set_xps_queue(struct net_device *dev,
Michael S. Tsirkin35735402013-10-02 09:14:06 +03003141 const struct cpumask *mask,
Alexander Duyck537c00d2013-01-10 08:57:02 +00003142 u16 index)
3143{
3144 return 0;
3145}
3146#endif
3147
Jiri Pirko5605c762015-05-12 14:56:12 +02003148u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3149 unsigned int num_tx_queues);
3150
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00003151/*
3152 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3153 * as a distribution range limit for the returned value.
3154 */
3155static inline u16 skb_tx_hash(const struct net_device *dev,
Tom Herbert0e001612014-07-01 21:32:27 -07003156 struct sk_buff *skb)
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00003157{
3158 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3159}
3160
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003161/**
3162 * netif_is_multiqueue - test if device has multiple transmit queues
3163 * @dev: network device
3164 *
3165 * Check if device has multiple transmit queues
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003166 */
David S. Miller4d295152012-03-07 21:02:35 -05003167static inline bool netif_is_multiqueue(const struct net_device *dev)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003168{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003169 return dev->num_tx_queues > 1;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003170}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003171
Joe Perchesf629d202013-09-26 14:48:15 -07003172int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
John Fastabendf0796d52010-07-01 13:21:57 +00003173
Michael Daltona953be52014-01-16 22:23:28 -08003174#ifdef CONFIG_SYSFS
Joe Perchesf629d202013-09-26 14:48:15 -07003175int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
Ben Hutchings62fe0b42010-09-27 08:24:33 +00003176#else
3177static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3178 unsigned int rxq)
3179{
3180 return 0;
3181}
3182#endif
3183
Michael Daltona953be52014-01-16 22:23:28 -08003184#ifdef CONFIG_SYSFS
3185static inline unsigned int get_netdev_rx_queue_index(
3186 struct netdev_rx_queue *queue)
3187{
3188 struct net_device *dev = queue->dev;
3189 int index = queue - dev->_rx;
3190
3191 BUG_ON(index >= dev->num_rx_queues);
3192 return index;
3193}
3194#endif
3195
Yuval Mintz16917b82012-07-01 03:18:50 +00003196#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
Joe Perchesf629d202013-09-26 14:48:15 -07003197int netif_get_num_default_rss_queues(void);
Yuval Mintz16917b82012-07-01 03:18:50 +00003198
Eric Dumazete6247022013-12-05 04:45:08 -08003199enum skb_free_reason {
3200 SKB_REASON_CONSUMED,
3201 SKB_REASON_DROPPED,
3202};
Linus Torvalds1da177e2005-04-16 15:20:36 -07003203
Eric Dumazete6247022013-12-05 04:45:08 -08003204void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3205void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3206
3207/*
3208 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3209 * interrupt context or with hardware interrupts being disabled.
3210 * (in_irq() || irqs_disabled())
3211 *
3212 * We provide four helpers that can be used in following contexts :
3213 *
3214 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3215 * replacing kfree_skb(skb)
3216 *
3217 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3218 * Typically used in place of consume_skb(skb) in TX completion path
3219 *
3220 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3221 * replacing kfree_skb(skb)
3222 *
3223 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3224 * and consumed a packet. Used in place of consume_skb(skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003225 */
Eric Dumazete6247022013-12-05 04:45:08 -08003226static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3227{
3228 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3229}
3230
3231static inline void dev_consume_skb_irq(struct sk_buff *skb)
3232{
3233 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3234}
3235
3236static inline void dev_kfree_skb_any(struct sk_buff *skb)
3237{
3238 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3239}
3240
3241static inline void dev_consume_skb_any(struct sk_buff *skb)
3242{
3243 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3244}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003245
Joe Perchesf629d202013-09-26 14:48:15 -07003246int netif_rx(struct sk_buff *skb);
3247int netif_rx_ni(struct sk_buff *skb);
Eric W. Biederman04eb4482015-09-15 20:04:15 -05003248int netif_receive_skb(struct sk_buff *skb);
Joe Perchesf629d202013-09-26 14:48:15 -07003249gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3250void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3251struct sk_buff *napi_get_frags(struct napi_struct *napi);
3252gro_result_t napi_gro_frags(struct napi_struct *napi);
Jerry Chubf5a7552014-01-07 10:23:19 -08003253struct packet_offload *gro_find_receive_by_type(__be16 type);
3254struct packet_offload *gro_find_complete_by_type(__be16 type);
Herbert Xu76620aa2009-04-16 02:02:07 -07003255
3256static inline void napi_free_frags(struct napi_struct *napi)
3257{
3258 kfree_skb(napi->skb);
3259 napi->skb = NULL;
3260}
3261
Mahesh Bandewar24b27fc2016-09-01 22:18:34 -07003262bool netdev_is_rx_handler_busy(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003263int netdev_rx_handler_register(struct net_device *dev,
3264 rx_handler_func_t *rx_handler,
3265 void *rx_handler_data);
3266void netdev_rx_handler_unregister(struct net_device *dev);
Jiri Pirkoab95bfe2010-06-01 21:52:08 +00003267
Joe Perchesf629d202013-09-26 14:48:15 -07003268bool dev_valid_name(const char *name);
3269int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3270int dev_ethtool(struct net *net, struct ifreq *);
3271unsigned int dev_get_flags(const struct net_device *);
3272int __dev_change_flags(struct net_device *, unsigned int flags);
3273int dev_change_flags(struct net_device *, unsigned int);
David S. Millercb178192013-09-30 15:36:45 -04003274void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3275 unsigned int gchanges);
Joe Perchesf629d202013-09-26 14:48:15 -07003276int dev_change_name(struct net_device *, const char *);
3277int dev_set_alias(struct net_device *, const char *, size_t);
3278int dev_change_net_namespace(struct net_device *, struct net *, const char *);
WANG Congf51048c2017-07-06 15:01:57 -07003279int __dev_set_mtu(struct net_device *, int);
Joe Perchesf629d202013-09-26 14:48:15 -07003280int dev_set_mtu(struct net_device *, int);
3281void dev_set_group(struct net_device *, int);
3282int dev_set_mac_address(struct net_device *, struct sockaddr *);
3283int dev_change_carrier(struct net_device *, bool new_carrier);
3284int dev_get_phys_port_id(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01003285 struct netdev_phys_item_id *ppid);
David Aherndb24a902015-03-17 20:23:15 -06003286int dev_get_phys_port_name(struct net_device *dev,
3287 char *name, size_t len);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07003288int dev_change_proto_down(struct net_device *dev, bool proto_down);
Eric Dumazet55a93b32014-10-03 15:31:07 -07003289struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
David S. Millerce937182014-08-30 19:22:20 -07003290struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3291 struct netdev_queue *txq, int *ret);
Daniel Borkmannd67b9cd2017-05-12 01:04:46 +02003292
3293typedef int (*xdp_op_t)(struct net_device *dev, struct netdev_xdp *xdp);
3294int dev_change_xdp_fd(struct net_device *dev, struct netlink_ext_ack *extack,
3295 int fd, u32 flags);
Jakub Kicinskice158e52017-06-21 18:25:09 -07003296u8 __dev_xdp_attached(struct net_device *dev, xdp_op_t xdp_op, u32 *prog_id);
Daniel Borkmannd67b9cd2017-05-12 01:04:46 +02003297
Herbert Xua0265d22014-04-17 13:45:03 +08003298int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Joe Perchesf629d202013-09-26 14:48:15 -07003299int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Nikolay Aleksandrovf4b05d22016-04-28 17:59:28 +02003300bool is_skb_forwardable(const struct net_device *dev,
3301 const struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003302
Martin KaFai Lau4e3264d2016-11-09 15:36:33 -08003303static __always_inline int ____dev_forward_skb(struct net_device *dev,
3304 struct sk_buff *skb)
3305{
3306 if (skb_orphan_frags(skb, GFP_ATOMIC) ||
3307 unlikely(!is_skb_forwardable(dev, skb))) {
3308 atomic_long_inc(&dev->rx_dropped);
3309 kfree_skb(skb);
3310 return NET_RX_DROP;
3311 }
3312
3313 skb_scrub_packet(skb, true);
3314 skb->priority = 0;
3315 return 0;
3316}
3317
David Ahern74b20582016-05-10 11:19:50 -07003318void dev_queue_xmit_nit(struct sk_buff *skb, struct net_device *dev);
3319
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003320extern int netdev_budget;
Matthew Whitehead7acf8a12017-04-19 12:37:10 -04003321extern unsigned int netdev_budget_usecs;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003322
3323/* Called by rtnetlink.c:rtnl_unlock() */
Joe Perchesf629d202013-09-26 14:48:15 -07003324void netdev_run_todo(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003325
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003326/**
3327 * dev_put - release reference to device
3328 * @dev: network device
3329 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07003330 * Release reference to device to allow it to be freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003331 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003332static inline void dev_put(struct net_device *dev)
3333{
Christoph Lameter933393f2011-12-22 11:58:51 -06003334 this_cpu_dec(*dev->pcpu_refcnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003335}
3336
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003337/**
3338 * dev_hold - get reference to device
3339 * @dev: network device
3340 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07003341 * Hold reference to device to keep it from being freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003342 */
Stephen Hemminger15333062006-03-20 22:32:28 -08003343static inline void dev_hold(struct net_device *dev)
3344{
Christoph Lameter933393f2011-12-22 11:58:51 -06003345 this_cpu_inc(*dev->pcpu_refcnt);
Stephen Hemminger15333062006-03-20 22:32:28 -08003346}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003347
3348/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3349 * and _off may be called from IRQ context, but it is caller
3350 * who is responsible for serialization of these calls.
Stefan Rompfb00055a2006-03-20 17:09:11 -08003351 *
3352 * The name carrier is inappropriate, these functions should really be
3353 * called netif_lowerlayer_*() because they represent the state of any
3354 * kind of lower layer not just hardware media.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003355 */
3356
Joe Perchesf629d202013-09-26 14:48:15 -07003357void linkwatch_init_dev(struct net_device *dev);
3358void linkwatch_fire_event(struct net_device *dev);
3359void linkwatch_forget_dev(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003360
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003361/**
3362 * netif_carrier_ok - test if carrier present
3363 * @dev: network device
3364 *
3365 * Check if carrier is present on device
3366 */
David S. Miller4d295152012-03-07 21:02:35 -05003367static inline bool netif_carrier_ok(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003368{
3369 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3370}
3371
Joe Perchesf629d202013-09-26 14:48:15 -07003372unsigned long dev_trans_start(struct net_device *dev);
Eric Dumazet9d214932009-05-17 20:55:16 -07003373
Joe Perchesf629d202013-09-26 14:48:15 -07003374void __netdev_watchdog_up(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003375
Joe Perchesf629d202013-09-26 14:48:15 -07003376void netif_carrier_on(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003377
Joe Perchesf629d202013-09-26 14:48:15 -07003378void netif_carrier_off(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003379
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003380/**
3381 * netif_dormant_on - mark device as dormant.
3382 * @dev: network device
3383 *
3384 * Mark device as dormant (as per RFC2863).
3385 *
3386 * The dormant state indicates that the relevant interface is not
3387 * actually in a condition to pass packets (i.e., it is not 'up') but is
3388 * in a "pending" state, waiting for some external event. For "on-
3389 * demand" interfaces, this new state identifies the situation where the
3390 * interface is waiting for events to place it in the up state.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003391 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003392static inline void netif_dormant_on(struct net_device *dev)
3393{
3394 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3395 linkwatch_fire_event(dev);
3396}
3397
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003398/**
3399 * netif_dormant_off - set device as not dormant.
3400 * @dev: network device
3401 *
3402 * Device is not in dormant state.
3403 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003404static inline void netif_dormant_off(struct net_device *dev)
3405{
3406 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3407 linkwatch_fire_event(dev);
3408}
3409
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003410/**
Zhang Shengju8ecbc402017-04-26 11:05:12 +08003411 * netif_dormant - test if device is dormant
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003412 * @dev: network device
3413 *
Zhang Shengju8ecbc402017-04-26 11:05:12 +08003414 * Check if device is dormant.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003415 */
David S. Miller4d295152012-03-07 21:02:35 -05003416static inline bool netif_dormant(const struct net_device *dev)
Stefan Rompfb00055a2006-03-20 17:09:11 -08003417{
3418 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3419}
3420
3421
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003422/**
3423 * netif_oper_up - test if device is operational
3424 * @dev: network device
3425 *
3426 * Check if carrier is operational
3427 */
David S. Miller4d295152012-03-07 21:02:35 -05003428static inline bool netif_oper_up(const struct net_device *dev)
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08003429{
Stefan Rompfb00055a2006-03-20 17:09:11 -08003430 return (dev->operstate == IF_OPER_UP ||
3431 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3432}
3433
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003434/**
3435 * netif_device_present - is device available or removed
3436 * @dev: network device
3437 *
3438 * Check if device has not been removed from system.
3439 */
David S. Miller4d295152012-03-07 21:02:35 -05003440static inline bool netif_device_present(struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003441{
3442 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3443}
3444
Joe Perchesf629d202013-09-26 14:48:15 -07003445void netif_device_detach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003446
Joe Perchesf629d202013-09-26 14:48:15 -07003447void netif_device_attach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003448
3449/*
3450 * Network interface message level settings
3451 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003452
3453enum {
3454 NETIF_MSG_DRV = 0x0001,
3455 NETIF_MSG_PROBE = 0x0002,
3456 NETIF_MSG_LINK = 0x0004,
3457 NETIF_MSG_TIMER = 0x0008,
3458 NETIF_MSG_IFDOWN = 0x0010,
3459 NETIF_MSG_IFUP = 0x0020,
3460 NETIF_MSG_RX_ERR = 0x0040,
3461 NETIF_MSG_TX_ERR = 0x0080,
3462 NETIF_MSG_TX_QUEUED = 0x0100,
3463 NETIF_MSG_INTR = 0x0200,
3464 NETIF_MSG_TX_DONE = 0x0400,
3465 NETIF_MSG_RX_STATUS = 0x0800,
3466 NETIF_MSG_PKTDATA = 0x1000,
3467 NETIF_MSG_HW = 0x2000,
3468 NETIF_MSG_WOL = 0x4000,
3469};
3470
3471#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3472#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3473#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3474#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3475#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3476#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3477#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3478#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3479#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3480#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3481#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3482#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3483#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3484#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3485#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3486
3487static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3488{
3489 /* use default */
3490 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3491 return default_msg_enable_bits;
3492 if (debug_value == 0) /* no output */
3493 return 0;
3494 /* set low N bits */
3495 return (1 << debug_value) - 1;
3496}
3497
David S. Millerc773e842008-07-08 23:13:53 -07003498static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
Herbert Xu932ff272006-06-09 12:20:56 -07003499{
David S. Millerc773e842008-07-08 23:13:53 -07003500 spin_lock(&txq->_xmit_lock);
3501 txq->xmit_lock_owner = cpu;
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003502}
3503
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003504static inline bool __netif_tx_acquire(struct netdev_queue *txq)
3505{
3506 __acquire(&txq->_xmit_lock);
3507 return true;
3508}
3509
3510static inline void __netif_tx_release(struct netdev_queue *txq)
3511{
3512 __release(&txq->_xmit_lock);
3513}
3514
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003515static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3516{
3517 spin_lock_bh(&txq->_xmit_lock);
3518 txq->xmit_lock_owner = smp_processor_id();
3519}
3520
David S. Miller4d295152012-03-07 21:02:35 -05003521static inline bool __netif_tx_trylock(struct netdev_queue *txq)
David S. Millerc773e842008-07-08 23:13:53 -07003522{
David S. Miller4d295152012-03-07 21:02:35 -05003523 bool ok = spin_trylock(&txq->_xmit_lock);
David S. Millerc773e842008-07-08 23:13:53 -07003524 if (likely(ok))
3525 txq->xmit_lock_owner = smp_processor_id();
3526 return ok;
Herbert Xu932ff272006-06-09 12:20:56 -07003527}
3528
David S. Millerc773e842008-07-08 23:13:53 -07003529static inline void __netif_tx_unlock(struct netdev_queue *txq)
3530{
3531 txq->xmit_lock_owner = -1;
3532 spin_unlock(&txq->_xmit_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003533}
3534
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003535static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3536{
3537 txq->xmit_lock_owner = -1;
3538 spin_unlock_bh(&txq->_xmit_lock);
3539}
3540
Eric Dumazet08baf562009-05-25 22:58:01 -07003541static inline void txq_trans_update(struct netdev_queue *txq)
3542{
3543 if (txq->xmit_lock_owner != -1)
3544 txq->trans_start = jiffies;
3545}
3546
Florian Westphalba162f82016-05-03 16:31:00 +02003547/* legacy drivers only, netdev_start_xmit() sets txq->trans_start */
3548static inline void netif_trans_update(struct net_device *dev)
3549{
Florian Westphal9b366272016-05-03 16:33:14 +02003550 struct netdev_queue *txq = netdev_get_tx_queue(dev, 0);
3551
3552 if (txq->trans_start != jiffies)
3553 txq->trans_start = jiffies;
Florian Westphalba162f82016-05-03 16:31:00 +02003554}
3555
David S. Millerc3f26a22008-07-31 16:58:50 -07003556/**
3557 * netif_tx_lock - grab network device transmit lock
3558 * @dev: network device
David S. Millerc3f26a22008-07-31 16:58:50 -07003559 *
3560 * Get network device transmit lock
3561 */
3562static inline void netif_tx_lock(struct net_device *dev)
3563{
3564 unsigned int i;
3565 int cpu;
3566
3567 spin_lock(&dev->tx_global_lock);
3568 cpu = smp_processor_id();
3569 for (i = 0; i < dev->num_tx_queues; i++) {
3570 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3571
3572 /* We are the only thread of execution doing a
3573 * freeze, but we have to grab the _xmit_lock in
3574 * order to synchronize with threads which are in
3575 * the ->hard_start_xmit() handler and already
3576 * checked the frozen bit.
3577 */
3578 __netif_tx_lock(txq, cpu);
3579 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3580 __netif_tx_unlock(txq);
3581 }
3582}
3583
3584static inline void netif_tx_lock_bh(struct net_device *dev)
3585{
3586 local_bh_disable();
3587 netif_tx_lock(dev);
3588}
3589
Herbert Xu932ff272006-06-09 12:20:56 -07003590static inline void netif_tx_unlock(struct net_device *dev)
3591{
David S. Millere8a04642008-07-17 00:34:19 -07003592 unsigned int i;
David S. Millerc773e842008-07-08 23:13:53 -07003593
David S. Millere8a04642008-07-17 00:34:19 -07003594 for (i = 0; i < dev->num_tx_queues; i++) {
3595 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millere8a04642008-07-17 00:34:19 -07003596
David S. Millerc3f26a22008-07-31 16:58:50 -07003597 /* No need to grab the _xmit_lock here. If the
3598 * queue is not stopped for another reason, we
3599 * force a schedule.
3600 */
3601 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003602 netif_schedule_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003603 }
3604 spin_unlock(&dev->tx_global_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003605}
3606
3607static inline void netif_tx_unlock_bh(struct net_device *dev)
3608{
David S. Millere8a04642008-07-17 00:34:19 -07003609 netif_tx_unlock(dev);
3610 local_bh_enable();
Herbert Xu932ff272006-06-09 12:20:56 -07003611}
3612
David S. Millerc773e842008-07-08 23:13:53 -07003613#define HARD_TX_LOCK(dev, txq, cpu) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003614 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003615 __netif_tx_lock(txq, cpu); \
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003616 } else { \
3617 __netif_tx_acquire(txq); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003618 } \
3619}
3620
Eric W. Biederman5efeac42014-03-27 15:42:20 -07003621#define HARD_TX_TRYLOCK(dev, txq) \
3622 (((dev->features & NETIF_F_LLTX) == 0) ? \
3623 __netif_tx_trylock(txq) : \
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003624 __netif_tx_acquire(txq))
Eric W. Biederman5efeac42014-03-27 15:42:20 -07003625
David S. Millerc773e842008-07-08 23:13:53 -07003626#define HARD_TX_UNLOCK(dev, txq) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003627 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003628 __netif_tx_unlock(txq); \
Michael S. Tsirkin5a717f42016-11-24 07:04:08 +02003629 } else { \
3630 __netif_tx_release(txq); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003631 } \
3632}
3633
Linus Torvalds1da177e2005-04-16 15:20:36 -07003634static inline void netif_tx_disable(struct net_device *dev)
3635{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003636 unsigned int i;
David S. Millerc3f26a22008-07-31 16:58:50 -07003637 int cpu;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003638
David S. Millerc3f26a22008-07-31 16:58:50 -07003639 local_bh_disable();
3640 cpu = smp_processor_id();
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003641 for (i = 0; i < dev->num_tx_queues; i++) {
3642 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millerc3f26a22008-07-31 16:58:50 -07003643
3644 __netif_tx_lock(txq, cpu);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003645 netif_tx_stop_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003646 __netif_tx_unlock(txq);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003647 }
David S. Millerc3f26a22008-07-31 16:58:50 -07003648 local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003649}
3650
David S. Millere308a5d2008-07-15 00:13:44 -07003651static inline void netif_addr_lock(struct net_device *dev)
3652{
3653 spin_lock(&dev->addr_list_lock);
3654}
3655
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003656static inline void netif_addr_lock_nested(struct net_device *dev)
3657{
Vlad Yasevich25175ba2014-05-16 17:04:54 -04003658 int subclass = SINGLE_DEPTH_NESTING;
3659
3660 if (dev->netdev_ops->ndo_get_lock_subclass)
3661 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3662
3663 spin_lock_nested(&dev->addr_list_lock, subclass);
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003664}
3665
David S. Millere308a5d2008-07-15 00:13:44 -07003666static inline void netif_addr_lock_bh(struct net_device *dev)
3667{
3668 spin_lock_bh(&dev->addr_list_lock);
3669}
3670
3671static inline void netif_addr_unlock(struct net_device *dev)
3672{
3673 spin_unlock(&dev->addr_list_lock);
3674}
3675
3676static inline void netif_addr_unlock_bh(struct net_device *dev)
3677{
3678 spin_unlock_bh(&dev->addr_list_lock);
3679}
3680
Jiri Pirkof001fde2009-05-05 02:48:28 +00003681/*
Jiri Pirko31278e72009-06-17 01:12:19 +00003682 * dev_addrs walker. Should be used only for read access. Call with
Jiri Pirkof001fde2009-05-05 02:48:28 +00003683 * rcu_read_lock held.
3684 */
3685#define for_each_dev_addr(dev, ha) \
Jiri Pirko31278e72009-06-17 01:12:19 +00003686 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
Jiri Pirkof001fde2009-05-05 02:48:28 +00003687
Linus Torvalds1da177e2005-04-16 15:20:36 -07003688/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3689
Joe Perchesf629d202013-09-26 14:48:15 -07003690void ether_setup(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003691
3692/* Support for loadable net-drivers */
Joe Perchesf629d202013-09-26 14:48:15 -07003693struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
Tom Gundersenc835a672014-07-14 16:37:24 +02003694 unsigned char name_assign_type,
Joe Perchesf629d202013-09-26 14:48:15 -07003695 void (*setup)(struct net_device *),
3696 unsigned int txqs, unsigned int rxqs);
Tom Gundersenc835a672014-07-14 16:37:24 +02003697#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3698 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
Tom Herbert36909ea2011-01-09 19:36:31 +00003699
Tom Gundersenc835a672014-07-14 16:37:24 +02003700#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3701 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3702 count)
Tom Herbert36909ea2011-01-09 19:36:31 +00003703
Joe Perchesf629d202013-09-26 14:48:15 -07003704int register_netdev(struct net_device *dev);
3705void unregister_netdev(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003706
Jiri Pirko22bedad32010-04-01 21:22:57 +00003707/* General hardware address lists handling functions */
Joe Perchesf629d202013-09-26 14:48:15 -07003708int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3709 struct netdev_hw_addr_list *from_list, int addr_len);
3710void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3711 struct netdev_hw_addr_list *from_list, int addr_len);
Alexander Duyck670e5b82014-05-28 18:44:46 -07003712int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3713 struct net_device *dev,
3714 int (*sync)(struct net_device *, const unsigned char *),
3715 int (*unsync)(struct net_device *,
3716 const unsigned char *));
3717void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3718 struct net_device *dev,
3719 int (*unsync)(struct net_device *,
3720 const unsigned char *));
Joe Perchesf629d202013-09-26 14:48:15 -07003721void __hw_addr_init(struct netdev_hw_addr_list *list);
Jiri Pirko22bedad32010-04-01 21:22:57 +00003722
Jiri Pirkof001fde2009-05-05 02:48:28 +00003723/* Functions used for device addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003724int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3725 unsigned char addr_type);
3726int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3727 unsigned char addr_type);
Joe Perchesf629d202013-09-26 14:48:15 -07003728void dev_addr_flush(struct net_device *dev);
3729int dev_addr_init(struct net_device *dev);
Jiri Pirkoa748ee22010-04-01 21:22:09 +00003730
3731/* Functions used for unicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003732int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3733int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3734int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3735int dev_uc_sync(struct net_device *to, struct net_device *from);
3736int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3737void dev_uc_unsync(struct net_device *to, struct net_device *from);
3738void dev_uc_flush(struct net_device *dev);
3739void dev_uc_init(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003740
Alexander Duyck670e5b82014-05-28 18:44:46 -07003741/**
3742 * __dev_uc_sync - Synchonize device's unicast list
3743 * @dev: device to sync
3744 * @sync: function to call if address should be added
3745 * @unsync: function to call if address should be removed
3746 *
3747 * Add newly added addresses to the interface, and release
3748 * addresses that have been deleted.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003749 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003750static inline int __dev_uc_sync(struct net_device *dev,
3751 int (*sync)(struct net_device *,
3752 const unsigned char *),
3753 int (*unsync)(struct net_device *,
3754 const unsigned char *))
3755{
3756 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3757}
3758
3759/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003760 * __dev_uc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003761 * @dev: device to sync
3762 * @unsync: function to call if address should be removed
3763 *
3764 * Remove all addresses that were added to the device by dev_uc_sync().
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003765 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003766static inline void __dev_uc_unsync(struct net_device *dev,
3767 int (*unsync)(struct net_device *,
3768 const unsigned char *))
3769{
3770 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3771}
3772
Jiri Pirko22bedad32010-04-01 21:22:57 +00003773/* Functions used for multicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003774int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3775int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3776int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3777int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3778int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3779int dev_mc_sync(struct net_device *to, struct net_device *from);
3780int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3781void dev_mc_unsync(struct net_device *to, struct net_device *from);
3782void dev_mc_flush(struct net_device *dev);
3783void dev_mc_init(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003784
Alexander Duyck670e5b82014-05-28 18:44:46 -07003785/**
3786 * __dev_mc_sync - Synchonize device's multicast list
3787 * @dev: device to sync
3788 * @sync: function to call if address should be added
3789 * @unsync: function to call if address should be removed
3790 *
3791 * Add newly added addresses to the interface, and release
3792 * addresses that have been deleted.
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003793 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003794static inline int __dev_mc_sync(struct net_device *dev,
3795 int (*sync)(struct net_device *,
3796 const unsigned char *),
3797 int (*unsync)(struct net_device *,
3798 const unsigned char *))
3799{
3800 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3801}
3802
3803/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003804 * __dev_mc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003805 * @dev: device to sync
3806 * @unsync: function to call if address should be removed
3807 *
3808 * Remove all addresses that were added to the device by dev_mc_sync().
Bjorn Helgaas5e82b4b2016-03-23 13:47:23 -05003809 */
Alexander Duyck670e5b82014-05-28 18:44:46 -07003810static inline void __dev_mc_unsync(struct net_device *dev,
3811 int (*unsync)(struct net_device *,
3812 const unsigned char *))
3813{
3814 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3815}
3816
Linus Torvalds1da177e2005-04-16 15:20:36 -07003817/* Functions used for secondary unicast and multicast support */
Joe Perchesf629d202013-09-26 14:48:15 -07003818void dev_set_rx_mode(struct net_device *dev);
3819void __dev_set_rx_mode(struct net_device *dev);
3820int dev_set_promiscuity(struct net_device *dev, int inc);
3821int dev_set_allmulti(struct net_device *dev, int inc);
3822void netdev_state_change(struct net_device *dev);
3823void netdev_notify_peers(struct net_device *dev);
3824void netdev_features_change(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003825/* Load a device via the kmod */
Joe Perchesf629d202013-09-26 14:48:15 -07003826void dev_load(struct net *net, const char *name);
3827struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3828 struct rtnl_link_stats64 *storage);
3829void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3830 const struct net_device_stats *netdev_stats);
Herbert Xufb286bb2005-11-10 13:01:24 -08003831
3832extern int netdev_max_backlog;
Eric Dumazet3b098e22010-05-15 23:57:10 -07003833extern int netdev_tstamp_prequeue;
Herbert Xufb286bb2005-11-10 13:01:24 -08003834extern int weight_p;
Matthias Tafelmeier3d48b532016-12-29 21:37:21 +01003835extern int dev_weight_rx_bias;
3836extern int dev_weight_tx_bias;
3837extern int dev_rx_weight;
3838extern int dev_tx_weight;
Jiri Pirko9ff162a2013-01-03 22:48:49 +00003839
Joe Perchesf629d202013-09-26 14:48:15 -07003840bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
Vlad Yasevich44a40852014-05-16 17:20:38 -04003841struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3842 struct list_head **iter);
Joe Perchesf629d202013-09-26 14:48:15 -07003843struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3844 struct list_head **iter);
Veaceslav Falico8b5be852013-08-28 23:25:08 +02003845
3846/* iterate through upper list, must be called under RCU read lock */
Vlad Yasevich44a40852014-05-16 17:20:38 -04003847#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3848 for (iter = &(dev)->adj_list.upper, \
3849 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3850 updev; \
3851 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3852
David Ahern1a3f0602016-10-17 19:15:44 -07003853int netdev_walk_all_upper_dev_rcu(struct net_device *dev,
3854 int (*fn)(struct net_device *upper_dev,
3855 void *data),
3856 void *data);
3857
3858bool netdev_has_upper_dev_all_rcu(struct net_device *dev,
3859 struct net_device *upper_dev);
3860
Joe Perchesf629d202013-09-26 14:48:15 -07003861void *netdev_lower_get_next_private(struct net_device *dev,
3862 struct list_head **iter);
3863void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3864 struct list_head **iter);
Veaceslav Falico31088a12013-09-25 09:20:12 +02003865
3866#define netdev_for_each_lower_private(dev, priv, iter) \
3867 for (iter = (dev)->adj_list.lower.next, \
3868 priv = netdev_lower_get_next_private(dev, &(iter)); \
3869 priv; \
3870 priv = netdev_lower_get_next_private(dev, &(iter)))
3871
3872#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3873 for (iter = &(dev)->adj_list.lower, \
3874 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3875 priv; \
3876 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3877
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003878void *netdev_lower_get_next(struct net_device *dev,
3879 struct list_head **iter);
Jiri Pirko7ce856a2016-07-04 08:23:12 +02003880
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003881#define netdev_for_each_lower_dev(dev, ldev, iter) \
Nikolay Aleksandrovcfdd28b2016-02-17 18:00:31 +01003882 for (iter = (dev)->adj_list.lower.next, \
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003883 ldev = netdev_lower_get_next(dev, &(iter)); \
3884 ldev; \
3885 ldev = netdev_lower_get_next(dev, &(iter)))
3886
Jiri Pirko7ce856a2016-07-04 08:23:12 +02003887struct net_device *netdev_all_lower_get_next(struct net_device *dev,
3888 struct list_head **iter);
3889struct net_device *netdev_all_lower_get_next_rcu(struct net_device *dev,
3890 struct list_head **iter);
3891
David Ahern1a3f0602016-10-17 19:15:44 -07003892int netdev_walk_all_lower_dev(struct net_device *dev,
3893 int (*fn)(struct net_device *lower_dev,
3894 void *data),
3895 void *data);
3896int netdev_walk_all_lower_dev_rcu(struct net_device *dev,
3897 int (*fn)(struct net_device *lower_dev,
3898 void *data),
3899 void *data);
3900
Joe Perchesf629d202013-09-26 14:48:15 -07003901void *netdev_adjacent_get_private(struct list_head *adj_list);
dingtianhonge001bfa2013-12-13 10:19:55 +08003902void *netdev_lower_get_first_private_rcu(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003903struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3904struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3905int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3906int netdev_master_upper_dev_link(struct net_device *dev,
Jiri Pirko6dffb042015-12-03 12:12:10 +01003907 struct net_device *upper_dev,
Jiri Pirko29bf24a2015-12-03 12:12:11 +01003908 void *upper_priv, void *upper_info);
Joe Perchesf629d202013-09-26 14:48:15 -07003909void netdev_upper_dev_unlink(struct net_device *dev,
3910 struct net_device *upper_dev);
Veaceslav Falico5bb025f2014-01-14 21:58:51 +01003911void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
Joe Perchesf629d202013-09-26 14:48:15 -07003912void *netdev_lower_dev_get_private(struct net_device *dev,
3913 struct net_device *lower_dev);
Jiri Pirko04d48262015-12-03 12:12:15 +01003914void netdev_lower_state_changed(struct net_device *lower_dev,
3915 void *lower_state_info);
Eric Dumazet960fb622014-11-16 06:23:05 -08003916
3917/* RSS keys are 40 or 52 bytes long */
3918#define NETDEV_RSS_KEY_LEN 52
Kim Jonesba905f52016-02-02 03:51:16 +00003919extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
Eric Dumazet960fb622014-11-16 06:23:05 -08003920void netdev_rss_key_fill(void *buffer, size_t len);
3921
Sabrina Dubroca952fcfd2016-08-12 16:10:33 +02003922int dev_get_nest_level(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003923int skb_checksum_help(struct sk_buff *skb);
Davide Carattib72b5bf2017-05-18 15:44:38 +02003924int skb_crc32c_csum_help(struct sk_buff *skb);
Davide Caratti43c26a12017-05-18 15:44:41 +02003925int skb_csum_hwoffload_help(struct sk_buff *skb,
3926 const netdev_features_t features);
3927
Joe Perchesf629d202013-09-26 14:48:15 -07003928struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3929 netdev_features_t features, bool tx_path);
3930struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3931 netdev_features_t features);
Cong Wang12b00042013-02-05 16:36:38 +00003932
Moni Shoua61bd3852015-02-03 16:48:29 +02003933struct netdev_bonding_info {
3934 ifslave slave;
3935 ifbond master;
3936};
3937
3938struct netdev_notifier_bonding_info {
3939 struct netdev_notifier_info info; /* must be first */
3940 struct netdev_bonding_info bonding_info;
3941};
3942
3943void netdev_bonding_info_change(struct net_device *dev,
3944 struct netdev_bonding_info *bonding_info);
3945
Cong Wang12b00042013-02-05 16:36:38 +00003946static inline
3947struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3948{
3949 return __skb_gso_segment(skb, features, true);
3950}
Vlad Yasevich53d64712014-03-27 17:26:18 -04003951__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
Pravin B Shelarec5f0612013-03-07 09:28:01 +00003952
3953static inline bool can_checksum_protocol(netdev_features_t features,
3954 __be16 protocol)
3955{
Tom Herbertc8cd0982015-12-14 11:19:44 -08003956 if (protocol == htons(ETH_P_FCOE))
3957 return !!(features & NETIF_F_FCOE_CRC);
3958
3959 /* Assume this is an IP checksum (not SCTP CRC) */
3960
3961 if (features & NETIF_F_HW_CSUM) {
3962 /* Can checksum everything */
3963 return true;
3964 }
3965
3966 switch (protocol) {
3967 case htons(ETH_P_IP):
3968 return !!(features & NETIF_F_IP_CSUM);
3969 case htons(ETH_P_IPV6):
3970 return !!(features & NETIF_F_IPV6_CSUM);
3971 default:
3972 return false;
3973 }
Pravin B Shelarec5f0612013-03-07 09:28:01 +00003974}
Cong Wang12b00042013-02-05 16:36:38 +00003975
Herbert Xufb286bb2005-11-10 13:01:24 -08003976#ifdef CONFIG_BUG
Joe Perchesf629d202013-09-26 14:48:15 -07003977void netdev_rx_csum_fault(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003978#else
3979static inline void netdev_rx_csum_fault(struct net_device *dev)
3980{
3981}
3982#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003983/* rx skb timestamps */
Joe Perchesf629d202013-09-26 14:48:15 -07003984void net_enable_timestamp(void);
3985void net_disable_timestamp(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003986
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003987#ifdef CONFIG_PROC_FS
Joe Perchesf629d202013-09-26 14:48:15 -07003988int __init dev_proc_init(void);
Cong Wang900ff8c2013-02-18 19:20:33 +00003989#else
3990#define dev_proc_init() 0
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003991#endif
3992
David S. Miller47982482014-08-22 16:21:53 -07003993static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003994 struct sk_buff *skb, struct net_device *dev,
3995 bool more)
David S. Miller47982482014-08-22 16:21:53 -07003996{
David S. Millerfa2dbdc2014-08-29 21:55:22 -07003997 skb->xmit_more = more ? 1 : 0;
David S. Miller0b725a22014-08-25 15:51:53 -07003998 return ops->ndo_start_xmit(skb, dev);
David S. Miller47982482014-08-22 16:21:53 -07003999}
4000
David S. Miller10b3ad82014-08-29 21:07:24 -07004001static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
David S. Millerfa2dbdc2014-08-29 21:55:22 -07004002 struct netdev_queue *txq, bool more)
David S. Miller47982482014-08-22 16:21:53 -07004003{
4004 const struct net_device_ops *ops = dev->netdev_ops;
David S. Miller10b3ad82014-08-29 21:07:24 -07004005 int rc;
David S. Miller47982482014-08-22 16:21:53 -07004006
David S. Millerfa2dbdc2014-08-29 21:55:22 -07004007 rc = __netdev_start_xmit(ops, skb, dev, more);
David S. Miller10b3ad82014-08-29 21:07:24 -07004008 if (rc == NETDEV_TX_OK)
4009 txq_trans_update(txq);
4010
4011 return rc;
David S. Miller47982482014-08-22 16:21:53 -07004012}
4013
Linus Torvalds42a2d922013-11-13 17:40:34 +09004014int netdev_class_create_file_ns(struct class_attribute *class_attr,
4015 const void *ns);
4016void netdev_class_remove_file_ns(struct class_attribute *class_attr,
4017 const void *ns);
Tejun Heo58292cbe2013-09-11 22:29:04 -04004018
4019static inline int netdev_class_create_file(struct class_attribute *class_attr)
4020{
4021 return netdev_class_create_file_ns(class_attr, NULL);
4022}
4023
4024static inline void netdev_class_remove_file(struct class_attribute *class_attr)
4025{
4026 netdev_class_remove_file_ns(class_attr, NULL);
4027}
Jay Vosburghb8a97872008-06-13 18:12:04 -07004028
Johannes Berg04600792010-08-05 17:45:15 +02004029extern struct kobj_ns_type_operations net_ns_type_operations;
4030
Joe Perchesf629d202013-09-26 14:48:15 -07004031const char *netdev_drivername(const struct net_device *dev);
Arjan van de Ven6579e572008-07-21 13:31:48 -07004032
Joe Perchesf629d202013-09-26 14:48:15 -07004033void linkwatch_run_queue(void);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03004034
Michal Kubečekda081432014-05-20 08:29:25 +02004035static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
4036 netdev_features_t f2)
4037{
Tom Herbertc8cd0982015-12-14 11:19:44 -08004038 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
4039 if (f1 & NETIF_F_HW_CSUM)
Tom Herbertb6a0e722016-01-11 10:19:10 -08004040 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
Tom Herbertc8cd0982015-12-14 11:19:44 -08004041 else
Tom Herbertb6a0e722016-01-11 10:19:10 -08004042 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
Tom Herbertc8cd0982015-12-14 11:19:44 -08004043 }
Michal Kubečekda081432014-05-20 08:29:25 +02004044
Tom Herbertc8cd0982015-12-14 11:19:44 -08004045 return f1 & f2;
Michal Kubečekda081432014-05-20 08:29:25 +02004046}
4047
Michał Mirosławc8f44af2011-11-15 15:29:55 +00004048static inline netdev_features_t netdev_get_wanted_features(
4049 struct net_device *dev)
Michał Mirosław5455c692011-02-15 16:59:17 +00004050{
4051 return (dev->features & ~dev->hw_features) | dev->wanted_features;
4052}
Michał Mirosławc8f44af2011-11-15 15:29:55 +00004053netdev_features_t netdev_increment_features(netdev_features_t all,
4054 netdev_features_t one, netdev_features_t mask);
Eric Dumazetb0ce3502013-05-16 07:34:53 +00004055
4056/* Allow TSO being used on stacked device :
4057 * Performing the GSO segmentation before last device
4058 * is a performance improvement.
4059 */
4060static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
4061 netdev_features_t mask)
4062{
4063 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
4064}
4065
Michał Mirosław6cb6a272011-04-02 22:48:47 -07004066int __netdev_update_features(struct net_device *dev);
Michał Mirosław5455c692011-02-15 16:59:17 +00004067void netdev_update_features(struct net_device *dev);
Michał Mirosławafe12cc2011-05-07 03:22:17 +00004068void netdev_change_features(struct net_device *dev);
Herbert Xu7f353bf2007-08-10 15:47:58 -07004069
Patrick Mullaneyfc4a7482009-12-03 15:59:22 -08004070void netif_stacked_transfer_operstate(const struct net_device *rootdev,
4071 struct net_device *dev);
4072
Toshiaki Makitae38f3022015-03-27 14:31:13 +09004073netdev_features_t passthru_features_check(struct sk_buff *skb,
4074 struct net_device *dev,
4075 netdev_features_t features);
Florian Westphalc1e756b2014-05-05 15:00:44 +02004076netdev_features_t netif_skb_features(struct sk_buff *skb);
Jesse Gross58e998c2010-10-29 12:14:55 +00004077
David S. Miller4d295152012-03-07 21:02:35 -05004078static inline bool net_gso_ok(netdev_features_t features, int gso_type)
Herbert Xubcd76112006-06-30 13:36:35 -07004079{
Marcelo Ricardo Leitner7b748342016-04-25 15:13:17 -03004080 netdev_features_t feature = (netdev_features_t)gso_type << NETIF_F_GSO_SHIFT;
Michał Mirosław0345e182011-11-16 14:05:33 +00004081
4082 /* check flags correspondence */
4083 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
Michał Mirosław0345e182011-11-16 14:05:33 +00004084 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4085 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
Alexander Duyckcbc53e02016-04-10 21:44:51 -04004086 BUILD_BUG_ON(SKB_GSO_TCP_FIXEDID != (NETIF_F_TSO_MANGLEID >> NETIF_F_GSO_SHIFT));
Michał Mirosław0345e182011-11-16 14:05:33 +00004087 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4088 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
Tom Herbert4b282522014-06-14 23:23:52 -07004089 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4090 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
Tom Herbert7e133182016-05-18 09:06:10 -07004091 BUILD_BUG_ON(SKB_GSO_IPXIP4 != (NETIF_F_GSO_IPXIP4 >> NETIF_F_GSO_SHIFT));
4092 BUILD_BUG_ON(SKB_GSO_IPXIP6 != (NETIF_F_GSO_IPXIP6 >> NETIF_F_GSO_SHIFT));
Tom Herbert4b282522014-06-14 23:23:52 -07004093 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4094 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 -04004095 BUILD_BUG_ON(SKB_GSO_PARTIAL != (NETIF_F_GSO_PARTIAL >> NETIF_F_GSO_SHIFT));
Tom Herberte585f232014-11-04 09:06:54 -08004096 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
Marcelo Ricardo Leitner90017ac2016-06-02 15:05:43 -03004097 BUILD_BUG_ON(SKB_GSO_SCTP != (NETIF_F_GSO_SCTP >> NETIF_F_GSO_SHIFT));
Steffen Klassertc7ef8f02017-04-14 10:05:36 +02004098 BUILD_BUG_ON(SKB_GSO_ESP != (NETIF_F_GSO_ESP >> NETIF_F_GSO_SHIFT));
Michał Mirosław0345e182011-11-16 14:05:33 +00004099
Herbert Xubcd76112006-06-30 13:36:35 -07004100 return (features & feature) == feature;
4101}
4102
David S. Miller4d295152012-03-07 21:02:35 -05004103static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
Herbert Xu576a30e2006-06-27 13:22:38 -07004104{
Herbert Xu278b2512009-06-03 21:20:51 -07004105 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
David S. Miller21dc3302010-08-23 00:13:46 -07004106 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
Herbert Xu576a30e2006-06-27 13:22:38 -07004107}
4108
Johannes Berg8b86a612015-04-17 15:45:04 +02004109static inline bool netif_needs_gso(struct sk_buff *skb,
David S. Miller4d295152012-03-07 21:02:35 -05004110 netdev_features_t features)
Herbert Xu79671682006-06-22 02:40:14 -07004111{
Jesse Grossfc741212011-01-09 06:23:32 +00004112 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
Yi Zoucdbee742012-03-16 23:08:11 +00004113 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4114 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
Herbert Xu79671682006-06-22 02:40:14 -07004115}
4116
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07004117static inline void netif_set_gso_max_size(struct net_device *dev,
4118 unsigned int size)
4119{
4120 dev->gso_max_size = size;
4121}
4122
Wei-Chun Chao7a7ffba2013-12-26 13:10:22 -08004123static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4124 int pulled_hlen, u16 mac_offset,
4125 int mac_len)
4126{
4127 skb->protocol = protocol;
4128 skb->encapsulation = 1;
4129 skb_push(skb, pulled_hlen);
4130 skb_reset_transport_header(skb);
4131 skb->mac_header = mac_offset;
4132 skb->network_header = skb->mac_header + mac_len;
4133 skb->mac_len = mac_len;
4134}
4135
Sabrina Dubroca3c175782016-03-11 18:07:32 +01004136static inline bool netif_is_macsec(const struct net_device *dev)
4137{
4138 return dev->priv_flags & IFF_MACSEC;
4139}
4140
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004141static inline bool netif_is_macvlan(const struct net_device *dev)
John Fastabenda6cc0cf2013-11-06 09:54:46 -08004142{
4143 return dev->priv_flags & IFF_MACVLAN;
4144}
4145
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004146static inline bool netif_is_macvlan_port(const struct net_device *dev)
Mahesh Bandewar2f33e7d2014-12-06 15:53:04 -08004147{
4148 return dev->priv_flags & IFF_MACVLAN_PORT;
4149}
4150
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004151static inline bool netif_is_ipvlan(const struct net_device *dev)
Mahesh Bandewar5933fea2014-12-06 15:53:33 -08004152{
4153 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4154}
4155
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004156static inline bool netif_is_ipvlan_port(const struct net_device *dev)
Mahesh Bandewar5933fea2014-12-06 15:53:33 -08004157{
4158 return dev->priv_flags & IFF_IPVLAN_MASTER;
4159}
4160
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004161static inline bool netif_is_bond_master(const struct net_device *dev)
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00004162{
4163 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4164}
4165
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004166static inline bool netif_is_bond_slave(const struct net_device *dev)
Jiri Pirko1765a572011-02-12 06:48:36 +00004167{
4168 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4169}
4170
Ben Greear3bdc0eb2012-02-11 15:39:30 +00004171static inline bool netif_supports_nofcs(struct net_device *dev)
4172{
4173 return dev->priv_flags & IFF_SUPP_NOFCS;
4174}
4175
David Ahern007979e2015-09-29 20:07:10 -07004176static inline bool netif_is_l3_master(const struct net_device *dev)
David Ahern4e3c8992015-08-13 14:59:00 -06004177{
David Ahern007979e2015-09-29 20:07:10 -07004178 return dev->priv_flags & IFF_L3MDEV_MASTER;
David Ahern4e3c8992015-08-13 14:59:00 -06004179}
4180
David Ahernfee6d4c2015-10-05 08:51:24 -07004181static inline bool netif_is_l3_slave(const struct net_device *dev)
4182{
4183 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4184}
4185
Jiri Pirko0894ae32015-08-27 09:31:19 +02004186static inline bool netif_is_bridge_master(const struct net_device *dev)
4187{
4188 return dev->priv_flags & IFF_EBRIDGE;
4189}
4190
Vlad Yasevich28f9ee22015-11-16 15:43:45 -05004191static inline bool netif_is_bridge_port(const struct net_device *dev)
4192{
4193 return dev->priv_flags & IFF_BRIDGE_PORT;
4194}
4195
Jiri Pirko35d4e172015-08-27 09:31:20 +02004196static inline bool netif_is_ovs_master(const struct net_device *dev)
4197{
4198 return dev->priv_flags & IFF_OPENVSWITCH;
4199}
4200
Jiri Pirko5be66142017-04-18 16:55:36 +02004201static inline bool netif_is_ovs_port(const struct net_device *dev)
4202{
4203 return dev->priv_flags & IFF_OVS_DATAPATH;
4204}
4205
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004206static inline bool netif_is_team_master(const struct net_device *dev)
Jiri Pirkoc981e422015-12-03 12:12:06 +01004207{
4208 return dev->priv_flags & IFF_TEAM;
4209}
4210
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004211static inline bool netif_is_team_port(const struct net_device *dev)
Jiri Pirkof7f019e2015-12-03 12:12:07 +01004212{
4213 return dev->priv_flags & IFF_TEAM_PORT;
4214}
4215
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004216static inline bool netif_is_lag_master(const struct net_device *dev)
Jiri Pirko7be61832015-12-03 12:12:08 +01004217{
4218 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4219}
4220
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004221static inline bool netif_is_lag_port(const struct net_device *dev)
Jiri Pirkoe0ba1412015-12-03 12:12:09 +01004222{
4223 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4224}
4225
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08004226static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4227{
4228 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4229}
4230
Eric Dumazet02875872014-10-05 18:38:35 -07004231/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4232static inline void netif_keep_dst(struct net_device *dev)
4233{
4234 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4235}
4236
Paolo Abeni18d3df32016-07-14 18:00:10 +02004237/* return true if dev can't cope with mtu frames that need vlan tag insertion */
4238static inline bool netif_reduces_vlan_mtu(struct net_device *dev)
4239{
4240 /* TODO: reserve and use an additional IFF bit, if we get more users */
4241 return dev->priv_flags & IFF_MACSEC;
4242}
4243
Eric W. Biederman505d4f72008-11-07 22:54:20 -08004244extern struct pernet_operations __net_initdata loopback_net_ops;
Patrick McHardyb1b67dd2009-04-20 04:49:28 +00004245
Joe Perches571ba422010-02-09 11:49:47 +00004246/* Logging, debugging and troubleshooting/diagnostic helpers. */
4247
4248/* netdev_printk helpers, similar to dev_printk */
4249
4250static inline const char *netdev_name(const struct net_device *dev)
4251{
Veaceslav Falicoc6f854d2014-07-17 19:46:09 +02004252 if (!dev->name[0] || strchr(dev->name, '%'))
4253 return "(unnamed net_device)";
Joe Perches571ba422010-02-09 11:49:47 +00004254 return dev->name;
4255}
4256
David Ahern8397ed32017-06-07 12:26:23 -06004257static inline bool netdev_unregistering(const struct net_device *dev)
4258{
4259 return dev->reg_state == NETREG_UNREGISTERING;
4260}
4261
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02004262static inline const char *netdev_reg_state(const struct net_device *dev)
4263{
4264 switch (dev->reg_state) {
4265 case NETREG_UNINITIALIZED: return " (uninitialized)";
4266 case NETREG_REGISTERED: return "";
4267 case NETREG_UNREGISTERING: return " (unregistering)";
4268 case NETREG_UNREGISTERED: return " (unregistered)";
4269 case NETREG_RELEASED: return " (released)";
4270 case NETREG_DUMMY: return " (dummy)";
4271 }
4272
4273 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4274 return " (unknown)";
4275}
4276
Joe Perchesf629d202013-09-26 14:48:15 -07004277__printf(3, 4)
Joe Perches6ea754e2014-09-22 11:10:50 -07004278void netdev_printk(const char *level, const struct net_device *dev,
4279 const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004280__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004281void netdev_emerg(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004282__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004283void netdev_alert(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004284__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004285void netdev_crit(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004286__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004287void netdev_err(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004288__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004289void netdev_warn(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004290__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004291void netdev_notice(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004292__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004293void netdev_info(const struct net_device *dev, const char *format, ...);
Joe Perches571ba422010-02-09 11:49:47 +00004294
Vasiliy Kulikov8909c9a2011-03-02 00:33:13 +03004295#define MODULE_ALIAS_NETDEV(device) \
4296 MODULE_ALIAS("netdev-" device)
4297
Jim Cromieb558c962011-12-19 17:11:18 -05004298#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perches571ba422010-02-09 11:49:47 +00004299#define netdev_dbg(__dev, format, args...) \
4300do { \
Jason Baronffa10cb2011-08-11 14:36:48 -04004301 dynamic_netdev_dbg(__dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00004302} while (0)
Jim Cromieb558c962011-12-19 17:11:18 -05004303#elif defined(DEBUG)
4304#define netdev_dbg(__dev, format, args...) \
4305 netdev_printk(KERN_DEBUG, __dev, format, ##args)
Joe Perches571ba422010-02-09 11:49:47 +00004306#else
4307#define netdev_dbg(__dev, format, args...) \
4308({ \
4309 if (0) \
4310 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00004311})
4312#endif
4313
4314#if defined(VERBOSE_DEBUG)
4315#define netdev_vdbg netdev_dbg
4316#else
4317
4318#define netdev_vdbg(dev, format, args...) \
4319({ \
4320 if (0) \
4321 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4322 0; \
4323})
4324#endif
4325
4326/*
4327 * netdev_WARN() acts like dev_printk(), but with the key difference
4328 * of using a WARN/WARN_ON to get the message out, including the
4329 * file/line information and a backtrace.
4330 */
4331#define netdev_WARN(dev, format, args...) \
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02004332 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4333 netdev_reg_state(dev), ##args)
Joe Perches571ba422010-02-09 11:49:47 +00004334
Joe Perchesb3d95c52010-02-09 11:49:49 +00004335/* netif printk helpers, similar to netdev_printk */
4336
4337#define netif_printk(priv, type, level, dev, fmt, args...) \
4338do { \
4339 if (netif_msg_##type(priv)) \
4340 netdev_printk(level, (dev), fmt, ##args); \
4341} while (0)
4342
Joe Perchesf45f4322010-06-27 01:02:36 +00004343#define netif_level(level, priv, type, dev, fmt, args...) \
4344do { \
4345 if (netif_msg_##type(priv)) \
4346 netdev_##level(dev, fmt, ##args); \
4347} while (0)
4348
Joe Perchesb3d95c52010-02-09 11:49:49 +00004349#define netif_emerg(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004350 netif_level(emerg, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004351#define netif_alert(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004352 netif_level(alert, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004353#define netif_crit(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004354 netif_level(crit, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004355#define netif_err(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004356 netif_level(err, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004357#define netif_warn(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004358 netif_level(warn, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004359#define netif_notice(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004360 netif_level(notice, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004361#define netif_info(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004362 netif_level(info, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004363
Joe Perches0053ea92012-05-30 07:43:34 +00004364#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004365#define netif_dbg(priv, type, netdev, format, args...) \
4366do { \
4367 if (netif_msg_##type(priv)) \
Jason Baronb5fb0a02011-08-11 14:36:53 -04004368 dynamic_netdev_dbg(netdev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00004369} while (0)
Joe Perches0053ea92012-05-30 07:43:34 +00004370#elif defined(DEBUG)
4371#define netif_dbg(priv, type, dev, format, args...) \
4372 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004373#else
4374#define netif_dbg(priv, type, dev, format, args...) \
4375({ \
4376 if (0) \
4377 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4378 0; \
4379})
4380#endif
4381
Edward Creef617f272017-01-27 15:02:26 +00004382/* if @cond then downgrade to debug, else print at @level */
4383#define netif_cond_dbg(priv, type, netdev, cond, level, fmt, args...) \
4384 do { \
4385 if (cond) \
4386 netif_dbg(priv, type, netdev, fmt, ##args); \
4387 else \
4388 netif_ ## level(priv, type, netdev, fmt, ##args); \
4389 } while (0)
4390
Joe Perchesb3d95c52010-02-09 11:49:49 +00004391#if defined(VERBOSE_DEBUG)
Ben Hutchingsbcfcc452010-07-02 07:08:44 +00004392#define netif_vdbg netif_dbg
Joe Perchesb3d95c52010-02-09 11:49:49 +00004393#else
4394#define netif_vdbg(priv, type, dev, format, args...) \
4395({ \
4396 if (0) \
Ben Hutchingsa4ed89c2010-05-18 06:56:32 +00004397 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00004398 0; \
4399})
4400#endif
Joe Perches571ba422010-02-09 11:49:47 +00004401
Cong Wang900ff8c2013-02-18 19:20:33 +00004402/*
4403 * The list of packet types we will receive (as opposed to discard)
4404 * and the routines to invoke.
4405 *
4406 * Why 16. Because with 16 the only overlap we get on a hash of the
4407 * low nibble of the protocol value is RARP/SNAP/X.25.
4408 *
4409 * NOTE: That is no longer true with the addition of VLAN tags. Not
4410 * sure which should go first, but I bet it won't make much
4411 * difference if we are running VLANs. The good news is that
4412 * this protocol won't be in the list unless compiled in, so
4413 * the average user (w/out VLANs) will not be adversely affected.
4414 * --BLG
4415 *
4416 * 0800 IP
4417 * 8100 802.1Q VLAN
4418 * 0001 802.3
4419 * 0002 AX.25
4420 * 0004 802.2
4421 * 8035 RARP
4422 * 0005 SNAP
4423 * 0805 X.25
4424 * 0806 ARP
4425 * 8137 IPX
4426 * 0009 Localtalk
4427 * 86DD IPv6
4428 */
4429#define PTYPE_HASH_SIZE (16)
4430#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4431
Jiri Pirko385a1542009-05-27 15:48:07 -07004432#endif /* _LINUX_NETDEVICE_H */