blob: f9eebd518545f2767e8f906d9d011f904853bd67 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * INET An implementation of the TCP/IP protocol suite for the LINUX
3 * operating system. INET is implemented using the BSD Socket
4 * interface as the means of communication with the user level.
5 *
6 * Definitions for the Interfaces handler.
7 *
8 * Version: @(#)dev.h 1.0.10 08/12/93
9 *
Jesper Juhl02c30a82005-05-05 16:16:16 -070010 * Authors: Ross Biro
Linus Torvalds1da177e2005-04-16 15:20:36 -070011 * Fred N. van Kempen, <waltje@uWalt.NL.Mugnet.ORG>
12 * Corey Minyard <wf-rch!minyard@relay.EU.net>
13 * Donald J. Becker, <becker@cesdis.gsfc.nasa.gov>
Alan Cox113aa832008-10-13 19:01:08 -070014 * Alan Cox, <alan@lxorguk.ukuu.org.uk>
Linus Torvalds1da177e2005-04-16 15:20:36 -070015 * Bjorn Ekwall. <bj0rn@blox.se>
16 * Pekka Riikonen <priikone@poseidon.pspt.fi>
17 *
18 * This program is free software; you can redistribute it and/or
19 * modify it under the terms of the GNU General Public License
20 * as published by the Free Software Foundation; either version
21 * 2 of the License, or (at your option) any later version.
22 *
23 * Moved to /usr/include/linux for NET3
24 */
25#ifndef _LINUX_NETDEVICE_H
26#define _LINUX_NETDEVICE_H
27
Al Virod7fe0f22006-12-03 23:15:30 -050028#include <linux/timer.h>
Paul Gortmaker187f1882011-11-23 20:12:59 -050029#include <linux/bug.h>
Stephen Hemmingerbea33482007-10-03 16:41:36 -070030#include <linux/delay.h>
Arun Sharma600634972011-07-26 16:09:06 -070031#include <linux/atomic.h>
Eric Dumazet53511452014-10-08 08:19:27 -070032#include <linux/prefetch.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070033#include <asm/cache.h>
34#include <asm/byteorder.h>
35
Linus Torvalds1da177e2005-04-16 15:20:36 -070036#include <linux/percpu.h>
David S. Miller4d5b78c2009-05-06 16:52:51 -070037#include <linux/rculist.h>
Chris Leechdb217332006-06-17 21:24:58 -070038#include <linux/dmaengine.h>
Stephen Hemmingerbea33482007-10-03 16:41:36 -070039#include <linux/workqueue.h>
Tom Herbert114cf582011-11-28 16:33:09 +000040#include <linux/dynamic_queue_limits.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070041
Patrick McHardyb1b67dd2009-04-20 04:49:28 +000042#include <linux/ethtool.h>
Daniel Lezcanoa050c332007-09-12 14:57:09 +020043#include <net/net_namespace.h>
Lennert Buytenhekcf85d082008-10-07 13:45:02 +000044#include <net/dsa.h>
Jeff Kirsher7a6b6f52008-11-25 01:02:08 -080045#ifdef CONFIG_DCB
Alexander Duyck2f90b862008-11-20 20:52:10 -080046#include <net/dcbnl.h>
47#endif
Neil Horman5bc14212011-11-22 05:10:51 +000048#include <net/netprio_cgroup.h>
Daniel Lezcanoa050c332007-09-12 14:57:09 +020049
Michał Mirosława59e2ec2011-11-15 15:29:55 +000050#include <linux/netdev_features.h>
John Fastabend77162022012-04-15 06:43:56 +000051#include <linux/neighbour.h>
David Howells607ca462012-10-13 10:46:48 +010052#include <uapi/linux/netdevice.h>
Moni Shoua61bd3852015-02-03 16:48:29 +020053#include <uapi/linux/if_bonding.h>
John Fastabende4c67342016-02-16 21:16:15 -080054#include <uapi/linux/pkt_cls.h>
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;
Johannes Berg704232c2007-04-23 12:20:05 -070059/* 802.11 specific */
60struct wireless_dev;
Alexander Aring98a18b62014-11-02 06:44:54 +010061/* 802.15.4 specific */
62struct wpan_dev;
Robert Shearman03c57742015-04-22 11:14:37 +010063struct mpls_dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -070064
Joe Perchesf629d202013-09-26 14:48:15 -070065void netdev_set_default_ethtool_ops(struct net_device *dev,
66 const struct ethtool_ops *ops);
Stanislaw Gruszkad07d7502013-01-10 23:19:10 +000067
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000068/* Backlog congestion levels */
69#define NET_RX_SUCCESS 0 /* keep 'em coming, baby */
70#define NET_RX_DROP 1 /* packet dropped */
71
Patrick McHardy572a9d72009-11-10 06:14:14 +000072/*
73 * Transmit return codes: transmit return codes originate from three different
74 * namespaces:
75 *
76 * - qdisc return codes
77 * - driver transmit return codes
78 * - errno values
79 *
80 * Drivers are allowed to return any one of those in their hard_start_xmit()
81 * function. Real network devices commonly used with qdiscs should only return
82 * the driver transmit return codes though - when qdiscs are used, the actual
83 * transmission happens asynchronously, so the value is not propagated to
84 * higher layers. Virtual network devices transmit synchronously, in this case
85 * the driver transmit return codes are consumed by dev_queue_xmit(), all
86 * others are propagated to higher layers.
87 */
88
89/* qdisc ->enqueue() return codes. */
90#define NET_XMIT_SUCCESS 0x00
Jarek Poplawski9a1654b2009-11-15 07:20:12 +000091#define NET_XMIT_DROP 0x01 /* skb dropped */
92#define NET_XMIT_CN 0x02 /* congestion notification */
93#define NET_XMIT_POLICED 0x03 /* skb is shot by police */
94#define NET_XMIT_MASK 0x0f /* qdisc flags in net/sch_generic.h */
Linus Torvalds1da177e2005-04-16 15:20:36 -070095
Gerrit Renkerb9df3cb2006-11-14 11:21:36 -020096/* NET_XMIT_CN is special. It does not guarantee that this packet is lost. It
97 * indicates that the device will soon be dropping packets, or already drops
98 * some packets of the same priority; prompting us to send less aggressively. */
Patrick McHardy572a9d72009-11-10 06:14:14 +000099#define net_xmit_eval(e) ((e) == NET_XMIT_CN ? 0 : (e))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100#define net_xmit_errno(e) ((e) != NET_XMIT_CN ? -ENOBUFS : 0)
101
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000102/* Driver transmit return codes */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000103#define NETDEV_TX_MASK 0xf0
Patrick McHardy572a9d72009-11-10 06:14:14 +0000104
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000105enum netdev_tx {
Patrick McHardy572a9d72009-11-10 06:14:14 +0000106 __NETDEV_TX_MIN = INT_MIN, /* make sure enum is signed */
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000107 NETDEV_TX_OK = 0x00, /* driver took care of packet */
108 NETDEV_TX_BUSY = 0x10, /* driver tx path was busy*/
109 NETDEV_TX_LOCKED = 0x20, /* driver tx lock was already taken */
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000110};
111typedef enum netdev_tx netdev_tx_t;
112
Jarek Poplawski9a1654b2009-11-15 07:20:12 +0000113/*
114 * Current order: NETDEV_TX_MASK > NET_XMIT_MASK >= 0 is significant;
115 * hard_start_xmit() return < NET_XMIT_MASK means skb was consumed.
116 */
117static inline bool dev_xmit_complete(int rc)
118{
119 /*
120 * Positive cases with an skb consumed by a driver:
121 * - successful transmission (rc == NETDEV_TX_OK)
122 * - error while transmitting (rc < 0)
123 * - error while queueing to a different device (rc & NET_XMIT_MASK)
124 */
125 if (likely(rc < NET_XMIT_MASK))
126 return true;
127
128 return false;
129}
130
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131/*
132 * Compute the worst case header length according to the protocols
133 * used.
134 */
Graf Yangfe2918b2009-02-05 21:26:19 -0800135
KY Srinivasanc0eb4542015-12-01 16:43:10 -0800136#if defined(CONFIG_HYPERV_NET)
137# define LL_MAX_HEADER 128
138#elif defined(CONFIG_WLAN) || IS_ENABLED(CONFIG_AX25)
David S. Miller8388e3d2008-05-12 20:17:33 -0700139# if defined(CONFIG_MAC80211_MESH)
140# define LL_MAX_HEADER 128
141# else
142# define LL_MAX_HEADER 96
143# endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700144#else
David S. Miller8388e3d2008-05-12 20:17:33 -0700145# define LL_MAX_HEADER 32
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146#endif
147
Ben Hutchingsd11ead72011-11-25 14:40:26 +0000148#if !IS_ENABLED(CONFIG_NET_IPIP) && !IS_ENABLED(CONFIG_NET_IPGRE) && \
149 !IS_ENABLED(CONFIG_IPV6_SIT) && !IS_ENABLED(CONFIG_IPV6_TUNNEL)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700150#define MAX_HEADER LL_MAX_HEADER
151#else
152#define MAX_HEADER (LL_MAX_HEADER + 48)
153#endif
154
155/*
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000156 * Old network device statistics. Fields are native words
157 * (unsigned long) so they can be read and written atomically.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158 */
Graf Yangfe2918b2009-02-05 21:26:19 -0800159
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800160struct net_device_stats {
Ben Hutchings3cfde792010-07-09 09:11:52 +0000161 unsigned long rx_packets;
162 unsigned long tx_packets;
163 unsigned long rx_bytes;
164 unsigned long tx_bytes;
165 unsigned long rx_errors;
166 unsigned long tx_errors;
167 unsigned long rx_dropped;
168 unsigned long tx_dropped;
169 unsigned long multicast;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170 unsigned long collisions;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171 unsigned long rx_length_errors;
Ben Hutchings3cfde792010-07-09 09:11:52 +0000172 unsigned long rx_over_errors;
173 unsigned long rx_crc_errors;
174 unsigned long rx_frame_errors;
175 unsigned long rx_fifo_errors;
176 unsigned long rx_missed_errors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700177 unsigned long tx_aborted_errors;
178 unsigned long tx_carrier_errors;
179 unsigned long tx_fifo_errors;
180 unsigned long tx_heartbeat_errors;
181 unsigned long tx_window_errors;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182 unsigned long rx_compressed;
183 unsigned long tx_compressed;
184};
185
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186
187#include <linux/cache.h>
188#include <linux/skbuff.h>
189
Eric Dumazetadc93002011-11-17 03:13:26 +0000190#ifdef CONFIG_RPS
Ingo Molnarc5905af2012-02-24 08:31:31 +0100191#include <linux/static_key.h>
192extern struct static_key rps_needed;
Eric Dumazetadc93002011-11-17 03:13:26 +0000193#endif
194
Linus Torvalds1da177e2005-04-16 15:20:36 -0700195struct neighbour;
196struct neigh_parms;
197struct sk_buff;
198
Jiri Pirkof001fde2009-05-05 02:48:28 +0000199struct netdev_hw_addr {
200 struct list_head list;
201 unsigned char addr[MAX_ADDR_LEN];
202 unsigned char type;
Jiri Pirkoccffad252009-05-22 23:22:17 +0000203#define NETDEV_HW_ADDR_T_LAN 1
204#define NETDEV_HW_ADDR_T_SAN 2
205#define NETDEV_HW_ADDR_T_SLAVE 3
206#define NETDEV_HW_ADDR_T_UNICAST 4
Jiri Pirko22bedad32010-04-01 21:22:57 +0000207#define NETDEV_HW_ADDR_T_MULTICAST 5
Jiri Pirko22bedad32010-04-01 21:22:57 +0000208 bool global_use;
Vlad Yasevich4cd729b02013-04-15 09:54:25 +0000209 int sync_cnt;
Eric Dumazet8f8f1032010-09-19 11:24:02 -0700210 int refcount;
Vlad Yasevich4543fbe2013-04-02 17:10:07 -0400211 int synced;
Jiri Pirkof001fde2009-05-05 02:48:28 +0000212 struct rcu_head rcu_head;
213};
214
Jiri Pirko31278e72009-06-17 01:12:19 +0000215struct netdev_hw_addr_list {
216 struct list_head list;
217 int count;
218};
219
Jiri Pirko22bedad32010-04-01 21:22:57 +0000220#define netdev_hw_addr_list_count(l) ((l)->count)
221#define netdev_hw_addr_list_empty(l) (netdev_hw_addr_list_count(l) == 0)
222#define netdev_hw_addr_list_for_each(ha, l) \
223 list_for_each_entry(ha, &(l)->list, list)
224
225#define netdev_uc_count(dev) netdev_hw_addr_list_count(&(dev)->uc)
226#define netdev_uc_empty(dev) netdev_hw_addr_list_empty(&(dev)->uc)
Jiri Pirko32e7bfc2010-01-25 13:36:10 -0800227#define netdev_for_each_uc_addr(ha, dev) \
Jiri Pirko22bedad32010-04-01 21:22:57 +0000228 netdev_hw_addr_list_for_each(ha, &(dev)->uc)
Jiri Pirko32e7bfc2010-01-25 13:36:10 -0800229
Jiri Pirko22bedad32010-04-01 21:22:57 +0000230#define netdev_mc_count(dev) netdev_hw_addr_list_count(&(dev)->mc)
231#define netdev_mc_empty(dev) netdev_hw_addr_list_empty(&(dev)->mc)
Pavel Roskin18e225f2010-04-07 16:40:09 -0700232#define netdev_for_each_mc_addr(ha, dev) \
Jiri Pirko22bedad32010-04-01 21:22:57 +0000233 netdev_hw_addr_list_for_each(ha, &(dev)->mc)
Jiri Pirko6683ece32010-02-04 10:22:25 -0800234
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800235struct hh_cache {
David S. Millerf6b72b622011-07-14 07:53:20 -0700236 u16 hh_len;
David S. Miller5c25f682011-07-13 00:51:10 -0700237 u16 __pad;
Stephen Hemminger3644f0c2006-12-07 15:08:17 -0800238 seqlock_t hh_lock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239
240 /* cached hardware header; allow for machine alignment needs. */
241#define HH_DATA_MOD 16
242#define HH_DATA_OFF(__len) \
Jiri Benc5ba0eac2005-06-02 16:48:05 -0700243 (HH_DATA_MOD - (((__len - 1) & (HH_DATA_MOD - 1)) + 1))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244#define HH_DATA_ALIGN(__len) \
245 (((__len)+(HH_DATA_MOD-1))&~(HH_DATA_MOD - 1))
246 unsigned long hh_data[HH_DATA_ALIGN(LL_MAX_HEADER) / sizeof(long)];
247};
248
249/* Reserve HH_DATA_MOD byte aligned hard_header_len, but at least that much.
250 * Alternative is:
251 * dev->hard_header_len ? (dev->hard_header_len +
252 * (HH_DATA_MOD - 1)) & ~(HH_DATA_MOD - 1) : 0
253 *
254 * We could use other alignment values, but we must maintain the
255 * relationship HH alignment <= LL alignment.
256 */
257#define LL_RESERVED_SPACE(dev) \
Johannes Bergf5184d22008-05-12 20:48:31 -0700258 ((((dev)->hard_header_len+(dev)->needed_headroom)&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259#define LL_RESERVED_SPACE_EXTRA(dev,extra) \
Johannes Bergf5184d22008-05-12 20:48:31 -0700260 ((((dev)->hard_header_len+(dev)->needed_headroom+(extra))&~(HH_DATA_MOD - 1)) + HH_DATA_MOD)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700262struct header_ops {
263 int (*create) (struct sk_buff *skb, struct net_device *dev,
264 unsigned short type, const void *daddr,
Eric Dumazet95c96172012-04-15 05:58:06 +0000265 const void *saddr, unsigned int len);
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700266 int (*parse)(const struct sk_buff *skb, unsigned char *haddr);
David S. Millere69dd332011-07-12 23:28:12 -0700267 int (*cache)(const struct neighbour *neigh, struct hh_cache *hh, __be16 type);
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700268 void (*cache_update)(struct hh_cache *hh,
269 const struct net_device *dev,
270 const unsigned char *haddr);
Willem de Bruijn2793a232016-03-09 21:58:32 -0500271 bool (*validate)(const char *ll_header, unsigned int len);
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -0700272};
273
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274/* These flag bits are private to the generic network queueing
275 * layer, they may not be explicitly referenced by any other
276 * code.
277 */
278
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800279enum netdev_state_t {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280 __LINK_STATE_START,
281 __LINK_STATE_PRESENT,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282 __LINK_STATE_NOCARRIER,
Stefan Rompfb00055a2006-03-20 17:09:11 -0800283 __LINK_STATE_LINKWATCH_PENDING,
284 __LINK_STATE_DORMANT,
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285};
286
287
288/*
289 * This structure holds at boot time configured netdevice settings. They
Graf Yangfe2918b2009-02-05 21:26:19 -0800290 * are then used in the device probing.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700291 */
292struct netdev_boot_setup {
293 char name[IFNAMSIZ];
294 struct ifmap map;
295};
296#define NETDEV_BOOT_SETUP_MAX 8
297
Joe Perchesf629d202013-09-26 14:48:15 -0700298int __init netdev_boot_setup(char *str);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299
300/*
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700301 * Structure for NAPI scheduling similar to tasklet but with weighting
302 */
303struct napi_struct {
304 /* The poll_list must only be managed by the entity which
305 * changes the state of the NAPI_STATE_SCHED bit. This means
306 * whoever atomically sets that bit can add this napi_struct
307 * to the per-cpu poll_list, and whoever clears that bit
308 * can remove from the list right before clearing the bit.
309 */
310 struct list_head poll_list;
311
312 unsigned long state;
313 int weight;
Eric Dumazet404f7c92012-09-26 07:07:47 +0000314 unsigned int gro_count;
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700315 int (*poll)(struct napi_struct *, int);
316#ifdef CONFIG_NETPOLL
317 spinlock_t poll_lock;
318 int poll_owner;
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700319#endif
Herbert Xu5d38a072009-01-04 16:13:40 -0800320 struct net_device *dev;
Herbert Xud565b0a2008-12-15 23:38:52 -0800321 struct sk_buff *gro_list;
Herbert Xu5d38a072009-01-04 16:13:40 -0800322 struct sk_buff *skb;
Eric Dumazet3b47d302014-11-06 21:09:44 -0800323 struct hrtimer timer;
Eric Dumazet404f7c92012-09-26 07:07:47 +0000324 struct list_head dev_list;
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300325 struct hlist_node napi_hash_node;
326 unsigned int napi_id;
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700327};
328
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800329enum {
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700330 NAPI_STATE_SCHED, /* Poll is scheduled */
David S. Millera0a46192008-01-07 20:35:07 -0800331 NAPI_STATE_DISABLE, /* Disable pending */
Neil Horman7b363e42008-12-09 23:22:26 -0800332 NAPI_STATE_NPSVC, /* Netpoll - don't dequeue from poll_list */
Eric Dumazetd64b5e82015-11-18 06:31:00 -0800333 NAPI_STATE_HASHED, /* In NAPI hash (busy polling possible) */
334 NAPI_STATE_NO_BUSY_POLL,/* Do not add in napi_hash, no busy polling */
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700335};
336
Ben Hutchings5b252f02009-10-29 07:17:09 +0000337enum gro_result {
Herbert Xud1c76af2009-03-16 10:50:02 -0700338 GRO_MERGED,
339 GRO_MERGED_FREE,
340 GRO_HELD,
341 GRO_NORMAL,
342 GRO_DROP,
343};
Ben Hutchings5b252f02009-10-29 07:17:09 +0000344typedef enum gro_result gro_result_t;
Herbert Xud1c76af2009-03-16 10:50:02 -0700345
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000346/*
347 * enum rx_handler_result - Possible return values for rx_handlers.
348 * @RX_HANDLER_CONSUMED: skb was consumed by rx_handler, do not process it
349 * further.
350 * @RX_HANDLER_ANOTHER: Do another round in receive path. This is indicated in
351 * case skb->dev was changed by rx_handler.
352 * @RX_HANDLER_EXACT: Force exact delivery, no wildcard.
353 * @RX_HANDLER_PASS: Do nothing, passe the skb as if no rx_handler was called.
354 *
355 * rx_handlers are functions called from inside __netif_receive_skb(), to do
356 * special processing of the skb, prior to delivery to protocol handlers.
357 *
358 * Currently, a net_device can only have a single rx_handler registered. Trying
359 * to register a second rx_handler will return -EBUSY.
360 *
361 * To register a rx_handler on a net_device, use netdev_rx_handler_register().
362 * To unregister a rx_handler on a net_device, use
363 * netdev_rx_handler_unregister().
364 *
365 * Upon return, rx_handler is expected to tell __netif_receive_skb() what to
366 * do with the skb.
367 *
368 * If the rx_handler consumed to skb in some way, it should return
369 * RX_HANDLER_CONSUMED. This is appropriate when the rx_handler arranged for
370 * the skb to be delivered in some other ways.
371 *
372 * If the rx_handler changed skb->dev, to divert the skb to another
373 * net_device, it should return RX_HANDLER_ANOTHER. The rx_handler for the
374 * new device will be called if it exists.
375 *
376 * If the rx_handler consider the skb should be ignored, it should return
377 * RX_HANDLER_EXACT. The skb will only be delivered to protocol handlers that
Adam Buchbinderd93cf062012-09-19 21:47:58 -0400378 * are registered on exact device (ptype->dev == skb->dev).
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000379 *
380 * If the rx_handler didn't changed skb->dev, but want the skb to be normally
381 * delivered, it should return RX_HANDLER_PASS.
382 *
383 * A device without a registered rx_handler will behave as if rx_handler
384 * returned RX_HANDLER_PASS.
385 */
386
387enum rx_handler_result {
388 RX_HANDLER_CONSUMED,
389 RX_HANDLER_ANOTHER,
390 RX_HANDLER_EXACT,
391 RX_HANDLER_PASS,
392};
393typedef enum rx_handler_result rx_handler_result_t;
394typedef rx_handler_result_t rx_handler_func_t(struct sk_buff **pskb);
Jiri Pirkoab95bfe2010-06-01 21:52:08 +0000395
Joe Perchesf629d202013-09-26 14:48:15 -0700396void __napi_schedule(struct napi_struct *n);
Eric Dumazetbc9ad162014-10-28 18:05:13 -0700397void __napi_schedule_irqoff(struct napi_struct *n);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700398
David S. Miller4d295152012-03-07 21:02:35 -0500399static inline bool napi_disable_pending(struct napi_struct *n)
David S. Millera0a46192008-01-07 20:35:07 -0800400{
401 return test_bit(NAPI_STATE_DISABLE, &n->state);
402}
403
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700404/**
405 * napi_schedule_prep - check if napi can be scheduled
406 * @n: napi context
407 *
408 * Test if NAPI routine is already running, and if not mark
409 * it as running. This is used as a condition variable
David S. Millera0a46192008-01-07 20:35:07 -0800410 * insure only one NAPI poll instance runs. We also make
411 * sure there is no pending NAPI disable.
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700412 */
David S. Miller4d295152012-03-07 21:02:35 -0500413static inline bool napi_schedule_prep(struct napi_struct *n)
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700414{
David S. Millera0a46192008-01-07 20:35:07 -0800415 return !napi_disable_pending(n) &&
416 !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700417}
418
419/**
420 * napi_schedule - schedule NAPI poll
421 * @n: napi context
422 *
423 * Schedule NAPI poll routine to be called if it is not already
424 * running.
425 */
426static inline void napi_schedule(struct napi_struct *n)
427{
428 if (napi_schedule_prep(n))
429 __napi_schedule(n);
430}
431
Eric Dumazetbc9ad162014-10-28 18:05:13 -0700432/**
433 * napi_schedule_irqoff - schedule NAPI poll
434 * @n: napi context
435 *
436 * Variant of napi_schedule(), assuming hard irqs are masked.
437 */
438static inline void napi_schedule_irqoff(struct napi_struct *n)
439{
440 if (napi_schedule_prep(n))
441 __napi_schedule_irqoff(n);
442}
443
Roland Dreierbfe13f52007-10-09 15:47:37 -0700444/* Try to reschedule poll. Called by dev->poll() after napi_complete(). */
David S. Miller4d295152012-03-07 21:02:35 -0500445static inline bool napi_reschedule(struct napi_struct *napi)
Roland Dreierbfe13f52007-10-09 15:47:37 -0700446{
447 if (napi_schedule_prep(napi)) {
448 __napi_schedule(napi);
David S. Miller4d295152012-03-07 21:02:35 -0500449 return true;
Roland Dreierbfe13f52007-10-09 15:47:37 -0700450 }
David S. Miller4d295152012-03-07 21:02:35 -0500451 return false;
Roland Dreierbfe13f52007-10-09 15:47:37 -0700452}
453
Eric Dumazet3b47d302014-11-06 21:09:44 -0800454void __napi_complete(struct napi_struct *n);
455void napi_complete_done(struct napi_struct *n, int work_done);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700456/**
457 * napi_complete - NAPI processing complete
458 * @n: napi context
459 *
460 * Mark NAPI processing as complete.
Eric Dumazet3b47d302014-11-06 21:09:44 -0800461 * Consider using napi_complete_done() instead.
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700462 */
Eric Dumazet3b47d302014-11-06 21:09:44 -0800463static inline void napi_complete(struct napi_struct *n)
464{
465 return napi_complete_done(n, 0);
466}
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700467
468/**
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300469 * napi_hash_add - add a NAPI to global hashtable
470 * @napi: napi context
471 *
472 * generate a new napi_id and store a @napi under it in napi_hash
Eric Dumazet93d05d42015-11-18 06:31:03 -0800473 * Used for busy polling (CONFIG_NET_RX_BUSY_POLL)
474 * Note: This is normally automatically done from netif_napi_add(),
475 * so might disappear in a future linux version.
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300476 */
Joe Perchesf629d202013-09-26 14:48:15 -0700477void napi_hash_add(struct napi_struct *napi);
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300478
479/**
480 * napi_hash_del - remove a NAPI from global table
481 * @napi: napi context
482 *
483 * Warning: caller must observe rcu grace period
Eric Dumazet34cbe272015-11-18 06:31:02 -0800484 * before freeing memory containing @napi, if
485 * this function returns true.
Eric Dumazet93d05d42015-11-18 06:31:03 -0800486 * Note: core networking stack automatically calls it
487 * from netif_napi_del()
488 * Drivers might want to call this helper to combine all
489 * the needed rcu grace periods into a single one.
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300490 */
Eric Dumazet34cbe272015-11-18 06:31:02 -0800491bool napi_hash_del(struct napi_struct *napi);
Eliezer Tamiraf12fa62013-06-10 11:39:41 +0300492
493/**
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700494 * napi_disable - prevent NAPI from scheduling
495 * @n: napi context
496 *
497 * Stop NAPI from being scheduled on this context.
498 * Waits till any outstanding processing completes.
499 */
Eric Dumazet3b47d302014-11-06 21:09:44 -0800500void napi_disable(struct napi_struct *n);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700501
502/**
503 * napi_enable - enable NAPI scheduling
504 * @n: napi context
505 *
506 * Resume NAPI from being scheduled on this context.
507 * Must be paired with napi_disable.
508 */
509static inline void napi_enable(struct napi_struct *n)
510{
511 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
Peter Zijlstra4e857c52014-03-17 18:06:10 +0100512 smp_mb__before_atomic();
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700513 clear_bit(NAPI_STATE_SCHED, &n->state);
Neil Horman2d8bff1262015-09-23 14:57:58 -0400514 clear_bit(NAPI_STATE_NPSVC, &n->state);
Stephen Hemmingerbea33482007-10-03 16:41:36 -0700515}
516
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700517/**
518 * napi_synchronize - wait until NAPI is not running
519 * @n: napi context
520 *
521 * Wait until NAPI is done being scheduled on this context.
522 * Waits till any outstanding processing completes but
523 * does not disable future activations.
524 */
525static inline void napi_synchronize(const struct napi_struct *n)
526{
Arnd Bergmannfacc4322016-01-22 11:43:44 +0100527 if (IS_ENABLED(CONFIG_SMP))
528 while (test_bit(NAPI_STATE_SCHED, &n->state))
529 msleep(1);
530 else
531 barrier();
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700532}
Stephen Hemmingerc264c3d2007-10-17 13:26:41 -0700533
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800534enum netdev_queue_state_t {
Tom Herbert734664982011-11-28 16:32:44 +0000535 __QUEUE_STATE_DRV_XOFF,
536 __QUEUE_STATE_STACK_XOFF,
David S. Millerc3f26a22008-07-31 16:58:50 -0700537 __QUEUE_STATE_FROZEN,
David S. Miller79d16382008-07-08 23:14:46 -0700538};
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +0200539
540#define QUEUE_STATE_DRV_XOFF (1 << __QUEUE_STATE_DRV_XOFF)
541#define QUEUE_STATE_STACK_XOFF (1 << __QUEUE_STATE_STACK_XOFF)
542#define QUEUE_STATE_FROZEN (1 << __QUEUE_STATE_FROZEN)
543
544#define QUEUE_STATE_ANY_XOFF (QUEUE_STATE_DRV_XOFF | QUEUE_STATE_STACK_XOFF)
545#define QUEUE_STATE_ANY_XOFF_OR_FROZEN (QUEUE_STATE_ANY_XOFF | \
546 QUEUE_STATE_FROZEN)
547#define QUEUE_STATE_DRV_XOFF_OR_FROZEN (QUEUE_STATE_DRV_XOFF | \
548 QUEUE_STATE_FROZEN)
549
Tom Herbert734664982011-11-28 16:32:44 +0000550/*
551 * __QUEUE_STATE_DRV_XOFF is used by drivers to stop the transmit queue. The
552 * netif_tx_* functions below are used to manipulate this flag. The
553 * __QUEUE_STATE_STACK_XOFF flag is used by the stack to stop the transmit
554 * queue independently. The netif_xmit_*stopped functions below are called
555 * to check if the queue has been stopped by the driver or stack (either
556 * of the XOFF bits are set in the state). Drivers should not need to call
557 * netif_xmit*stopped functions, they should only be using netif_tx_*.
558 */
David S. Miller79d16382008-07-08 23:14:46 -0700559
David S. Millerbb949fb2008-07-08 16:55:56 -0700560struct netdev_queue {
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700561/*
562 * read mostly part
563 */
David S. Millerbb949fb2008-07-08 16:55:56 -0700564 struct net_device *dev;
John Fastabend46e5da40a2014-09-12 20:04:52 -0700565 struct Qdisc __rcu *qdisc;
David S. Millerb0e1e642008-07-08 17:42:10 -0700566 struct Qdisc *qdisc_sleeping;
david decotignyccf5ff62011-11-16 12:15:10 +0000567#ifdef CONFIG_SYSFS
Tom Herbert1d24eb42010-11-21 13:17:27 +0000568 struct kobject kobj;
569#endif
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000570#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
571 int numa_node;
572#endif
Eric Dumazet6a321cb2009-04-28 04:43:42 -0700573/*
574 * write mostly part
575 */
576 spinlock_t _xmit_lock ____cacheline_aligned_in_smp;
577 int xmit_lock_owner;
Eric Dumazet9d214932009-05-17 20:55:16 -0700578 /*
579 * please use this field instead of dev->trans_start
580 */
581 unsigned long trans_start;
david decotignyccf5ff62011-11-16 12:15:10 +0000582
583 /*
584 * Number of TX timeouts for this queue
585 * (/sys/class/net/DEV/Q/trans_timeout)
586 */
587 unsigned long trans_timeout;
Tom Herbert114cf582011-11-28 16:33:09 +0000588
589 unsigned long state;
590
591#ifdef CONFIG_BQL
592 struct dql dql;
593#endif
John Fastabend822b3b22015-03-18 14:57:33 +0200594 unsigned long tx_maxrate;
David S. Millere8a04642008-07-17 00:34:19 -0700595} ____cacheline_aligned_in_smp;
David S. Millerbb949fb2008-07-08 16:55:56 -0700596
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000597static inline int netdev_queue_numa_node_read(const struct netdev_queue *q)
598{
599#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
600 return q->numa_node;
601#else
Changli Gaob236da62010-12-14 03:09:15 +0000602 return NUMA_NO_NODE;
Eric Dumazetf2cd2d32010-11-29 08:14:37 +0000603#endif
604}
605
606static inline void netdev_queue_numa_node_write(struct netdev_queue *q, int node)
607{
608#if defined(CONFIG_XPS) && defined(CONFIG_NUMA)
609 q->numa_node = node;
610#endif
611}
612
Eric Dumazetdf334542010-03-24 19:13:54 +0000613#ifdef CONFIG_RPS
Tom Herbert0a9627f2010-03-16 08:03:29 +0000614/*
615 * This structure holds an RPS map which can be of variable length. The
616 * map is an array of CPUs.
617 */
618struct rps_map {
619 unsigned int len;
620 struct rcu_head rcu;
621 u16 cpus[0];
622};
Eric Dumazet60b778c2011-12-24 06:56:49 +0000623#define RPS_MAP_SIZE(_num) (sizeof(struct rps_map) + ((_num) * sizeof(u16)))
Tom Herbert0a9627f2010-03-16 08:03:29 +0000624
Tom Herbertfec5e652010-04-16 16:01:27 -0700625/*
Ben Hutchingsc4454772011-01-19 11:03:53 +0000626 * The rps_dev_flow structure contains the mapping of a flow to a CPU, the
627 * tail pointer for that CPU's input queue at the time of last enqueue, and
628 * a hardware filter index.
Tom Herbertfec5e652010-04-16 16:01:27 -0700629 */
630struct rps_dev_flow {
631 u16 cpu;
Ben Hutchingsc4454772011-01-19 11:03:53 +0000632 u16 filter;
Tom Herbertfec5e652010-04-16 16:01:27 -0700633 unsigned int last_qtail;
634};
Ben Hutchingsc4454772011-01-19 11:03:53 +0000635#define RPS_NO_FILTER 0xffff
Tom Herbertfec5e652010-04-16 16:01:27 -0700636
637/*
638 * The rps_dev_flow_table structure contains a table of flow mappings.
639 */
640struct rps_dev_flow_table {
641 unsigned int mask;
642 struct rcu_head rcu;
Tom Herbertfec5e652010-04-16 16:01:27 -0700643 struct rps_dev_flow flows[0];
644};
645#define RPS_DEV_FLOW_TABLE_SIZE(_num) (sizeof(struct rps_dev_flow_table) + \
Eric Dumazet60b778c2011-12-24 06:56:49 +0000646 ((_num) * sizeof(struct rps_dev_flow)))
Tom Herbertfec5e652010-04-16 16:01:27 -0700647
648/*
649 * The rps_sock_flow_table contains mappings of flows to the last CPU
650 * on which they were processed by the application (set in recvmsg).
Eric Dumazet567e4b72015-02-06 12:59:01 -0800651 * Each entry is a 32bit value. Upper part is the high order bits
652 * of flow hash, lower part is cpu number.
653 * rps_cpu_mask is used to partition the space, depending on number of
654 * possible cpus : rps_cpu_mask = roundup_pow_of_two(nr_cpu_ids) - 1
655 * For example, if 64 cpus are possible, rps_cpu_mask = 0x3f,
656 * meaning we use 32-6=26 bits for the hash.
Tom Herbertfec5e652010-04-16 16:01:27 -0700657 */
658struct rps_sock_flow_table {
Eric Dumazet567e4b72015-02-06 12:59:01 -0800659 u32 mask;
Eric Dumazet93c1af62015-02-08 20:39:13 -0800660
661 u32 ents[0] ____cacheline_aligned_in_smp;
Tom Herbertfec5e652010-04-16 16:01:27 -0700662};
Eric Dumazet567e4b72015-02-06 12:59:01 -0800663#define RPS_SOCK_FLOW_TABLE_SIZE(_num) (offsetof(struct rps_sock_flow_table, ents[_num]))
Tom Herbertfec5e652010-04-16 16:01:27 -0700664
665#define RPS_NO_CPU 0xffff
666
Eric Dumazet567e4b72015-02-06 12:59:01 -0800667extern u32 rps_cpu_mask;
668extern struct rps_sock_flow_table __rcu *rps_sock_flow_table;
669
Tom Herbertfec5e652010-04-16 16:01:27 -0700670static inline void rps_record_sock_flow(struct rps_sock_flow_table *table,
671 u32 hash)
672{
673 if (table && hash) {
Eric Dumazet567e4b72015-02-06 12:59:01 -0800674 unsigned int index = hash & table->mask;
675 u32 val = hash & ~rps_cpu_mask;
Tom Herbertfec5e652010-04-16 16:01:27 -0700676
677 /* We only give a hint, preemption can change cpu under us */
Eric Dumazet567e4b72015-02-06 12:59:01 -0800678 val |= raw_smp_processor_id();
Tom Herbertfec5e652010-04-16 16:01:27 -0700679
Eric Dumazet567e4b72015-02-06 12:59:01 -0800680 if (table->ents[index] != val)
681 table->ents[index] = val;
Tom Herbertfec5e652010-04-16 16:01:27 -0700682 }
683}
684
Ben Hutchingsc4454772011-01-19 11:03:53 +0000685#ifdef CONFIG_RFS_ACCEL
Joe Perchesf629d202013-09-26 14:48:15 -0700686bool rps_may_expire_flow(struct net_device *dev, u16 rxq_index, u32 flow_id,
687 u16 filter_id);
Ben Hutchingsc4454772011-01-19 11:03:53 +0000688#endif
Michael Daltona953be52014-01-16 22:23:28 -0800689#endif /* CONFIG_RPS */
Ben Hutchingsc4454772011-01-19 11:03:53 +0000690
Tom Herbert0a9627f2010-03-16 08:03:29 +0000691/* This structure contains an instance of an RX queue. */
692struct netdev_rx_queue {
Michael Daltona953be52014-01-16 22:23:28 -0800693#ifdef CONFIG_RPS
Eric Dumazet6e3f7fa2010-10-25 03:02:02 +0000694 struct rps_map __rcu *rps_map;
695 struct rps_dev_flow_table __rcu *rps_flow_table;
Michael Daltona953be52014-01-16 22:23:28 -0800696#endif
Eric Dumazet6e3f7fa2010-10-25 03:02:02 +0000697 struct kobject kobj;
Tom Herbertfe822242010-11-09 10:47:38 +0000698 struct net_device *dev;
Tom Herbert0a9627f2010-03-16 08:03:29 +0000699} ____cacheline_aligned_in_smp;
Michael Daltona953be52014-01-16 22:23:28 -0800700
701/*
702 * RX queue sysfs structures and functions.
703 */
704struct rx_queue_attribute {
705 struct attribute attr;
706 ssize_t (*show)(struct netdev_rx_queue *queue,
707 struct rx_queue_attribute *attr, char *buf);
708 ssize_t (*store)(struct netdev_rx_queue *queue,
709 struct rx_queue_attribute *attr, const char *buf, size_t len);
710};
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800711
Tom Herbertbf264142010-11-26 08:36:09 +0000712#ifdef CONFIG_XPS
713/*
714 * This structure holds an XPS map which can be of variable length. The
715 * map is an array of queues.
716 */
717struct xps_map {
718 unsigned int len;
719 unsigned int alloc_len;
720 struct rcu_head rcu;
721 u16 queues[0];
722};
Eric Dumazet60b778c2011-12-24 06:56:49 +0000723#define XPS_MAP_SIZE(_num) (sizeof(struct xps_map) + ((_num) * sizeof(u16)))
Helge Dellerc59f4192015-10-25 10:00:32 +0100724#define XPS_MIN_MAP_ALLOC ((L1_CACHE_ALIGN(offsetof(struct xps_map, queues[1])) \
725 - sizeof(struct xps_map)) / sizeof(u16))
Tom Herbertbf264142010-11-26 08:36:09 +0000726
727/*
728 * This structure holds all XPS maps for device. Maps are indexed by CPU.
729 */
730struct xps_dev_maps {
731 struct rcu_head rcu;
Eric Dumazeta4177862010-11-28 21:43:02 +0000732 struct xps_map __rcu *cpu_map[0];
Tom Herbertbf264142010-11-26 08:36:09 +0000733};
734#define XPS_DEV_MAPS_SIZE (sizeof(struct xps_dev_maps) + \
735 (nr_cpu_ids * sizeof(struct xps_map *)))
736#endif /* CONFIG_XPS */
737
John Fastabend4f57c082011-01-17 08:06:04 +0000738#define TC_MAX_QUEUE 16
739#define TC_BITMASK 15
740/* HW offloaded queuing disciplines txq count and offset maps */
741struct netdev_tc_txq {
742 u16 count;
743 u16 offset;
744};
745
Neerav Parikh68bad942012-01-04 20:23:39 +0000746#if defined(CONFIG_FCOE) || defined(CONFIG_FCOE_MODULE)
747/*
748 * This structure is to hold information about the device
749 * configured to run FCoE protocol stack.
750 */
751struct netdev_fcoe_hbainfo {
752 char manufacturer[64];
753 char serial_number[64];
754 char hardware_version[64];
755 char driver_version[64];
756 char optionrom_version[64];
757 char firmware_version[64];
758 char model[256];
759 char model_description[256];
760};
761#endif
762
Jiri Pirko02637fc2014-11-28 14:34:16 +0100763#define MAX_PHYS_ITEM_ID_LEN 32
Jiri Pirko66b52b02013-07-29 18:16:49 +0200764
Jiri Pirko02637fc2014-11-28 14:34:16 +0100765/* This structure holds a unique identifier to identify some
766 * physical item (port for example) used by a netdevice.
Jiri Pirko66b52b02013-07-29 18:16:49 +0200767 */
Jiri Pirko02637fc2014-11-28 14:34:16 +0100768struct netdev_phys_item_id {
769 unsigned char id[MAX_PHYS_ITEM_ID_LEN];
Jiri Pirko66b52b02013-07-29 18:16:49 +0200770 unsigned char id_len;
771};
772
Scott Feldmand754f982015-07-18 18:24:49 -0700773static inline bool netdev_phys_item_id_same(struct netdev_phys_item_id *a,
774 struct netdev_phys_item_id *b)
775{
776 return a->id_len == b->id_len &&
777 memcmp(a->id, b->id, a->id_len) == 0;
778}
779
Daniel Borkmann99932d42014-02-16 15:55:20 +0100780typedef u16 (*select_queue_fallback_t)(struct net_device *dev,
781 struct sk_buff *skb);
782
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800783/* These structures hold the attributes of qdisc and classifiers
John Fastabend16e5cc62016-02-16 21:16:43 -0800784 * that are being passed to the netdevice through the setup_tc op.
785 */
786enum {
787 TC_SETUP_MQPRIO,
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800788 TC_SETUP_CLSU32,
Amir Vadai5b33f482016-03-08 12:42:29 +0200789 TC_SETUP_CLSFLOWER,
John Fastabend16e5cc62016-02-16 21:16:43 -0800790};
791
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800792struct tc_cls_u32_offload;
793
John Fastabend16e5cc62016-02-16 21:16:43 -0800794struct tc_to_netdev {
795 unsigned int type;
796 union {
797 u8 tc;
John Fastabenda1b7c5f2016-02-16 21:17:09 -0800798 struct tc_cls_u32_offload *cls_u32;
Amir Vadai5b33f482016-03-08 12:42:29 +0200799 struct tc_cls_flower_offload *cls_flower;
John Fastabend16e5cc62016-02-16 21:16:43 -0800800 };
801};
802
803
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800804/*
805 * This structure defines the management hooks for network devices.
Stephen Hemminger00829822008-11-20 20:14:53 -0800806 * The following hooks can be defined; unless noted otherwise, they are
807 * optional and can be filled with a null pointer.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800808 *
809 * int (*ndo_init)(struct net_device *dev);
810 * This function is called once when network device is registered.
811 * The network device can use this to any late stage initializaton
812 * or semantic validattion. It can fail with an error code which will
813 * be propogated back to register_netdev
814 *
815 * void (*ndo_uninit)(struct net_device *dev);
816 * This function is called when device is unregistered or when registration
817 * fails. It is not called if init fails.
818 *
819 * int (*ndo_open)(struct net_device *dev);
820 * This function is called when network device transistions to the up
821 * state.
822 *
823 * int (*ndo_stop)(struct net_device *dev);
824 * This function is called when network device transistions to the down
825 * state.
826 *
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000827 * netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
828 * struct net_device *dev);
Stephen Hemminger00829822008-11-20 20:14:53 -0800829 * Called when a packet needs to be transmitted.
Rusty Russelle79d8422015-04-03 22:17:17 +1030830 * Returns NETDEV_TX_OK. Can return NETDEV_TX_BUSY, but you should stop
831 * the queue before that can happen; it's for obsolete devices and weird
832 * corner cases, but the stack really does a non-trivial amount
833 * of useless work if you return NETDEV_TX_BUSY.
Stephen Hemmingerdc1f8bf2009-08-31 19:50:40 +0000834 * (can also return NETDEV_TX_LOCKED iff NETIF_F_LLTX)
Stephen Hemminger00829822008-11-20 20:14:53 -0800835 * Required can not be NULL.
836 *
Eric Dumazetcdba7562016-01-06 06:53:50 -0800837 * netdev_features_t (*ndo_fix_features)(struct net_device *dev,
838 * netdev_features_t features);
839 * Adjusts the requested feature flags according to device-specific
840 * constraints, and returns the resulting flags. Must not modify
841 * the device state.
842 *
Jason Wangf663dd92014-01-10 16:18:26 +0800843 * u16 (*ndo_select_queue)(struct net_device *dev, struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +0100844 * void *accel_priv, select_queue_fallback_t fallback);
Stephen Hemminger00829822008-11-20 20:14:53 -0800845 * Called to decide which queue to when device supports multiple
846 * transmit queues.
847 *
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800848 * void (*ndo_change_rx_flags)(struct net_device *dev, int flags);
849 * This function is called to allow device receiver to make
850 * changes to configuration when multicast or promiscious is enabled.
851 *
852 * void (*ndo_set_rx_mode)(struct net_device *dev);
853 * This function is called device changes address list filtering.
Jiri Pirko01789342011-08-16 06:29:00 +0000854 * If driver handles unicast address filtering, it should set
855 * IFF_UNICAST_FLT to its priv_flags.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800856 *
857 * int (*ndo_set_mac_address)(struct net_device *dev, void *addr);
858 * This function is called when the Media Access Control address
Mike Rapoport37b607c2009-04-27 05:45:54 -0700859 * needs to be changed. If this interface is not defined, the
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800860 * mac address can not be changed.
861 *
862 * int (*ndo_validate_addr)(struct net_device *dev);
863 * Test if Media Access Control address is valid for the device.
864 *
865 * int (*ndo_do_ioctl)(struct net_device *dev, struct ifreq *ifr, int cmd);
866 * Called when a user request an ioctl which can't be handled by
867 * the generic interface code. If not defined ioctl's return
868 * not supported error code.
869 *
870 * int (*ndo_set_config)(struct net_device *dev, struct ifmap *map);
871 * Used to set network devices bus interface parameters. This interface
872 * is retained for legacy reason, new devices should use the bus
873 * interface (PCI) for low level management.
874 *
875 * int (*ndo_change_mtu)(struct net_device *dev, int new_mtu);
876 * Called when a user wants to change the Maximum Transfer Unit
877 * of a device. If not defined, any request to change MTU will
878 * will return an error.
879 *
Stephen Hemminger00829822008-11-20 20:14:53 -0800880 * void (*ndo_tx_timeout)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800881 * Callback uses when the transmitter has not made any progress
882 * for dev->watchdog ticks.
883 *
Ben Hutchings3cfde792010-07-09 09:11:52 +0000884 * struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
Eric Dumazet28172732010-07-07 14:58:56 -0700885 * struct rtnl_link_stats64 *storage);
Wolfram Sangd308e382009-10-07 13:53:11 -0700886 * struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800887 * Called when a user wants to get the network device usage
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000888 * statistics. Drivers must do one of the following:
Ben Hutchings3cfde792010-07-09 09:11:52 +0000889 * 1. Define @ndo_get_stats64 to fill in a zero-initialised
890 * rtnl_link_stats64 structure passed by the caller.
Ben Hutchings82695d92010-06-15 15:08:48 -0700891 * 2. Define @ndo_get_stats to update a net_device_stats structure
Ben Hutchingsbe1f3c22010-06-08 07:19:54 +0000892 * (which should normally be dev->stats) and return a pointer to
893 * it. The structure may be changed asynchronously only if each
894 * field is written atomically.
895 * 3. Update dev->stats asynchronously and atomically, and define
896 * neither operation.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800897 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530898 * int (*ndo_vlan_rx_add_vid)(struct net_device *dev, __be16 proto, u16 vid);
Patrick McHardy80d5c362013-04-19 02:04:28 +0000899 * If device support VLAN filtering this function is called when a
900 * VLAN id is registered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800901 *
B Viswanath5d632cb2015-01-12 14:46:25 +0530902 * int (*ndo_vlan_rx_kill_vid)(struct net_device *dev, __be16 proto, u16 vid);
Patrick McHardy80d5c362013-04-19 02:04:28 +0000903 * If device support VLAN filtering this function is called when a
904 * VLAN id is unregistered.
Stephen Hemmingerd3147742008-11-19 21:32:24 -0800905 *
906 * void (*ndo_poll_controller)(struct net_device *dev);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000907 *
908 * SR-IOV management functions.
909 * int (*ndo_set_vf_mac)(struct net_device *dev, int vf, u8* mac);
910 * int (*ndo_set_vf_vlan)(struct net_device *dev, int vf, u16 vlan, u8 qos);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -0400911 * int (*ndo_set_vf_rate)(struct net_device *dev, int vf, int min_tx_rate,
912 * int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +0000913 * int (*ndo_set_vf_spoofchk)(struct net_device *dev, int vf, bool setting);
Hiroshi Shimamotodd461d62015-08-28 06:57:55 +0000914 * int (*ndo_set_vf_trust)(struct net_device *dev, int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +0000915 * int (*ndo_get_vf_config)(struct net_device *dev,
916 * int vf, struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +0300917 * int (*ndo_set_vf_link_state)(struct net_device *dev, int vf, int link_state);
Scott Feldman57b61082010-05-17 22:49:55 -0700918 * int (*ndo_set_vf_port)(struct net_device *dev, int vf,
919 * struct nlattr *port[]);
Vlad Zolotarov01a3d792015-03-30 21:35:23 +0300920 *
921 * Enable or disable the VF ability to query its RSS Redirection Table and
922 * Hash Key. This is needed since on some devices VF share this information
923 * with PF and querying it may adduce a theoretical security risk.
924 * int (*ndo_set_vf_rss_query_en)(struct net_device *dev, int vf, bool setting);
Scott Feldman57b61082010-05-17 22:49:55 -0700925 * int (*ndo_get_vf_port)(struct net_device *dev, int vf, struct sk_buff *skb);
John Fastabend4f57c082011-01-17 08:06:04 +0000926 * int (*ndo_setup_tc)(struct net_device *dev, u8 tc)
927 * Called to setup 'tc' number of traffic classes in the net device. This
928 * is always called from the stack with the rtnl lock held and netif tx
929 * queues stopped. This allows the netdevice to perform queue management
930 * safely.
Ben Hutchingsc4454772011-01-19 11:03:53 +0000931 *
Yi Zoue9bce842011-03-09 08:48:03 +0000932 * Fiber Channel over Ethernet (FCoE) offload functions.
933 * int (*ndo_fcoe_enable)(struct net_device *dev);
934 * Called when the FCoE protocol stack wants to start using LLD for FCoE
935 * so the underlying device can perform whatever needed configuration or
936 * initialization to support acceleration of FCoE traffic.
937 *
938 * int (*ndo_fcoe_disable)(struct net_device *dev);
939 * Called when the FCoE protocol stack wants to stop using LLD for FCoE
940 * so the underlying device can perform whatever needed clean-ups to
941 * stop supporting acceleration of FCoE traffic.
942 *
943 * int (*ndo_fcoe_ddp_setup)(struct net_device *dev, u16 xid,
944 * struct scatterlist *sgl, unsigned int sgc);
945 * Called when the FCoE Initiator wants to initialize an I/O that
946 * is a possible candidate for Direct Data Placement (DDP). The LLD can
947 * perform necessary setup and returns 1 to indicate the device is set up
948 * successfully to perform DDP on this I/O, otherwise this returns 0.
949 *
950 * int (*ndo_fcoe_ddp_done)(struct net_device *dev, u16 xid);
951 * Called when the FCoE Initiator/Target is done with the DDPed I/O as
952 * indicated by the FC exchange id 'xid', so the underlying device can
953 * clean up and reuse resources for later DDP requests.
954 *
955 * int (*ndo_fcoe_ddp_target)(struct net_device *dev, u16 xid,
956 * struct scatterlist *sgl, unsigned int sgc);
957 * Called when the FCoE Target wants to initialize an I/O that
958 * is a possible candidate for Direct Data Placement (DDP). The LLD can
959 * perform necessary setup and returns 1 to indicate the device is set up
960 * successfully to perform DDP on this I/O, otherwise this returns 0.
961 *
Neerav Parikh68bad942012-01-04 20:23:39 +0000962 * int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
963 * struct netdev_fcoe_hbainfo *hbainfo);
964 * Called when the FCoE Protocol stack wants information on the underlying
965 * device. This information is utilized by the FCoE protocol stack to
966 * register attributes with Fiber Channel management service as per the
967 * FC-GS Fabric Device Management Information(FDMI) specification.
968 *
Yi Zoue9bce842011-03-09 08:48:03 +0000969 * int (*ndo_fcoe_get_wwn)(struct net_device *dev, u64 *wwn, int type);
970 * Called when the underlying device wants to override default World Wide
971 * Name (WWN) generation mechanism in FCoE protocol stack to pass its own
972 * World Wide Port Name (WWPN) or World Wide Node Name (WWNN) to the FCoE
973 * protocol stack to use.
974 *
Ben Hutchingsc4454772011-01-19 11:03:53 +0000975 * RFS acceleration.
976 * int (*ndo_rx_flow_steer)(struct net_device *dev, const struct sk_buff *skb,
977 * u16 rxq_index, u32 flow_id);
978 * Set hardware filter for RFS. rxq_index is the target queue index;
979 * flow_id is a flow ID to be passed to rps_may_expire_flow() later.
980 * Return the filter ID on success, or a negative error code.
Jiri Pirkofbaec0e2011-02-13 10:15:37 +0000981 *
Jiri Pirko8b98a702013-01-03 22:49:02 +0000982 * Slave management functions (for bridge, bonding, etc).
Jiri Pirkofbaec0e2011-02-13 10:15:37 +0000983 * int (*ndo_add_slave)(struct net_device *dev, struct net_device *slave_dev);
984 * Called to make another netdev an underling.
985 *
986 * int (*ndo_del_slave)(struct net_device *dev, struct net_device *slave_dev);
987 * Called to release previously enslaved netdev.
Michał Mirosław5455c692011-02-15 16:59:17 +0000988 *
989 * Feature/offload setting functions.
Michał Mirosławc8f44af2011-11-15 15:29:55 +0000990 * int (*ndo_set_features)(struct net_device *dev, netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +0000991 * Called to update device configuration to new features. Passed
992 * feature set might be less than what was returned by ndo_fix_features()).
993 * Must return >0 or -errno if it changed dev->features itself.
994 *
stephen hemmingeredc7d572012-10-01 12:32:33 +0000995 * int (*ndo_fdb_add)(struct ndmsg *ndm, struct nlattr *tb[],
996 * struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +0100997 * const unsigned char *addr, u16 vid, u16 flags)
John Fastabend77162022012-04-15 06:43:56 +0000998 * Adds an FDB entry to dev for addr.
Vlad Yasevich1690be62013-02-13 12:00:18 +0000999 * int (*ndo_fdb_del)(struct ndmsg *ndm, struct nlattr *tb[],
1000 * struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001001 * const unsigned char *addr, u16 vid)
John Fastabend77162022012-04-15 06:43:56 +00001002 * Deletes the FDB entry from dev coresponding to addr.
1003 * int (*ndo_fdb_dump)(struct sk_buff *skb, struct netlink_callback *cb,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -04001004 * struct net_device *dev, struct net_device *filter_dev,
1005 * int idx)
John Fastabend77162022012-04-15 06:43:56 +00001006 * Used to add FDB entries to dump requests. Implementers should add
1007 * entries to skb and update idx with the number of entries.
John Fastabende5a55a82012-10-24 08:12:57 +00001008 *
Nicolas Dichtelad41faa2015-03-17 11:16:00 +01001009 * int (*ndo_bridge_setlink)(struct net_device *dev, struct nlmsghdr *nlh,
1010 * u16 flags)
John Fastabende5a55a82012-10-24 08:12:57 +00001011 * int (*ndo_bridge_getlink)(struct sk_buff *skb, u32 pid, u32 seq,
Nicolas Dichtel46c264d2015-04-28 18:33:49 +02001012 * struct net_device *dev, u32 filter_mask,
1013 * int nlflags)
Nicolas Dichtelad41faa2015-03-17 11:16:00 +01001014 * int (*ndo_bridge_dellink)(struct net_device *dev, struct nlmsghdr *nlh,
1015 * u16 flags);
Jiri Pirko4bf84c32012-12-27 23:49:37 +00001016 *
1017 * int (*ndo_change_carrier)(struct net_device *dev, bool new_carrier);
1018 * Called to change device carrier. Soft-devices (like dummy, team, etc)
1019 * which do not represent real hardware may define this to allow their
1020 * userspace components to manage their virtual carrier state. Devices
1021 * that determine carrier state from physical hardware properties (eg
1022 * network cables) or protocol-dependent mechanisms (eg
1023 * USB_CDC_NOTIFY_NETWORK_CONNECTION) should NOT implement this function.
Jiri Pirko66b52b02013-07-29 18:16:49 +02001024 *
1025 * int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01001026 * struct netdev_phys_item_id *ppid);
Jiri Pirko66b52b02013-07-29 18:16:49 +02001027 * Called to get ID of physical port of this device. If driver does
1028 * not implement this, it is assumed that the hw is not able to have
1029 * multiple net devices on single physical port.
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07001030 *
1031 * void (*ndo_add_vxlan_port)(struct net_device *dev,
Joseph Gasparakis35e42372013-09-13 07:34:13 -07001032 * sa_family_t sa_family, __be16 port);
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07001033 * Called by vxlan to notiy a driver about the UDP port and socket
1034 * address family that vxlan is listnening to. It is called only when
1035 * a new port starts listening. The operation is protected by the
1036 * vxlan_net->sock_lock.
1037 *
Singhai, Anjalia8170d22015-12-14 12:21:17 -08001038 * void (*ndo_add_geneve_port)(struct net_device *dev,
1039 * sa_family_t sa_family, __be16 port);
1040 * Called by geneve to notify a driver about the UDP port and socket
1041 * address family that geneve is listnening to. It is called only when
1042 * a new port starts listening. The operation is protected by the
1043 * geneve_net->sock_lock.
1044 *
1045 * void (*ndo_del_geneve_port)(struct net_device *dev,
1046 * sa_family_t sa_family, __be16 port);
1047 * Called by geneve to notify the driver about a UDP port and socket
1048 * address family that geneve is not listening to anymore. The operation
1049 * is protected by the geneve_net->sock_lock.
1050 *
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07001051 * void (*ndo_del_vxlan_port)(struct net_device *dev,
Joseph Gasparakis35e42372013-09-13 07:34:13 -07001052 * sa_family_t sa_family, __be16 port);
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07001053 * Called by vxlan to notify the driver about a UDP port and socket
1054 * address family that vxlan is not listening to anymore. The operation
1055 * is protected by the vxlan_net->sock_lock.
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001056 *
1057 * void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1058 * struct net_device *dev)
1059 * Called by upper layer devices to accelerate switching or other
1060 * station functionality into hardware. 'pdev is the lowerdev
1061 * to use for the offload and 'dev' is the net device that will
1062 * back the offload. Returns a pointer to the private structure
1063 * the upper layer will maintain.
1064 * void (*ndo_dfwd_del_station)(struct net_device *pdev, void *priv)
1065 * Called by upper layer device to delete the station created
1066 * by 'ndo_dfwd_add_station'. 'pdev' is the net device backing
1067 * the station and priv is the structure returned by the add
1068 * operation.
1069 * netdev_tx_t (*ndo_dfwd_start_xmit)(struct sk_buff *skb,
1070 * struct net_device *dev,
1071 * void *priv);
1072 * Callback to use for xmit over the accelerated station. This
1073 * is used in place of ndo_start_xmit on accelerated net
1074 * devices.
Jesse Gross5f352272014-12-23 22:37:26 -08001075 * netdev_features_t (*ndo_features_check) (struct sk_buff *skb,
1076 * struct net_device *dev
1077 * netdev_features_t features);
Tom Herbert04ffcb22014-10-14 15:19:06 -07001078 * Called by core transmit path to determine if device is capable of
Jesse Gross5f352272014-12-23 22:37:26 -08001079 * performing offload operations on a given packet. This is to give
1080 * the device an opportunity to implement any restrictions that cannot
1081 * be otherwise expressed by feature flags. The check is called with
1082 * the set of features that the stack has calculated and it returns
1083 * those the driver believes to be appropriate.
John Fastabend822b3b22015-03-18 14:57:33 +02001084 * int (*ndo_set_tx_maxrate)(struct net_device *dev,
1085 * int queue_index, u32 maxrate);
1086 * Called when a user wants to set a max-rate limitation of specific
1087 * TX queue.
Nicolas Dichtela54acb32015-04-02 17:07:00 +02001088 * int (*ndo_get_iflink)(const struct net_device *dev);
1089 * Called to get the iflink value of this device.
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001090 * void (*ndo_change_proto_down)(struct net_device *dev,
1091 * bool proto_down);
1092 * This function is used to pass protocol port error state information
1093 * to the switch driver. The switch driver can react to the proto_down
1094 * by doing a phys down on the associated switch port.
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07001095 * int (*ndo_fill_metadata_dst)(struct net_device *dev, struct sk_buff *skb);
1096 * This function is used to get egress tunnel information for given skb.
1097 * This is useful for retrieving outer tunnel header parameters while
1098 * sampling packet.
Paolo Abeni871b6422016-02-26 10:45:37 +01001099 * void (*ndo_set_rx_headroom)(struct net_device *dev, int needed_headroom);
1100 * This function is used to specify the headroom that the skb must
1101 * consider when allocation skb during packet reception. Setting
1102 * appropriate rx headroom value allows avoiding skb head copy on
1103 * forward. Setting a negative value reset the rx headroom to the
1104 * default value.
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001105 *
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001106 */
1107struct net_device_ops {
1108 int (*ndo_init)(struct net_device *dev);
1109 void (*ndo_uninit)(struct net_device *dev);
1110 int (*ndo_open)(struct net_device *dev);
1111 int (*ndo_stop)(struct net_device *dev);
Eric Dumazetcdba7562016-01-06 06:53:50 -08001112 netdev_tx_t (*ndo_start_xmit)(struct sk_buff *skb,
1113 struct net_device *dev);
1114 netdev_features_t (*ndo_features_check)(struct sk_buff *skb,
1115 struct net_device *dev,
1116 netdev_features_t features);
Stephen Hemminger00829822008-11-20 20:14:53 -08001117 u16 (*ndo_select_queue)(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08001118 struct sk_buff *skb,
Daniel Borkmann99932d42014-02-16 15:55:20 +01001119 void *accel_priv,
1120 select_queue_fallback_t fallback);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001121 void (*ndo_change_rx_flags)(struct net_device *dev,
1122 int flags);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001123 void (*ndo_set_rx_mode)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001124 int (*ndo_set_mac_address)(struct net_device *dev,
1125 void *addr);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001126 int (*ndo_validate_addr)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001127 int (*ndo_do_ioctl)(struct net_device *dev,
1128 struct ifreq *ifr, int cmd);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001129 int (*ndo_set_config)(struct net_device *dev,
1130 struct ifmap *map);
Stephen Hemminger00829822008-11-20 20:14:53 -08001131 int (*ndo_change_mtu)(struct net_device *dev,
1132 int new_mtu);
1133 int (*ndo_neigh_setup)(struct net_device *dev,
1134 struct neigh_parms *);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001135 void (*ndo_tx_timeout) (struct net_device *dev);
1136
Eric Dumazet28172732010-07-07 14:58:56 -07001137 struct rtnl_link_stats64* (*ndo_get_stats64)(struct net_device *dev,
1138 struct rtnl_link_stats64 *storage);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001139 struct net_device_stats* (*ndo_get_stats)(struct net_device *dev);
1140
Jiri Pirko8e586132011-12-08 19:52:37 -05001141 int (*ndo_vlan_rx_add_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001142 __be16 proto, u16 vid);
Jiri Pirko8e586132011-12-08 19:52:37 -05001143 int (*ndo_vlan_rx_kill_vid)(struct net_device *dev,
Patrick McHardy80d5c362013-04-19 02:04:28 +00001144 __be16 proto, u16 vid);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001145#ifdef CONFIG_NET_POLL_CONTROLLER
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001146 void (*ndo_poll_controller)(struct net_device *dev);
Herbert Xu4247e162010-06-10 16:12:47 +00001147 int (*ndo_netpoll_setup)(struct net_device *dev,
Eric W. Biedermana8779ec2014-03-27 15:36:38 -07001148 struct netpoll_info *info);
WANG Cong0e34e932010-05-06 00:47:21 -07001149 void (*ndo_netpoll_cleanup)(struct net_device *dev);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001150#endif
Cong Wange0d10952013-08-01 11:10:25 +08001151#ifdef CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir8b80cda2013-07-10 17:13:26 +03001152 int (*ndo_busy_poll)(struct napi_struct *dev);
Eliezer Tamir06021292013-06-10 11:39:50 +03001153#endif
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001154 int (*ndo_set_vf_mac)(struct net_device *dev,
1155 int queue, u8 *mac);
1156 int (*ndo_set_vf_vlan)(struct net_device *dev,
1157 int queue, u16 vlan, u8 qos);
Sucheta Chakrabortyed616682014-05-22 09:59:05 -04001158 int (*ndo_set_vf_rate)(struct net_device *dev,
1159 int vf, int min_tx_rate,
1160 int max_tx_rate);
Greg Rose5f8444a2011-10-08 03:05:24 +00001161 int (*ndo_set_vf_spoofchk)(struct net_device *dev,
1162 int vf, bool setting);
Hiroshi Shimamotodd461d62015-08-28 06:57:55 +00001163 int (*ndo_set_vf_trust)(struct net_device *dev,
1164 int vf, bool setting);
Williams, Mitch A95c26df2010-02-10 01:43:46 +00001165 int (*ndo_get_vf_config)(struct net_device *dev,
1166 int vf,
1167 struct ifla_vf_info *ivf);
Rony Efraim1d8faf42013-06-13 13:19:10 +03001168 int (*ndo_set_vf_link_state)(struct net_device *dev,
1169 int vf, int link_state);
Eran Ben Elisha3b766cd2015-06-15 17:59:07 +03001170 int (*ndo_get_vf_stats)(struct net_device *dev,
1171 int vf,
1172 struct ifla_vf_stats
1173 *vf_stats);
Scott Feldman57b61082010-05-17 22:49:55 -07001174 int (*ndo_set_vf_port)(struct net_device *dev,
1175 int vf,
1176 struct nlattr *port[]);
1177 int (*ndo_get_vf_port)(struct net_device *dev,
1178 int vf, struct sk_buff *skb);
Vlad Zolotarov01a3d792015-03-30 21:35:23 +03001179 int (*ndo_set_vf_rss_query_en)(
1180 struct net_device *dev,
1181 int vf, bool setting);
John Fastabend16e5cc62016-02-16 21:16:43 -08001182 int (*ndo_setup_tc)(struct net_device *dev,
1183 u32 handle,
1184 __be16 protocol,
1185 struct tc_to_netdev *tc);
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001186#if IS_ENABLED(CONFIG_FCOE)
Yi Zoucb454392009-08-31 12:31:36 +00001187 int (*ndo_fcoe_enable)(struct net_device *dev);
1188 int (*ndo_fcoe_disable)(struct net_device *dev);
Yi Zou4d288d52009-02-27 14:06:59 -08001189 int (*ndo_fcoe_ddp_setup)(struct net_device *dev,
1190 u16 xid,
1191 struct scatterlist *sgl,
1192 unsigned int sgc);
1193 int (*ndo_fcoe_ddp_done)(struct net_device *dev,
1194 u16 xid);
Yi Zou6247e082011-02-01 07:22:06 +00001195 int (*ndo_fcoe_ddp_target)(struct net_device *dev,
1196 u16 xid,
1197 struct scatterlist *sgl,
1198 unsigned int sgc);
Neerav Parikh68bad942012-01-04 20:23:39 +00001199 int (*ndo_fcoe_get_hbainfo)(struct net_device *dev,
1200 struct netdev_fcoe_hbainfo *hbainfo);
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001201#endif
1202
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001203#if IS_ENABLED(CONFIG_LIBFCOE)
Yi Zoudf5c7942009-10-28 18:24:35 +00001204#define NETDEV_FCOE_WWNN 0
1205#define NETDEV_FCOE_WWPN 1
1206 int (*ndo_fcoe_get_wwn)(struct net_device *dev,
1207 u64 *wwn, int type);
Yi Zou4d288d52009-02-27 14:06:59 -08001208#endif
Bhanu Prakash Gollapudi3c9c36bc2011-08-26 09:45:41 +00001209
Ben Hutchingsc4454772011-01-19 11:03:53 +00001210#ifdef CONFIG_RFS_ACCEL
1211 int (*ndo_rx_flow_steer)(struct net_device *dev,
1212 const struct sk_buff *skb,
1213 u16 rxq_index,
1214 u32 flow_id);
1215#endif
Jiri Pirkofbaec0e2011-02-13 10:15:37 +00001216 int (*ndo_add_slave)(struct net_device *dev,
1217 struct net_device *slave_dev);
1218 int (*ndo_del_slave)(struct net_device *dev,
1219 struct net_device *slave_dev);
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001220 netdev_features_t (*ndo_fix_features)(struct net_device *dev,
1221 netdev_features_t features);
Michał Mirosław5455c692011-02-15 16:59:17 +00001222 int (*ndo_set_features)(struct net_device *dev,
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001223 netdev_features_t features);
David Millerda6a8fa2011-07-25 00:01:38 +00001224 int (*ndo_neigh_construct)(struct neighbour *n);
David S. Miller447f2192011-12-19 15:04:41 -05001225 void (*ndo_neigh_destroy)(struct neighbour *n);
John Fastabend77162022012-04-15 06:43:56 +00001226
1227 int (*ndo_fdb_add)(struct ndmsg *ndm,
stephen hemmingeredc7d572012-10-01 12:32:33 +00001228 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001229 struct net_device *dev,
stephen hemminger6b6e2722012-09-17 10:03:26 +00001230 const unsigned char *addr,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001231 u16 vid,
John Fastabend77162022012-04-15 06:43:56 +00001232 u16 flags);
1233 int (*ndo_fdb_del)(struct ndmsg *ndm,
Vlad Yasevich1690be62013-02-13 12:00:18 +00001234 struct nlattr *tb[],
John Fastabend77162022012-04-15 06:43:56 +00001235 struct net_device *dev,
Jiri Pirkof6f64242014-11-28 14:34:15 +01001236 const unsigned char *addr,
1237 u16 vid);
John Fastabend77162022012-04-15 06:43:56 +00001238 int (*ndo_fdb_dump)(struct sk_buff *skb,
1239 struct netlink_callback *cb,
1240 struct net_device *dev,
Jamal Hadi Salim5d5eacb2014-07-10 07:01:58 -04001241 struct net_device *filter_dev,
John Fastabend77162022012-04-15 06:43:56 +00001242 int idx);
John Fastabende5a55a82012-10-24 08:12:57 +00001243
1244 int (*ndo_bridge_setlink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001245 struct nlmsghdr *nlh,
1246 u16 flags);
John Fastabende5a55a82012-10-24 08:12:57 +00001247 int (*ndo_bridge_getlink)(struct sk_buff *skb,
1248 u32 pid, u32 seq,
Vlad Yasevich6cbdcee2013-02-13 12:00:13 +00001249 struct net_device *dev,
Nicolas Dichtel46c264d2015-04-28 18:33:49 +02001250 u32 filter_mask,
1251 int nlflags);
Vlad Yasevich407af322013-02-13 12:00:12 +00001252 int (*ndo_bridge_dellink)(struct net_device *dev,
Roopa Prabhuadd511b2015-01-29 22:40:12 -08001253 struct nlmsghdr *nlh,
1254 u16 flags);
Jiri Pirko4bf84c32012-12-27 23:49:37 +00001255 int (*ndo_change_carrier)(struct net_device *dev,
1256 bool new_carrier);
Jiri Pirko66b52b02013-07-29 18:16:49 +02001257 int (*ndo_get_phys_port_id)(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01001258 struct netdev_phys_item_id *ppid);
David Aherndb24a902015-03-17 20:23:15 -06001259 int (*ndo_get_phys_port_name)(struct net_device *dev,
1260 char *name, size_t len);
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07001261 void (*ndo_add_vxlan_port)(struct net_device *dev,
1262 sa_family_t sa_family,
Joseph Gasparakis35e42372013-09-13 07:34:13 -07001263 __be16 port);
Joseph Gasparakis53cf52752013-09-04 02:13:38 -07001264 void (*ndo_del_vxlan_port)(struct net_device *dev,
1265 sa_family_t sa_family,
Joseph Gasparakis35e42372013-09-13 07:34:13 -07001266 __be16 port);
Singhai, Anjalia8170d22015-12-14 12:21:17 -08001267 void (*ndo_add_geneve_port)(struct net_device *dev,
1268 sa_family_t sa_family,
1269 __be16 port);
1270 void (*ndo_del_geneve_port)(struct net_device *dev,
1271 sa_family_t sa_family,
1272 __be16 port);
John Fastabenda6cc0cf2013-11-06 09:54:46 -08001273 void* (*ndo_dfwd_add_station)(struct net_device *pdev,
1274 struct net_device *dev);
1275 void (*ndo_dfwd_del_station)(struct net_device *pdev,
1276 void *priv);
1277
1278 netdev_tx_t (*ndo_dfwd_start_xmit) (struct sk_buff *skb,
1279 struct net_device *dev,
1280 void *priv);
Vlad Yasevich25175ba2014-05-16 17:04:54 -04001281 int (*ndo_get_lock_subclass)(struct net_device *dev);
John Fastabend822b3b22015-03-18 14:57:33 +02001282 int (*ndo_set_tx_maxrate)(struct net_device *dev,
1283 int queue_index,
1284 u32 maxrate);
Nicolas Dichtela54acb32015-04-02 17:07:00 +02001285 int (*ndo_get_iflink)(const struct net_device *dev);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001286 int (*ndo_change_proto_down)(struct net_device *dev,
1287 bool proto_down);
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07001288 int (*ndo_fill_metadata_dst)(struct net_device *dev,
1289 struct sk_buff *skb);
Paolo Abeni871b6422016-02-26 10:45:37 +01001290 void (*ndo_set_rx_headroom)(struct net_device *dev,
1291 int needed_headroom);
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001292};
1293
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001294/**
1295 * enum net_device_priv_flags - &struct net_device priv_flags
1296 *
1297 * These are the &struct net_device, they are only set internally
1298 * by drivers and used in the kernel. These flags are invisible to
1299 * userspace, this means that the order of these flags can change
1300 * during any kernel release.
1301 *
1302 * You should have a pretty good reason to be extending these flags.
1303 *
1304 * @IFF_802_1Q_VLAN: 802.1Q VLAN device
1305 * @IFF_EBRIDGE: Ethernet bridging device
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001306 * @IFF_BONDING: bonding master or slave
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001307 * @IFF_ISATAP: ISATAP interface (RFC4214)
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001308 * @IFF_WAN_HDLC: WAN HDLC device
1309 * @IFF_XMIT_DST_RELEASE: dev_hard_start_xmit() is allowed to
1310 * release skb->dst
1311 * @IFF_DONT_BRIDGE: disallow bridging this ether dev
1312 * @IFF_DISABLE_NETPOLL: disable netpoll at run-time
1313 * @IFF_MACVLAN_PORT: device used as macvlan port
1314 * @IFF_BRIDGE_PORT: device used as bridge port
1315 * @IFF_OVS_DATAPATH: device used as Open vSwitch datapath port
1316 * @IFF_TX_SKB_SHARING: The interface supports sharing skbs on transmit
1317 * @IFF_UNICAST_FLT: Supports unicast filtering
1318 * @IFF_TEAM_PORT: device used as team port
1319 * @IFF_SUPP_NOFCS: device supports sending custom FCS
1320 * @IFF_LIVE_ADDR_CHANGE: device supports hardware address
1321 * change when it's running
1322 * @IFF_MACVLAN: Macvlan device
David Ahern007979e2015-09-29 20:07:10 -07001323 * @IFF_L3MDEV_MASTER: device is an L3 master device
Phil Sutterfa8187c2015-08-13 19:01:06 +02001324 * @IFF_NO_QUEUE: device can run without qdisc attached
Jiri Pirko35d4e172015-08-27 09:31:20 +02001325 * @IFF_OPENVSWITCH: device is a Open vSwitch master
David Ahernfee6d4c2015-10-05 08:51:24 -07001326 * @IFF_L3MDEV_SLAVE: device is enslaved to an L3 master device
Jiri Pirkoc981e422015-12-03 12:12:06 +01001327 * @IFF_TEAM: device is a team device
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001328 * @IFF_RXFH_CONFIGURED: device has had Rx Flow indirection table configured
Paolo Abeni871b6422016-02-26 10:45:37 +01001329 * @IFF_PHONY_HEADROOM: the headroom value is controlled by an external
1330 * entity (i.e. the master device for bridged veth)
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001331 * @IFF_MACSEC: device is a MACsec device
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001332 */
1333enum netdev_priv_flags {
1334 IFF_802_1Q_VLAN = 1<<0,
1335 IFF_EBRIDGE = 1<<1,
Jiri Pirko0dc15492015-08-27 09:31:21 +02001336 IFF_BONDING = 1<<2,
1337 IFF_ISATAP = 1<<3,
1338 IFF_WAN_HDLC = 1<<4,
1339 IFF_XMIT_DST_RELEASE = 1<<5,
1340 IFF_DONT_BRIDGE = 1<<6,
1341 IFF_DISABLE_NETPOLL = 1<<7,
1342 IFF_MACVLAN_PORT = 1<<8,
1343 IFF_BRIDGE_PORT = 1<<9,
1344 IFF_OVS_DATAPATH = 1<<10,
1345 IFF_TX_SKB_SHARING = 1<<11,
1346 IFF_UNICAST_FLT = 1<<12,
1347 IFF_TEAM_PORT = 1<<13,
1348 IFF_SUPP_NOFCS = 1<<14,
1349 IFF_LIVE_ADDR_CHANGE = 1<<15,
1350 IFF_MACVLAN = 1<<16,
1351 IFF_XMIT_DST_RELEASE_PERM = 1<<17,
1352 IFF_IPVLAN_MASTER = 1<<18,
1353 IFF_IPVLAN_SLAVE = 1<<19,
David Ahern007979e2015-09-29 20:07:10 -07001354 IFF_L3MDEV_MASTER = 1<<20,
Jiri Pirko0dc15492015-08-27 09:31:21 +02001355 IFF_NO_QUEUE = 1<<21,
1356 IFF_OPENVSWITCH = 1<<22,
David Ahernfee6d4c2015-10-05 08:51:24 -07001357 IFF_L3MDEV_SLAVE = 1<<23,
Jiri Pirkoc981e422015-12-03 12:12:06 +01001358 IFF_TEAM = 1<<24,
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001359 IFF_RXFH_CONFIGURED = 1<<25,
Paolo Abeni871b6422016-02-26 10:45:37 +01001360 IFF_PHONY_HEADROOM = 1<<26,
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001361 IFF_MACSEC = 1<<27,
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001362};
1363
1364#define IFF_802_1Q_VLAN IFF_802_1Q_VLAN
1365#define IFF_EBRIDGE IFF_EBRIDGE
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001366#define IFF_BONDING IFF_BONDING
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001367#define IFF_ISATAP IFF_ISATAP
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001368#define IFF_WAN_HDLC IFF_WAN_HDLC
1369#define IFF_XMIT_DST_RELEASE IFF_XMIT_DST_RELEASE
1370#define IFF_DONT_BRIDGE IFF_DONT_BRIDGE
1371#define IFF_DISABLE_NETPOLL IFF_DISABLE_NETPOLL
1372#define IFF_MACVLAN_PORT IFF_MACVLAN_PORT
1373#define IFF_BRIDGE_PORT IFF_BRIDGE_PORT
1374#define IFF_OVS_DATAPATH IFF_OVS_DATAPATH
1375#define IFF_TX_SKB_SHARING IFF_TX_SKB_SHARING
1376#define IFF_UNICAST_FLT IFF_UNICAST_FLT
1377#define IFF_TEAM_PORT IFF_TEAM_PORT
1378#define IFF_SUPP_NOFCS IFF_SUPP_NOFCS
1379#define IFF_LIVE_ADDR_CHANGE IFF_LIVE_ADDR_CHANGE
1380#define IFF_MACVLAN IFF_MACVLAN
Eric Dumazet02875872014-10-05 18:38:35 -07001381#define IFF_XMIT_DST_RELEASE_PERM IFF_XMIT_DST_RELEASE_PERM
Mahesh Bandewar2ad7bf32014-11-23 23:07:46 -08001382#define IFF_IPVLAN_MASTER IFF_IPVLAN_MASTER
1383#define IFF_IPVLAN_SLAVE IFF_IPVLAN_SLAVE
David Ahern007979e2015-09-29 20:07:10 -07001384#define IFF_L3MDEV_MASTER IFF_L3MDEV_MASTER
Phil Sutterfa8187c2015-08-13 19:01:06 +02001385#define IFF_NO_QUEUE IFF_NO_QUEUE
Jiri Pirko35d4e172015-08-27 09:31:20 +02001386#define IFF_OPENVSWITCH IFF_OPENVSWITCH
Jiri Pirko8f253482015-11-04 14:59:06 +01001387#define IFF_L3MDEV_SLAVE IFF_L3MDEV_SLAVE
Jiri Pirkoc981e422015-12-03 12:12:06 +01001388#define IFF_TEAM IFF_TEAM
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08001389#define IFF_RXFH_CONFIGURED IFF_RXFH_CONFIGURED
Sabrina Dubroca3c175782016-03-11 18:07:32 +01001390#define IFF_MACSEC IFF_MACSEC
Luis R. Rodriguez7aa98042014-02-25 17:15:13 -08001391
Karoly Kemeny536721b2014-07-30 20:27:36 +02001392/**
1393 * struct net_device - The DEVICE structure.
1394 * Actually, this whole structure is a big mistake. It mixes I/O
1395 * data with strictly "high-level" data, and it has to know about
1396 * almost every data structure used in the INET module.
1397 *
1398 * @name: This is the first field of the "visible" part of this structure
1399 * (i.e. as seen by users in the "Space.c" file). It is the name
1400 * of the interface.
1401 *
1402 * @name_hlist: Device name hash chain, please keep it close to name[]
1403 * @ifalias: SNMP alias
1404 * @mem_end: Shared memory end
1405 * @mem_start: Shared memory start
1406 * @base_addr: Device I/O address
1407 * @irq: Device IRQ number
1408 *
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001409 * @carrier_changes: Stats to monitor carrier on<->off transitions
1410 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001411 * @state: Generic network queuing layer state, see netdev_state_t
1412 * @dev_list: The global list of network devices
1413 * @napi_list: List entry, that is used for polling napi devices
1414 * @unreg_list: List entry, that is used, when we are unregistering the
1415 * device, see the function unregister_netdev
1416 * @close_list: List entry, that is used, when we are closing the device
1417 *
1418 * @adj_list: Directly linked devices, like slaves for bonding
1419 * @all_adj_list: All linked devices, *including* neighbours
1420 * @features: Currently active device features
1421 * @hw_features: User-changeable features
1422 *
1423 * @wanted_features: User-requested features
1424 * @vlan_features: Mask of features inheritable by VLAN devices
1425 *
1426 * @hw_enc_features: Mask of features inherited by encapsulating devices
1427 * This field indicates what encapsulation
1428 * offloads the hardware is capable of doing,
1429 * and drivers will need to set them appropriately.
1430 *
1431 * @mpls_features: Mask of features inheritable by MPLS
1432 *
1433 * @ifindex: interface index
Nicolas Dichtel388069d2015-04-07 12:10:16 +02001434 * @group: The group, that the device belongs to
Karoly Kemeny536721b2014-07-30 20:27:36 +02001435 *
1436 * @stats: Statistics struct, which was left as a legacy, use
1437 * rtnl_link_stats64 instead
1438 *
1439 * @rx_dropped: Dropped packets by core network,
1440 * do not use this in drivers
1441 * @tx_dropped: Dropped packets by core network,
1442 * do not use this in drivers
Jarod Wilson6e7333d2016-02-01 18:51:05 -05001443 * @rx_nohandler: nohandler dropped packets by core network on
1444 * inactive devices, do not use this in drivers
Karoly Kemeny536721b2014-07-30 20:27:36 +02001445 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001446 * @wireless_handlers: List of functions to handle Wireless Extensions,
1447 * instead of ioctl,
1448 * see <net/iw_handler.h> for details.
1449 * @wireless_data: Instance data managed by the core of wireless extensions
1450 *
1451 * @netdev_ops: Includes several pointers to callbacks,
1452 * if one wants to override the ndo_*() functions
1453 * @ethtool_ops: Management operations
Eric W. Biedermand4760592015-03-02 00:11:09 -06001454 * @header_ops: Includes callbacks for creating,parsing,caching,etc
Karoly Kemeny536721b2014-07-30 20:27:36 +02001455 * of Layer 2 headers.
1456 *
1457 * @flags: Interface flags (a la BSD)
1458 * @priv_flags: Like 'flags' but invisible to userspace,
1459 * see if.h for the definitions
1460 * @gflags: Global flags ( kept as legacy )
1461 * @padded: How much padding added by alloc_netdev()
1462 * @operstate: RFC2863 operstate
1463 * @link_mode: Mapping policy to operstate
1464 * @if_port: Selectable AUI, TP, ...
1465 * @dma: DMA channel
1466 * @mtu: Interface MTU value
1467 * @type: Interface hardware type
Willem de Bruijn2793a232016-03-09 21:58:32 -05001468 * @hard_header_len: Maximum hardware header length.
Karoly Kemeny536721b2014-07-30 20:27:36 +02001469 *
1470 * @needed_headroom: Extra headroom the hardware may need, but not in all
1471 * cases can this be guaranteed
1472 * @needed_tailroom: Extra tailroom the hardware may need, but not in all
1473 * cases can this be guaranteed. Some cases also use
1474 * LL_MAX_HEADER instead to allocate the skb
1475 *
1476 * interface address info:
1477 *
1478 * @perm_addr: Permanent hw address
1479 * @addr_assign_type: Hw address assignment type
1480 * @addr_len: Hardware address length
1481 * @neigh_priv_len; Used in neigh_alloc(),
1482 * initialized only in atm/clip.c
1483 * @dev_id: Used to differentiate devices that share
1484 * the same link layer address
1485 * @dev_port: Used to differentiate devices that share
1486 * the same function
1487 * @addr_list_lock: XXX: need comments on this one
Karoly Kemeny536721b2014-07-30 20:27:36 +02001488 * @uc_promisc: Counter, that indicates, that promiscuous mode
1489 * has been enabled due to the need to listen to
1490 * additional unicast addresses in a device that
1491 * does not implement ndo_set_rx_mode()
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001492 * @uc: unicast mac addresses
1493 * @mc: multicast mac addresses
1494 * @dev_addrs: list of device hw addresses
1495 * @queues_kset: Group of all Kobjects in the Tx and RX queues
Karoly Kemeny536721b2014-07-30 20:27:36 +02001496 * @promiscuity: Number of times, the NIC is told to work in
1497 * Promiscuous mode, if it becomes 0 the NIC will
1498 * exit from working in Promiscuous mode
1499 * @allmulti: Counter, enables or disables allmulticast mode
1500 *
1501 * @vlan_info: VLAN info
1502 * @dsa_ptr: dsa specific data
1503 * @tipc_ptr: TIPC specific data
1504 * @atalk_ptr: AppleTalk link
1505 * @ip_ptr: IPv4 specific data
1506 * @dn_ptr: DECnet specific data
1507 * @ip6_ptr: IPv6 specific data
1508 * @ax25_ptr: AX.25 specific data
1509 * @ieee80211_ptr: IEEE 802.11 specific data, assign before registering
1510 *
1511 * @last_rx: Time of last Rx
1512 * @dev_addr: Hw address (before bcast,
1513 * because most packets are unicast)
1514 *
1515 * @_rx: Array of RX queues
1516 * @num_rx_queues: Number of RX queues
1517 * allocated at register_netdev() time
1518 * @real_num_rx_queues: Number of RX queues currently active in device
1519 *
1520 * @rx_handler: handler for received packets
1521 * @rx_handler_data: XXX: need comments on this one
1522 * @ingress_queue: XXX: need comments on this one
1523 * @broadcast: hw bcast address
1524 *
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001525 * @rx_cpu_rmap: CPU reverse-mapping for RX completion interrupts,
1526 * indexed by RX queue number. Assigned by driver.
1527 * This must only be set if the ndo_rx_flow_steer
1528 * operation is defined
1529 * @index_hlist: Device index hash chain
1530 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001531 * @_tx: Array of TX queues
1532 * @num_tx_queues: Number of TX queues allocated at alloc_netdev_mq() time
1533 * @real_num_tx_queues: Number of TX queues currently active in device
1534 * @qdisc: Root qdisc from userspace point of view
1535 * @tx_queue_len: Max frames per queue allowed
1536 * @tx_global_lock: XXX: need comments on this one
1537 *
1538 * @xps_maps: XXX: need comments on this one
1539 *
Scott Feldman0c4f6912015-07-18 18:24:48 -07001540 * @offload_fwd_mark: Offload device fwding mark
1541 *
Karoly Kemeny536721b2014-07-30 20:27:36 +02001542 * @trans_start: Time (in jiffies) of last Tx
1543 * @watchdog_timeo: Represents the timeout that is used by
1544 * the watchdog ( see dev_watchdog() )
1545 * @watchdog_timer: List of timers
1546 *
1547 * @pcpu_refcnt: Number of references to this device
1548 * @todo_list: Delayed register/unregister
Karoly Kemeny536721b2014-07-30 20:27:36 +02001549 * @link_watch_list: XXX: need comments on this one
1550 *
1551 * @reg_state: Register/unregister state machine
1552 * @dismantle: Device is going to be freed
1553 * @rtnl_link_state: This enum represents the phases of creating
1554 * a new link
1555 *
1556 * @destructor: Called from unregister,
1557 * can be used to call free_netdev
1558 * @npinfo: XXX: need comments on this one
1559 * @nd_net: Network namespace this network device is inside
1560 *
1561 * @ml_priv: Mid-layer private
1562 * @lstats: Loopback statistics
1563 * @tstats: Tunnel statistics
1564 * @dstats: Dummy statistics
1565 * @vstats: Virtual ethernet statistics
1566 *
1567 * @garp_port: GARP
1568 * @mrp_port: MRP
1569 *
1570 * @dev: Class/net/name entry
1571 * @sysfs_groups: Space for optional device, statistics and wireless
1572 * sysfs groups
1573 *
1574 * @sysfs_rx_queue_group: Space for optional per-rx queue attributes
1575 * @rtnl_link_ops: Rtnl_link_ops
1576 *
1577 * @gso_max_size: Maximum size of generic segmentation offload
1578 * @gso_max_segs: Maximum number of segments that can be passed to the
1579 * NIC for GSO
Eric Dumazetfcbeb972014-10-05 10:11:27 -07001580 * @gso_min_segs: Minimum number of segments that can be passed to the
1581 * NIC for GSO
Karoly Kemeny536721b2014-07-30 20:27:36 +02001582 *
1583 * @dcbnl_ops: Data Center Bridging netlink ops
1584 * @num_tc: Number of traffic classes in the net device
1585 * @tc_to_txq: XXX: need comments on this one
1586 * @prio_tc_map XXX: need comments on this one
1587 *
1588 * @fcoe_ddp_xid: Max exchange id for FCoE LRO by ddp
1589 *
1590 * @priomap: XXX: need comments on this one
1591 * @phydev: Physical device may attach itself
1592 * for hardware timestamping
1593 *
1594 * @qdisc_tx_busylock: XXX: need comments on this one
1595 *
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001596 * @proto_down: protocol port state information can be sent to the
1597 * switch driver and used to set the phys state of the
1598 * switch port.
1599 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001600 * FIXME: cleanup struct net_device such that network protocol info
1601 * moves out.
1602 */
1603
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08001604struct net_device {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001605 char name[IFNAMSIZ];
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001606 struct hlist_node name_hlist;
Stephen Hemminger0b815a12008-09-22 21:28:11 -07001607 char *ifalias;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001608 /*
1609 * I/O specific fields
1610 * FIXME: Merge these and struct ifmap into one
1611 */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001612 unsigned long mem_end;
1613 unsigned long mem_start;
1614 unsigned long base_addr;
1615 int irq;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001616
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001617 atomic_t carrier_changes;
1618
Linus Torvalds1da177e2005-04-16 15:20:36 -07001619 /*
Karoly Kemeny536721b2014-07-30 20:27:36 +02001620 * Some hardware also needs these fields (state,dev_list,
1621 * napi_list,unreg_list,close_list) but they are not
Linus Torvalds1da177e2005-04-16 15:20:36 -07001622 * part of the usual set specified in Space.c.
1623 */
1624
Linus Torvalds1da177e2005-04-16 15:20:36 -07001625 unsigned long state;
1626
Pavel Emelianov7562f872007-05-03 15:13:45 -07001627 struct list_head dev_list;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07001628 struct list_head napi_list;
Eric Dumazet44a08732009-10-27 07:03:04 +00001629 struct list_head unreg_list;
Eric W. Biederman5cde2822013-10-05 19:26:05 -07001630 struct list_head close_list;
Salam Noureddine7866a622015-01-27 11:35:48 -08001631 struct list_head ptype_all;
1632 struct list_head ptype_specific;
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001633
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001634 struct {
1635 struct list_head upper;
1636 struct list_head lower;
1637 } adj_list;
1638
Veaceslav Falico2f268f12013-09-25 09:20:07 +02001639 struct {
1640 struct list_head upper;
1641 struct list_head lower;
1642 } all_adj_list;
Eric Dumazet4c3d5e72013-03-30 06:31:03 +00001643
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001644 netdev_features_t features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001645 netdev_features_t hw_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001646 netdev_features_t wanted_features;
Michał Mirosławc8f44af2011-11-15 15:29:55 +00001647 netdev_features_t vlan_features;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001648 netdev_features_t hw_enc_features;
Simon Horman0d89d202013-05-23 21:02:52 +00001649 netdev_features_t mpls_features;
Michał Mirosław04ed3e72011-01-24 15:32:47 -08001650
Linus Torvalds1da177e2005-04-16 15:20:36 -07001651 int ifindex;
Nicolas Dichtel7a66bbc2015-04-02 17:07:09 +02001652 int group;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653
Rusty Russellc45d2862007-03-28 14:29:08 -07001654 struct net_device_stats stats;
Eric Dumazet015f0682014-03-27 08:45:56 -07001655
Eric Dumazet015f0682014-03-27 08:45:56 -07001656 atomic_long_t rx_dropped;
1657 atomic_long_t tx_dropped;
Jarod Wilson6e7333d2016-02-01 18:51:05 -05001658 atomic_long_t rx_nohandler;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001659
Johannes Bergb86e0282007-04-26 20:48:23 -07001660#ifdef CONFIG_WIRELESS_EXT
Linus Torvalds1da177e2005-04-16 15:20:36 -07001661 const struct iw_handler_def * wireless_handlers;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001662 struct iw_public_data * wireless_data;
Johannes Bergb86e0282007-04-26 20:48:23 -07001663#endif
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001664 const struct net_device_ops *netdev_ops;
Stephen Hemminger76fd8592006-09-08 11:16:13 -07001665 const struct ethtool_ops *ethtool_ops;
Scott Feldman41706042015-03-15 21:07:14 -07001666#ifdef CONFIG_NET_SWITCHDEV
Jiri Pirko9d47c0a2015-05-10 09:47:47 -07001667 const struct switchdev_ops *switchdev_ops;
Scott Feldman41706042015-03-15 21:07:14 -07001668#endif
David Ahern1b69c6d2015-09-29 20:07:11 -07001669#ifdef CONFIG_NET_L3_MASTER_DEV
1670 const struct l3mdev_ops *l3mdev_ops;
1671#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001672
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07001673 const struct header_ops *header_ops;
1674
Karoly Kemeny536721b2014-07-30 20:27:36 +02001675 unsigned int flags;
1676 unsigned int priv_flags;
1677
Linus Torvalds1da177e2005-04-16 15:20:36 -07001678 unsigned short gflags;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001679 unsigned short padded;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001680
Karoly Kemeny536721b2014-07-30 20:27:36 +02001681 unsigned char operstate;
1682 unsigned char link_mode;
Stefan Rompfb00055a2006-03-20 17:09:11 -08001683
Karoly Kemeny536721b2014-07-30 20:27:36 +02001684 unsigned char if_port;
1685 unsigned char dma;
Joe Perchesbdc220d2011-05-09 17:42:46 +00001686
Karoly Kemeny536721b2014-07-30 20:27:36 +02001687 unsigned int mtu;
1688 unsigned short type;
1689 unsigned short hard_header_len;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001690
Johannes Bergf5184d22008-05-12 20:48:31 -07001691 unsigned short needed_headroom;
1692 unsigned short needed_tailroom;
1693
Linus Torvalds1da177e2005-04-16 15:20:36 -07001694 /* Interface address info. */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001695 unsigned char perm_addr[MAX_ADDR_LEN];
1696 unsigned char addr_assign_type;
1697 unsigned char addr_len;
Sebastian Siewiora0a96632013-12-12 10:15:59 +01001698 unsigned short neigh_priv_len;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001699 unsigned short dev_id;
1700 unsigned short dev_port;
Jiri Pirkoccffad252009-05-22 23:22:17 +00001701 spinlock_t addr_list_lock;
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001702 unsigned char name_assign_type;
1703 bool uc_promisc;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001704 struct netdev_hw_addr_list uc;
1705 struct netdev_hw_addr_list mc;
1706 struct netdev_hw_addr_list dev_addrs;
1707
Eric Dumazet4c3d5e72013-03-30 06:31:03 +00001708#ifdef CONFIG_SYSFS
1709 struct kset *queues_kset;
1710#endif
Wang Chen9d45abe2008-06-17 21:12:48 -07001711 unsigned int promiscuity;
1712 unsigned int allmulti;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001713
Linus Torvalds1da177e2005-04-16 15:20:36 -07001714
1715 /* Protocol specific pointers */
Jesse Gross65ac6a52010-10-20 13:56:05 +00001716
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001717#if IS_ENABLED(CONFIG_VLAN_8021Q)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001718 struct vlan_info __rcu *vlan_info;
Jesse Gross65ac6a52010-10-20 13:56:05 +00001719#endif
Ben Hutchings34a430d2011-11-25 14:38:38 +00001720#if IS_ENABLED(CONFIG_NET_DSA)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001721 struct dsa_switch_tree *dsa_ptr;
Lennert Buytenhek91da11f2008-10-07 13:44:02 +00001722#endif
Ying Xue37cb0622013-12-10 20:45:41 -08001723#if IS_ENABLED(CONFIG_TIPC)
Karoly Kemeny536721b2014-07-30 20:27:36 +02001724 struct tipc_bearer __rcu *tipc_ptr;
Ying Xue37cb0622013-12-10 20:45:41 -08001725#endif
Karoly Kemeny536721b2014-07-30 20:27:36 +02001726 void *atalk_ptr;
1727 struct in_device __rcu *ip_ptr;
1728 struct dn_dev __rcu *dn_ptr;
1729 struct inet6_dev __rcu *ip6_ptr;
1730 void *ax25_ptr;
1731 struct wireless_dev *ieee80211_ptr;
Alexander Aring98a18b62014-11-02 06:44:54 +01001732 struct wpan_dev *ieee802154_ptr;
Robert Shearman03c57742015-04-22 11:14:37 +01001733#if IS_ENABLED(CONFIG_MPLS_ROUTING)
1734 struct mpls_dev __rcu *mpls_ptr;
1735#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001736
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001737/*
Eric Dumazetcd135392010-09-16 02:58:13 +00001738 * Cache lines mostly used on receive path (including eth_type_trans())
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001739 */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001740 unsigned long last_rx;
Eric Dumazet4dc89132010-08-31 07:40:16 +00001741
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001742 /* Interface address info used in eth_type_trans() */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001743 unsigned char *dev_addr;
Jiri Pirkof001fde2009-05-05 02:48:28 +00001744
Tom Herbert0a9627f2010-03-16 08:03:29 +00001745
Michael Daltona953be52014-01-16 22:23:28 -08001746#ifdef CONFIG_SYSFS
Tom Herbert0a9627f2010-03-16 08:03:29 +00001747 struct netdev_rx_queue *_rx;
1748
Tom Herbert0a9627f2010-03-16 08:03:29 +00001749 unsigned int num_rx_queues;
Ben Hutchings62fe0b42010-09-27 08:24:33 +00001750 unsigned int real_num_rx_queues;
Ben Hutchingsc4454772011-01-19 11:03:53 +00001751
Eric Dumazetdf334542010-03-24 19:13:54 +00001752#endif
Tom Herbert0a9627f2010-03-16 08:03:29 +00001753
Eric Dumazet3b47d302014-11-06 21:09:44 -08001754 unsigned long gro_flush_timeout;
stephen hemminger61391cd2010-11-15 06:38:12 +00001755 rx_handler_func_t __rcu *rx_handler;
1756 void __rcu *rx_handler_data;
David S. Millere8a04642008-07-17 00:34:19 -07001757
Daniel Borkmann4cda01e2015-05-11 19:28:49 +02001758#ifdef CONFIG_NET_CLS_ACT
Daniel Borkmannd2788d32015-05-09 22:51:32 +02001759 struct tcf_proto __rcu *ingress_cl_list;
1760#endif
Eric Dumazet24824a02010-10-02 06:11:55 +00001761 struct netdev_queue __rcu *ingress_queue;
Pablo Neirae687ad62015-05-13 18:19:38 +02001762#ifdef CONFIG_NETFILTER_INGRESS
1763 struct list_head nf_hooks_ingress;
1764#endif
Daniel Borkmannd2788d32015-05-09 22:51:32 +02001765
Karoly Kemeny536721b2014-07-30 20:27:36 +02001766 unsigned char broadcast[MAX_ADDR_LEN];
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001767#ifdef CONFIG_RFS_ACCEL
1768 struct cpu_rmap *rx_cpu_rmap;
1769#endif
1770 struct hlist_node index_hlist;
Eric Dumazetcd135392010-09-16 02:58:13 +00001771
1772/*
1773 * Cache lines mostly used on transmit path
1774 */
David S. Millere8a04642008-07-17 00:34:19 -07001775 struct netdev_queue *_tx ____cacheline_aligned_in_smp;
1776 unsigned int num_tx_queues;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07001777 unsigned int real_num_tx_queues;
Patrick McHardyaf356af2009-09-04 06:41:18 +00001778 struct Qdisc *qdisc;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001779 unsigned long tx_queue_len;
David S. Millerc3f26a22008-07-31 16:58:50 -07001780 spinlock_t tx_global_lock;
Thomas Graf14ffbbb2015-04-10 15:52:38 +02001781 int watchdog_timeo;
Eric Dumazetcd135392010-09-16 02:58:13 +00001782
Tom Herbertbf264142010-11-26 08:36:09 +00001783#ifdef CONFIG_XPS
Eric Dumazeta4177862010-11-28 21:43:02 +00001784 struct xps_dev_maps __rcu *xps_maps;
Tom Herbertbf264142010-11-26 08:36:09 +00001785#endif
Daniel Borkmann1f211a12016-01-07 22:29:47 +01001786#ifdef CONFIG_NET_CLS_ACT
1787 struct tcf_proto __rcu *egress_cl_list;
1788#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -07001789#ifdef CONFIG_NET_SWITCHDEV
1790 u32 offload_fwd_mark;
1791#endif
1792
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001793 /* These may be needed for future network-power-down code. */
Eric Dumazet9d214932009-05-17 20:55:16 -07001794
1795 /*
1796 * trans_start here is expensive for high speed devices on SMP,
1797 * please use netdev_queue->trans_start instead.
1798 */
Karoly Kemeny536721b2014-07-30 20:27:36 +02001799 unsigned long trans_start;
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001800
Eric Dumazet9356b8f2005-09-27 15:23:16 -07001801 struct timer_list watchdog_timer;
1802
Eric Dumazet29b44332010-10-11 10:22:12 +00001803 int __percpu *pcpu_refcnt;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001804 struct list_head todo_list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001805
Eric Dumazete014deb2009-11-17 05:59:21 +00001806 struct list_head link_watch_list;
Herbert Xu572a1032007-05-08 18:34:17 -07001807
Linus Torvalds1da177e2005-04-16 15:20:36 -07001808 enum { NETREG_UNINITIALIZED=0,
Stephen Hemmingerb17a7c12006-05-10 13:21:17 -07001809 NETREG_REGISTERED, /* completed register_netdevice */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001810 NETREG_UNREGISTERING, /* called unregister_netdevice */
1811 NETREG_UNREGISTERED, /* completed unregister todo */
1812 NETREG_RELEASED, /* called free_netdev */
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08001813 NETREG_DUMMY, /* dummy device for NAPI poll */
Eric Dumazet449f4542011-05-19 12:24:16 +00001814 } reg_state:8;
1815
Karoly Kemeny536721b2014-07-30 20:27:36 +02001816 bool dismantle;
Patrick McHardya2835762010-02-26 06:34:51 +00001817
1818 enum {
1819 RTNL_LINK_INITIALIZED,
1820 RTNL_LINK_INITIALIZING,
1821 } rtnl_link_state:16;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001822
Stephen Hemmingerd3147742008-11-19 21:32:24 -08001823 void (*destructor)(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001824
Linus Torvalds1da177e2005-04-16 15:20:36 -07001825#ifdef CONFIG_NETPOLL
Cong Wang5fbee842013-01-22 21:29:39 +00001826 struct netpoll_info __rcu *npinfo;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001827#endif
David S. Millereae792b2008-07-15 03:03:33 -07001828
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -05001829 possible_net_t nd_net;
Eric W. Biederman4a1c5372007-09-12 11:56:32 +02001830
David S. Miller49517042008-05-12 03:29:11 -07001831 /* mid-layer private */
Eric Dumazeta7855c72010-09-23 23:51:51 +00001832 union {
Karoly Kemeny536721b2014-07-30 20:27:36 +02001833 void *ml_priv;
1834 struct pcpu_lstats __percpu *lstats;
Li RongQing8f849852014-01-04 13:57:59 +08001835 struct pcpu_sw_netstats __percpu *tstats;
Karoly Kemeny536721b2014-07-30 20:27:36 +02001836 struct pcpu_dstats __percpu *dstats;
1837 struct pcpu_vstats __percpu *vstats;
Eric Dumazeta7855c72010-09-23 23:51:51 +00001838 };
Karoly Kemeny536721b2014-07-30 20:27:36 +02001839
Eric Dumazet3cc77ec2010-10-24 21:32:36 +00001840 struct garp_port __rcu *garp_port;
David Wardfebf0182013-02-08 17:17:06 +00001841 struct mrp_port __rcu *mrp_port;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001842
Karoly Kemeny536721b2014-07-30 20:27:36 +02001843 struct device dev;
Eric W. Biederman0c509a62009-10-29 14:18:21 +00001844 const struct attribute_group *sysfs_groups[4];
Michael Daltona953be52014-01-16 22:23:28 -08001845 const struct attribute_group *sysfs_rx_queue_group;
Patrick McHardy38f7b872007-06-13 12:03:51 -07001846
Patrick McHardy38f7b872007-06-13 12:03:51 -07001847 const struct rtnl_link_ops *rtnl_link_ops;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001848
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07001849 /* for setting kernel sock attribute on TCP connection setup */
1850#define GSO_MAX_SIZE 65536
1851 unsigned int gso_max_size;
Ben Hutchings30b678d2012-07-30 15:57:00 +00001852#define GSO_MAX_SEGS 65535
1853 u16 gso_max_segs;
Eric Dumazetfcbeb972014-10-05 10:11:27 -07001854 u16 gso_min_segs;
Jeff Kirsher7a6b6f52008-11-25 01:02:08 -08001855#ifdef CONFIG_DCB
Stephen Hemminger32953542009-10-05 06:01:03 +00001856 const struct dcbnl_rtnl_ops *dcbnl_ops;
Alexander Duyck2f90b862008-11-20 20:52:10 -08001857#endif
John Fastabend4f57c082011-01-17 08:06:04 +00001858 u8 num_tc;
1859 struct netdev_tc_txq tc_to_txq[TC_MAX_QUEUE];
1860 u8 prio_tc_map[TC_BITMASK + 1];
Alexander Duyck2f90b862008-11-20 20:52:10 -08001861
Ben Hutchingsd11ead72011-11-25 14:40:26 +00001862#if IS_ENABLED(CONFIG_FCOE)
Yi Zou4d288d52009-02-27 14:06:59 -08001863 unsigned int fcoe_ddp_xid;
1864#endif
Daniel Borkmann86f85152013-12-29 17:27:11 +01001865#if IS_ENABLED(CONFIG_CGROUP_NET_PRIO)
Neil Horman5bc14212011-11-22 05:10:51 +00001866 struct netprio_map __rcu *priomap;
1867#endif
Richard Cochranc1f19b52010-07-17 08:49:36 +00001868 struct phy_device *phydev;
Eric Dumazet23d3b8b2012-09-05 01:02:56 +00001869 struct lock_class_key *qdisc_tx_busylock;
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07001870 bool proto_down;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001871};
Greg Kroah-Hartman43cb76d2002-04-09 12:14:34 -07001872#define to_net_dev(d) container_of(d, struct net_device, dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001873
1874#define NETDEV_ALIGN 32
Linus Torvalds1da177e2005-04-16 15:20:36 -07001875
David S. Millere8a04642008-07-17 00:34:19 -07001876static inline
John Fastabend4f57c082011-01-17 08:06:04 +00001877int netdev_get_prio_tc_map(const struct net_device *dev, u32 prio)
1878{
1879 return dev->prio_tc_map[prio & TC_BITMASK];
1880}
1881
1882static inline
1883int netdev_set_prio_tc_map(struct net_device *dev, u8 prio, u8 tc)
1884{
1885 if (tc >= dev->num_tc)
1886 return -EINVAL;
1887
1888 dev->prio_tc_map[prio & TC_BITMASK] = tc & TC_BITMASK;
1889 return 0;
1890}
1891
1892static inline
1893void netdev_reset_tc(struct net_device *dev)
1894{
1895 dev->num_tc = 0;
1896 memset(dev->tc_to_txq, 0, sizeof(dev->tc_to_txq));
1897 memset(dev->prio_tc_map, 0, sizeof(dev->prio_tc_map));
1898}
1899
1900static inline
1901int netdev_set_tc_queue(struct net_device *dev, u8 tc, u16 count, u16 offset)
1902{
1903 if (tc >= dev->num_tc)
1904 return -EINVAL;
1905
1906 dev->tc_to_txq[tc].count = count;
1907 dev->tc_to_txq[tc].offset = offset;
1908 return 0;
1909}
1910
1911static inline
1912int netdev_set_num_tc(struct net_device *dev, u8 num_tc)
1913{
1914 if (num_tc > TC_MAX_QUEUE)
1915 return -EINVAL;
1916
1917 dev->num_tc = num_tc;
1918 return 0;
1919}
1920
1921static inline
1922int netdev_get_num_tc(struct net_device *dev)
1923{
1924 return dev->num_tc;
1925}
1926
1927static inline
David S. Millere8a04642008-07-17 00:34:19 -07001928struct netdev_queue *netdev_get_tx_queue(const struct net_device *dev,
1929 unsigned int index)
1930{
1931 return &dev->_tx[index];
1932}
1933
Daniel Borkmann10c51b56232014-08-27 11:11:27 +02001934static inline struct netdev_queue *skb_get_tx_queue(const struct net_device *dev,
1935 const struct sk_buff *skb)
1936{
1937 return netdev_get_tx_queue(dev, skb_get_queue_mapping(skb));
1938}
1939
David S. Millere8a04642008-07-17 00:34:19 -07001940static inline void netdev_for_each_tx_queue(struct net_device *dev,
1941 void (*f)(struct net_device *,
1942 struct netdev_queue *,
1943 void *),
1944 void *arg)
1945{
1946 unsigned int i;
1947
1948 for (i = 0; i < dev->num_tx_queues; i++)
1949 f(dev, &dev->_tx[i], arg);
1950}
1951
Joe Perchesf629d202013-09-26 14:48:15 -07001952struct netdev_queue *netdev_pick_tx(struct net_device *dev,
Jason Wangf663dd92014-01-10 16:18:26 +08001953 struct sk_buff *skb,
1954 void *accel_priv);
Amerigo Wang8c4c49d2012-09-17 20:16:31 +00001955
Paolo Abeni871b6422016-02-26 10:45:37 +01001956/* returns the headroom that the master device needs to take in account
1957 * when forwarding to this dev
1958 */
1959static inline unsigned netdev_get_fwd_headroom(struct net_device *dev)
1960{
1961 return dev->priv_flags & IFF_PHONY_HEADROOM ? 0 : dev->needed_headroom;
1962}
1963
1964static inline void netdev_set_rx_headroom(struct net_device *dev, int new_hr)
1965{
1966 if (dev->netdev_ops->ndo_set_rx_headroom)
1967 dev->netdev_ops->ndo_set_rx_headroom(dev, new_hr);
1968}
1969
1970/* set the device rx headroom to the dev's default */
1971static inline void netdev_reset_rx_headroom(struct net_device *dev)
1972{
1973 netdev_set_rx_headroom(dev, -1);
1974}
1975
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001976/*
1977 * Net namespace inlines
1978 */
1979static inline
1980struct net *dev_net(const struct net_device *dev)
1981{
Eric Dumazetc2d9ba92010-06-01 06:51:19 +00001982 return read_pnet(&dev->nd_net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001983}
1984
1985static inline
Denis V. Lunevf5aa23f2008-03-26 00:48:17 -07001986void dev_net_set(struct net_device *dev, struct net *net)
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001987{
Eric W. Biederman0c5c9fb2015-03-11 23:06:44 -05001988 write_pnet(&dev->nd_net, net);
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09001989}
1990
Florian Fainelli3e8a72d2014-08-27 17:04:46 -07001991static inline bool netdev_uses_dsa(struct net_device *dev)
Lennert Buytenhekcf85d082008-10-07 13:45:02 +00001992{
Alexander Duyck3fc88672014-09-12 23:12:46 -07001993#if IS_ENABLED(CONFIG_NET_DSA)
Florian Fainelli5aed85c2014-08-27 17:04:52 -07001994 if (dev->dsa_ptr != NULL)
1995 return dsa_uses_tagged_protocol(dev->dsa_ptr);
Lennert Buytenhekcf85d082008-10-07 13:45:02 +00001996#endif
Florian Fainelli5aed85c2014-08-27 17:04:52 -07001997 return false;
Lennert Buytenhek396138f2008-10-07 13:46:07 +00001998}
1999
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002000/**
2001 * netdev_priv - access network device private data
2002 * @dev: network device
2003 *
2004 * Get network device private data
2005 */
Patrick McHardy6472ce62007-06-13 12:03:21 -07002006static inline void *netdev_priv(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002007{
Eric Dumazet1ce8e7b2009-05-27 04:42:37 +00002008 return (char *)dev + ALIGN(sizeof(struct net_device), NETDEV_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009}
2010
Linus Torvalds1da177e2005-04-16 15:20:36 -07002011/* Set the sysfs physical device reference for the network logical device
2012 * if set prior to registration will cause a symlink during initialization.
2013 */
Greg Kroah-Hartman43cb76d2002-04-09 12:14:34 -07002014#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002015
Marcel Holtmann384912e2009-08-31 21:08:19 +00002016/* Set the sysfs device type for the network logical device to allow
Maxime Jayat3f794102013-10-12 01:29:46 +02002017 * fine-grained identification of different network device types. For
Marcel Holtmann384912e2009-08-31 21:08:19 +00002018 * example Ethernet, Wirelss LAN, Bluetooth, WiMAX etc.
2019 */
2020#define SET_NETDEV_DEVTYPE(net, devtype) ((net)->dev.type = (devtype))
2021
Eric Dumazet82dc3c63c2013-03-05 15:57:22 +00002022/* Default NAPI poll() weight
2023 * Device drivers are strongly advised to not use bigger value
2024 */
2025#define NAPI_POLL_WEIGHT 64
2026
Stephen Hemminger3b582cc2007-11-01 02:21:47 -07002027/**
2028 * netif_napi_add - initialize a napi context
2029 * @dev: network device
2030 * @napi: napi context
2031 * @poll: polling function
2032 * @weight: default weight
2033 *
2034 * netif_napi_add() must be used to initialize a napi context prior to calling
2035 * *any* of the other napi related functions.
2036 */
Herbert Xud565b0a2008-12-15 23:38:52 -08002037void netif_napi_add(struct net_device *dev, struct napi_struct *napi,
2038 int (*poll)(struct napi_struct *, int), int weight);
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002039
Alexander Duyckd8156532008-07-08 15:13:05 -07002040/**
Eric Dumazetd64b5e82015-11-18 06:31:00 -08002041 * netif_tx_napi_add - initialize a napi context
2042 * @dev: network device
2043 * @napi: napi context
2044 * @poll: polling function
2045 * @weight: default weight
2046 *
2047 * This variant of netif_napi_add() should be used from drivers using NAPI
2048 * to exclusively poll a TX queue.
2049 * This will avoid we add it into napi_hash[], thus polluting this hash table.
2050 */
2051static inline void netif_tx_napi_add(struct net_device *dev,
2052 struct napi_struct *napi,
2053 int (*poll)(struct napi_struct *, int),
2054 int weight)
2055{
2056 set_bit(NAPI_STATE_NO_BUSY_POLL, &napi->state);
2057 netif_napi_add(dev, napi, poll, weight);
2058}
2059
2060/**
Alexander Duyckd8156532008-07-08 15:13:05 -07002061 * netif_napi_del - remove a napi context
2062 * @napi: napi context
2063 *
2064 * netif_napi_del() removes a napi context from the network device napi list
2065 */
Herbert Xud565b0a2008-12-15 23:38:52 -08002066void netif_napi_del(struct napi_struct *napi);
2067
2068struct napi_gro_cb {
Herbert Xu78a478d2009-05-26 18:50:21 +00002069 /* Virtual address of skb_shinfo(skb)->frags[0].page + offset. */
2070 void *frag0;
2071
Herbert Xu74895942009-05-26 18:50:27 +00002072 /* Length of frag0. */
2073 unsigned int frag0_len;
2074
Herbert Xu86911732009-01-29 14:19:50 +00002075 /* This indicates where we are processing relative to skb->data. */
2076 int data_offset;
2077
Herbert Xud565b0a2008-12-15 23:38:52 -08002078 /* This is non-zero if the packet cannot be merged with the new skb. */
Jerry Chubf5a7552014-01-07 10:23:19 -08002079 u16 flush;
2080
2081 /* Save the IP ID here and check when we get to the transport layer */
2082 u16 flush_id;
Herbert Xud565b0a2008-12-15 23:38:52 -08002083
2084 /* Number of segments aggregated. */
Eric Dumazet2e71a6f2012-10-06 08:08:49 +00002085 u16 count;
2086
Tom Herbert15e23962015-02-10 16:30:31 -08002087 /* Start offset for remote checksum offload */
2088 u16 gro_remcsum_start;
2089
Eric Dumazet2e71a6f2012-10-06 08:08:49 +00002090 /* jiffies when first packet was created/queued */
2091 unsigned long age;
Eric Dumazet86347242012-10-08 21:38:50 +02002092
Tom Herbertafe93322014-09-17 12:25:57 -07002093 /* Used in ipv6_gro_receive() and foo-over-udp */
Or Gerlitzb582ef02014-01-20 13:59:19 +02002094 u16 proto;
2095
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002096 /* This is non-zero if the packet may be of the same flow. */
2097 u8 same_flow:1;
2098
Jesse Grossfac8e0f2016-03-19 09:32:01 -07002099 /* Used in tunnel GRO receive */
2100 u8 encap_mark:1;
Tom Herbert573e8fc2014-08-22 13:33:47 -07002101
2102 /* GRO checksum is valid */
2103 u8 csum_valid:1;
2104
Tom Herbert662880f2014-08-27 21:26:56 -07002105 /* Number of checksums via CHECKSUM_UNNECESSARY */
2106 u8 csum_cnt:3;
Eric Dumazetc3c7c252012-12-06 13:54:59 +00002107
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002108 /* Free the skb? */
2109 u8 free:2;
2110#define NAPI_GRO_FREE 1
2111#define NAPI_GRO_FREE_STOLEN_HEAD 2
2112
Tom Herbertefc98d02014-10-03 15:48:08 -07002113 /* Used in foo-over-udp, set in udp[46]_gro_receive */
2114 u8 is_ipv6:1;
2115
Tom Herbertbaa32ff2015-02-10 16:30:30 -08002116 /* 7 bit hole */
2117
Jerry Chubf5a7552014-01-07 10:23:19 -08002118 /* used to support CHECKSUM_COMPLETE for tunneling protocols */
2119 __wsum csum;
2120
Eric Dumazetc3c7c252012-12-06 13:54:59 +00002121 /* used in skb_gro_receive() slow path */
2122 struct sk_buff *last;
Herbert Xud565b0a2008-12-15 23:38:52 -08002123};
2124
2125#define NAPI_GRO_CB(skb) ((struct napi_gro_cb *)(skb)->cb)
Alexander Duyckd8156532008-07-08 15:13:05 -07002126
Linus Torvalds1da177e2005-04-16 15:20:36 -07002127struct packet_type {
David S. Millerf2ccd8f2005-08-09 19:34:12 -07002128 __be16 type; /* This is really htons(ether_type). */
2129 struct net_device *dev; /* NULL is wildcarded here */
2130 int (*func) (struct sk_buff *,
2131 struct net_device *,
2132 struct packet_type *,
2133 struct net_device *);
Eric Leblondc0de08d2012-08-16 22:02:58 +00002134 bool (*id_match)(struct packet_type *ptype,
2135 struct sock *sk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002136 void *af_packet_priv;
2137 struct list_head list;
2138};
2139
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002140struct offload_callbacks {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002141 struct sk_buff *(*gso_segment)(struct sk_buff *skb,
2142 netdev_features_t features);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002143 struct sk_buff **(*gro_receive)(struct sk_buff **head,
Tom Herberta2b12f32015-01-12 17:00:37 -08002144 struct sk_buff *skb);
Jerry Chu299603e82013-12-11 20:53:45 -08002145 int (*gro_complete)(struct sk_buff *skb, int nhoff);
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002146};
2147
2148struct packet_offload {
2149 __be16 type; /* This is really htons(ether_type). */
David S. Millerbdef7de2015-06-01 14:56:09 -07002150 u16 priority;
Vlad Yasevichf191a1d2012-11-15 08:49:23 +00002151 struct offload_callbacks callbacks;
2152 struct list_head list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002153};
2154
Tom Herberta2b12f32015-01-12 17:00:37 -08002155struct udp_offload;
2156
2157struct udp_offload_callbacks {
2158 struct sk_buff **(*gro_receive)(struct sk_buff **head,
2159 struct sk_buff *skb,
2160 struct udp_offload *uoff);
2161 int (*gro_complete)(struct sk_buff *skb,
2162 int nhoff,
2163 struct udp_offload *uoff);
2164};
2165
Or Gerlitzb582ef02014-01-20 13:59:19 +02002166struct udp_offload {
2167 __be16 port;
Tom Herbertafe93322014-09-17 12:25:57 -07002168 u8 ipproto;
Tom Herberta2b12f32015-01-12 17:00:37 -08002169 struct udp_offload_callbacks callbacks;
Or Gerlitzb582ef02014-01-20 13:59:19 +02002170};
2171
Li RongQing8f849852014-01-04 13:57:59 +08002172/* often modified stats are per cpu, other are shared (netdev->stats) */
2173struct pcpu_sw_netstats {
2174 u64 rx_packets;
2175 u64 rx_bytes;
2176 u64 tx_packets;
2177 u64 tx_bytes;
2178 struct u64_stats_sync syncp;
2179};
2180
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002181#define __netdev_alloc_pcpu_stats(type, gfp) \
2182({ \
2183 typeof(type) __percpu *pcpu_stats = alloc_percpu_gfp(type, gfp);\
2184 if (pcpu_stats) { \
2185 int __cpu; \
2186 for_each_possible_cpu(__cpu) { \
2187 typeof(type) *stat; \
2188 stat = per_cpu_ptr(pcpu_stats, __cpu); \
2189 u64_stats_init(&stat->syncp); \
2190 } \
2191 } \
2192 pcpu_stats; \
WANG Cong1c213bd2014-02-13 11:46:28 -08002193})
2194
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002195#define netdev_alloc_pcpu_stats(type) \
Felix Fietkau326fcfa2015-12-05 13:58:11 +01002196 __netdev_alloc_pcpu_stats(type, GFP_KERNEL)
Pablo Neira Ayusoaabc92b2015-11-10 14:31:18 +01002197
Jiri Pirko764f5e52015-12-03 12:12:12 +01002198enum netdev_lag_tx_type {
2199 NETDEV_LAG_TX_TYPE_UNKNOWN,
2200 NETDEV_LAG_TX_TYPE_RANDOM,
2201 NETDEV_LAG_TX_TYPE_BROADCAST,
2202 NETDEV_LAG_TX_TYPE_ROUNDROBIN,
2203 NETDEV_LAG_TX_TYPE_ACTIVEBACKUP,
2204 NETDEV_LAG_TX_TYPE_HASH,
2205};
2206
2207struct netdev_lag_upper_info {
2208 enum netdev_lag_tx_type tx_type;
2209};
2210
Jiri Pirkofb1b2e32015-12-03 12:12:16 +01002211struct netdev_lag_lower_state_info {
2212 u8 link_up : 1,
2213 tx_enabled : 1;
2214};
2215
Linus Torvalds1da177e2005-04-16 15:20:36 -07002216#include <linux/notifier.h>
2217
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002218/* netdevice notifier chain. Please remember to update the rtnetlink
2219 * notification exclusion list in rtnetlink_event() when adding new
2220 * types.
2221 */
2222#define NETDEV_UP 0x0001 /* For now you can't veto a device up/down */
2223#define NETDEV_DOWN 0x0002
2224#define NETDEV_REBOOT 0x0003 /* Tell a protocol stack a network interface
2225 detected a hardware crash and restarted
2226 - we can use this eg to kick tcp sessions
2227 once done */
2228#define NETDEV_CHANGE 0x0004 /* Notify device state change */
2229#define NETDEV_REGISTER 0x0005
2230#define NETDEV_UNREGISTER 0x0006
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002231#define NETDEV_CHANGEMTU 0x0007 /* notify after mtu change happened */
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002232#define NETDEV_CHANGEADDR 0x0008
2233#define NETDEV_GOING_DOWN 0x0009
2234#define NETDEV_CHANGENAME 0x000A
2235#define NETDEV_FEAT_CHANGE 0x000B
2236#define NETDEV_BONDING_FAILOVER 0x000C
2237#define NETDEV_PRE_UP 0x000D
2238#define NETDEV_PRE_TYPE_CHANGE 0x000E
2239#define NETDEV_POST_TYPE_CHANGE 0x000F
2240#define NETDEV_POST_INIT 0x0010
Eric Dumazet0115e8e2012-08-22 17:19:46 +00002241#define NETDEV_UNREGISTER_FINAL 0x0011
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002242#define NETDEV_RELEASE 0x0012
2243#define NETDEV_NOTIFY_PEERS 0x0013
2244#define NETDEV_JOIN 0x0014
Jiri Pirko42e52bf2013-05-25 04:12:10 +00002245#define NETDEV_CHANGEUPPER 0x0015
Jiri Pirko4aa5dee2013-07-20 12:13:53 +02002246#define NETDEV_RESEND_IGMP 0x0016
Veaceslav Falico1d486bf2014-01-16 00:02:18 +01002247#define NETDEV_PRECHANGEMTU 0x0017 /* notify before mtu change happened */
Jiri Pirkod4261e52014-08-19 16:02:12 +02002248#define NETDEV_CHANGEINFODATA 0x0018
Moni Shoua61bd3852015-02-03 16:48:29 +02002249#define NETDEV_BONDING_INFO 0x0019
Jiri Pirko573c7ba2015-10-16 14:01:22 +02002250#define NETDEV_PRECHANGEUPPER 0x001A
Jiri Pirko04d48262015-12-03 12:12:15 +01002251#define NETDEV_CHANGELOWERSTATE 0x001B
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002252
Joe Perchesf629d202013-09-26 14:48:15 -07002253int register_netdevice_notifier(struct notifier_block *nb);
2254int unregister_netdevice_notifier(struct notifier_block *nb);
Jiri Pirko351638e2013-05-28 01:30:21 +00002255
2256struct netdev_notifier_info {
2257 struct net_device *dev;
2258};
2259
Jiri Pirkobe9efd32013-05-28 01:30:22 +00002260struct netdev_notifier_change_info {
2261 struct netdev_notifier_info info; /* must be first */
2262 unsigned int flags_changed;
2263};
2264
Jiri Pirko0e4ead92015-08-27 09:31:18 +02002265struct netdev_notifier_changeupper_info {
2266 struct netdev_notifier_info info; /* must be first */
2267 struct net_device *upper_dev; /* new upper dev */
2268 bool master; /* is upper dev master */
2269 bool linking; /* is the nofication for link or unlink */
Jiri Pirko29bf24a2015-12-03 12:12:11 +01002270 void *upper_info; /* upper dev info */
Jiri Pirko0e4ead92015-08-27 09:31:18 +02002271};
2272
Jiri Pirko04d48262015-12-03 12:12:15 +01002273struct netdev_notifier_changelowerstate_info {
2274 struct netdev_notifier_info info; /* must be first */
2275 void *lower_state_info; /* is lower dev state */
2276};
2277
Cong Wang75538c22013-05-29 11:30:50 +08002278static inline void netdev_notifier_info_init(struct netdev_notifier_info *info,
2279 struct net_device *dev)
2280{
2281 info->dev = dev;
2282}
2283
Jiri Pirko351638e2013-05-28 01:30:21 +00002284static inline struct net_device *
2285netdev_notifier_info_to_dev(const struct netdev_notifier_info *info)
2286{
2287 return info->dev;
2288}
2289
Joe Perchesf629d202013-09-26 14:48:15 -07002290int call_netdevice_notifiers(unsigned long val, struct net_device *dev);
Amerigo Wangdcfe1422011-07-25 17:13:09 -07002291
2292
Linus Torvalds1da177e2005-04-16 15:20:36 -07002293extern rwlock_t dev_base_lock; /* Device list lock */
2294
Eric W. Biederman881d9662007-09-17 11:56:21 -07002295#define for_each_netdev(net, d) \
2296 list_for_each_entry(d, &(net)->dev_base_head, dev_list)
Eric W. Biedermandcbccbd42009-11-29 22:25:26 +00002297#define for_each_netdev_reverse(net, d) \
2298 list_for_each_entry_reverse(d, &(net)->dev_base_head, dev_list)
Eric Dumazetc6d14c82009-11-04 05:43:23 -08002299#define for_each_netdev_rcu(net, d) \
2300 list_for_each_entry_rcu(d, &(net)->dev_base_head, dev_list)
Eric W. Biederman881d9662007-09-17 11:56:21 -07002301#define for_each_netdev_safe(net, d, n) \
2302 list_for_each_entry_safe(d, n, &(net)->dev_base_head, dev_list)
2303#define for_each_netdev_continue(net, d) \
2304 list_for_each_entry_continue(d, &(net)->dev_base_head, dev_list)
stephen hemminger254245d2009-11-10 07:54:47 +00002305#define for_each_netdev_continue_rcu(net, d) \
2306 list_for_each_entry_continue_rcu(d, &(net)->dev_base_head, dev_list)
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00002307#define for_each_netdev_in_bond_rcu(bond, slave) \
2308 for_each_netdev_rcu(&init_net, slave) \
Benjamin Poirier4ccce022015-01-14 16:52:35 +09002309 if (netdev_master_upper_dev_get_rcu(slave) == (bond))
Pavel Emelianov7562f872007-05-03 15:13:45 -07002310#define net_device_entry(lh) list_entry(lh, struct net_device, dev_list)
2311
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002312static inline struct net_device *next_net_device(struct net_device *dev)
2313{
2314 struct list_head *lh;
2315 struct net *net;
Pavel Emelianov7562f872007-05-03 15:13:45 -07002316
YOSHIFUJI Hideakic346dca2008-03-25 21:47:49 +09002317 net = dev_net(dev);
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002318 lh = dev->dev_list.next;
2319 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2320}
2321
Eric Dumazetce81b762009-11-11 17:34:30 +00002322static inline struct net_device *next_net_device_rcu(struct net_device *dev)
2323{
2324 struct list_head *lh;
2325 struct net *net;
2326
2327 net = dev_net(dev);
Eric Dumazetccf43432011-01-26 18:08:02 +00002328 lh = rcu_dereference(list_next_rcu(&dev->dev_list));
Eric Dumazetce81b762009-11-11 17:34:30 +00002329 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2330}
2331
Daniel Lezcanoa050c332007-09-12 14:57:09 +02002332static inline struct net_device *first_net_device(struct net *net)
2333{
2334 return list_empty(&net->dev_base_head) ? NULL :
2335 net_device_entry(net->dev_base_head.next);
2336}
Pavel Emelianov7562f872007-05-03 15:13:45 -07002337
Eric Dumazetccf43432011-01-26 18:08:02 +00002338static inline struct net_device *first_net_device_rcu(struct net *net)
2339{
2340 struct list_head *lh = rcu_dereference(list_next_rcu(&net->dev_base_head));
2341
2342 return lh == &net->dev_base_head ? NULL : net_device_entry(lh);
2343}
2344
Joe Perchesf629d202013-09-26 14:48:15 -07002345int netdev_boot_setup_check(struct net_device *dev);
2346unsigned long netdev_boot_base(const char *prefix, int unit);
2347struct net_device *dev_getbyhwaddr_rcu(struct net *net, unsigned short type,
2348 const char *hwaddr);
2349struct net_device *dev_getfirstbyhwtype(struct net *net, unsigned short type);
2350struct net_device *__dev_getfirstbyhwtype(struct net *net, unsigned short type);
2351void dev_add_pack(struct packet_type *pt);
2352void dev_remove_pack(struct packet_type *pt);
2353void __dev_remove_pack(struct packet_type *pt);
2354void dev_add_offload(struct packet_offload *po);
2355void dev_remove_offload(struct packet_offload *po);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002356
Nicolas Dichtela54acb32015-04-02 17:07:00 +02002357int dev_get_iflink(const struct net_device *dev);
Pravin B Shelarfc4099f2015-10-22 18:17:16 -07002358int dev_fill_metadata_dst(struct net_device *dev, struct sk_buff *skb);
WANG Cong6c555492014-09-11 15:35:09 -07002359struct net_device *__dev_get_by_flags(struct net *net, unsigned short flags,
2360 unsigned short mask);
Joe Perchesf629d202013-09-26 14:48:15 -07002361struct net_device *dev_get_by_name(struct net *net, const char *name);
2362struct net_device *dev_get_by_name_rcu(struct net *net, const char *name);
2363struct net_device *__dev_get_by_name(struct net *net, const char *name);
2364int dev_alloc_name(struct net_device *dev, const char *name);
2365int dev_open(struct net_device *dev);
2366int dev_close(struct net_device *dev);
David S. Miller99c4a262015-03-18 22:52:33 -04002367int dev_close_many(struct list_head *head, bool unlink);
Joe Perchesf629d202013-09-26 14:48:15 -07002368void dev_disable_lro(struct net_device *dev);
Eric W. Biederman0c4b51f2015-09-15 20:04:18 -05002369int dev_loopback_xmit(struct net *net, struct sock *sk, struct sk_buff *newskb);
Eric W. Biederman2b4aa3c2015-09-15 20:04:07 -05002370int dev_queue_xmit(struct sk_buff *skb);
Jason Wangf663dd92014-01-10 16:18:26 +08002371int dev_queue_xmit_accel(struct sk_buff *skb, void *accel_priv);
Joe Perchesf629d202013-09-26 14:48:15 -07002372int register_netdevice(struct net_device *dev);
2373void unregister_netdevice_queue(struct net_device *dev, struct list_head *head);
2374void unregister_netdevice_many(struct list_head *head);
Eric Dumazet44a08732009-10-27 07:03:04 +00002375static inline void unregister_netdevice(struct net_device *dev)
2376{
2377 unregister_netdevice_queue(dev, NULL);
2378}
2379
Joe Perchesf629d202013-09-26 14:48:15 -07002380int netdev_refcnt_read(const struct net_device *dev);
2381void free_netdev(struct net_device *dev);
Eric Dumazet74d332c2013-10-30 13:10:44 -07002382void netdev_freemem(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002383void synchronize_net(void);
2384int init_dummy_netdev(struct net_device *dev);
Benjamin Herrenschmidt937f1ba2009-01-14 21:05:05 -08002385
hannes@stressinduktion.orgf60e5992015-04-01 17:07:44 +02002386DECLARE_PER_CPU(int, xmit_recursion);
2387static inline int dev_recursion_level(void)
2388{
2389 return this_cpu_read(xmit_recursion);
2390}
2391
Joe Perchesf629d202013-09-26 14:48:15 -07002392struct net_device *dev_get_by_index(struct net *net, int ifindex);
2393struct net_device *__dev_get_by_index(struct net *net, int ifindex);
2394struct net_device *dev_get_by_index_rcu(struct net *net, int ifindex);
2395int netdev_get_name(struct net *net, char *name, int ifindex);
2396int dev_restart(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07002397int skb_gro_receive(struct sk_buff **head, struct sk_buff *skb);
Herbert Xu86911732009-01-29 14:19:50 +00002398
2399static inline unsigned int skb_gro_offset(const struct sk_buff *skb)
2400{
2401 return NAPI_GRO_CB(skb)->data_offset;
2402}
2403
2404static inline unsigned int skb_gro_len(const struct sk_buff *skb)
2405{
2406 return skb->len - NAPI_GRO_CB(skb)->data_offset;
2407}
2408
2409static inline void skb_gro_pull(struct sk_buff *skb, unsigned int len)
2410{
2411 NAPI_GRO_CB(skb)->data_offset += len;
2412}
2413
Herbert Xua5b1cf22009-05-26 18:50:28 +00002414static inline void *skb_gro_header_fast(struct sk_buff *skb,
2415 unsigned int offset)
Herbert Xu86911732009-01-29 14:19:50 +00002416{
Herbert Xu78a478d2009-05-26 18:50:21 +00002417 return NAPI_GRO_CB(skb)->frag0 + offset;
Herbert Xu86911732009-01-29 14:19:50 +00002418}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002419
Herbert Xua5b1cf22009-05-26 18:50:28 +00002420static inline int skb_gro_header_hard(struct sk_buff *skb, unsigned int hlen)
2421{
2422 return NAPI_GRO_CB(skb)->frag0_len < hlen;
2423}
2424
2425static inline void *skb_gro_header_slow(struct sk_buff *skb, unsigned int hlen,
2426 unsigned int offset)
2427{
Herbert Xu17dd7592011-07-27 06:16:28 -07002428 if (!pskb_may_pull(skb, hlen))
2429 return NULL;
2430
Herbert Xua5b1cf22009-05-26 18:50:28 +00002431 NAPI_GRO_CB(skb)->frag0 = NULL;
2432 NAPI_GRO_CB(skb)->frag0_len = 0;
Herbert Xu17dd7592011-07-27 06:16:28 -07002433 return skb->data + offset;
Herbert Xua5b1cf22009-05-26 18:50:28 +00002434}
2435
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002436static inline void *skb_gro_network_header(struct sk_buff *skb)
2437{
Herbert Xu78d3fd02009-05-26 18:50:23 +00002438 return (NAPI_GRO_CB(skb)->frag0 ?: skb->data) +
2439 skb_network_offset(skb);
Herbert Xu36e7b1b2009-04-27 05:44:45 -07002440}
2441
Jerry Chubf5a7552014-01-07 10:23:19 -08002442static inline void skb_gro_postpull_rcsum(struct sk_buff *skb,
2443 const void *start, unsigned int len)
2444{
Tom Herbert573e8fc2014-08-22 13:33:47 -07002445 if (NAPI_GRO_CB(skb)->csum_valid)
Jerry Chubf5a7552014-01-07 10:23:19 -08002446 NAPI_GRO_CB(skb)->csum = csum_sub(NAPI_GRO_CB(skb)->csum,
2447 csum_partial(start, len, 0));
2448}
2449
Tom Herbert573e8fc2014-08-22 13:33:47 -07002450/* GRO checksum functions. These are logical equivalents of the normal
2451 * checksum functions (in skbuff.h) except that they operate on the GRO
2452 * offsets and fields in sk_buff.
2453 */
2454
2455__sum16 __skb_gro_checksum_complete(struct sk_buff *skb);
2456
Tom Herbert15e23962015-02-10 16:30:31 -08002457static inline bool skb_at_gro_remcsum_start(struct sk_buff *skb)
2458{
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002459 return (NAPI_GRO_CB(skb)->gro_remcsum_start == skb_gro_offset(skb));
Tom Herbert15e23962015-02-10 16:30:31 -08002460}
2461
Tom Herbert573e8fc2014-08-22 13:33:47 -07002462static inline bool __skb_gro_checksum_validate_needed(struct sk_buff *skb,
2463 bool zero_okay,
2464 __sum16 check)
2465{
Tom Herbert6edec0e2015-02-10 16:30:28 -08002466 return ((skb->ip_summed != CHECKSUM_PARTIAL ||
2467 skb_checksum_start_offset(skb) <
2468 skb_gro_offset(skb)) &&
Tom Herbert15e23962015-02-10 16:30:31 -08002469 !skb_at_gro_remcsum_start(skb) &&
Tom Herbert662880f2014-08-27 21:26:56 -07002470 NAPI_GRO_CB(skb)->csum_cnt == 0 &&
Tom Herbert573e8fc2014-08-22 13:33:47 -07002471 (!zero_okay || check));
2472}
2473
2474static inline __sum16 __skb_gro_checksum_validate_complete(struct sk_buff *skb,
2475 __wsum psum)
2476{
2477 if (NAPI_GRO_CB(skb)->csum_valid &&
2478 !csum_fold(csum_add(psum, NAPI_GRO_CB(skb)->csum)))
2479 return 0;
2480
2481 NAPI_GRO_CB(skb)->csum = psum;
2482
2483 return __skb_gro_checksum_complete(skb);
2484}
2485
Tom Herbert573e8fc2014-08-22 13:33:47 -07002486static inline void skb_gro_incr_csum_unnecessary(struct sk_buff *skb)
2487{
Tom Herbert662880f2014-08-27 21:26:56 -07002488 if (NAPI_GRO_CB(skb)->csum_cnt > 0) {
2489 /* Consume a checksum from CHECKSUM_UNNECESSARY */
2490 NAPI_GRO_CB(skb)->csum_cnt--;
2491 } else {
2492 /* Update skb for CHECKSUM_UNNECESSARY and csum_level when we
2493 * verified a new top level checksum or an encapsulated one
2494 * during GRO. This saves work if we fallback to normal path.
2495 */
2496 __skb_incr_checksum_unnecessary(skb);
Tom Herbert573e8fc2014-08-22 13:33:47 -07002497 }
2498}
2499
2500#define __skb_gro_checksum_validate(skb, proto, zero_okay, check, \
2501 compute_pseudo) \
2502({ \
2503 __sum16 __ret = 0; \
2504 if (__skb_gro_checksum_validate_needed(skb, zero_okay, check)) \
2505 __ret = __skb_gro_checksum_validate_complete(skb, \
2506 compute_pseudo(skb, proto)); \
Tom Herbert5a212322014-08-31 15:12:41 -07002507 if (__ret) \
2508 __skb_mark_checksum_bad(skb); \
2509 else \
Tom Herbert573e8fc2014-08-22 13:33:47 -07002510 skb_gro_incr_csum_unnecessary(skb); \
2511 __ret; \
2512})
2513
2514#define skb_gro_checksum_validate(skb, proto, compute_pseudo) \
2515 __skb_gro_checksum_validate(skb, proto, false, 0, compute_pseudo)
2516
2517#define skb_gro_checksum_validate_zero_check(skb, proto, check, \
2518 compute_pseudo) \
2519 __skb_gro_checksum_validate(skb, proto, true, check, compute_pseudo)
2520
2521#define skb_gro_checksum_simple_validate(skb) \
2522 __skb_gro_checksum_validate(skb, 0, false, 0, null_compute_pseudo)
2523
Tom Herbertd96535a2014-08-31 15:12:42 -07002524static inline bool __skb_gro_checksum_convert_check(struct sk_buff *skb)
2525{
2526 return (NAPI_GRO_CB(skb)->csum_cnt == 0 &&
2527 !NAPI_GRO_CB(skb)->csum_valid);
2528}
2529
2530static inline void __skb_gro_checksum_convert(struct sk_buff *skb,
2531 __sum16 check, __wsum pseudo)
2532{
2533 NAPI_GRO_CB(skb)->csum = ~pseudo;
2534 NAPI_GRO_CB(skb)->csum_valid = 1;
2535}
2536
2537#define skb_gro_checksum_try_convert(skb, proto, check, compute_pseudo) \
2538do { \
2539 if (__skb_gro_checksum_convert_check(skb)) \
2540 __skb_gro_checksum_convert(skb, check, \
2541 compute_pseudo(skb, proto)); \
2542} while (0)
2543
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002544struct gro_remcsum {
2545 int offset;
2546 __wsum delta;
2547};
2548
2549static inline void skb_gro_remcsum_init(struct gro_remcsum *grc)
2550{
Geert Uytterhoeven846cd662015-02-18 11:38:06 +01002551 grc->offset = 0;
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002552 grc->delta = 0;
2553}
2554
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002555static inline void *skb_gro_remcsum_process(struct sk_buff *skb, void *ptr,
2556 unsigned int off, size_t hdrlen,
2557 int start, int offset,
2558 struct gro_remcsum *grc,
2559 bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08002560{
2561 __wsum delta;
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002562 size_t plen = hdrlen + max_t(size_t, offset + sizeof(u16), start);
Tom Herbertdcdc8992015-02-02 16:07:34 -08002563
2564 BUG_ON(!NAPI_GRO_CB(skb)->csum_valid);
2565
Tom Herbert15e23962015-02-10 16:30:31 -08002566 if (!nopartial) {
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002567 NAPI_GRO_CB(skb)->gro_remcsum_start = off + hdrlen + start;
2568 return ptr;
Tom Herbert15e23962015-02-10 16:30:31 -08002569 }
2570
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002571 ptr = skb_gro_header_fast(skb, off);
2572 if (skb_gro_header_hard(skb, off + plen)) {
2573 ptr = skb_gro_header_slow(skb, off + plen, off);
2574 if (!ptr)
2575 return NULL;
2576 }
2577
2578 delta = remcsum_adjust(ptr + hdrlen, NAPI_GRO_CB(skb)->csum,
2579 start, offset);
Tom Herbertdcdc8992015-02-02 16:07:34 -08002580
2581 /* Adjust skb->csum since we changed the packet */
Tom Herbertdcdc8992015-02-02 16:07:34 -08002582 NAPI_GRO_CB(skb)->csum = csum_add(NAPI_GRO_CB(skb)->csum, delta);
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002583
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002584 grc->offset = off + hdrlen + offset;
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002585 grc->delta = delta;
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002586
2587 return ptr;
Tom Herbertdcdc8992015-02-02 16:07:34 -08002588}
2589
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002590static inline void skb_gro_remcsum_cleanup(struct sk_buff *skb,
2591 struct gro_remcsum *grc)
2592{
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002593 void *ptr;
2594 size_t plen = grc->offset + sizeof(u16);
2595
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002596 if (!grc->delta)
2597 return;
2598
Tom Herbertb7fe10e2015-08-19 17:07:32 -07002599 ptr = skb_gro_header_fast(skb, grc->offset);
2600 if (skb_gro_header_hard(skb, grc->offset + sizeof(u16))) {
2601 ptr = skb_gro_header_slow(skb, plen, grc->offset);
2602 if (!ptr)
2603 return;
2604 }
2605
2606 remcsum_unadjust((__sum16 *)ptr, grc->delta);
Tom Herbert26c4f7d2015-02-10 16:30:27 -08002607}
Tom Herbertdcdc8992015-02-02 16:07:34 -08002608
Tom Herbert6ae23ad2015-12-14 11:19:46 -08002609struct skb_csum_offl_spec {
2610 __u16 ipv4_okay:1,
2611 ipv6_okay:1,
2612 encap_okay:1,
2613 ip_options_okay:1,
2614 ext_hdrs_okay:1,
2615 tcp_okay:1,
2616 udp_okay:1,
2617 sctp_okay:1,
2618 vlan_okay:1,
2619 no_encapped_ipv6:1,
2620 no_not_encapped:1;
2621};
2622
2623bool __skb_csum_offload_chk(struct sk_buff *skb,
2624 const struct skb_csum_offl_spec *spec,
2625 bool *csum_encapped,
2626 bool csum_help);
2627
2628static inline bool skb_csum_offload_chk(struct sk_buff *skb,
2629 const struct skb_csum_offl_spec *spec,
2630 bool *csum_encapped,
2631 bool csum_help)
2632{
2633 if (skb->ip_summed != CHECKSUM_PARTIAL)
2634 return false;
2635
2636 return __skb_csum_offload_chk(skb, spec, csum_encapped, csum_help);
2637}
2638
2639static inline bool skb_csum_offload_chk_help(struct sk_buff *skb,
2640 const struct skb_csum_offl_spec *spec)
2641{
2642 bool csum_encapped;
2643
2644 return skb_csum_offload_chk(skb, spec, &csum_encapped, true);
2645}
2646
2647static inline bool skb_csum_off_chk_help_cmn(struct sk_buff *skb)
2648{
2649 static const struct skb_csum_offl_spec csum_offl_spec = {
2650 .ipv4_okay = 1,
2651 .ip_options_okay = 1,
2652 .ipv6_okay = 1,
2653 .vlan_okay = 1,
2654 .tcp_okay = 1,
2655 .udp_okay = 1,
2656 };
2657
2658 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2659}
2660
2661static inline bool skb_csum_off_chk_help_cmn_v4_only(struct sk_buff *skb)
2662{
2663 static const struct skb_csum_offl_spec csum_offl_spec = {
2664 .ipv4_okay = 1,
2665 .ip_options_okay = 1,
2666 .tcp_okay = 1,
2667 .udp_okay = 1,
2668 .vlan_okay = 1,
2669 };
2670
2671 return skb_csum_offload_chk_help(skb, &csum_offl_spec);
2672}
2673
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002674static inline int dev_hard_header(struct sk_buff *skb, struct net_device *dev,
2675 unsigned short type,
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002676 const void *daddr, const void *saddr,
Eric Dumazet95c96172012-04-15 05:58:06 +00002677 unsigned int len)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002678{
Ursula Braunf1ecfd52007-10-22 16:16:14 +02002679 if (!dev->header_ops || !dev->header_ops->create)
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002680 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002681
2682 return dev->header_ops->create(skb, dev, type, daddr, saddr, len);
Stephen Hemminger0c4e8582007-10-09 01:36:32 -07002683}
2684
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002685static inline int dev_parse_header(const struct sk_buff *skb,
2686 unsigned char *haddr)
2687{
2688 const struct net_device *dev = skb->dev;
2689
Patrick McHardy1b833362007-10-18 05:09:28 -07002690 if (!dev->header_ops || !dev->header_ops->parse)
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002691 return 0;
Stephen Hemminger3b04ddd2007-10-09 01:40:57 -07002692 return dev->header_ops->parse(skb, haddr);
Stephen Hemmingerb95cce32007-09-26 22:13:38 -07002693}
2694
Willem de Bruijn2793a232016-03-09 21:58:32 -05002695/* ll_header must have at least hard_header_len allocated */
2696static inline bool dev_validate_header(const struct net_device *dev,
2697 char *ll_header, int len)
2698{
2699 if (likely(len >= dev->hard_header_len))
2700 return true;
2701
2702 if (capable(CAP_SYS_RAWIO)) {
2703 memset(ll_header + len, 0, dev->hard_header_len - len);
2704 return true;
2705 }
2706
2707 if (dev->header_ops && dev->header_ops->validate)
2708 return dev->header_ops->validate(ll_header, len);
2709
2710 return false;
2711}
2712
Linus Torvalds1da177e2005-04-16 15:20:36 -07002713typedef int gifconf_func_t(struct net_device * dev, char __user * bufptr, int len);
Joe Perchesf629d202013-09-26 14:48:15 -07002714int register_gifconf(unsigned int family, gifconf_func_t *gifconf);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002715static inline int unregister_gifconf(unsigned int family)
2716{
2717 return register_gifconf(family, NULL);
2718}
2719
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002720#ifdef CONFIG_NET_FLOW_LIMIT
Willem de Bruijn5f121b92013-06-13 15:29:38 -04002721#define FLOW_LIMIT_HISTORY (1 << 7) /* must be ^2 and !overflow buckets */
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002722struct sd_flow_limit {
2723 u64 count;
2724 unsigned int num_buckets;
2725 unsigned int history_head;
2726 u16 history[FLOW_LIMIT_HISTORY];
2727 u8 buckets[];
2728};
2729
2730extern int netdev_flow_limit_table_len;
2731#endif /* CONFIG_NET_FLOW_LIMIT */
2732
Linus Torvalds1da177e2005-04-16 15:20:36 -07002733/*
Eric Dumazet88751272010-04-19 05:07:33 +00002734 * Incoming packets are placed on per-cpu queues
Linus Torvalds1da177e2005-04-16 15:20:36 -07002735 */
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08002736struct softnet_data {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002737 struct list_head poll_list;
Changli Gao6e7676c2010-04-27 15:07:33 -07002738 struct sk_buff_head process_queue;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002739
Changli Gaodee42872010-05-02 05:42:16 +00002740 /* stats */
David S. Millercd7b5392010-05-02 22:27:59 -07002741 unsigned int processed;
2742 unsigned int time_squeeze;
2743 unsigned int cpu_collision;
2744 unsigned int received_rps;
Changli Gaofd793d82010-04-15 00:16:59 -07002745#ifdef CONFIG_RPS
Eric Dumazet88751272010-04-19 05:07:33 +00002746 struct softnet_data *rps_ipi_list;
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002747#endif
2748#ifdef CONFIG_NET_FLOW_LIMIT
2749 struct sd_flow_limit __rcu *flow_limit;
2750#endif
2751 struct Qdisc *output_queue;
2752 struct Qdisc **output_queue_tailp;
2753 struct sk_buff *completion_queue;
Eric Dumazet88751272010-04-19 05:07:33 +00002754
Eric Dumazet4cdb1e22014-11-02 06:00:12 -08002755#ifdef CONFIG_RPS
Eric Dumazet88751272010-04-19 05:07:33 +00002756 /* Elements below can be accessed between CPUs for RPS */
Tom Herbert0a9627f2010-03-16 08:03:29 +00002757 struct call_single_data csd ____cacheline_aligned_in_smp;
Eric Dumazet88751272010-04-19 05:07:33 +00002758 struct softnet_data *rps_ipi_next;
2759 unsigned int cpu;
Tom Herbertfec5e652010-04-16 16:01:27 -07002760 unsigned int input_queue_head;
Tom Herbert76cc8b12010-05-20 18:37:59 +00002761 unsigned int input_queue_tail;
Tom Herbert1e94d722010-03-18 17:45:44 -07002762#endif
Eric Dumazet95c96172012-04-15 05:58:06 +00002763 unsigned int dropped;
Tom Herbert0a9627f2010-03-16 08:03:29 +00002764 struct sk_buff_head input_pkt_queue;
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002765 struct napi_struct backlog;
Willem de Bruijn99bbc702013-05-20 04:02:32 +00002766
Linus Torvalds1da177e2005-04-16 15:20:36 -07002767};
2768
Tom Herbert76cc8b12010-05-20 18:37:59 +00002769static inline void input_queue_head_incr(struct softnet_data *sd)
Tom Herbertfec5e652010-04-16 16:01:27 -07002770{
2771#ifdef CONFIG_RPS
Tom Herbert76cc8b12010-05-20 18:37:59 +00002772 sd->input_queue_head++;
2773#endif
2774}
2775
2776static inline void input_queue_tail_incr_save(struct softnet_data *sd,
2777 unsigned int *qtail)
2778{
2779#ifdef CONFIG_RPS
2780 *qtail = ++sd->input_queue_tail;
Tom Herbertfec5e652010-04-16 16:01:27 -07002781#endif
2782}
2783
Tom Herbert0a9627f2010-03-16 08:03:29 +00002784DECLARE_PER_CPU_ALIGNED(struct softnet_data, softnet_data);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002785
Joe Perchesf629d202013-09-26 14:48:15 -07002786void __netif_schedule(struct Qdisc *q);
John Fastabend46e5da40a2014-09-12 20:04:52 -07002787void netif_schedule_queue(struct netdev_queue *txq);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002788
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002789static inline void netif_tx_schedule_all(struct net_device *dev)
2790{
2791 unsigned int i;
2792
2793 for (i = 0; i < dev->num_tx_queues; i++)
2794 netif_schedule_queue(netdev_get_tx_queue(dev, i));
2795}
2796
Dave Jonesd29f7492008-07-22 14:09:06 -07002797static inline void netif_tx_start_queue(struct netdev_queue *dev_queue)
2798{
Tom Herbert734664982011-11-28 16:32:44 +00002799 clear_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002800}
2801
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002802/**
2803 * netif_start_queue - allow transmit
2804 * @dev: network device
2805 *
2806 * Allow upper layers to call the device hard_start_xmit routine.
2807 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002808static inline void netif_start_queue(struct net_device *dev)
2809{
David S. Millere8a04642008-07-17 00:34:19 -07002810 netif_tx_start_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002811}
2812
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002813static inline void netif_tx_start_all_queues(struct net_device *dev)
2814{
2815 unsigned int i;
2816
2817 for (i = 0; i < dev->num_tx_queues; i++) {
2818 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2819 netif_tx_start_queue(txq);
2820 }
2821}
2822
John Fastabend46e5da40a2014-09-12 20:04:52 -07002823void netif_tx_wake_queue(struct netdev_queue *dev_queue);
David S. Miller79d16382008-07-08 23:14:46 -07002824
Dave Jonesd29f7492008-07-22 14:09:06 -07002825/**
2826 * netif_wake_queue - restart transmit
2827 * @dev: network device
2828 *
2829 * Allow upper layers to call the device hard_start_xmit routine.
2830 * Used for flow control when transmit resources are available.
2831 */
David S. Miller79d16382008-07-08 23:14:46 -07002832static inline void netif_wake_queue(struct net_device *dev)
2833{
David S. Millere8a04642008-07-17 00:34:19 -07002834 netif_tx_wake_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002835}
2836
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002837static inline void netif_tx_wake_all_queues(struct net_device *dev)
2838{
2839 unsigned int i;
2840
2841 for (i = 0; i < dev->num_tx_queues; i++) {
2842 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
2843 netif_tx_wake_queue(txq);
2844 }
2845}
2846
Dave Jonesd29f7492008-07-22 14:09:06 -07002847static inline void netif_tx_stop_queue(struct netdev_queue *dev_queue)
2848{
Tom Herbert734664982011-11-28 16:32:44 +00002849 set_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002850}
2851
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002852/**
2853 * netif_stop_queue - stop transmitted packets
2854 * @dev: network device
2855 *
2856 * Stop upper layers calling the device hard_start_xmit routine.
2857 * Used for flow control when transmit resources are unavailable.
2858 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002859static inline void netif_stop_queue(struct net_device *dev)
2860{
David S. Millere8a04642008-07-17 00:34:19 -07002861 netif_tx_stop_queue(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002862}
2863
Denys Vlasenkoa2029242015-05-11 21:17:53 +02002864void netif_tx_stop_all_queues(struct net_device *dev);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07002865
David S. Miller4d295152012-03-07 21:02:35 -05002866static inline bool netif_tx_queue_stopped(const struct netdev_queue *dev_queue)
Dave Jonesd29f7492008-07-22 14:09:06 -07002867{
Tom Herbert734664982011-11-28 16:32:44 +00002868 return test_bit(__QUEUE_STATE_DRV_XOFF, &dev_queue->state);
Dave Jonesd29f7492008-07-22 14:09:06 -07002869}
2870
Stephen Hemmingerbea33482007-10-03 16:41:36 -07002871/**
2872 * netif_queue_stopped - test if transmit queue is flowblocked
2873 * @dev: network device
2874 *
2875 * Test if transmit queue on device is currently unable to send.
2876 */
David S. Miller4d295152012-03-07 21:02:35 -05002877static inline bool netif_queue_stopped(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002878{
David S. Millere8a04642008-07-17 00:34:19 -07002879 return netif_tx_queue_stopped(netdev_get_tx_queue(dev, 0));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002880}
2881
David S. Miller4d295152012-03-07 21:02:35 -05002882static inline bool netif_xmit_stopped(const struct netdev_queue *dev_queue)
David S. Millerc3f26a22008-07-31 16:58:50 -07002883{
Tom Herbert734664982011-11-28 16:32:44 +00002884 return dev_queue->state & QUEUE_STATE_ANY_XOFF;
2885}
2886
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002887static inline bool
2888netif_xmit_frozen_or_stopped(const struct netdev_queue *dev_queue)
Tom Herbert734664982011-11-28 16:32:44 +00002889{
2890 return dev_queue->state & QUEUE_STATE_ANY_XOFF_OR_FROZEN;
2891}
2892
Daniel Borkmann8e2f1a62014-04-02 20:52:57 +02002893static inline bool
2894netif_xmit_frozen_or_drv_stopped(const struct netdev_queue *dev_queue)
2895{
2896 return dev_queue->state & QUEUE_STATE_DRV_XOFF_OR_FROZEN;
2897}
2898
Eric Dumazet53511452014-10-08 08:19:27 -07002899/**
2900 * netdev_txq_bql_enqueue_prefetchw - prefetch bql data for write
2901 * @dev_queue: pointer to transmit queue
2902 *
2903 * BQL enabled drivers might use this helper in their ndo_start_xmit(),
2904 * to give appropriate hint to the cpu.
2905 */
2906static inline void netdev_txq_bql_enqueue_prefetchw(struct netdev_queue *dev_queue)
2907{
2908#ifdef CONFIG_BQL
2909 prefetchw(&dev_queue->dql.num_queued);
2910#endif
2911}
2912
2913/**
2914 * netdev_txq_bql_complete_prefetchw - prefetch bql data for write
2915 * @dev_queue: pointer to transmit queue
2916 *
2917 * BQL enabled drivers might use this helper in their TX completion path,
2918 * to give appropriate hint to the cpu.
2919 */
2920static inline void netdev_txq_bql_complete_prefetchw(struct netdev_queue *dev_queue)
2921{
2922#ifdef CONFIG_BQL
2923 prefetchw(&dev_queue->dql.limit);
2924#endif
2925}
2926
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002927static inline void netdev_tx_sent_queue(struct netdev_queue *dev_queue,
2928 unsigned int bytes)
2929{
Tom Herbert114cf582011-11-28 16:33:09 +00002930#ifdef CONFIG_BQL
2931 dql_queued(&dev_queue->dql, bytes);
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00002932
2933 if (likely(dql_avail(&dev_queue->dql) >= 0))
2934 return;
2935
2936 set_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
2937
2938 /*
2939 * The XOFF flag must be set before checking the dql_avail below,
2940 * because in netdev_tx_completed_queue we update the dql_completed
2941 * before checking the XOFF flag.
2942 */
2943 smp_mb();
2944
2945 /* check again in case another CPU has just made room avail */
2946 if (unlikely(dql_avail(&dev_queue->dql) >= 0))
2947 clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state);
Tom Herbert114cf582011-11-28 16:33:09 +00002948#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002949}
2950
Florian Fainelli0042d0c2013-09-06 16:58:00 +01002951/**
2952 * netdev_sent_queue - report the number of bytes queued to hardware
2953 * @dev: network device
2954 * @bytes: number of bytes queued to the hardware device queue
2955 *
2956 * Report the number of bytes queued for sending/completion to the network
2957 * device hardware queue. @bytes should be a good approximation and should
2958 * exactly match netdev_completed_queue() @bytes
2959 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002960static inline void netdev_sent_queue(struct net_device *dev, unsigned int bytes)
2961{
2962 netdev_tx_sent_queue(netdev_get_tx_queue(dev, 0), bytes);
2963}
2964
2965static inline void netdev_tx_completed_queue(struct netdev_queue *dev_queue,
Eric Dumazet95c96172012-04-15 05:58:06 +00002966 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002967{
Tom Herbert114cf582011-11-28 16:33:09 +00002968#ifdef CONFIG_BQL
Alexander Duyckb37c0fb2012-02-07 02:29:06 +00002969 if (unlikely(!bytes))
2970 return;
2971
2972 dql_completed(&dev_queue->dql, bytes);
2973
2974 /*
2975 * Without the memory barrier there is a small possiblity that
2976 * netdev_tx_sent_queue will miss the update and cause the queue to
2977 * be stopped forever
2978 */
2979 smp_mb();
2980
2981 if (dql_avail(&dev_queue->dql) < 0)
2982 return;
2983
2984 if (test_and_clear_bit(__QUEUE_STATE_STACK_XOFF, &dev_queue->state))
2985 netif_schedule_queue(dev_queue);
Tom Herbert114cf582011-11-28 16:33:09 +00002986#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002987}
2988
Florian Fainelli0042d0c2013-09-06 16:58:00 +01002989/**
2990 * netdev_completed_queue - report bytes and packets completed by device
2991 * @dev: network device
2992 * @pkts: actual number of packets sent over the medium
2993 * @bytes: actual number of bytes sent over the medium
2994 *
2995 * Report the number of bytes and packets transmitted by the network device
2996 * hardware queue over the physical medium, @bytes must exactly match the
2997 * @bytes amount passed to netdev_sent_queue()
2998 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00002999static inline void netdev_completed_queue(struct net_device *dev,
Eric Dumazet95c96172012-04-15 05:58:06 +00003000 unsigned int pkts, unsigned int bytes)
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003001{
3002 netdev_tx_completed_queue(netdev_get_tx_queue(dev, 0), pkts, bytes);
3003}
3004
3005static inline void netdev_tx_reset_queue(struct netdev_queue *q)
3006{
Tom Herbert114cf582011-11-28 16:33:09 +00003007#ifdef CONFIG_BQL
Alexander Duyck5c490352012-02-07 02:29:01 +00003008 clear_bit(__QUEUE_STATE_STACK_XOFF, &q->state);
Tom Herbert114cf582011-11-28 16:33:09 +00003009 dql_reset(&q->dql);
3010#endif
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003011}
3012
Florian Fainelli0042d0c2013-09-06 16:58:00 +01003013/**
3014 * netdev_reset_queue - reset the packets and bytes count of a network device
3015 * @dev_queue: network device
3016 *
3017 * Reset the bytes and packet count of a network device and clear the
3018 * software flow control OFF bit for this network device
3019 */
Tom Herbertc5d67bd2011-11-28 16:32:52 +00003020static inline void netdev_reset_queue(struct net_device *dev_queue)
3021{
3022 netdev_tx_reset_queue(netdev_get_tx_queue(dev_queue, 0));
David S. Millerc3f26a22008-07-31 16:58:50 -07003023}
3024
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003025/**
Daniel Borkmannb9507bd2014-02-16 15:55:21 +01003026 * netdev_cap_txqueue - check if selected tx queue exceeds device queues
3027 * @dev: network device
3028 * @queue_index: given tx queue index
3029 *
3030 * Returns 0 if given tx queue index >= number of device tx queues,
3031 * otherwise returns the originally passed tx queue index.
3032 */
3033static inline u16 netdev_cap_txqueue(struct net_device *dev, u16 queue_index)
3034{
3035 if (unlikely(queue_index >= dev->real_num_tx_queues)) {
3036 net_warn_ratelimited("%s selects TX queue %d, but real number of TX queues is %d\n",
3037 dev->name, queue_index,
3038 dev->real_num_tx_queues);
3039 return 0;
3040 }
3041
3042 return queue_index;
3043}
3044
3045/**
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003046 * netif_running - test if up
3047 * @dev: network device
3048 *
3049 * Test if the device has been brought up.
3050 */
David S. Miller4d295152012-03-07 21:02:35 -05003051static inline bool netif_running(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003052{
3053 return test_bit(__LINK_STATE_START, &dev->state);
3054}
3055
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003056/*
3057 * Routines to manage the subqueues on a device. We only need start
3058 * stop, and a check if it's stopped. All other device management is
3059 * done at the overall netdevice level.
3060 * Also test the device if we're multiqueue.
3061 */
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003062
3063/**
3064 * netif_start_subqueue - allow sending packets on subqueue
3065 * @dev: network device
3066 * @queue_index: sub queue index
3067 *
3068 * Start individual transmit queue of a device with multiple transmit queues.
3069 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003070static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
3071{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003072 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003073
3074 netif_tx_start_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003075}
3076
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003077/**
3078 * netif_stop_subqueue - stop sending packets on subqueue
3079 * @dev: network device
3080 * @queue_index: sub queue index
3081 *
3082 * Stop individual transmit queue of a device with multiple transmit queues.
3083 */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003084static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
3085{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003086 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003087 netif_tx_stop_queue(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003088}
3089
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003090/**
3091 * netif_subqueue_stopped - test status of subqueue
3092 * @dev: network device
3093 * @queue_index: sub queue index
3094 *
3095 * Check individual transmit queue of a device with multiple transmit queues.
3096 */
David S. Miller4d295152012-03-07 21:02:35 -05003097static inline bool __netif_subqueue_stopped(const struct net_device *dev,
3098 u16 queue_index)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003099{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003100 struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003101
3102 return netif_tx_queue_stopped(txq);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003103}
3104
David S. Miller4d295152012-03-07 21:02:35 -05003105static inline bool netif_subqueue_stopped(const struct net_device *dev,
3106 struct sk_buff *skb)
Pavel Emelyanov668f8952007-10-21 17:01:56 -07003107{
3108 return __netif_subqueue_stopped(dev, skb_get_queue_mapping(skb));
3109}
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003110
John Fastabend46e5da40a2014-09-12 20:04:52 -07003111void netif_wake_subqueue(struct net_device *dev, u16 queue_index);
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003112
Alexander Duyck537c00d2013-01-10 08:57:02 +00003113#ifdef CONFIG_XPS
David S. Miller53af53a2013-10-08 23:07:53 -04003114int netif_set_xps_queue(struct net_device *dev, const struct cpumask *mask,
Joe Perchesf629d202013-09-26 14:48:15 -07003115 u16 index);
Alexander Duyck537c00d2013-01-10 08:57:02 +00003116#else
3117static inline int netif_set_xps_queue(struct net_device *dev,
Michael S. Tsirkin35735402013-10-02 09:14:06 +03003118 const struct cpumask *mask,
Alexander Duyck537c00d2013-01-10 08:57:02 +00003119 u16 index)
3120{
3121 return 0;
3122}
3123#endif
3124
Jiri Pirko5605c762015-05-12 14:56:12 +02003125u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
3126 unsigned int num_tx_queues);
3127
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00003128/*
3129 * Returns a Tx hash for the given packet when dev->real_num_tx_queues is used
3130 * as a distribution range limit for the returned value.
3131 */
3132static inline u16 skb_tx_hash(const struct net_device *dev,
Tom Herbert0e001612014-07-01 21:32:27 -07003133 struct sk_buff *skb)
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00003134{
3135 return __skb_tx_hash(dev, skb, dev->real_num_tx_queues);
3136}
3137
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003138/**
3139 * netif_is_multiqueue - test if device has multiple transmit queues
3140 * @dev: network device
3141 *
3142 * Check if device has multiple transmit queues
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003143 */
David S. Miller4d295152012-03-07 21:02:35 -05003144static inline bool netif_is_multiqueue(const struct net_device *dev)
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003145{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003146 return dev->num_tx_queues > 1;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003147}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003148
Joe Perchesf629d202013-09-26 14:48:15 -07003149int netif_set_real_num_tx_queues(struct net_device *dev, unsigned int txq);
John Fastabendf0796d52010-07-01 13:21:57 +00003150
Michael Daltona953be52014-01-16 22:23:28 -08003151#ifdef CONFIG_SYSFS
Joe Perchesf629d202013-09-26 14:48:15 -07003152int netif_set_real_num_rx_queues(struct net_device *dev, unsigned int rxq);
Ben Hutchings62fe0b42010-09-27 08:24:33 +00003153#else
3154static inline int netif_set_real_num_rx_queues(struct net_device *dev,
3155 unsigned int rxq)
3156{
3157 return 0;
3158}
3159#endif
3160
Michael Daltona953be52014-01-16 22:23:28 -08003161#ifdef CONFIG_SYSFS
3162static inline unsigned int get_netdev_rx_queue_index(
3163 struct netdev_rx_queue *queue)
3164{
3165 struct net_device *dev = queue->dev;
3166 int index = queue - dev->_rx;
3167
3168 BUG_ON(index >= dev->num_rx_queues);
3169 return index;
3170}
3171#endif
3172
Yuval Mintz16917b82012-07-01 03:18:50 +00003173#define DEFAULT_MAX_NUM_RSS_QUEUES (8)
Joe Perchesf629d202013-09-26 14:48:15 -07003174int netif_get_num_default_rss_queues(void);
Yuval Mintz16917b82012-07-01 03:18:50 +00003175
Eric Dumazete6247022013-12-05 04:45:08 -08003176enum skb_free_reason {
3177 SKB_REASON_CONSUMED,
3178 SKB_REASON_DROPPED,
3179};
Linus Torvalds1da177e2005-04-16 15:20:36 -07003180
Eric Dumazete6247022013-12-05 04:45:08 -08003181void __dev_kfree_skb_irq(struct sk_buff *skb, enum skb_free_reason reason);
3182void __dev_kfree_skb_any(struct sk_buff *skb, enum skb_free_reason reason);
3183
3184/*
3185 * It is not allowed to call kfree_skb() or consume_skb() from hardware
3186 * interrupt context or with hardware interrupts being disabled.
3187 * (in_irq() || irqs_disabled())
3188 *
3189 * We provide four helpers that can be used in following contexts :
3190 *
3191 * dev_kfree_skb_irq(skb) when caller drops a packet from irq context,
3192 * replacing kfree_skb(skb)
3193 *
3194 * dev_consume_skb_irq(skb) when caller consumes a packet from irq context.
3195 * Typically used in place of consume_skb(skb) in TX completion path
3196 *
3197 * dev_kfree_skb_any(skb) when caller doesn't know its current irq context,
3198 * replacing kfree_skb(skb)
3199 *
3200 * dev_consume_skb_any(skb) when caller doesn't know its current irq context,
3201 * and consumed a packet. Used in place of consume_skb(skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003202 */
Eric Dumazete6247022013-12-05 04:45:08 -08003203static inline void dev_kfree_skb_irq(struct sk_buff *skb)
3204{
3205 __dev_kfree_skb_irq(skb, SKB_REASON_DROPPED);
3206}
3207
3208static inline void dev_consume_skb_irq(struct sk_buff *skb)
3209{
3210 __dev_kfree_skb_irq(skb, SKB_REASON_CONSUMED);
3211}
3212
3213static inline void dev_kfree_skb_any(struct sk_buff *skb)
3214{
3215 __dev_kfree_skb_any(skb, SKB_REASON_DROPPED);
3216}
3217
3218static inline void dev_consume_skb_any(struct sk_buff *skb)
3219{
3220 __dev_kfree_skb_any(skb, SKB_REASON_CONSUMED);
3221}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003222
Joe Perchesf629d202013-09-26 14:48:15 -07003223int netif_rx(struct sk_buff *skb);
3224int netif_rx_ni(struct sk_buff *skb);
Eric W. Biederman04eb4482015-09-15 20:04:15 -05003225int netif_receive_skb(struct sk_buff *skb);
Joe Perchesf629d202013-09-26 14:48:15 -07003226gro_result_t napi_gro_receive(struct napi_struct *napi, struct sk_buff *skb);
3227void napi_gro_flush(struct napi_struct *napi, bool flush_old);
3228struct sk_buff *napi_get_frags(struct napi_struct *napi);
3229gro_result_t napi_gro_frags(struct napi_struct *napi);
Jerry Chubf5a7552014-01-07 10:23:19 -08003230struct packet_offload *gro_find_receive_by_type(__be16 type);
3231struct packet_offload *gro_find_complete_by_type(__be16 type);
Herbert Xu76620aa2009-04-16 02:02:07 -07003232
3233static inline void napi_free_frags(struct napi_struct *napi)
3234{
3235 kfree_skb(napi->skb);
3236 napi->skb = NULL;
3237}
3238
Joe Perchesf629d202013-09-26 14:48:15 -07003239int netdev_rx_handler_register(struct net_device *dev,
3240 rx_handler_func_t *rx_handler,
3241 void *rx_handler_data);
3242void netdev_rx_handler_unregister(struct net_device *dev);
Jiri Pirkoab95bfe2010-06-01 21:52:08 +00003243
Joe Perchesf629d202013-09-26 14:48:15 -07003244bool dev_valid_name(const char *name);
3245int dev_ioctl(struct net *net, unsigned int cmd, void __user *);
3246int dev_ethtool(struct net *net, struct ifreq *);
3247unsigned int dev_get_flags(const struct net_device *);
3248int __dev_change_flags(struct net_device *, unsigned int flags);
3249int dev_change_flags(struct net_device *, unsigned int);
David S. Millercb178192013-09-30 15:36:45 -04003250void __dev_notify_flags(struct net_device *, unsigned int old_flags,
3251 unsigned int gchanges);
Joe Perchesf629d202013-09-26 14:48:15 -07003252int dev_change_name(struct net_device *, const char *);
3253int dev_set_alias(struct net_device *, const char *, size_t);
3254int dev_change_net_namespace(struct net_device *, struct net *, const char *);
3255int dev_set_mtu(struct net_device *, int);
3256void dev_set_group(struct net_device *, int);
3257int dev_set_mac_address(struct net_device *, struct sockaddr *);
3258int dev_change_carrier(struct net_device *, bool new_carrier);
3259int dev_get_phys_port_id(struct net_device *dev,
Jiri Pirko02637fc2014-11-28 14:34:16 +01003260 struct netdev_phys_item_id *ppid);
David Aherndb24a902015-03-17 20:23:15 -06003261int dev_get_phys_port_name(struct net_device *dev,
3262 char *name, size_t len);
Anuradha Karuppiahd746d702015-07-14 13:43:19 -07003263int dev_change_proto_down(struct net_device *dev, bool proto_down);
Eric Dumazet55a93b32014-10-03 15:31:07 -07003264struct sk_buff *validate_xmit_skb_list(struct sk_buff *skb, struct net_device *dev);
David S. Millerce937182014-08-30 19:22:20 -07003265struct sk_buff *dev_hard_start_xmit(struct sk_buff *skb, struct net_device *dev,
3266 struct netdev_queue *txq, int *ret);
Herbert Xua0265d22014-04-17 13:45:03 +08003267int __dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Joe Perchesf629d202013-09-26 14:48:15 -07003268int dev_forward_skb(struct net_device *dev, struct sk_buff *skb);
Vlad Yasevich1ee481f2014-03-27 17:32:29 -04003269bool is_skb_forwardable(struct net_device *dev, struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003270
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003271extern int netdev_budget;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003272
3273/* Called by rtnetlink.c:rtnl_unlock() */
Joe Perchesf629d202013-09-26 14:48:15 -07003274void netdev_run_todo(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003275
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003276/**
3277 * dev_put - release reference to device
3278 * @dev: network device
3279 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07003280 * Release reference to device to allow it to be freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003281 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003282static inline void dev_put(struct net_device *dev)
3283{
Christoph Lameter933393f2011-12-22 11:58:51 -06003284 this_cpu_dec(*dev->pcpu_refcnt);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003285}
3286
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003287/**
3288 * dev_hold - get reference to device
3289 * @dev: network device
3290 *
Benjamin Thery9ef44292007-10-10 21:18:17 -07003291 * Hold reference to device to keep it from being freed.
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003292 */
Stephen Hemminger15333062006-03-20 22:32:28 -08003293static inline void dev_hold(struct net_device *dev)
3294{
Christoph Lameter933393f2011-12-22 11:58:51 -06003295 this_cpu_inc(*dev->pcpu_refcnt);
Stephen Hemminger15333062006-03-20 22:32:28 -08003296}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003297
3298/* Carrier loss detection, dial on demand. The functions netif_carrier_on
3299 * and _off may be called from IRQ context, but it is caller
3300 * who is responsible for serialization of these calls.
Stefan Rompfb00055a2006-03-20 17:09:11 -08003301 *
3302 * The name carrier is inappropriate, these functions should really be
3303 * called netif_lowerlayer_*() because they represent the state of any
3304 * kind of lower layer not just hardware media.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003305 */
3306
Joe Perchesf629d202013-09-26 14:48:15 -07003307void linkwatch_init_dev(struct net_device *dev);
3308void linkwatch_fire_event(struct net_device *dev);
3309void linkwatch_forget_dev(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003310
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003311/**
3312 * netif_carrier_ok - test if carrier present
3313 * @dev: network device
3314 *
3315 * Check if carrier is present on device
3316 */
David S. Miller4d295152012-03-07 21:02:35 -05003317static inline bool netif_carrier_ok(const struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003318{
3319 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
3320}
3321
Joe Perchesf629d202013-09-26 14:48:15 -07003322unsigned long dev_trans_start(struct net_device *dev);
Eric Dumazet9d214932009-05-17 20:55:16 -07003323
Joe Perchesf629d202013-09-26 14:48:15 -07003324void __netdev_watchdog_up(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003325
Joe Perchesf629d202013-09-26 14:48:15 -07003326void netif_carrier_on(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003327
Joe Perchesf629d202013-09-26 14:48:15 -07003328void netif_carrier_off(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003329
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003330/**
3331 * netif_dormant_on - mark device as dormant.
3332 * @dev: network device
3333 *
3334 * Mark device as dormant (as per RFC2863).
3335 *
3336 * The dormant state indicates that the relevant interface is not
3337 * actually in a condition to pass packets (i.e., it is not 'up') but is
3338 * in a "pending" state, waiting for some external event. For "on-
3339 * demand" interfaces, this new state identifies the situation where the
3340 * interface is waiting for events to place it in the up state.
3341 *
3342 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003343static inline void netif_dormant_on(struct net_device *dev)
3344{
3345 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
3346 linkwatch_fire_event(dev);
3347}
3348
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003349/**
3350 * netif_dormant_off - set device as not dormant.
3351 * @dev: network device
3352 *
3353 * Device is not in dormant state.
3354 */
Stefan Rompfb00055a2006-03-20 17:09:11 -08003355static inline void netif_dormant_off(struct net_device *dev)
3356{
3357 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
3358 linkwatch_fire_event(dev);
3359}
3360
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003361/**
3362 * netif_dormant - 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_dormant(const struct net_device *dev)
Stefan Rompfb00055a2006-03-20 17:09:11 -08003368{
3369 return test_bit(__LINK_STATE_DORMANT, &dev->state);
3370}
3371
3372
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003373/**
3374 * netif_oper_up - test if device is operational
3375 * @dev: network device
3376 *
3377 * Check if carrier is operational
3378 */
David S. Miller4d295152012-03-07 21:02:35 -05003379static inline bool netif_oper_up(const struct net_device *dev)
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08003380{
Stefan Rompfb00055a2006-03-20 17:09:11 -08003381 return (dev->operstate == IF_OPER_UP ||
3382 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
3383}
3384
Stephen Hemmingerbea33482007-10-03 16:41:36 -07003385/**
3386 * netif_device_present - is device available or removed
3387 * @dev: network device
3388 *
3389 * Check if device has not been removed from system.
3390 */
David S. Miller4d295152012-03-07 21:02:35 -05003391static inline bool netif_device_present(struct net_device *dev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003392{
3393 return test_bit(__LINK_STATE_PRESENT, &dev->state);
3394}
3395
Joe Perchesf629d202013-09-26 14:48:15 -07003396void netif_device_detach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003397
Joe Perchesf629d202013-09-26 14:48:15 -07003398void netif_device_attach(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003399
3400/*
3401 * Network interface message level settings
3402 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003403
3404enum {
3405 NETIF_MSG_DRV = 0x0001,
3406 NETIF_MSG_PROBE = 0x0002,
3407 NETIF_MSG_LINK = 0x0004,
3408 NETIF_MSG_TIMER = 0x0008,
3409 NETIF_MSG_IFDOWN = 0x0010,
3410 NETIF_MSG_IFUP = 0x0020,
3411 NETIF_MSG_RX_ERR = 0x0040,
3412 NETIF_MSG_TX_ERR = 0x0080,
3413 NETIF_MSG_TX_QUEUED = 0x0100,
3414 NETIF_MSG_INTR = 0x0200,
3415 NETIF_MSG_TX_DONE = 0x0400,
3416 NETIF_MSG_RX_STATUS = 0x0800,
3417 NETIF_MSG_PKTDATA = 0x1000,
3418 NETIF_MSG_HW = 0x2000,
3419 NETIF_MSG_WOL = 0x4000,
3420};
3421
3422#define netif_msg_drv(p) ((p)->msg_enable & NETIF_MSG_DRV)
3423#define netif_msg_probe(p) ((p)->msg_enable & NETIF_MSG_PROBE)
3424#define netif_msg_link(p) ((p)->msg_enable & NETIF_MSG_LINK)
3425#define netif_msg_timer(p) ((p)->msg_enable & NETIF_MSG_TIMER)
3426#define netif_msg_ifdown(p) ((p)->msg_enable & NETIF_MSG_IFDOWN)
3427#define netif_msg_ifup(p) ((p)->msg_enable & NETIF_MSG_IFUP)
3428#define netif_msg_rx_err(p) ((p)->msg_enable & NETIF_MSG_RX_ERR)
3429#define netif_msg_tx_err(p) ((p)->msg_enable & NETIF_MSG_TX_ERR)
3430#define netif_msg_tx_queued(p) ((p)->msg_enable & NETIF_MSG_TX_QUEUED)
3431#define netif_msg_intr(p) ((p)->msg_enable & NETIF_MSG_INTR)
3432#define netif_msg_tx_done(p) ((p)->msg_enable & NETIF_MSG_TX_DONE)
3433#define netif_msg_rx_status(p) ((p)->msg_enable & NETIF_MSG_RX_STATUS)
3434#define netif_msg_pktdata(p) ((p)->msg_enable & NETIF_MSG_PKTDATA)
3435#define netif_msg_hw(p) ((p)->msg_enable & NETIF_MSG_HW)
3436#define netif_msg_wol(p) ((p)->msg_enable & NETIF_MSG_WOL)
3437
3438static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
3439{
3440 /* use default */
3441 if (debug_value < 0 || debug_value >= (sizeof(u32) * 8))
3442 return default_msg_enable_bits;
3443 if (debug_value == 0) /* no output */
3444 return 0;
3445 /* set low N bits */
3446 return (1 << debug_value) - 1;
3447}
3448
David S. Millerc773e842008-07-08 23:13:53 -07003449static inline void __netif_tx_lock(struct netdev_queue *txq, int cpu)
Herbert Xu932ff272006-06-09 12:20:56 -07003450{
David S. Millerc773e842008-07-08 23:13:53 -07003451 spin_lock(&txq->_xmit_lock);
3452 txq->xmit_lock_owner = cpu;
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003453}
3454
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003455static inline void __netif_tx_lock_bh(struct netdev_queue *txq)
3456{
3457 spin_lock_bh(&txq->_xmit_lock);
3458 txq->xmit_lock_owner = smp_processor_id();
3459}
3460
David S. Miller4d295152012-03-07 21:02:35 -05003461static inline bool __netif_tx_trylock(struct netdev_queue *txq)
David S. Millerc773e842008-07-08 23:13:53 -07003462{
David S. Miller4d295152012-03-07 21:02:35 -05003463 bool ok = spin_trylock(&txq->_xmit_lock);
David S. Millerc773e842008-07-08 23:13:53 -07003464 if (likely(ok))
3465 txq->xmit_lock_owner = smp_processor_id();
3466 return ok;
Herbert Xu932ff272006-06-09 12:20:56 -07003467}
3468
David S. Millerc773e842008-07-08 23:13:53 -07003469static inline void __netif_tx_unlock(struct netdev_queue *txq)
3470{
3471 txq->xmit_lock_owner = -1;
3472 spin_unlock(&txq->_xmit_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003473}
3474
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003475static inline void __netif_tx_unlock_bh(struct netdev_queue *txq)
3476{
3477 txq->xmit_lock_owner = -1;
3478 spin_unlock_bh(&txq->_xmit_lock);
3479}
3480
Eric Dumazet08baf562009-05-25 22:58:01 -07003481static inline void txq_trans_update(struct netdev_queue *txq)
3482{
3483 if (txq->xmit_lock_owner != -1)
3484 txq->trans_start = jiffies;
3485}
3486
David S. Millerc3f26a22008-07-31 16:58:50 -07003487/**
3488 * netif_tx_lock - grab network device transmit lock
3489 * @dev: network device
David S. Millerc3f26a22008-07-31 16:58:50 -07003490 *
3491 * Get network device transmit lock
3492 */
3493static inline void netif_tx_lock(struct net_device *dev)
3494{
3495 unsigned int i;
3496 int cpu;
3497
3498 spin_lock(&dev->tx_global_lock);
3499 cpu = smp_processor_id();
3500 for (i = 0; i < dev->num_tx_queues; i++) {
3501 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
3502
3503 /* We are the only thread of execution doing a
3504 * freeze, but we have to grab the _xmit_lock in
3505 * order to synchronize with threads which are in
3506 * the ->hard_start_xmit() handler and already
3507 * checked the frozen bit.
3508 */
3509 __netif_tx_lock(txq, cpu);
3510 set_bit(__QUEUE_STATE_FROZEN, &txq->state);
3511 __netif_tx_unlock(txq);
3512 }
3513}
3514
3515static inline void netif_tx_lock_bh(struct net_device *dev)
3516{
3517 local_bh_disable();
3518 netif_tx_lock(dev);
3519}
3520
Herbert Xu932ff272006-06-09 12:20:56 -07003521static inline void netif_tx_unlock(struct net_device *dev)
3522{
David S. Millere8a04642008-07-17 00:34:19 -07003523 unsigned int i;
David S. Millerc773e842008-07-08 23:13:53 -07003524
David S. Millere8a04642008-07-17 00:34:19 -07003525 for (i = 0; i < dev->num_tx_queues; i++) {
3526 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millere8a04642008-07-17 00:34:19 -07003527
David S. Millerc3f26a22008-07-31 16:58:50 -07003528 /* No need to grab the _xmit_lock here. If the
3529 * queue is not stopped for another reason, we
3530 * force a schedule.
3531 */
3532 clear_bit(__QUEUE_STATE_FROZEN, &txq->state);
Krishna Kumar7b3d3e42009-08-29 20:21:21 +00003533 netif_schedule_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003534 }
3535 spin_unlock(&dev->tx_global_lock);
Herbert Xu932ff272006-06-09 12:20:56 -07003536}
3537
3538static inline void netif_tx_unlock_bh(struct net_device *dev)
3539{
David S. Millere8a04642008-07-17 00:34:19 -07003540 netif_tx_unlock(dev);
3541 local_bh_enable();
Herbert Xu932ff272006-06-09 12:20:56 -07003542}
3543
David S. Millerc773e842008-07-08 23:13:53 -07003544#define HARD_TX_LOCK(dev, txq, cpu) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003545 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003546 __netif_tx_lock(txq, cpu); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003547 } \
3548}
3549
Eric W. Biederman5efeac42014-03-27 15:42:20 -07003550#define HARD_TX_TRYLOCK(dev, txq) \
3551 (((dev->features & NETIF_F_LLTX) == 0) ? \
3552 __netif_tx_trylock(txq) : \
3553 true )
3554
David S. Millerc773e842008-07-08 23:13:53 -07003555#define HARD_TX_UNLOCK(dev, txq) { \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003556 if ((dev->features & NETIF_F_LLTX) == 0) { \
David S. Millerc773e842008-07-08 23:13:53 -07003557 __netif_tx_unlock(txq); \
Jamal Hadi Salim22dd7492007-09-16 14:40:49 -07003558 } \
3559}
3560
Linus Torvalds1da177e2005-04-16 15:20:36 -07003561static inline void netif_tx_disable(struct net_device *dev)
3562{
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003563 unsigned int i;
David S. Millerc3f26a22008-07-31 16:58:50 -07003564 int cpu;
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003565
David S. Millerc3f26a22008-07-31 16:58:50 -07003566 local_bh_disable();
3567 cpu = smp_processor_id();
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003568 for (i = 0; i < dev->num_tx_queues; i++) {
3569 struct netdev_queue *txq = netdev_get_tx_queue(dev, i);
David S. Millerc3f26a22008-07-31 16:58:50 -07003570
3571 __netif_tx_lock(txq, cpu);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003572 netif_tx_stop_queue(txq);
David S. Millerc3f26a22008-07-31 16:58:50 -07003573 __netif_tx_unlock(txq);
David S. Millerfd2ea0a2008-07-17 01:56:23 -07003574 }
David S. Millerc3f26a22008-07-31 16:58:50 -07003575 local_bh_enable();
Linus Torvalds1da177e2005-04-16 15:20:36 -07003576}
3577
David S. Millere308a5d2008-07-15 00:13:44 -07003578static inline void netif_addr_lock(struct net_device *dev)
3579{
3580 spin_lock(&dev->addr_list_lock);
3581}
3582
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003583static inline void netif_addr_lock_nested(struct net_device *dev)
3584{
Vlad Yasevich25175ba2014-05-16 17:04:54 -04003585 int subclass = SINGLE_DEPTH_NESTING;
3586
3587 if (dev->netdev_ops->ndo_get_lock_subclass)
3588 subclass = dev->netdev_ops->ndo_get_lock_subclass(dev);
3589
3590 spin_lock_nested(&dev->addr_list_lock, subclass);
Jiri Pirko2429f7a2012-01-09 06:36:54 +00003591}
3592
David S. Millere308a5d2008-07-15 00:13:44 -07003593static inline void netif_addr_lock_bh(struct net_device *dev)
3594{
3595 spin_lock_bh(&dev->addr_list_lock);
3596}
3597
3598static inline void netif_addr_unlock(struct net_device *dev)
3599{
3600 spin_unlock(&dev->addr_list_lock);
3601}
3602
3603static inline void netif_addr_unlock_bh(struct net_device *dev)
3604{
3605 spin_unlock_bh(&dev->addr_list_lock);
3606}
3607
Jiri Pirkof001fde2009-05-05 02:48:28 +00003608/*
Jiri Pirko31278e72009-06-17 01:12:19 +00003609 * dev_addrs walker. Should be used only for read access. Call with
Jiri Pirkof001fde2009-05-05 02:48:28 +00003610 * rcu_read_lock held.
3611 */
3612#define for_each_dev_addr(dev, ha) \
Jiri Pirko31278e72009-06-17 01:12:19 +00003613 list_for_each_entry_rcu(ha, &dev->dev_addrs.list, list)
Jiri Pirkof001fde2009-05-05 02:48:28 +00003614
Linus Torvalds1da177e2005-04-16 15:20:36 -07003615/* These functions live elsewhere (drivers/net/net_init.c, but related) */
3616
Joe Perchesf629d202013-09-26 14:48:15 -07003617void ether_setup(struct net_device *dev);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003618
3619/* Support for loadable net-drivers */
Joe Perchesf629d202013-09-26 14:48:15 -07003620struct net_device *alloc_netdev_mqs(int sizeof_priv, const char *name,
Tom Gundersenc835a672014-07-14 16:37:24 +02003621 unsigned char name_assign_type,
Joe Perchesf629d202013-09-26 14:48:15 -07003622 void (*setup)(struct net_device *),
3623 unsigned int txqs, unsigned int rxqs);
Tom Gundersenc835a672014-07-14 16:37:24 +02003624#define alloc_netdev(sizeof_priv, name, name_assign_type, setup) \
3625 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, 1, 1)
Tom Herbert36909ea2011-01-09 19:36:31 +00003626
Tom Gundersenc835a672014-07-14 16:37:24 +02003627#define alloc_netdev_mq(sizeof_priv, name, name_assign_type, setup, count) \
3628 alloc_netdev_mqs(sizeof_priv, name, name_assign_type, setup, count, \
3629 count)
Tom Herbert36909ea2011-01-09 19:36:31 +00003630
Joe Perchesf629d202013-09-26 14:48:15 -07003631int register_netdev(struct net_device *dev);
3632void unregister_netdev(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003633
Jiri Pirko22bedad32010-04-01 21:22:57 +00003634/* General hardware address lists handling functions */
Joe Perchesf629d202013-09-26 14:48:15 -07003635int __hw_addr_sync(struct netdev_hw_addr_list *to_list,
3636 struct netdev_hw_addr_list *from_list, int addr_len);
3637void __hw_addr_unsync(struct netdev_hw_addr_list *to_list,
3638 struct netdev_hw_addr_list *from_list, int addr_len);
Alexander Duyck670e5b82014-05-28 18:44:46 -07003639int __hw_addr_sync_dev(struct netdev_hw_addr_list *list,
3640 struct net_device *dev,
3641 int (*sync)(struct net_device *, const unsigned char *),
3642 int (*unsync)(struct net_device *,
3643 const unsigned char *));
3644void __hw_addr_unsync_dev(struct netdev_hw_addr_list *list,
3645 struct net_device *dev,
3646 int (*unsync)(struct net_device *,
3647 const unsigned char *));
Joe Perchesf629d202013-09-26 14:48:15 -07003648void __hw_addr_init(struct netdev_hw_addr_list *list);
Jiri Pirko22bedad32010-04-01 21:22:57 +00003649
Jiri Pirkof001fde2009-05-05 02:48:28 +00003650/* Functions used for device addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003651int dev_addr_add(struct net_device *dev, const unsigned char *addr,
3652 unsigned char addr_type);
3653int dev_addr_del(struct net_device *dev, const unsigned char *addr,
3654 unsigned char addr_type);
Joe Perchesf629d202013-09-26 14:48:15 -07003655void dev_addr_flush(struct net_device *dev);
3656int dev_addr_init(struct net_device *dev);
Jiri Pirkoa748ee22010-04-01 21:22:09 +00003657
3658/* Functions used for unicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003659int dev_uc_add(struct net_device *dev, const unsigned char *addr);
3660int dev_uc_add_excl(struct net_device *dev, const unsigned char *addr);
3661int dev_uc_del(struct net_device *dev, const unsigned char *addr);
3662int dev_uc_sync(struct net_device *to, struct net_device *from);
3663int dev_uc_sync_multiple(struct net_device *to, struct net_device *from);
3664void dev_uc_unsync(struct net_device *to, struct net_device *from);
3665void dev_uc_flush(struct net_device *dev);
3666void dev_uc_init(struct net_device *dev);
Jiri Pirkof001fde2009-05-05 02:48:28 +00003667
Alexander Duyck670e5b82014-05-28 18:44:46 -07003668/**
3669 * __dev_uc_sync - Synchonize device's unicast list
3670 * @dev: device to sync
3671 * @sync: function to call if address should be added
3672 * @unsync: function to call if address should be removed
3673 *
3674 * Add newly added addresses to the interface, and release
3675 * addresses that have been deleted.
3676 **/
3677static inline int __dev_uc_sync(struct net_device *dev,
3678 int (*sync)(struct net_device *,
3679 const unsigned char *),
3680 int (*unsync)(struct net_device *,
3681 const unsigned char *))
3682{
3683 return __hw_addr_sync_dev(&dev->uc, dev, sync, unsync);
3684}
3685
3686/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003687 * __dev_uc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003688 * @dev: device to sync
3689 * @unsync: function to call if address should be removed
3690 *
3691 * Remove all addresses that were added to the device by dev_uc_sync().
3692 **/
3693static inline void __dev_uc_unsync(struct net_device *dev,
3694 int (*unsync)(struct net_device *,
3695 const unsigned char *))
3696{
3697 __hw_addr_unsync_dev(&dev->uc, dev, unsync);
3698}
3699
Jiri Pirko22bedad32010-04-01 21:22:57 +00003700/* Functions used for multicast addresses handling */
Joe Perchesf629d202013-09-26 14:48:15 -07003701int dev_mc_add(struct net_device *dev, const unsigned char *addr);
3702int dev_mc_add_global(struct net_device *dev, const unsigned char *addr);
3703int dev_mc_add_excl(struct net_device *dev, const unsigned char *addr);
3704int dev_mc_del(struct net_device *dev, const unsigned char *addr);
3705int dev_mc_del_global(struct net_device *dev, const unsigned char *addr);
3706int dev_mc_sync(struct net_device *to, struct net_device *from);
3707int dev_mc_sync_multiple(struct net_device *to, struct net_device *from);
3708void dev_mc_unsync(struct net_device *to, struct net_device *from);
3709void dev_mc_flush(struct net_device *dev);
3710void dev_mc_init(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003711
Alexander Duyck670e5b82014-05-28 18:44:46 -07003712/**
3713 * __dev_mc_sync - Synchonize device's multicast list
3714 * @dev: device to sync
3715 * @sync: function to call if address should be added
3716 * @unsync: function to call if address should be removed
3717 *
3718 * Add newly added addresses to the interface, and release
3719 * addresses that have been deleted.
3720 **/
3721static inline int __dev_mc_sync(struct net_device *dev,
3722 int (*sync)(struct net_device *,
3723 const unsigned char *),
3724 int (*unsync)(struct net_device *,
3725 const unsigned char *))
3726{
3727 return __hw_addr_sync_dev(&dev->mc, dev, sync, unsync);
3728}
3729
3730/**
Masanari Iidae793c0f2014-09-04 23:44:36 +09003731 * __dev_mc_unsync - Remove synchronized addresses from device
Alexander Duyck670e5b82014-05-28 18:44:46 -07003732 * @dev: device to sync
3733 * @unsync: function to call if address should be removed
3734 *
3735 * Remove all addresses that were added to the device by dev_mc_sync().
3736 **/
3737static inline void __dev_mc_unsync(struct net_device *dev,
3738 int (*unsync)(struct net_device *,
3739 const unsigned char *))
3740{
3741 __hw_addr_unsync_dev(&dev->mc, dev, unsync);
3742}
3743
Linus Torvalds1da177e2005-04-16 15:20:36 -07003744/* Functions used for secondary unicast and multicast support */
Joe Perchesf629d202013-09-26 14:48:15 -07003745void dev_set_rx_mode(struct net_device *dev);
3746void __dev_set_rx_mode(struct net_device *dev);
3747int dev_set_promiscuity(struct net_device *dev, int inc);
3748int dev_set_allmulti(struct net_device *dev, int inc);
3749void netdev_state_change(struct net_device *dev);
3750void netdev_notify_peers(struct net_device *dev);
3751void netdev_features_change(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003752/* Load a device via the kmod */
Joe Perchesf629d202013-09-26 14:48:15 -07003753void dev_load(struct net *net, const char *name);
3754struct rtnl_link_stats64 *dev_get_stats(struct net_device *dev,
3755 struct rtnl_link_stats64 *storage);
3756void netdev_stats_to_stats64(struct rtnl_link_stats64 *stats64,
3757 const struct net_device_stats *netdev_stats);
Herbert Xufb286bb2005-11-10 13:01:24 -08003758
3759extern int netdev_max_backlog;
Eric Dumazet3b098e22010-05-15 23:57:10 -07003760extern int netdev_tstamp_prequeue;
Herbert Xufb286bb2005-11-10 13:01:24 -08003761extern int weight_p;
Eric Dumazet0a148422011-04-20 09:27:32 +00003762extern int bpf_jit_enable;
Jiri Pirko9ff162a2013-01-03 22:48:49 +00003763
Joe Perchesf629d202013-09-26 14:48:15 -07003764bool netdev_has_upper_dev(struct net_device *dev, struct net_device *upper_dev);
Vlad Yasevich44a40852014-05-16 17:20:38 -04003765struct net_device *netdev_upper_get_next_dev_rcu(struct net_device *dev,
3766 struct list_head **iter);
Joe Perchesf629d202013-09-26 14:48:15 -07003767struct net_device *netdev_all_upper_get_next_dev_rcu(struct net_device *dev,
3768 struct list_head **iter);
Veaceslav Falico8b5be852013-08-28 23:25:08 +02003769
3770/* iterate through upper list, must be called under RCU read lock */
Vlad Yasevich44a40852014-05-16 17:20:38 -04003771#define netdev_for_each_upper_dev_rcu(dev, updev, iter) \
3772 for (iter = &(dev)->adj_list.upper, \
3773 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)); \
3774 updev; \
3775 updev = netdev_upper_get_next_dev_rcu(dev, &(iter)))
3776
3777/* iterate through upper list, must be called under RCU read lock */
Veaceslav Falico2f268f12013-09-25 09:20:07 +02003778#define netdev_for_each_all_upper_dev_rcu(dev, updev, iter) \
3779 for (iter = &(dev)->all_adj_list.upper, \
3780 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)); \
3781 updev; \
3782 updev = netdev_all_upper_get_next_dev_rcu(dev, &(iter)))
Veaceslav Falico8b5be852013-08-28 23:25:08 +02003783
Joe Perchesf629d202013-09-26 14:48:15 -07003784void *netdev_lower_get_next_private(struct net_device *dev,
3785 struct list_head **iter);
3786void *netdev_lower_get_next_private_rcu(struct net_device *dev,
3787 struct list_head **iter);
Veaceslav Falico31088a12013-09-25 09:20:12 +02003788
3789#define netdev_for_each_lower_private(dev, priv, iter) \
3790 for (iter = (dev)->adj_list.lower.next, \
3791 priv = netdev_lower_get_next_private(dev, &(iter)); \
3792 priv; \
3793 priv = netdev_lower_get_next_private(dev, &(iter)))
3794
3795#define netdev_for_each_lower_private_rcu(dev, priv, iter) \
3796 for (iter = &(dev)->adj_list.lower, \
3797 priv = netdev_lower_get_next_private_rcu(dev, &(iter)); \
3798 priv; \
3799 priv = netdev_lower_get_next_private_rcu(dev, &(iter)))
3800
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003801void *netdev_lower_get_next(struct net_device *dev,
3802 struct list_head **iter);
3803#define netdev_for_each_lower_dev(dev, ldev, iter) \
Nikolay Aleksandrovcfdd28b2016-02-17 18:00:31 +01003804 for (iter = (dev)->adj_list.lower.next, \
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003805 ldev = netdev_lower_get_next(dev, &(iter)); \
3806 ldev; \
3807 ldev = netdev_lower_get_next(dev, &(iter)))
3808
Joe Perchesf629d202013-09-26 14:48:15 -07003809void *netdev_adjacent_get_private(struct list_head *adj_list);
dingtianhonge001bfa2013-12-13 10:19:55 +08003810void *netdev_lower_get_first_private_rcu(struct net_device *dev);
Joe Perchesf629d202013-09-26 14:48:15 -07003811struct net_device *netdev_master_upper_dev_get(struct net_device *dev);
3812struct net_device *netdev_master_upper_dev_get_rcu(struct net_device *dev);
3813int netdev_upper_dev_link(struct net_device *dev, struct net_device *upper_dev);
3814int netdev_master_upper_dev_link(struct net_device *dev,
Jiri Pirko6dffb042015-12-03 12:12:10 +01003815 struct net_device *upper_dev,
Jiri Pirko29bf24a2015-12-03 12:12:11 +01003816 void *upper_priv, void *upper_info);
Joe Perchesf629d202013-09-26 14:48:15 -07003817void netdev_upper_dev_unlink(struct net_device *dev,
3818 struct net_device *upper_dev);
Veaceslav Falico5bb025f2014-01-14 21:58:51 +01003819void netdev_adjacent_rename_links(struct net_device *dev, char *oldname);
Joe Perchesf629d202013-09-26 14:48:15 -07003820void *netdev_lower_dev_get_private(struct net_device *dev,
3821 struct net_device *lower_dev);
Jiri Pirko04d48262015-12-03 12:12:15 +01003822void netdev_lower_state_changed(struct net_device *lower_dev,
3823 void *lower_state_info);
Eric Dumazet960fb622014-11-16 06:23:05 -08003824
3825/* RSS keys are 40 or 52 bytes long */
3826#define NETDEV_RSS_KEY_LEN 52
Kim Jonesba905f52016-02-02 03:51:16 +00003827extern u8 netdev_rss_key[NETDEV_RSS_KEY_LEN] __read_mostly;
Eric Dumazet960fb622014-11-16 06:23:05 -08003828void netdev_rss_key_fill(void *buffer, size_t len);
3829
Vlad Yasevich4085ebe2014-05-16 17:04:53 -04003830int dev_get_nest_level(struct net_device *dev,
Jiri Pirkob618aaa2015-12-04 15:01:31 +01003831 bool (*type_check)(const struct net_device *dev));
Joe Perchesf629d202013-09-26 14:48:15 -07003832int skb_checksum_help(struct sk_buff *skb);
3833struct sk_buff *__skb_gso_segment(struct sk_buff *skb,
3834 netdev_features_t features, bool tx_path);
3835struct sk_buff *skb_mac_gso_segment(struct sk_buff *skb,
3836 netdev_features_t features);
Cong Wang12b00042013-02-05 16:36:38 +00003837
Moni Shoua61bd3852015-02-03 16:48:29 +02003838struct netdev_bonding_info {
3839 ifslave slave;
3840 ifbond master;
3841};
3842
3843struct netdev_notifier_bonding_info {
3844 struct netdev_notifier_info info; /* must be first */
3845 struct netdev_bonding_info bonding_info;
3846};
3847
3848void netdev_bonding_info_change(struct net_device *dev,
3849 struct netdev_bonding_info *bonding_info);
3850
Cong Wang12b00042013-02-05 16:36:38 +00003851static inline
3852struct sk_buff *skb_gso_segment(struct sk_buff *skb, netdev_features_t features)
3853{
3854 return __skb_gso_segment(skb, features, true);
3855}
Vlad Yasevich53d64712014-03-27 17:26:18 -04003856__be16 skb_network_protocol(struct sk_buff *skb, int *depth);
Pravin B Shelarec5f0612013-03-07 09:28:01 +00003857
3858static inline bool can_checksum_protocol(netdev_features_t features,
3859 __be16 protocol)
3860{
Tom Herbertc8cd0982015-12-14 11:19:44 -08003861 if (protocol == htons(ETH_P_FCOE))
3862 return !!(features & NETIF_F_FCOE_CRC);
3863
3864 /* Assume this is an IP checksum (not SCTP CRC) */
3865
3866 if (features & NETIF_F_HW_CSUM) {
3867 /* Can checksum everything */
3868 return true;
3869 }
3870
3871 switch (protocol) {
3872 case htons(ETH_P_IP):
3873 return !!(features & NETIF_F_IP_CSUM);
3874 case htons(ETH_P_IPV6):
3875 return !!(features & NETIF_F_IPV6_CSUM);
3876 default:
3877 return false;
3878 }
Pravin B Shelarec5f0612013-03-07 09:28:01 +00003879}
Cong Wang12b00042013-02-05 16:36:38 +00003880
Tom Herbert6ae23ad2015-12-14 11:19:46 -08003881/* Map an ethertype into IP protocol if possible */
3882static inline int eproto_to_ipproto(int eproto)
3883{
3884 switch (eproto) {
3885 case htons(ETH_P_IP):
3886 return IPPROTO_IP;
3887 case htons(ETH_P_IPV6):
3888 return IPPROTO_IPV6;
3889 default:
3890 return -1;
3891 }
3892}
3893
Herbert Xufb286bb2005-11-10 13:01:24 -08003894#ifdef CONFIG_BUG
Joe Perchesf629d202013-09-26 14:48:15 -07003895void netdev_rx_csum_fault(struct net_device *dev);
Herbert Xufb286bb2005-11-10 13:01:24 -08003896#else
3897static inline void netdev_rx_csum_fault(struct net_device *dev)
3898{
3899}
3900#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003901/* rx skb timestamps */
Joe Perchesf629d202013-09-26 14:48:15 -07003902void net_enable_timestamp(void);
3903void net_disable_timestamp(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003904
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003905#ifdef CONFIG_PROC_FS
Joe Perchesf629d202013-09-26 14:48:15 -07003906int __init dev_proc_init(void);
Cong Wang900ff8c2013-02-18 19:20:33 +00003907#else
3908#define dev_proc_init() 0
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003909#endif
3910
David S. Miller47982482014-08-22 16:21:53 -07003911static inline netdev_tx_t __netdev_start_xmit(const struct net_device_ops *ops,
David S. Millerfa2dbdc22014-08-29 21:55:22 -07003912 struct sk_buff *skb, struct net_device *dev,
3913 bool more)
David S. Miller47982482014-08-22 16:21:53 -07003914{
David S. Millerfa2dbdc22014-08-29 21:55:22 -07003915 skb->xmit_more = more ? 1 : 0;
David S. Miller0b725a22014-08-25 15:51:53 -07003916 return ops->ndo_start_xmit(skb, dev);
David S. Miller47982482014-08-22 16:21:53 -07003917}
3918
David S. Miller10b3ad82014-08-29 21:07:24 -07003919static inline netdev_tx_t netdev_start_xmit(struct sk_buff *skb, struct net_device *dev,
David S. Millerfa2dbdc22014-08-29 21:55:22 -07003920 struct netdev_queue *txq, bool more)
David S. Miller47982482014-08-22 16:21:53 -07003921{
3922 const struct net_device_ops *ops = dev->netdev_ops;
David S. Miller10b3ad82014-08-29 21:07:24 -07003923 int rc;
David S. Miller47982482014-08-22 16:21:53 -07003924
David S. Millerfa2dbdc22014-08-29 21:55:22 -07003925 rc = __netdev_start_xmit(ops, skb, dev, more);
David S. Miller10b3ad82014-08-29 21:07:24 -07003926 if (rc == NETDEV_TX_OK)
3927 txq_trans_update(txq);
3928
3929 return rc;
David S. Miller47982482014-08-22 16:21:53 -07003930}
3931
Linus Torvalds42a2d922013-11-13 17:40:34 +09003932int netdev_class_create_file_ns(struct class_attribute *class_attr,
3933 const void *ns);
3934void netdev_class_remove_file_ns(struct class_attribute *class_attr,
3935 const void *ns);
Tejun Heo58292cbe2013-09-11 22:29:04 -04003936
3937static inline int netdev_class_create_file(struct class_attribute *class_attr)
3938{
3939 return netdev_class_create_file_ns(class_attr, NULL);
3940}
3941
3942static inline void netdev_class_remove_file(struct class_attribute *class_attr)
3943{
3944 netdev_class_remove_file_ns(class_attr, NULL);
3945}
Jay Vosburghb8a97872008-06-13 18:12:04 -07003946
Johannes Berg04600792010-08-05 17:45:15 +02003947extern struct kobj_ns_type_operations net_ns_type_operations;
3948
Joe Perchesf629d202013-09-26 14:48:15 -07003949const char *netdev_drivername(const struct net_device *dev);
Arjan van de Ven6579e572008-07-21 13:31:48 -07003950
Joe Perchesf629d202013-09-26 14:48:15 -07003951void linkwatch_run_queue(void);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003952
Michal Kubečekda081432014-05-20 08:29:25 +02003953static inline netdev_features_t netdev_intersect_features(netdev_features_t f1,
3954 netdev_features_t f2)
3955{
Tom Herbertc8cd0982015-12-14 11:19:44 -08003956 if ((f1 ^ f2) & NETIF_F_HW_CSUM) {
3957 if (f1 & NETIF_F_HW_CSUM)
Tom Herbertb6a0e722016-01-11 10:19:10 -08003958 f1 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
Tom Herbertc8cd0982015-12-14 11:19:44 -08003959 else
Tom Herbertb6a0e722016-01-11 10:19:10 -08003960 f2 |= (NETIF_F_IP_CSUM|NETIF_F_IPV6_CSUM);
Tom Herbertc8cd0982015-12-14 11:19:44 -08003961 }
Michal Kubečekda081432014-05-20 08:29:25 +02003962
Tom Herbertc8cd0982015-12-14 11:19:44 -08003963 return f1 & f2;
Michal Kubečekda081432014-05-20 08:29:25 +02003964}
3965
Michał Mirosławc8f44af2011-11-15 15:29:55 +00003966static inline netdev_features_t netdev_get_wanted_features(
3967 struct net_device *dev)
Michał Mirosław5455c692011-02-15 16:59:17 +00003968{
3969 return (dev->features & ~dev->hw_features) | dev->wanted_features;
3970}
Michał Mirosławc8f44af2011-11-15 15:29:55 +00003971netdev_features_t netdev_increment_features(netdev_features_t all,
3972 netdev_features_t one, netdev_features_t mask);
Eric Dumazetb0ce3502013-05-16 07:34:53 +00003973
3974/* Allow TSO being used on stacked device :
3975 * Performing the GSO segmentation before last device
3976 * is a performance improvement.
3977 */
3978static inline netdev_features_t netdev_add_tso_features(netdev_features_t features,
3979 netdev_features_t mask)
3980{
3981 return netdev_increment_features(features, NETIF_F_ALL_TSO, mask);
3982}
3983
Michał Mirosław6cb6a272011-04-02 22:48:47 -07003984int __netdev_update_features(struct net_device *dev);
Michał Mirosław5455c692011-02-15 16:59:17 +00003985void netdev_update_features(struct net_device *dev);
Michał Mirosławafe12cc2011-05-07 03:22:17 +00003986void netdev_change_features(struct net_device *dev);
Herbert Xu7f353bf2007-08-10 15:47:58 -07003987
Patrick Mullaneyfc4a7482009-12-03 15:59:22 -08003988void netif_stacked_transfer_operstate(const struct net_device *rootdev,
3989 struct net_device *dev);
3990
Toshiaki Makitae38f3022015-03-27 14:31:13 +09003991netdev_features_t passthru_features_check(struct sk_buff *skb,
3992 struct net_device *dev,
3993 netdev_features_t features);
Florian Westphalc1e756b2014-05-05 15:00:44 +02003994netdev_features_t netif_skb_features(struct sk_buff *skb);
Jesse Gross58e998c2010-10-29 12:14:55 +00003995
David S. Miller4d295152012-03-07 21:02:35 -05003996static inline bool net_gso_ok(netdev_features_t features, int gso_type)
Herbert Xubcd76112006-06-30 13:36:35 -07003997{
Michał Mirosławc8f44af2011-11-15 15:29:55 +00003998 netdev_features_t feature = gso_type << NETIF_F_GSO_SHIFT;
Michał Mirosław0345e182011-11-16 14:05:33 +00003999
4000 /* check flags correspondence */
4001 BUILD_BUG_ON(SKB_GSO_TCPV4 != (NETIF_F_TSO >> NETIF_F_GSO_SHIFT));
4002 BUILD_BUG_ON(SKB_GSO_UDP != (NETIF_F_UFO >> NETIF_F_GSO_SHIFT));
4003 BUILD_BUG_ON(SKB_GSO_DODGY != (NETIF_F_GSO_ROBUST >> NETIF_F_GSO_SHIFT));
4004 BUILD_BUG_ON(SKB_GSO_TCP_ECN != (NETIF_F_TSO_ECN >> NETIF_F_GSO_SHIFT));
4005 BUILD_BUG_ON(SKB_GSO_TCPV6 != (NETIF_F_TSO6 >> NETIF_F_GSO_SHIFT));
4006 BUILD_BUG_ON(SKB_GSO_FCOE != (NETIF_F_FSO >> NETIF_F_GSO_SHIFT));
Tom Herbert4b282522014-06-14 23:23:52 -07004007 BUILD_BUG_ON(SKB_GSO_GRE != (NETIF_F_GSO_GRE >> NETIF_F_GSO_SHIFT));
4008 BUILD_BUG_ON(SKB_GSO_GRE_CSUM != (NETIF_F_GSO_GRE_CSUM >> NETIF_F_GSO_SHIFT));
4009 BUILD_BUG_ON(SKB_GSO_IPIP != (NETIF_F_GSO_IPIP >> NETIF_F_GSO_SHIFT));
4010 BUILD_BUG_ON(SKB_GSO_SIT != (NETIF_F_GSO_SIT >> NETIF_F_GSO_SHIFT));
4011 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL != (NETIF_F_GSO_UDP_TUNNEL >> NETIF_F_GSO_SHIFT));
4012 BUILD_BUG_ON(SKB_GSO_UDP_TUNNEL_CSUM != (NETIF_F_GSO_UDP_TUNNEL_CSUM >> NETIF_F_GSO_SHIFT));
Tom Herberte585f232014-11-04 09:06:54 -08004013 BUILD_BUG_ON(SKB_GSO_TUNNEL_REMCSUM != (NETIF_F_GSO_TUNNEL_REMCSUM >> NETIF_F_GSO_SHIFT));
Michał Mirosław0345e182011-11-16 14:05:33 +00004014
Herbert Xubcd76112006-06-30 13:36:35 -07004015 return (features & feature) == feature;
4016}
4017
David S. Miller4d295152012-03-07 21:02:35 -05004018static inline bool skb_gso_ok(struct sk_buff *skb, netdev_features_t features)
Herbert Xu576a30e2006-06-27 13:22:38 -07004019{
Herbert Xu278b2512009-06-03 21:20:51 -07004020 return net_gso_ok(features, skb_shinfo(skb)->gso_type) &&
David S. Miller21dc3302010-08-23 00:13:46 -07004021 (!skb_has_frag_list(skb) || (features & NETIF_F_FRAGLIST));
Herbert Xu576a30e2006-06-27 13:22:38 -07004022}
4023
Johannes Berg8b86a612015-04-17 15:45:04 +02004024static inline bool netif_needs_gso(struct sk_buff *skb,
David S. Miller4d295152012-03-07 21:02:35 -05004025 netdev_features_t features)
Herbert Xu79671682006-06-22 02:40:14 -07004026{
Jesse Grossfc741212011-01-09 06:23:32 +00004027 return skb_is_gso(skb) && (!skb_gso_ok(skb, features) ||
Yi Zoucdbee742012-03-16 23:08:11 +00004028 unlikely((skb->ip_summed != CHECKSUM_PARTIAL) &&
4029 (skb->ip_summed != CHECKSUM_UNNECESSARY)));
Herbert Xu79671682006-06-22 02:40:14 -07004030}
4031
Peter P Waskiewicz Jr82cc1a72008-03-21 03:43:19 -07004032static inline void netif_set_gso_max_size(struct net_device *dev,
4033 unsigned int size)
4034{
4035 dev->gso_max_size = size;
4036}
4037
Wei-Chun Chao7a7ffba2013-12-26 13:10:22 -08004038static inline void skb_gso_error_unwind(struct sk_buff *skb, __be16 protocol,
4039 int pulled_hlen, u16 mac_offset,
4040 int mac_len)
4041{
4042 skb->protocol = protocol;
4043 skb->encapsulation = 1;
4044 skb_push(skb, pulled_hlen);
4045 skb_reset_transport_header(skb);
4046 skb->mac_header = mac_offset;
4047 skb->network_header = skb->mac_header + mac_len;
4048 skb->mac_len = mac_len;
4049}
4050
Sabrina Dubroca3c175782016-03-11 18:07:32 +01004051static inline bool netif_is_macsec(const struct net_device *dev)
4052{
4053 return dev->priv_flags & IFF_MACSEC;
4054}
4055
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004056static inline bool netif_is_macvlan(const struct net_device *dev)
John Fastabenda6cc0cf2013-11-06 09:54:46 -08004057{
4058 return dev->priv_flags & IFF_MACVLAN;
4059}
4060
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004061static inline bool netif_is_macvlan_port(const struct net_device *dev)
Mahesh Bandewar2f33e7d2014-12-06 15:53:04 -08004062{
4063 return dev->priv_flags & IFF_MACVLAN_PORT;
4064}
4065
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004066static inline bool netif_is_ipvlan(const struct net_device *dev)
Mahesh Bandewar5933fea2014-12-06 15:53:33 -08004067{
4068 return dev->priv_flags & IFF_IPVLAN_SLAVE;
4069}
4070
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004071static inline bool netif_is_ipvlan_port(const struct net_device *dev)
Mahesh Bandewar5933fea2014-12-06 15:53:33 -08004072{
4073 return dev->priv_flags & IFF_IPVLAN_MASTER;
4074}
4075
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004076static inline bool netif_is_bond_master(const struct net_device *dev)
nikolay@redhat.com8a7fbfa2013-03-12 02:49:01 +00004077{
4078 return dev->flags & IFF_MASTER && dev->priv_flags & IFF_BONDING;
4079}
4080
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004081static inline bool netif_is_bond_slave(const struct net_device *dev)
Jiri Pirko1765a572011-02-12 06:48:36 +00004082{
4083 return dev->flags & IFF_SLAVE && dev->priv_flags & IFF_BONDING;
4084}
4085
Ben Greear3bdc0eb2012-02-11 15:39:30 +00004086static inline bool netif_supports_nofcs(struct net_device *dev)
4087{
4088 return dev->priv_flags & IFF_SUPP_NOFCS;
4089}
4090
David Ahern007979e2015-09-29 20:07:10 -07004091static inline bool netif_is_l3_master(const struct net_device *dev)
David Ahern4e3c8992015-08-13 14:59:00 -06004092{
David Ahern007979e2015-09-29 20:07:10 -07004093 return dev->priv_flags & IFF_L3MDEV_MASTER;
David Ahern4e3c8992015-08-13 14:59:00 -06004094}
4095
David Ahernfee6d4c2015-10-05 08:51:24 -07004096static inline bool netif_is_l3_slave(const struct net_device *dev)
4097{
4098 return dev->priv_flags & IFF_L3MDEV_SLAVE;
4099}
4100
Jiri Pirko0894ae32015-08-27 09:31:19 +02004101static inline bool netif_is_bridge_master(const struct net_device *dev)
4102{
4103 return dev->priv_flags & IFF_EBRIDGE;
4104}
4105
Vlad Yasevich28f9ee22015-11-16 15:43:45 -05004106static inline bool netif_is_bridge_port(const struct net_device *dev)
4107{
4108 return dev->priv_flags & IFF_BRIDGE_PORT;
4109}
4110
Jiri Pirko35d4e172015-08-27 09:31:20 +02004111static inline bool netif_is_ovs_master(const struct net_device *dev)
4112{
4113 return dev->priv_flags & IFF_OPENVSWITCH;
4114}
4115
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004116static inline bool netif_is_team_master(const struct net_device *dev)
Jiri Pirkoc981e422015-12-03 12:12:06 +01004117{
4118 return dev->priv_flags & IFF_TEAM;
4119}
4120
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004121static inline bool netif_is_team_port(const struct net_device *dev)
Jiri Pirkof7f019e2015-12-03 12:12:07 +01004122{
4123 return dev->priv_flags & IFF_TEAM_PORT;
4124}
4125
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004126static inline bool netif_is_lag_master(const struct net_device *dev)
Jiri Pirko7be61832015-12-03 12:12:08 +01004127{
4128 return netif_is_bond_master(dev) || netif_is_team_master(dev);
4129}
4130
Jiri Pirkob618aaa2015-12-04 15:01:31 +01004131static inline bool netif_is_lag_port(const struct net_device *dev)
Jiri Pirkoe0ba1412015-12-03 12:12:09 +01004132{
4133 return netif_is_bond_slave(dev) || netif_is_team_port(dev);
4134}
4135
Keller, Jacob Ed4ab4282016-02-08 16:05:03 -08004136static inline bool netif_is_rxfh_configured(const struct net_device *dev)
4137{
4138 return dev->priv_flags & IFF_RXFH_CONFIGURED;
4139}
4140
Eric Dumazet02875872014-10-05 18:38:35 -07004141/* This device needs to keep skb dst for qdisc enqueue or ndo_start_xmit() */
4142static inline void netif_keep_dst(struct net_device *dev)
4143{
4144 dev->priv_flags &= ~(IFF_XMIT_DST_RELEASE | IFF_XMIT_DST_RELEASE_PERM);
4145}
4146
Eric W. Biederman505d4f72008-11-07 22:54:20 -08004147extern struct pernet_operations __net_initdata loopback_net_ops;
Patrick McHardyb1b67dd2009-04-20 04:49:28 +00004148
Joe Perches571ba422010-02-09 11:49:47 +00004149/* Logging, debugging and troubleshooting/diagnostic helpers. */
4150
4151/* netdev_printk helpers, similar to dev_printk */
4152
4153static inline const char *netdev_name(const struct net_device *dev)
4154{
Veaceslav Falicoc6f854d2014-07-17 19:46:09 +02004155 if (!dev->name[0] || strchr(dev->name, '%'))
4156 return "(unnamed net_device)";
Joe Perches571ba422010-02-09 11:49:47 +00004157 return dev->name;
4158}
4159
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02004160static inline const char *netdev_reg_state(const struct net_device *dev)
4161{
4162 switch (dev->reg_state) {
4163 case NETREG_UNINITIALIZED: return " (uninitialized)";
4164 case NETREG_REGISTERED: return "";
4165 case NETREG_UNREGISTERING: return " (unregistering)";
4166 case NETREG_UNREGISTERED: return " (unregistered)";
4167 case NETREG_RELEASED: return " (released)";
4168 case NETREG_DUMMY: return " (dummy)";
4169 }
4170
4171 WARN_ONCE(1, "%s: unknown reg_state %d\n", dev->name, dev->reg_state);
4172 return " (unknown)";
4173}
4174
Joe Perchesf629d202013-09-26 14:48:15 -07004175__printf(3, 4)
Joe Perches6ea754e2014-09-22 11:10:50 -07004176void netdev_printk(const char *level, const struct net_device *dev,
4177 const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004178__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004179void netdev_emerg(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004180__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004181void netdev_alert(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004182__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004183void netdev_crit(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004184__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004185void netdev_err(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004186__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004187void netdev_warn(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004188__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004189void netdev_notice(const struct net_device *dev, const char *format, ...);
Joe Perchesf629d202013-09-26 14:48:15 -07004190__printf(2, 3)
Joe Perches6ea754e2014-09-22 11:10:50 -07004191void netdev_info(const struct net_device *dev, const char *format, ...);
Joe Perches571ba422010-02-09 11:49:47 +00004192
Vasiliy Kulikov8909c9a2011-03-02 00:33:13 +03004193#define MODULE_ALIAS_NETDEV(device) \
4194 MODULE_ALIAS("netdev-" device)
4195
Jim Cromieb558c962011-12-19 17:11:18 -05004196#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perches571ba422010-02-09 11:49:47 +00004197#define netdev_dbg(__dev, format, args...) \
4198do { \
Jason Baronffa10cb2011-08-11 14:36:48 -04004199 dynamic_netdev_dbg(__dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00004200} while (0)
Jim Cromieb558c962011-12-19 17:11:18 -05004201#elif defined(DEBUG)
4202#define netdev_dbg(__dev, format, args...) \
4203 netdev_printk(KERN_DEBUG, __dev, format, ##args)
Joe Perches571ba422010-02-09 11:49:47 +00004204#else
4205#define netdev_dbg(__dev, format, args...) \
4206({ \
4207 if (0) \
4208 netdev_printk(KERN_DEBUG, __dev, format, ##args); \
Joe Perches571ba422010-02-09 11:49:47 +00004209})
4210#endif
4211
4212#if defined(VERBOSE_DEBUG)
4213#define netdev_vdbg netdev_dbg
4214#else
4215
4216#define netdev_vdbg(dev, format, args...) \
4217({ \
4218 if (0) \
4219 netdev_printk(KERN_DEBUG, dev, format, ##args); \
4220 0; \
4221})
4222#endif
4223
4224/*
4225 * netdev_WARN() acts like dev_printk(), but with the key difference
4226 * of using a WARN/WARN_ON to get the message out, including the
4227 * file/line information and a backtrace.
4228 */
4229#define netdev_WARN(dev, format, args...) \
Veaceslav Falicoccc7f492014-07-17 19:46:10 +02004230 WARN(1, "netdevice: %s%s\n" format, netdev_name(dev), \
4231 netdev_reg_state(dev), ##args)
Joe Perches571ba422010-02-09 11:49:47 +00004232
Joe Perchesb3d95c52010-02-09 11:49:49 +00004233/* netif printk helpers, similar to netdev_printk */
4234
4235#define netif_printk(priv, type, level, dev, fmt, args...) \
4236do { \
4237 if (netif_msg_##type(priv)) \
4238 netdev_printk(level, (dev), fmt, ##args); \
4239} while (0)
4240
Joe Perchesf45f4322010-06-27 01:02:36 +00004241#define netif_level(level, priv, type, dev, fmt, args...) \
4242do { \
4243 if (netif_msg_##type(priv)) \
4244 netdev_##level(dev, fmt, ##args); \
4245} while (0)
4246
Joe Perchesb3d95c52010-02-09 11:49:49 +00004247#define netif_emerg(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004248 netif_level(emerg, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004249#define netif_alert(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004250 netif_level(alert, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004251#define netif_crit(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004252 netif_level(crit, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004253#define netif_err(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004254 netif_level(err, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004255#define netif_warn(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004256 netif_level(warn, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004257#define netif_notice(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004258 netif_level(notice, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004259#define netif_info(priv, type, dev, fmt, args...) \
Joe Perchesf45f4322010-06-27 01:02:36 +00004260 netif_level(info, priv, type, dev, fmt, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004261
Joe Perches0053ea92012-05-30 07:43:34 +00004262#if defined(CONFIG_DYNAMIC_DEBUG)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004263#define netif_dbg(priv, type, netdev, format, args...) \
4264do { \
4265 if (netif_msg_##type(priv)) \
Jason Baronb5fb0a02011-08-11 14:36:53 -04004266 dynamic_netdev_dbg(netdev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00004267} while (0)
Joe Perches0053ea92012-05-30 07:43:34 +00004268#elif defined(DEBUG)
4269#define netif_dbg(priv, type, dev, format, args...) \
4270 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args)
Joe Perchesb3d95c52010-02-09 11:49:49 +00004271#else
4272#define netif_dbg(priv, type, dev, format, args...) \
4273({ \
4274 if (0) \
4275 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
4276 0; \
4277})
4278#endif
4279
4280#if defined(VERBOSE_DEBUG)
Ben Hutchingsbcfcc452010-07-02 07:08:44 +00004281#define netif_vdbg netif_dbg
Joe Perchesb3d95c52010-02-09 11:49:49 +00004282#else
4283#define netif_vdbg(priv, type, dev, format, args...) \
4284({ \
4285 if (0) \
Ben Hutchingsa4ed89c2010-05-18 06:56:32 +00004286 netif_printk(priv, type, KERN_DEBUG, dev, format, ##args); \
Joe Perchesb3d95c52010-02-09 11:49:49 +00004287 0; \
4288})
4289#endif
Joe Perches571ba422010-02-09 11:49:47 +00004290
Cong Wang900ff8c2013-02-18 19:20:33 +00004291/*
4292 * The list of packet types we will receive (as opposed to discard)
4293 * and the routines to invoke.
4294 *
4295 * Why 16. Because with 16 the only overlap we get on a hash of the
4296 * low nibble of the protocol value is RARP/SNAP/X.25.
4297 *
4298 * NOTE: That is no longer true with the addition of VLAN tags. Not
4299 * sure which should go first, but I bet it won't make much
4300 * difference if we are running VLANs. The good news is that
4301 * this protocol won't be in the list unless compiled in, so
4302 * the average user (w/out VLANs) will not be adversely affected.
4303 * --BLG
4304 *
4305 * 0800 IP
4306 * 8100 802.1Q VLAN
4307 * 0001 802.3
4308 * 0002 AX.25
4309 * 0004 802.2
4310 * 8035 RARP
4311 * 0005 SNAP
4312 * 0805 X.25
4313 * 0806 ARP
4314 * 8137 IPX
4315 * 0009 Localtalk
4316 * 86DD IPv6
4317 */
4318#define PTYPE_HASH_SIZE (16)
4319#define PTYPE_HASH_MASK (PTYPE_HASH_SIZE - 1)
4320
Jiri Pirko385a1542009-05-27 15:48:07 -07004321#endif /* _LINUX_NETDEVICE_H */