blob: d89876b806a0eb5a79adc0cce7546fca1f95b164 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * Definitions for the 'struct sk_buff' memory handlers.
3 *
4 * Authors:
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14#ifndef _LINUX_SKBUFF_H
15#define _LINUX_SKBUFF_H
16
Linus Torvalds1da177e2005-04-16 15:20:36 -070017#include <linux/kernel.h>
Vegard Nossumfe55f6d2008-08-30 12:16:35 +020018#include <linux/kmemcheck.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070019#include <linux/compiler.h>
20#include <linux/time.h>
21#include <linux/cache.h>
22
23#include <asm/atomic.h>
24#include <asm/types.h>
25#include <linux/spinlock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070026#include <linux/net.h>
Thomas Graf3fc7e8a2005-06-23 21:00:17 -070027#include <linux/textsearch.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070028#include <net/checksum.h>
Al Viroa80958f2006-12-04 20:41:19 +000029#include <linux/rcupdate.h>
Chris Leech97fc2f02006-05-23 17:55:33 -070030#include <linux/dmaengine.h>
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -070031#include <linux/hrtimer.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070032
Herbert Xu60476372007-04-09 11:59:39 -070033/* Don't change this without changing skb_csum_unnecessary! */
Linus Torvalds1da177e2005-04-16 15:20:36 -070034#define CHECKSUM_NONE 0
Herbert Xu60476372007-04-09 11:59:39 -070035#define CHECKSUM_UNNECESSARY 1
36#define CHECKSUM_COMPLETE 2
37#define CHECKSUM_PARTIAL 3
Linus Torvalds1da177e2005-04-16 15:20:36 -070038
39#define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
40 ~(SMP_CACHE_BYTES - 1))
David S. Millerfc910a22007-03-25 20:27:59 -070041#define SKB_WITH_OVERHEAD(X) \
Herbert Xudeea84b2007-10-21 16:27:46 -070042 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
David S. Millerfc910a22007-03-25 20:27:59 -070043#define SKB_MAX_ORDER(X, ORDER) \
44 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
Linus Torvalds1da177e2005-04-16 15:20:36 -070045#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
46#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
47
48/* A. Checksumming of received packets by device.
49 *
50 * NONE: device failed to checksum this packet.
51 * skb->csum is undefined.
52 *
53 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
54 * skb->csum is undefined.
55 * It is bad option, but, unfortunately, many of vendors do this.
56 * Apparently with secret goal to sell you new device, when you
57 * will add new protocol to your host. F.e. IPv6. 8)
58 *
Patrick McHardy84fa7932006-08-29 16:44:56 -070059 * COMPLETE: the most generic way. Device supplied checksum of _all_
Linus Torvalds1da177e2005-04-16 15:20:36 -070060 * the packet as seen by netif_rx in skb->csum.
61 * NOTE: Even if device supports only some protocols, but
Patrick McHardy84fa7932006-08-29 16:44:56 -070062 * is able to produce some skb->csum, it MUST use COMPLETE,
Linus Torvalds1da177e2005-04-16 15:20:36 -070063 * not UNNECESSARY.
64 *
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070065 * PARTIAL: identical to the case for output below. This may occur
66 * on a packet received directly from another Linux OS, e.g.,
67 * a virtualised Linux kernel on the same host. The packet can
68 * be treated in the same way as UNNECESSARY except that on
69 * output (i.e., forwarding) the checksum must be filled in
70 * by the OS or the hardware.
71 *
Linus Torvalds1da177e2005-04-16 15:20:36 -070072 * B. Checksumming on output.
73 *
74 * NONE: skb is checksummed by protocol or csum is not required.
75 *
Patrick McHardy84fa7932006-08-29 16:44:56 -070076 * PARTIAL: device is required to csum packet as seen by hard_start_xmit
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070077 * from skb->csum_start to the end and to record the checksum
78 * at skb->csum_start + skb->csum_offset.
Linus Torvalds1da177e2005-04-16 15:20:36 -070079 *
80 * Device must show its capabilities in dev->features, set
81 * at device setup time.
82 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
83 * everything.
84 * NETIF_F_NO_CSUM - loopback or reliable single hop media.
85 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
86 * TCP/UDP over IPv4. Sigh. Vendors like this
87 * way by an unknown reason. Though, see comment above
88 * about CHECKSUM_UNNECESSARY. 8)
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070089 * NETIF_F_IPV6_CSUM about as dumb as the last one but does IPv6 instead.
Linus Torvalds1da177e2005-04-16 15:20:36 -070090 *
91 * Any questions? No questions, good. --ANK
92 */
93
Linus Torvalds1da177e2005-04-16 15:20:36 -070094struct net_device;
David Howells716ea3a2007-04-02 20:19:53 -070095struct scatterlist;
Jens Axboe9c55e012007-11-06 23:30:13 -080096struct pipe_inode_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -070097
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -070098#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Linus Torvalds1da177e2005-04-16 15:20:36 -070099struct nf_conntrack {
100 atomic_t use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101};
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700102#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103
104#ifdef CONFIG_BRIDGE_NETFILTER
105struct nf_bridge_info {
106 atomic_t use;
107 struct net_device *physindev;
108 struct net_device *physoutdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700109 unsigned int mask;
110 unsigned long data[32 / sizeof(unsigned long)];
111};
112#endif
113
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114struct sk_buff_head {
115 /* These two members must be first. */
116 struct sk_buff *next;
117 struct sk_buff *prev;
118
119 __u32 qlen;
120 spinlock_t lock;
121};
122
123struct sk_buff;
124
125/* To allow 64K frame to be packed as single skb without frag_list */
126#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
127
128typedef struct skb_frag_struct skb_frag_t;
129
130struct skb_frag_struct {
131 struct page *page;
David S. Millera309bb02007-07-30 18:47:03 -0700132 __u32 page_offset;
133 __u32 size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134};
135
Patrick Ohlyac45f602009-02-12 05:03:37 +0000136#define HAVE_HW_TIME_STAMP
137
138/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800139 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000140 * @hwtstamp: hardware time stamp transformed into duration
141 * since arbitrary point in time
142 * @syststamp: hwtstamp transformed to system time base
143 *
144 * Software time stamps generated by ktime_get_real() are stored in
145 * skb->tstamp. The relation between the different kinds of time
146 * stamps is as follows:
147 *
148 * syststamp and tstamp can be compared against each other in
149 * arbitrary combinations. The accuracy of a
150 * syststamp/tstamp/"syststamp from other device" comparison is
151 * limited by the accuracy of the transformation into system time
152 * base. This depends on the device driver and its underlying
153 * hardware.
154 *
155 * hwtstamps can only be compared against other hwtstamps from
156 * the same device.
157 *
158 * This structure is attached to packets as part of the
159 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
160 */
161struct skb_shared_hwtstamps {
162 ktime_t hwtstamp;
163 ktime_t syststamp;
164};
165
166/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800167 * struct skb_shared_tx - instructions for time stamping of outgoing packets
Patrick Ohlyac45f602009-02-12 05:03:37 +0000168 * @hardware: generate hardware time stamp
169 * @software: generate software time stamp
170 * @in_progress: device driver is going to provide
171 * hardware time stamp
Randy Dunlap4b21cd42009-03-28 23:38:40 -0700172 * @flags: all shared_tx flags
Patrick Ohlyac45f602009-02-12 05:03:37 +0000173 *
174 * These flags are attached to packets as part of the
175 * &skb_shared_info. Use skb_tx() to get a pointer.
176 */
177union skb_shared_tx {
178 struct {
179 __u8 hardware:1,
180 software:1,
181 in_progress:1;
182 };
183 __u8 flags;
184};
185
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186/* This data is invariant across clones and lives at
187 * the end of the header data, ie. at skb->end.
188 */
189struct skb_shared_info {
Benjamin LaHaise4947d3e2006-01-03 14:06:50 -0800190 unsigned short nr_frags;
Herbert Xu79671682006-06-22 02:40:14 -0700191 unsigned short gso_size;
192 /* Warning: this field is not always filled in (UFO)! */
193 unsigned short gso_segs;
194 unsigned short gso_type;
Al Viroae08e1f2006-11-08 00:27:11 -0800195 __be32 ip6_frag_id;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000196 union skb_shared_tx tx_flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000198 struct skb_shared_hwtstamps hwtstamps;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700199
200 /*
201 * Warning : all fields before dataref are cleared in __alloc_skb()
202 */
203 atomic_t dataref;
204
Johann Baudy69e3c752009-05-18 22:11:22 -0700205 /* Intermediate layers must ensure that destructor_arg
206 * remains valid until skb destructor */
207 void * destructor_arg;
Eric Dumazetfed66382010-07-22 19:09:08 +0000208 /* must be last field, see pskb_expand_head() */
209 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210};
211
212/* We divide dataref into two halves. The higher 16 bits hold references
213 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700214 * the entire skb->data. A clone of a headerless skb holds the length of
215 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216 *
217 * All users must obey the rule that the skb->data reference count must be
218 * greater than or equal to the payload reference count.
219 *
220 * Holding a reference to the payload part means that the user does not
221 * care about modifications to the header part of skb->data.
222 */
223#define SKB_DATAREF_SHIFT 16
224#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
225
David S. Millerd179cd12005-08-17 14:57:30 -0700226
227enum {
228 SKB_FCLONE_UNAVAILABLE,
229 SKB_FCLONE_ORIG,
230 SKB_FCLONE_CLONE,
231};
232
Herbert Xu79671682006-06-22 02:40:14 -0700233enum {
234 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700235 SKB_GSO_UDP = 1 << 1,
Herbert Xu576a30e2006-06-27 13:22:38 -0700236
237 /* This indicates the skb is from an untrusted source. */
238 SKB_GSO_DODGY = 1 << 2,
Michael Chanb0da85372006-06-29 12:30:00 -0700239
240 /* This indicates the tcp segment has CWR set. */
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700241 SKB_GSO_TCP_ECN = 1 << 3,
242
243 SKB_GSO_TCPV6 = 1 << 4,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800244
245 SKB_GSO_FCOE = 1 << 5,
Herbert Xu79671682006-06-22 02:40:14 -0700246};
247
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700248#if BITS_PER_LONG > 32
249#define NET_SKBUFF_DATA_USES_OFFSET 1
250#endif
251
252#ifdef NET_SKBUFF_DATA_USES_OFFSET
253typedef unsigned int sk_buff_data_t;
254#else
255typedef unsigned char *sk_buff_data_t;
256#endif
257
Linus Torvalds1da177e2005-04-16 15:20:36 -0700258/**
259 * struct sk_buff - socket buffer
260 * @next: Next buffer in list
261 * @prev: Previous buffer in list
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262 * @sk: Socket we are owned by
Herbert Xu325ed822005-10-03 13:57:23 -0700263 * @tstamp: Time we arrived
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264 * @dev: Device we arrived on/are leaving by
Randy Dunlapbe52178b2007-05-03 03:16:20 -0700265 * @transport_header: Transport layer header
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -0700266 * @network_header: Network layer header
267 * @mac_header: Link layer header
Eric Dumazet7fee2262010-05-11 23:19:48 +0000268 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700269 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270 * @cb: Control buffer. Free for use by every layer. Put private vars here
271 * @len: Length of actual data
272 * @data_len: Data length
273 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700274 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700275 * @csum: Checksum (must include start/offset pair)
276 * @csum_start: Offset from skb->head where checksumming should start
277 * @csum_offset: Offset from csum_start where checksum should be stored
Martin Waitz67be2dd2005-05-01 08:59:26 -0700278 * @local_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279 * @cloned: Head may be cloned (check refcnt to be sure)
280 * @nohdr: Payload reference only, must not modify header
281 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700282 * @fclone: skbuff clone status
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283 * @ip_summed: Driver fed us an IP checksum
284 * @priority: Packet queueing priority
285 * @users: User count - see {datagram,tcp}.c
286 * @protocol: Packet protocol from driver
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287 * @truesize: Buffer size
288 * @head: Head of buffer
289 * @data: Data head pointer
290 * @tail: Tail pointer
291 * @end: End pointer
292 * @destructor: Destruct function
Thomas Graf82e91ff2006-11-09 15:19:14 -0800293 * @mark: Generic packet mark
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294 * @nfct: Associated connection, if any
Randy Dunlapc83c2482005-10-18 22:07:41 -0700295 * @ipvs_property: skbuff is owned by ipvs
Randy Dunlap31729362008-02-18 20:52:13 -0800296 * @peeked: this packet has been seen already, so stats have been
297 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700298 * @nf_trace: netfilter packet trace flag
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299 * @nfctinfo: Relationship of this skb to the connection
Randy Dunlap461ddf32005-11-20 21:25:15 -0800300 * @nfct_reasm: netfilter conntrack re-assembly pointer
Linus Torvalds1da177e2005-04-16 15:20:36 -0700301 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800302 * @skb_iif: ifindex of device we arrived on
Tom Herbert0a9627f2010-03-16 08:03:29 +0000303 * @rxhash: the packet hash computed on receive
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700304 * @queue_mapping: Queue mapping for multiqueue devices
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305 * @tc_index: Traffic control index
306 * @tc_verd: traffic control verdict
Randy Dunlap553a5672008-04-20 10:51:01 -0700307 * @ndisc_nodetype: router type (from link layer)
Randy Dunlapf4b8ea72006-06-22 16:00:11 -0700308 * @dma_cookie: a cookie to one of several possible DMA operations
309 * done by skb DMA functions
James Morris984bc162006-06-09 00:29:17 -0700310 * @secmark: security marking
Patrick McHardy6aa895b02008-07-14 22:49:06 -0700311 * @vlan_tci: vlan tag control information
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312 */
313
314struct sk_buff {
315 /* These two members must be first. */
316 struct sk_buff *next;
317 struct sk_buff *prev;
318
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700319 ktime_t tstamp;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000320
321 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324 /*
325 * This is the control buffer. It is free to use for every
326 * layer. Please put your private variables there. If you
327 * want to keep them across layers you have to do a skb_clone()
328 * first. This is owned by whoever has the skb queued ATM.
329 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000330 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331
Eric Dumazet7fee2262010-05-11 23:19:48 +0000332 unsigned long _skb_refdst;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000333#ifdef CONFIG_XFRM
334 struct sec_path *sp;
335#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700337 data_len;
338 __u16 mac_len,
339 hdr_len;
Al Viroff1dcad2006-11-20 18:07:29 -0800340 union {
341 __wsum csum;
Herbert Xu663ead32007-04-09 11:59:07 -0700342 struct {
343 __u16 csum_start;
344 __u16 csum_offset;
345 };
Al Viroff1dcad2006-11-20 18:07:29 -0800346 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347 __u32 priority;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200348 kmemcheck_bitfield_begin(flags1);
Thomas Graf1cbb3382005-07-05 14:13:41 -0700349 __u8 local_df:1,
350 cloned:1,
351 ip_summed:2,
Harald Welte6869c4d2005-08-09 19:24:19 -0700352 nohdr:1,
353 nfctinfo:3;
David S. Millerd179cd12005-08-17 14:57:30 -0700354 __u8 pkt_type:3,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800355 fclone:2,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700356 ipvs_property:1,
Herbert Xua59322b2007-12-05 01:53:40 -0800357 peeked:1,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700358 nf_trace:1;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200359 kmemcheck_bitfield_end(flags1);
Eric Dumazet4ab408d2010-03-01 03:09:26 +0000360 __be16 protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361
362 void (*destructor)(struct sk_buff *skb);
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800363#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700364 struct nf_conntrack *nfct;
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800365 struct sk_buff *nfct_reasm;
366#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367#ifdef CONFIG_BRIDGE_NETFILTER
368 struct nf_bridge_info *nf_bridge;
369#endif
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700370
Eric Dumazet8964be42009-11-20 15:35:04 -0800371 int skb_iif;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700373 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700375 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700377#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200378
Tom Herbert0a9627f2010-03-16 08:03:29 +0000379 __u32 rxhash;
380
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200381 kmemcheck_bitfield_begin(flags2);
Eric Dumazet14d18a82009-10-29 00:10:37 +0000382 __u16 queue_mapping:16;
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400383#ifdef CONFIG_IPV6_NDISC_NODETYPE
John Fastabend597a2642010-06-03 09:30:11 +0000384 __u8 ndisc_nodetype:2,
385 deliver_no_wcard:1;
386#else
387 __u8 deliver_no_wcard:1;
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400388#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200389 kmemcheck_bitfield_end(flags2);
390
Johannes Berg72bce622009-06-17 17:45:28 +0200391 /* 0/14 bit hole */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700392
Chris Leech97fc2f02006-05-23 17:55:33 -0700393#ifdef CONFIG_NET_DMA
394 dma_cookie_t dma_cookie;
395#endif
James Morris984bc162006-06-09 00:29:17 -0700396#ifdef CONFIG_NETWORK_SECMARK
397 __u32 secmark;
398#endif
Neil Horman3b885782009-10-12 13:26:31 -0700399 union {
400 __u32 mark;
401 __u32 dropcount;
402 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403
Patrick McHardy6aa895b02008-07-14 22:49:06 -0700404 __u16 vlan_tci;
405
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700406 sk_buff_data_t transport_header;
407 sk_buff_data_t network_header;
408 sk_buff_data_t mac_header;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700410 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700411 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700413 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700414 unsigned int truesize;
415 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416};
417
418#ifdef __KERNEL__
419/*
420 * Handling routines are only of interest to the kernel
421 */
422#include <linux/slab.h>
423
424#include <asm/system.h>
425
Eric Dumazet7fee2262010-05-11 23:19:48 +0000426/*
427 * skb might have a dst pointer attached, refcounted or not.
428 * _skb_refdst low order bit is set if refcount was _not_ taken
429 */
430#define SKB_DST_NOREF 1UL
431#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
432
433/**
434 * skb_dst - returns skb dst_entry
435 * @skb: buffer
436 *
437 * Returns skb dst_entry, regardless of reference taken or not.
438 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000439static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
440{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000441 /* If refdst was not refcounted, check we still are in a
442 * rcu_read_lock section
443 */
444 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
445 !rcu_read_lock_held() &&
446 !rcu_read_lock_bh_held());
447 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000448}
449
Eric Dumazet7fee2262010-05-11 23:19:48 +0000450/**
451 * skb_dst_set - sets skb dst
452 * @skb: buffer
453 * @dst: dst entry
454 *
455 * Sets skb dst, assuming a reference was taken on dst and should
456 * be released by skb_dst_drop()
457 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000458static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
459{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000460 skb->_skb_refdst = (unsigned long)dst;
461}
462
463/**
464 * skb_dst_set_noref - sets skb dst, without a reference
465 * @skb: buffer
466 * @dst: dst entry
467 *
468 * Sets skb dst, assuming a reference was not taken on dst
469 * skb_dst_drop() should not dst_release() this dst
470 */
471static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
472{
473 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
474 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
475}
476
477/**
478 * skb_dst_is_noref - Test if skb dst isnt refcounted
479 * @skb: buffer
480 */
481static inline bool skb_dst_is_noref(const struct sk_buff *skb)
482{
483 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000484}
485
Eric Dumazet511c3f92009-06-02 05:14:27 +0000486static inline struct rtable *skb_rtable(const struct sk_buff *skb)
487{
Eric Dumazetadf30902009-06-02 05:19:30 +0000488 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000489}
490
Jörn Engel231d06a2006-03-20 21:28:35 -0800491extern void kfree_skb(struct sk_buff *skb);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000492extern void consume_skb(struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493extern void __kfree_skb(struct sk_buff *skb);
David S. Millerd179cd12005-08-17 14:57:30 -0700494extern struct sk_buff *__alloc_skb(unsigned int size,
Christoph Hellwigb30973f2006-12-06 20:32:36 -0800495 gfp_t priority, int fclone, int node);
David S. Millerd179cd12005-08-17 14:57:30 -0700496static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100497 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700498{
Christoph Hellwigb30973f2006-12-06 20:32:36 -0800499 return __alloc_skb(size, priority, 0, -1);
David S. Millerd179cd12005-08-17 14:57:30 -0700500}
501
502static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100503 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700504{
Christoph Hellwigb30973f2006-12-06 20:32:36 -0800505 return __alloc_skb(size, priority, 1, -1);
David S. Millerd179cd12005-08-17 14:57:30 -0700506}
507
Changli Gao5b0daa32010-05-29 00:12:13 -0700508extern bool skb_recycle_check(struct sk_buff *skb, int skb_size);
Lennert Buytenhek04a4bb52008-10-01 02:33:12 -0700509
Herbert Xue0053ec2007-10-14 00:37:52 -0700510extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700511extern struct sk_buff *skb_clone(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100512 gfp_t priority);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700513extern struct sk_buff *skb_copy(const struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100514 gfp_t priority);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700515extern struct sk_buff *pskb_copy(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100516 gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517extern int pskb_expand_head(struct sk_buff *skb,
Victor Fusco86a76ca2005-07-08 14:57:47 -0700518 int nhead, int ntail,
Al Virodd0fc662005-10-07 07:46:04 +0100519 gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
521 unsigned int headroom);
522extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
523 int newheadroom, int newtailroom,
Al Virodd0fc662005-10-07 07:46:04 +0100524 gfp_t priority);
David Howells716ea3a2007-04-02 20:19:53 -0700525extern int skb_to_sgvec(struct sk_buff *skb,
526 struct scatterlist *sg, int offset,
527 int len);
528extern int skb_cow_data(struct sk_buff *skb, int tailbits,
529 struct sk_buff **trailer);
Herbert Xu5b057c62006-06-23 02:06:41 -0700530extern int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000531#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700533extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
534 int getfrag(void *from, char *to, int offset,
535 int len,int odd, struct sk_buff *skb),
536 void *from, int length);
537
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800538struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700539 __u32 lower_offset;
540 __u32 upper_offset;
541 __u32 frag_idx;
542 __u32 stepped_offset;
543 struct sk_buff *root_skb;
544 struct sk_buff *cur_skb;
545 __u8 *frag_data;
546};
547
548extern void skb_prepare_seq_read(struct sk_buff *skb,
549 unsigned int from, unsigned int to,
550 struct skb_seq_state *st);
551extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
552 struct skb_seq_state *st);
553extern void skb_abort_seq_read(struct skb_seq_state *st);
554
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700555extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
556 unsigned int to, struct ts_config *config,
557 struct ts_state *state);
558
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700559#ifdef NET_SKBUFF_DATA_USES_OFFSET
560static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
561{
562 return skb->head + skb->end;
563}
564#else
565static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
566{
567 return skb->end;
568}
569#endif
570
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700572#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700573
Patrick Ohlyac45f602009-02-12 05:03:37 +0000574static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
575{
576 return &skb_shinfo(skb)->hwtstamps;
577}
578
579static inline union skb_shared_tx *skb_tx(struct sk_buff *skb)
580{
581 return &skb_shinfo(skb)->tx_flags;
582}
583
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584/**
585 * skb_queue_empty - check if a queue is empty
586 * @list: queue head
587 *
588 * Returns true if the queue is empty, false otherwise.
589 */
590static inline int skb_queue_empty(const struct sk_buff_head *list)
591{
592 return list->next == (struct sk_buff *)list;
593}
594
595/**
David S. Millerfc7ebb22008-09-23 00:34:07 -0700596 * skb_queue_is_last - check if skb is the last entry in the queue
597 * @list: queue head
598 * @skb: buffer
599 *
600 * Returns true if @skb is the last buffer on the list.
601 */
602static inline bool skb_queue_is_last(const struct sk_buff_head *list,
603 const struct sk_buff *skb)
604{
605 return (skb->next == (struct sk_buff *) list);
606}
607
608/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800609 * skb_queue_is_first - check if skb is the first entry in the queue
610 * @list: queue head
611 * @skb: buffer
612 *
613 * Returns true if @skb is the first buffer on the list.
614 */
615static inline bool skb_queue_is_first(const struct sk_buff_head *list,
616 const struct sk_buff *skb)
617{
618 return (skb->prev == (struct sk_buff *) list);
619}
620
621/**
David S. Miller249c8b42008-09-23 00:44:42 -0700622 * skb_queue_next - return the next packet in the queue
623 * @list: queue head
624 * @skb: current buffer
625 *
626 * Return the next packet in @list after @skb. It is only valid to
627 * call this if skb_queue_is_last() evaluates to false.
628 */
629static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
630 const struct sk_buff *skb)
631{
632 /* This BUG_ON may seem severe, but if we just return then we
633 * are going to dereference garbage.
634 */
635 BUG_ON(skb_queue_is_last(list, skb));
636 return skb->next;
637}
638
639/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800640 * skb_queue_prev - return the prev packet in the queue
641 * @list: queue head
642 * @skb: current buffer
643 *
644 * Return the prev packet in @list before @skb. It is only valid to
645 * call this if skb_queue_is_first() evaluates to false.
646 */
647static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
648 const struct sk_buff *skb)
649{
650 /* This BUG_ON may seem severe, but if we just return then we
651 * are going to dereference garbage.
652 */
653 BUG_ON(skb_queue_is_first(list, skb));
654 return skb->prev;
655}
656
657/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658 * skb_get - reference buffer
659 * @skb: buffer to reference
660 *
661 * Makes another reference to a socket buffer and returns a pointer
662 * to the buffer.
663 */
664static inline struct sk_buff *skb_get(struct sk_buff *skb)
665{
666 atomic_inc(&skb->users);
667 return skb;
668}
669
670/*
671 * If users == 1, we are the only owner and are can avoid redundant
672 * atomic change.
673 */
674
675/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700676 * skb_cloned - is the buffer a clone
677 * @skb: buffer to check
678 *
679 * Returns true if the buffer was generated with skb_clone() and is
680 * one of multiple shared copies of the buffer. Cloned buffers are
681 * shared data so must not be written to under normal circumstances.
682 */
683static inline int skb_cloned(const struct sk_buff *skb)
684{
685 return skb->cloned &&
686 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
687}
688
689/**
690 * skb_header_cloned - is the header a clone
691 * @skb: buffer to check
692 *
693 * Returns true if modifying the header part of the buffer requires
694 * the data to be copied.
695 */
696static inline int skb_header_cloned(const struct sk_buff *skb)
697{
698 int dataref;
699
700 if (!skb->cloned)
701 return 0;
702
703 dataref = atomic_read(&skb_shinfo(skb)->dataref);
704 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
705 return dataref != 1;
706}
707
708/**
709 * skb_header_release - release reference to header
710 * @skb: buffer to operate on
711 *
712 * Drop a reference to the header part of the buffer. This is done
713 * by acquiring a payload reference. You must not read from the header
714 * part of skb->data after this.
715 */
716static inline void skb_header_release(struct sk_buff *skb)
717{
718 BUG_ON(skb->nohdr);
719 skb->nohdr = 1;
720 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
721}
722
723/**
724 * skb_shared - is the buffer shared
725 * @skb: buffer to check
726 *
727 * Returns true if more than one person has a reference to this
728 * buffer.
729 */
730static inline int skb_shared(const struct sk_buff *skb)
731{
732 return atomic_read(&skb->users) != 1;
733}
734
735/**
736 * skb_share_check - check if buffer is shared and if so clone it
737 * @skb: buffer to check
738 * @pri: priority for memory allocation
739 *
740 * If the buffer is shared the buffer is cloned and the old copy
741 * drops a reference. A new clone with a single reference is returned.
742 * If the buffer is not shared the original buffer is returned. When
743 * being called from interrupt status or with spinlocks held pri must
744 * be GFP_ATOMIC.
745 *
746 * NULL is returned on a memory allocation failure.
747 */
Victor Fusco86a76ca2005-07-08 14:57:47 -0700748static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100749 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750{
751 might_sleep_if(pri & __GFP_WAIT);
752 if (skb_shared(skb)) {
753 struct sk_buff *nskb = skb_clone(skb, pri);
754 kfree_skb(skb);
755 skb = nskb;
756 }
757 return skb;
758}
759
760/*
761 * Copy shared buffers into a new sk_buff. We effectively do COW on
762 * packets to handle cases where we have a local reader and forward
763 * and a couple of other messy ones. The normal one is tcpdumping
764 * a packet thats being forwarded.
765 */
766
767/**
768 * skb_unshare - make a copy of a shared buffer
769 * @skb: buffer to check
770 * @pri: priority for memory allocation
771 *
772 * If the socket buffer is a clone then this function creates a new
773 * copy of the data, drops a reference count on the old copy and returns
774 * the new copy with the reference count at 1. If the buffer is not a clone
775 * the original buffer is returned. When called with a spinlock held or
776 * from interrupt state @pri must be %GFP_ATOMIC
777 *
778 * %NULL is returned on a memory allocation failure.
779 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -0700780static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100781 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700782{
783 might_sleep_if(pri & __GFP_WAIT);
784 if (skb_cloned(skb)) {
785 struct sk_buff *nskb = skb_copy(skb, pri);
786 kfree_skb(skb); /* Free our shared copy */
787 skb = nskb;
788 }
789 return skb;
790}
791
792/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -0800793 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794 * @list_: list to peek at
795 *
796 * Peek an &sk_buff. Unlike most other operations you _MUST_
797 * be careful with this one. A peek leaves the buffer on the
798 * list and someone else may run off with it. You must hold
799 * the appropriate locks or have a private queue to do this.
800 *
801 * Returns %NULL for an empty list or a pointer to the head element.
802 * The reference count is not incremented and the reference is therefore
803 * volatile. Use with caution.
804 */
805static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
806{
807 struct sk_buff *list = ((struct sk_buff *)list_)->next;
808 if (list == (struct sk_buff *)list_)
809 list = NULL;
810 return list;
811}
812
813/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -0800814 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815 * @list_: list to peek at
816 *
817 * Peek an &sk_buff. Unlike most other operations you _MUST_
818 * be careful with this one. A peek leaves the buffer on the
819 * list and someone else may run off with it. You must hold
820 * the appropriate locks or have a private queue to do this.
821 *
822 * Returns %NULL for an empty list or a pointer to the tail element.
823 * The reference count is not incremented and the reference is therefore
824 * volatile. Use with caution.
825 */
826static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
827{
828 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
829 if (list == (struct sk_buff *)list_)
830 list = NULL;
831 return list;
832}
833
834/**
835 * skb_queue_len - get queue length
836 * @list_: list to measure
837 *
838 * Return the length of an &sk_buff queue.
839 */
840static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
841{
842 return list_->qlen;
843}
844
David S. Miller67fed452008-09-21 22:36:24 -0700845/**
846 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
847 * @list: queue to initialize
848 *
849 * This initializes only the list and queue length aspects of
850 * an sk_buff_head object. This allows to initialize the list
851 * aspects of an sk_buff_head without reinitializing things like
852 * the spinlock. It can also be used for on-stack sk_buff_head
853 * objects where the spinlock is known to not be used.
854 */
855static inline void __skb_queue_head_init(struct sk_buff_head *list)
856{
857 list->prev = list->next = (struct sk_buff *)list;
858 list->qlen = 0;
859}
860
Arjan van de Ven76f10ad2006-08-02 14:06:55 -0700861/*
862 * This function creates a split out lock class for each invocation;
863 * this is needed for now since a whole lot of users of the skb-queue
864 * infrastructure in drivers have different locking usage (in hardirq)
865 * than the networking core (in softirq only). In the long run either the
866 * network layer or drivers should need annotation to consolidate the
867 * main types of usage into 3 classes.
868 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869static inline void skb_queue_head_init(struct sk_buff_head *list)
870{
871 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -0700872 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700873}
874
Pavel Emelianovc2ecba72007-04-17 12:45:31 -0700875static inline void skb_queue_head_init_class(struct sk_buff_head *list,
876 struct lock_class_key *class)
877{
878 skb_queue_head_init(list);
879 lockdep_set_class(&list->lock, class);
880}
881
Linus Torvalds1da177e2005-04-16 15:20:36 -0700882/*
Gerrit Renkerbf299272008-04-14 00:04:51 -0700883 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884 *
885 * The "__skb_xxxx()" functions are the non-atomic ones that
886 * can only be called with interrupts disabled.
887 */
Gerrit Renkerbf299272008-04-14 00:04:51 -0700888extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
889static inline void __skb_insert(struct sk_buff *newsk,
890 struct sk_buff *prev, struct sk_buff *next,
891 struct sk_buff_head *list)
892{
893 newsk->next = next;
894 newsk->prev = prev;
895 next->prev = prev->next = newsk;
896 list->qlen++;
897}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898
David S. Miller67fed452008-09-21 22:36:24 -0700899static inline void __skb_queue_splice(const struct sk_buff_head *list,
900 struct sk_buff *prev,
901 struct sk_buff *next)
902{
903 struct sk_buff *first = list->next;
904 struct sk_buff *last = list->prev;
905
906 first->prev = prev;
907 prev->next = first;
908
909 last->next = next;
910 next->prev = last;
911}
912
913/**
914 * skb_queue_splice - join two skb lists, this is designed for stacks
915 * @list: the new list to add
916 * @head: the place to add it in the first list
917 */
918static inline void skb_queue_splice(const struct sk_buff_head *list,
919 struct sk_buff_head *head)
920{
921 if (!skb_queue_empty(list)) {
922 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -0700923 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -0700924 }
925}
926
927/**
928 * skb_queue_splice - join two skb lists and reinitialise the emptied list
929 * @list: the new list to add
930 * @head: the place to add it in the first list
931 *
932 * The list at @list is reinitialised
933 */
934static inline void skb_queue_splice_init(struct sk_buff_head *list,
935 struct sk_buff_head *head)
936{
937 if (!skb_queue_empty(list)) {
938 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -0700939 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -0700940 __skb_queue_head_init(list);
941 }
942}
943
944/**
945 * skb_queue_splice_tail - join two skb lists, each list being a queue
946 * @list: the new list to add
947 * @head: the place to add it in the first list
948 */
949static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
950 struct sk_buff_head *head)
951{
952 if (!skb_queue_empty(list)) {
953 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -0700954 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -0700955 }
956}
957
958/**
959 * skb_queue_splice_tail - join two skb lists and reinitialise the emptied list
960 * @list: the new list to add
961 * @head: the place to add it in the first list
962 *
963 * Each of the lists is a queue.
964 * The list at @list is reinitialised
965 */
966static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
967 struct sk_buff_head *head)
968{
969 if (!skb_queue_empty(list)) {
970 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -0700971 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -0700972 __skb_queue_head_init(list);
973 }
974}
975
Linus Torvalds1da177e2005-04-16 15:20:36 -0700976/**
Stephen Hemminger300ce172005-10-30 13:47:34 -0800977 * __skb_queue_after - queue a buffer at the list head
978 * @list: list to use
979 * @prev: place after this buffer
980 * @newsk: buffer to queue
981 *
982 * Queue a buffer int the middle of a list. This function takes no locks
983 * and you must therefore hold required locks before calling it.
984 *
985 * A buffer cannot be placed on two lists at the same time.
986 */
987static inline void __skb_queue_after(struct sk_buff_head *list,
988 struct sk_buff *prev,
989 struct sk_buff *newsk)
990{
Gerrit Renkerbf299272008-04-14 00:04:51 -0700991 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -0800992}
993
Gerrit Renker7de6c032008-04-14 00:05:09 -0700994extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
995 struct sk_buff_head *list);
996
Gerrit Renkerf5572852008-04-14 00:05:28 -0700997static inline void __skb_queue_before(struct sk_buff_head *list,
998 struct sk_buff *next,
999 struct sk_buff *newsk)
1000{
1001 __skb_insert(newsk, next->prev, next, list);
1002}
1003
Stephen Hemminger300ce172005-10-30 13:47:34 -08001004/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005 * __skb_queue_head - queue a buffer at the list head
1006 * @list: list to use
1007 * @newsk: buffer to queue
1008 *
1009 * Queue a buffer at the start of a list. This function takes no locks
1010 * and you must therefore hold required locks before calling it.
1011 *
1012 * A buffer cannot be placed on two lists at the same time.
1013 */
1014extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1015static inline void __skb_queue_head(struct sk_buff_head *list,
1016 struct sk_buff *newsk)
1017{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001018 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019}
1020
1021/**
1022 * __skb_queue_tail - queue a buffer at the list tail
1023 * @list: list to use
1024 * @newsk: buffer to queue
1025 *
1026 * Queue a buffer at the end of a list. This function takes no locks
1027 * and you must therefore hold required locks before calling it.
1028 *
1029 * A buffer cannot be placed on two lists at the same time.
1030 */
1031extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1032static inline void __skb_queue_tail(struct sk_buff_head *list,
1033 struct sk_buff *newsk)
1034{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001035 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036}
1037
Linus Torvalds1da177e2005-04-16 15:20:36 -07001038/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001039 * remove sk_buff from list. _Must_ be called atomically, and with
1040 * the list known..
1041 */
David S. Miller8728b832005-08-09 19:25:21 -07001042extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1044{
1045 struct sk_buff *next, *prev;
1046
1047 list->qlen--;
1048 next = skb->next;
1049 prev = skb->prev;
1050 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051 next->prev = prev;
1052 prev->next = next;
1053}
1054
Gerrit Renkerf525c062008-04-14 00:04:12 -07001055/**
1056 * __skb_dequeue - remove from the head of the queue
1057 * @list: list to dequeue from
1058 *
1059 * Remove the head of the list. This function does not take any locks
1060 * so must be used with appropriate locks held only. The head item is
1061 * returned or %NULL if the list is empty.
1062 */
1063extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1064static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1065{
1066 struct sk_buff *skb = skb_peek(list);
1067 if (skb)
1068 __skb_unlink(skb, list);
1069 return skb;
1070}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071
1072/**
1073 * __skb_dequeue_tail - remove from the tail of the queue
1074 * @list: list to dequeue from
1075 *
1076 * Remove the tail of the list. This function does not take any locks
1077 * so must be used with appropriate locks held only. The tail item is
1078 * returned or %NULL if the list is empty.
1079 */
1080extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1081static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1082{
1083 struct sk_buff *skb = skb_peek_tail(list);
1084 if (skb)
1085 __skb_unlink(skb, list);
1086 return skb;
1087}
1088
1089
1090static inline int skb_is_nonlinear(const struct sk_buff *skb)
1091{
1092 return skb->data_len;
1093}
1094
1095static inline unsigned int skb_headlen(const struct sk_buff *skb)
1096{
1097 return skb->len - skb->data_len;
1098}
1099
1100static inline int skb_pagelen(const struct sk_buff *skb)
1101{
1102 int i, len = 0;
1103
1104 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1105 len += skb_shinfo(skb)->frags[i].size;
1106 return len + skb_headlen(skb);
1107}
1108
1109static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1110 struct page *page, int off, int size)
1111{
1112 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1113
1114 frag->page = page;
1115 frag->page_offset = off;
1116 frag->size = size;
1117 skb_shinfo(skb)->nr_frags = i + 1;
1118}
1119
Peter Zijlstra654bed12008-10-07 14:22:33 -07001120extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
1121 int off, int size);
1122
Linus Torvalds1da177e2005-04-16 15:20:36 -07001123#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Milleree039872009-06-09 00:17:13 -07001124#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frags(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001125#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1126
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001127#ifdef NET_SKBUFF_DATA_USES_OFFSET
1128static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1129{
1130 return skb->head + skb->tail;
1131}
1132
1133static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1134{
1135 skb->tail = skb->data - skb->head;
1136}
1137
1138static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1139{
1140 skb_reset_tail_pointer(skb);
1141 skb->tail += offset;
1142}
1143#else /* NET_SKBUFF_DATA_USES_OFFSET */
1144static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1145{
1146 return skb->tail;
1147}
1148
1149static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1150{
1151 skb->tail = skb->data;
1152}
1153
1154static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1155{
1156 skb->tail = skb->data + offset;
1157}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001158
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001159#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1160
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161/*
1162 * Add data to an sk_buff
1163 */
Ilpo Järvinen0dde3e12008-03-27 17:43:41 -07001164extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001165static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1166{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001167 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168 SKB_LINEAR_ASSERT(skb);
1169 skb->tail += len;
1170 skb->len += len;
1171 return tmp;
1172}
1173
Ilpo Järvinenc2aa2702008-03-27 17:52:40 -07001174extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1176{
1177 skb->data -= len;
1178 skb->len += len;
1179 return skb->data;
1180}
1181
Ilpo Järvinen6be8ac22008-03-27 17:47:24 -07001182extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1184{
1185 skb->len -= len;
1186 BUG_ON(skb->len < skb->data_len);
1187 return skb->data += len;
1188}
1189
David S. Miller47d29642010-05-02 02:21:44 -07001190static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1191{
1192 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1193}
1194
Linus Torvalds1da177e2005-04-16 15:20:36 -07001195extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1196
1197static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1198{
1199 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001200 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001201 return NULL;
1202 skb->len -= len;
1203 return skb->data += len;
1204}
1205
1206static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1207{
1208 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1209}
1210
1211static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1212{
1213 if (likely(len <= skb_headlen(skb)))
1214 return 1;
1215 if (unlikely(len > skb->len))
1216 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001217 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218}
1219
1220/**
1221 * skb_headroom - bytes at buffer head
1222 * @skb: buffer to check
1223 *
1224 * Return the number of bytes of free space at the head of an &sk_buff.
1225 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001226static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001227{
1228 return skb->data - skb->head;
1229}
1230
1231/**
1232 * skb_tailroom - bytes at buffer end
1233 * @skb: buffer to check
1234 *
1235 * Return the number of bytes of free space at the tail of an sk_buff
1236 */
1237static inline int skb_tailroom(const struct sk_buff *skb)
1238{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001239 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001240}
1241
1242/**
1243 * skb_reserve - adjust headroom
1244 * @skb: buffer to alter
1245 * @len: bytes to move
1246 *
1247 * Increase the headroom of an empty &sk_buff by reducing the tail
1248 * room. This is only allowed for an empty buffer.
1249 */
David S. Miller8243126c2006-01-17 02:54:21 -08001250static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001251{
1252 skb->data += len;
1253 skb->tail += len;
1254}
1255
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001256#ifdef NET_SKBUFF_DATA_USES_OFFSET
1257static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1258{
1259 return skb->head + skb->transport_header;
1260}
1261
1262static inline void skb_reset_transport_header(struct sk_buff *skb)
1263{
1264 skb->transport_header = skb->data - skb->head;
1265}
1266
1267static inline void skb_set_transport_header(struct sk_buff *skb,
1268 const int offset)
1269{
1270 skb_reset_transport_header(skb);
1271 skb->transport_header += offset;
1272}
1273
1274static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1275{
1276 return skb->head + skb->network_header;
1277}
1278
1279static inline void skb_reset_network_header(struct sk_buff *skb)
1280{
1281 skb->network_header = skb->data - skb->head;
1282}
1283
1284static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1285{
1286 skb_reset_network_header(skb);
1287 skb->network_header += offset;
1288}
1289
1290static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1291{
1292 return skb->head + skb->mac_header;
1293}
1294
1295static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1296{
1297 return skb->mac_header != ~0U;
1298}
1299
1300static inline void skb_reset_mac_header(struct sk_buff *skb)
1301{
1302 skb->mac_header = skb->data - skb->head;
1303}
1304
1305static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1306{
1307 skb_reset_mac_header(skb);
1308 skb->mac_header += offset;
1309}
1310
1311#else /* NET_SKBUFF_DATA_USES_OFFSET */
1312
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001313static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1314{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001315 return skb->transport_header;
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001316}
1317
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001318static inline void skb_reset_transport_header(struct sk_buff *skb)
1319{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001320 skb->transport_header = skb->data;
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001321}
1322
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001323static inline void skb_set_transport_header(struct sk_buff *skb,
1324 const int offset)
1325{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001326 skb->transport_header = skb->data + offset;
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001327}
1328
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001329static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1330{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001331 return skb->network_header;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001332}
1333
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001334static inline void skb_reset_network_header(struct sk_buff *skb)
1335{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001336 skb->network_header = skb->data;
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001337}
1338
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001339static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1340{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001341 skb->network_header = skb->data + offset;
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001342}
1343
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001344static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1345{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001346 return skb->mac_header;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001347}
1348
1349static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1350{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001351 return skb->mac_header != NULL;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001352}
1353
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001354static inline void skb_reset_mac_header(struct sk_buff *skb)
1355{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001356 skb->mac_header = skb->data;
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001357}
1358
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001359static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1360{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001361 skb->mac_header = skb->data + offset;
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001362}
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001363#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1364
1365static inline int skb_transport_offset(const struct sk_buff *skb)
1366{
1367 return skb_transport_header(skb) - skb->data;
1368}
1369
1370static inline u32 skb_network_header_len(const struct sk_buff *skb)
1371{
1372 return skb->transport_header - skb->network_header;
1373}
1374
1375static inline int skb_network_offset(const struct sk_buff *skb)
1376{
1377 return skb_network_header(skb) - skb->data;
1378}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001379
Linus Torvalds1da177e2005-04-16 15:20:36 -07001380/*
1381 * CPUs often take a performance hit when accessing unaligned memory
1382 * locations. The actual performance hit varies, it can be small if the
1383 * hardware handles it or large if we have to take an exception and fix it
1384 * in software.
1385 *
1386 * Since an ethernet header is 14 bytes network drivers often end up with
1387 * the IP header at an unaligned offset. The IP header can be aligned by
1388 * shifting the start of the packet by 2 bytes. Drivers should do this
1389 * with:
1390 *
Tobias Klauser8660c122009-07-13 22:48:16 +00001391 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392 *
1393 * The downside to this alignment of the IP header is that the DMA is now
1394 * unaligned. On some architectures the cost of an unaligned DMA is high
1395 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00001396 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001397 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1398 * to be overridden.
1399 */
1400#ifndef NET_IP_ALIGN
1401#define NET_IP_ALIGN 2
1402#endif
1403
Anton Blanchard025be812006-03-31 02:27:06 -08001404/*
1405 * The networking layer reserves some headroom in skb data (via
1406 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1407 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08001408 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08001409 *
1410 * Unfortunately this headroom changes the DMA alignment of the resulting
1411 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1412 * on some architectures. An architecture can override this value,
1413 * perhaps setting it to a cacheline in size (since that will maintain
1414 * cacheline alignment of the DMA). It must be a power of 2.
1415 *
David S. Millerd6301d32009-02-08 19:24:13 -08001416 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08001417 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07001418 *
1419 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1420 * to reduce average number of cache lines per packet.
1421 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07001422 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08001423 */
1424#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07001425#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08001426#endif
1427
Herbert Xu3cc0e872006-06-09 16:13:38 -07001428extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001429
1430static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1431{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001432 if (unlikely(skb->data_len)) {
1433 WARN_ON(1);
1434 return;
1435 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001436 skb->len = len;
1437 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001438}
1439
Ilpo Järvinen419ae742008-03-27 17:54:01 -07001440extern void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001441
1442static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1443{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001444 if (skb->data_len)
1445 return ___pskb_trim(skb, len);
1446 __skb_trim(skb, len);
1447 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001448}
1449
1450static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1451{
1452 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1453}
1454
1455/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07001456 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1457 * @skb: buffer to alter
1458 * @len: new length
1459 *
1460 * This is identical to pskb_trim except that the caller knows that
1461 * the skb is not cloned so we should never get an error due to out-
1462 * of-memory.
1463 */
1464static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1465{
1466 int err = pskb_trim(skb, len);
1467 BUG_ON(err);
1468}
1469
1470/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001471 * skb_orphan - orphan a buffer
1472 * @skb: buffer to orphan
1473 *
1474 * If a buffer currently has an owner then we call the owner's
1475 * destructor function and make the @skb unowned. The buffer continues
1476 * to exist but is no longer charged to its former owner.
1477 */
1478static inline void skb_orphan(struct sk_buff *skb)
1479{
1480 if (skb->destructor)
1481 skb->destructor(skb);
1482 skb->destructor = NULL;
1483 skb->sk = NULL;
1484}
1485
1486/**
1487 * __skb_queue_purge - empty a list
1488 * @list: list to empty
1489 *
1490 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1491 * the list and one reference dropped. This function does not take the
1492 * list lock and the caller must hold the relevant locks to use it.
1493 */
1494extern void skb_queue_purge(struct sk_buff_head *list);
1495static inline void __skb_queue_purge(struct sk_buff_head *list)
1496{
1497 struct sk_buff *skb;
1498 while ((skb = __skb_dequeue(list)) != NULL)
1499 kfree_skb(skb);
1500}
1501
1502/**
Christoph Hellwigb4e54de2006-07-24 15:31:14 -07001503 * __dev_alloc_skb - allocate an skbuff for receiving
Linus Torvalds1da177e2005-04-16 15:20:36 -07001504 * @length: length to allocate
1505 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1506 *
1507 * Allocate a new &sk_buff and assign it a usage count of one. The
1508 * buffer has unspecified headroom built in. Users should allocate
1509 * the headroom they think they need without accounting for the
1510 * built in space. The built in space is used for optimisations.
1511 *
Christoph Hellwig766ea8c2006-08-07 15:49:53 -07001512 * %NULL is returned if there is no free memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001513 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001514static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
Al Virodd0fc662005-10-07 07:46:04 +01001515 gfp_t gfp_mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001516{
Anton Blanchard025be812006-03-31 02:27:06 -08001517 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001518 if (likely(skb))
Anton Blanchard025be812006-03-31 02:27:06 -08001519 skb_reserve(skb, NET_SKB_PAD);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520 return skb;
1521}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001522
Ilpo Järvinenf58518e2008-03-27 17:51:31 -07001523extern struct sk_buff *dev_alloc_skb(unsigned int length);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524
Christoph Hellwig8af27452006-07-31 22:35:23 -07001525extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1526 unsigned int length, gfp_t gfp_mask);
1527
1528/**
1529 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1530 * @dev: network device to receive on
1531 * @length: length to allocate
1532 *
1533 * Allocate a new &sk_buff and assign it a usage count of one. The
1534 * buffer has unspecified headroom built in. Users should allocate
1535 * the headroom they think they need without accounting for the
1536 * built in space. The built in space is used for optimisations.
1537 *
1538 * %NULL is returned if there is no free memory. Although this function
1539 * allocates memory it can be called from an interrupt.
1540 */
1541static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1542 unsigned int length)
1543{
1544 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1545}
1546
Eric Dumazet61321bb2009-10-07 17:11:23 +00001547static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1548 unsigned int length)
1549{
1550 struct sk_buff *skb = netdev_alloc_skb(dev, length + NET_IP_ALIGN);
1551
1552 if (NET_IP_ALIGN && skb)
1553 skb_reserve(skb, NET_IP_ALIGN);
1554 return skb;
1555}
1556
Peter Zijlstra654bed12008-10-07 14:22:33 -07001557extern struct page *__netdev_alloc_page(struct net_device *dev, gfp_t gfp_mask);
1558
1559/**
1560 * netdev_alloc_page - allocate a page for ps-rx on a specific device
1561 * @dev: network device to receive on
1562 *
1563 * Allocate a new page node local to the specified device.
1564 *
1565 * %NULL is returned if there is no free memory.
1566 */
1567static inline struct page *netdev_alloc_page(struct net_device *dev)
1568{
1569 return __netdev_alloc_page(dev, GFP_ATOMIC);
1570}
1571
1572static inline void netdev_free_page(struct net_device *dev, struct page *page)
1573{
1574 __free_page(page);
1575}
1576
Linus Torvalds1da177e2005-04-16 15:20:36 -07001577/**
Patrick McHardy334a8132007-06-25 04:35:20 -07001578 * skb_clone_writable - is the header of a clone writable
1579 * @skb: buffer to check
1580 * @len: length up to which to write
1581 *
1582 * Returns true if modifying the header part of the cloned buffer
1583 * does not requires the data to be copied.
1584 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001585static inline int skb_clone_writable(struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07001586{
1587 return !skb_header_cloned(skb) &&
1588 skb_headroom(skb) + len <= skb->hdr_len;
1589}
1590
Herbert Xud9cc2042007-09-16 16:21:16 -07001591static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
1592 int cloned)
1593{
1594 int delta = 0;
1595
1596 if (headroom < NET_SKB_PAD)
1597 headroom = NET_SKB_PAD;
1598 if (headroom > skb_headroom(skb))
1599 delta = headroom - skb_headroom(skb);
1600
1601 if (delta || cloned)
1602 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
1603 GFP_ATOMIC);
1604 return 0;
1605}
1606
Patrick McHardy334a8132007-06-25 04:35:20 -07001607/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001608 * skb_cow - copy header of skb when it is required
1609 * @skb: buffer to cow
1610 * @headroom: needed headroom
1611 *
1612 * If the skb passed lacks sufficient headroom or its data part
1613 * is shared, data is reallocated. If reallocation fails, an error
1614 * is returned and original skb is not changed.
1615 *
1616 * The result is skb with writable area skb->head...skb->tail
1617 * and at least @headroom of space at head.
1618 */
1619static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1620{
Herbert Xud9cc2042007-09-16 16:21:16 -07001621 return __skb_cow(skb, headroom, skb_cloned(skb));
1622}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001623
Herbert Xud9cc2042007-09-16 16:21:16 -07001624/**
1625 * skb_cow_head - skb_cow but only making the head writable
1626 * @skb: buffer to cow
1627 * @headroom: needed headroom
1628 *
1629 * This function is identical to skb_cow except that we replace the
1630 * skb_cloned check by skb_header_cloned. It should be used when
1631 * you only need to push on some header and do not need to modify
1632 * the data.
1633 */
1634static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
1635{
1636 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001637}
1638
1639/**
1640 * skb_padto - pad an skbuff up to a minimal size
1641 * @skb: buffer to pad
1642 * @len: minimal length
1643 *
1644 * Pads up a buffer to ensure the trailing bytes exist and are
1645 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07001646 * is untouched. Otherwise it is extended. Returns zero on
1647 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001648 */
1649
Herbert Xu5b057c62006-06-23 02:06:41 -07001650static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001651{
1652 unsigned int size = skb->len;
1653 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07001654 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001655 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001656}
1657
1658static inline int skb_add_data(struct sk_buff *skb,
1659 char __user *from, int copy)
1660{
1661 const int off = skb->len;
1662
1663 if (skb->ip_summed == CHECKSUM_NONE) {
1664 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08001665 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
Linus Torvalds1da177e2005-04-16 15:20:36 -07001666 copy, 0, &err);
1667 if (!err) {
1668 skb->csum = csum_block_add(skb->csum, csum, off);
1669 return 0;
1670 }
1671 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1672 return 0;
1673
1674 __skb_trim(skb, off);
1675 return -EFAULT;
1676}
1677
1678static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1679 struct page *page, int off)
1680{
1681 if (i) {
1682 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1683
1684 return page == frag->page &&
1685 off == frag->page_offset + frag->size;
1686 }
1687 return 0;
1688}
1689
Herbert Xu364c6ba2006-06-09 16:10:40 -07001690static inline int __skb_linearize(struct sk_buff *skb)
1691{
1692 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1693}
1694
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695/**
1696 * skb_linearize - convert paged skb to linear one
1697 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07001698 *
1699 * If there is no free memory -ENOMEM is returned, otherwise zero
1700 * is returned and the old skb data released.
1701 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07001702static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001703{
Herbert Xu364c6ba2006-06-09 16:10:40 -07001704 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1705}
1706
1707/**
1708 * skb_linearize_cow - make sure skb is linear and writable
1709 * @skb: buffer to process
1710 *
1711 * If there is no free memory -ENOMEM is returned, otherwise zero
1712 * is returned and the old skb data released.
1713 */
1714static inline int skb_linearize_cow(struct sk_buff *skb)
1715{
1716 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1717 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001718}
1719
1720/**
1721 * skb_postpull_rcsum - update checksum for received skb after pull
1722 * @skb: buffer to update
1723 * @start: start of data before pull
1724 * @len: length of data pulled
1725 *
1726 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07001727 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1728 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001729 */
1730
1731static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08001732 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733{
Patrick McHardy84fa7932006-08-29 16:44:56 -07001734 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001735 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1736}
1737
Herbert Xucbb042f2006-03-20 22:43:56 -08001738unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1739
Linus Torvalds1da177e2005-04-16 15:20:36 -07001740/**
1741 * pskb_trim_rcsum - trim received skb and update checksum
1742 * @skb: buffer to trim
1743 * @len: new length
1744 *
1745 * This is exactly the same as pskb_trim except that it ensures the
1746 * checksum of received packets are still valid after the operation.
1747 */
1748
1749static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1750{
Stephen Hemminger0e4e4222005-09-08 12:32:03 -07001751 if (likely(len >= skb->len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001752 return 0;
Patrick McHardy84fa7932006-08-29 16:44:56 -07001753 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001754 skb->ip_summed = CHECKSUM_NONE;
1755 return __pskb_trim(skb, len);
1756}
1757
Linus Torvalds1da177e2005-04-16 15:20:36 -07001758#define skb_queue_walk(queue, skb) \
1759 for (skb = (queue)->next; \
1760 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1761 skb = skb->next)
1762
James Chapman46f89142007-04-30 00:07:31 -07001763#define skb_queue_walk_safe(queue, skb, tmp) \
1764 for (skb = (queue)->next, tmp = skb->next; \
1765 skb != (struct sk_buff *)(queue); \
1766 skb = tmp, tmp = skb->next)
1767
David S. Miller1164f522008-09-23 00:49:44 -07001768#define skb_queue_walk_from(queue, skb) \
1769 for (; prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1770 skb = skb->next)
1771
1772#define skb_queue_walk_from_safe(queue, skb, tmp) \
1773 for (tmp = skb->next; \
1774 skb != (struct sk_buff *)(queue); \
1775 skb = tmp, tmp = skb->next)
1776
Stephen Hemminger300ce172005-10-30 13:47:34 -08001777#define skb_queue_reverse_walk(queue, skb) \
1778 for (skb = (queue)->prev; \
1779 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1780 skb = skb->prev)
1781
Linus Torvalds1da177e2005-04-16 15:20:36 -07001782
David S. Milleree039872009-06-09 00:17:13 -07001783static inline bool skb_has_frags(const struct sk_buff *skb)
1784{
1785 return skb_shinfo(skb)->frag_list != NULL;
1786}
1787
1788static inline void skb_frag_list_init(struct sk_buff *skb)
1789{
1790 skb_shinfo(skb)->frag_list = NULL;
1791}
1792
1793static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
1794{
1795 frag->next = skb_shinfo(skb)->frag_list;
1796 skb_shinfo(skb)->frag_list = frag;
1797}
1798
1799#define skb_walk_frags(skb, iter) \
1800 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
1801
Herbert Xua59322b2007-12-05 01:53:40 -08001802extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
1803 int *peeked, int *err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001804extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1805 int noblock, int *err);
1806extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1807 struct poll_table_struct *wait);
1808extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1809 int offset, struct iovec *to,
1810 int size);
Herbert Xufb286bb2005-11-10 13:01:24 -08001811extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001812 int hlen,
1813 struct iovec *iov);
Rusty Russelldb543c12008-08-15 15:13:53 -07001814extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
1815 int offset,
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +00001816 const struct iovec *from,
1817 int from_offset,
Rusty Russelldb543c12008-08-15 15:13:53 -07001818 int len);
Michael S. Tsirkin0a1ec072009-04-20 01:25:46 +00001819extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
1820 int offset,
1821 const struct iovec *to,
1822 int to_offset,
1823 int size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001824extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
Eric Dumazet9d410c72009-10-30 05:03:53 +00001825extern void skb_free_datagram_locked(struct sock *sk,
1826 struct sk_buff *skb);
Herbert Xu27ab2562007-12-05 01:51:58 -08001827extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
Herbert Xu3305b802005-12-13 23:16:37 -08001828 unsigned int flags);
Al Viro2bbbc862006-11-14 21:37:14 -08001829extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
1830 int len, __wsum csum);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001831extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1832 void *to, int len);
Stephen Hemminger0c6fcc82007-04-20 16:40:01 -07001833extern int skb_store_bits(struct sk_buff *skb, int offset,
1834 const void *from, int len);
Al Viro81d77662006-11-14 21:37:33 -08001835extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001836 int offset, u8 *to, int len,
Al Viro81d77662006-11-14 21:37:33 -08001837 __wsum csum);
Jens Axboe9c55e012007-11-06 23:30:13 -08001838extern int skb_splice_bits(struct sk_buff *skb,
1839 unsigned int offset,
1840 struct pipe_inode_info *pipe,
1841 unsigned int len,
1842 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001843extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1844extern void skb_split(struct sk_buff *skb,
1845 struct sk_buff *skb1, const u32 len);
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001846extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
1847 int shiftlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848
Herbert Xu576a30e2006-06-27 13:22:38 -07001849extern struct sk_buff *skb_segment(struct sk_buff *skb, int features);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001850
Linus Torvalds1da177e2005-04-16 15:20:36 -07001851static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1852 int len, void *buffer)
1853{
1854 int hlen = skb_headlen(skb);
1855
Patrick McHardy55820ee2005-07-05 14:08:10 -07001856 if (hlen - offset >= len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001857 return skb->data + offset;
1858
1859 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1860 return NULL;
1861
1862 return buffer;
1863}
1864
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03001865static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
1866 void *to,
1867 const unsigned int len)
1868{
1869 memcpy(to, skb->data, len);
1870}
1871
1872static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
1873 const int offset, void *to,
1874 const unsigned int len)
1875{
1876 memcpy(to, skb->data + offset, len);
1877}
1878
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03001879static inline void skb_copy_to_linear_data(struct sk_buff *skb,
1880 const void *from,
1881 const unsigned int len)
1882{
1883 memcpy(skb->data, from, len);
1884}
1885
1886static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
1887 const int offset,
1888 const void *from,
1889 const unsigned int len)
1890{
1891 memcpy(skb->data + offset, from, len);
1892}
1893
Linus Torvalds1da177e2005-04-16 15:20:36 -07001894extern void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001895
Patrick Ohlyac45f602009-02-12 05:03:37 +00001896static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
1897{
1898 return skb->tstamp;
1899}
1900
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001901/**
1902 * skb_get_timestamp - get timestamp from a skb
1903 * @skb: skb to get stamp from
1904 * @stamp: pointer to struct timeval to store stamp in
1905 *
1906 * Timestamps are stored in the skb as offsets to a base timestamp.
1907 * This function converts the offset back to a struct timeval and stores
1908 * it in stamp.
1909 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00001910static inline void skb_get_timestamp(const struct sk_buff *skb,
1911 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001912{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001913 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001914}
1915
Patrick Ohlyac45f602009-02-12 05:03:37 +00001916static inline void skb_get_timestampns(const struct sk_buff *skb,
1917 struct timespec *stamp)
1918{
1919 *stamp = ktime_to_timespec(skb->tstamp);
1920}
1921
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001922static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001923{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001924 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001925}
1926
Stephen Hemminger164891a2007-04-23 22:26:16 -07001927static inline ktime_t net_timedelta(ktime_t t)
1928{
1929 return ktime_sub(ktime_get_real(), t);
1930}
1931
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07001932static inline ktime_t net_invalid_timestamp(void)
1933{
1934 return ktime_set(0, 0);
1935}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001936
Richard Cochranc1f19b52010-07-17 08:49:36 +00001937extern void skb_timestamping_init(void);
1938
1939#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
1940
1941extern void skb_clone_tx_timestamp(struct sk_buff *skb);
1942extern bool skb_defer_rx_timestamp(struct sk_buff *skb);
1943
1944#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
1945
1946static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
1947{
1948}
1949
1950static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
1951{
1952 return false;
1953}
1954
1955#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
1956
1957/**
1958 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
1959 *
1960 * @skb: clone of the the original outgoing packet
1961 * @hwtstamps: hardware time stamps
1962 *
1963 */
1964void skb_complete_tx_timestamp(struct sk_buff *skb,
1965 struct skb_shared_hwtstamps *hwtstamps);
1966
Patrick Ohlyac45f602009-02-12 05:03:37 +00001967/**
1968 * skb_tstamp_tx - queue clone of skb with send time stamps
1969 * @orig_skb: the original outgoing packet
1970 * @hwtstamps: hardware time stamps, may be NULL if not available
1971 *
1972 * If the skb has a socket associated, then this function clones the
1973 * skb (thus sharing the actual data and optional structures), stores
1974 * the optional hardware time stamping information (if non NULL) or
1975 * generates a software time stamp (otherwise), then queues the clone
1976 * to the error queue of the socket. Errors are silently ignored.
1977 */
1978extern void skb_tstamp_tx(struct sk_buff *orig_skb,
1979 struct skb_shared_hwtstamps *hwtstamps);
1980
Richard Cochran4507a712010-07-17 08:48:28 +00001981static inline void sw_tx_timestamp(struct sk_buff *skb)
1982{
1983 union skb_shared_tx *shtx = skb_tx(skb);
1984 if (shtx->software && !shtx->in_progress)
1985 skb_tstamp_tx(skb, NULL);
1986}
1987
1988/**
1989 * skb_tx_timestamp() - Driver hook for transmit timestamping
1990 *
1991 * Ethernet MAC Drivers should call this function in their hard_xmit()
1992 * function as soon as possible after giving the sk_buff to the MAC
1993 * hardware, but before freeing the sk_buff.
1994 *
1995 * @skb: A socket buffer.
1996 */
1997static inline void skb_tx_timestamp(struct sk_buff *skb)
1998{
Richard Cochranc1f19b52010-07-17 08:49:36 +00001999 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002000 sw_tx_timestamp(skb);
2001}
2002
Herbert Xu759e5d02007-03-25 20:10:56 -07002003extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
Al Virob51655b2006-11-14 21:40:42 -08002004extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002005
Herbert Xu60476372007-04-09 11:59:39 -07002006static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2007{
2008 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2009}
2010
Herbert Xufb286bb2005-11-10 13:01:24 -08002011/**
2012 * skb_checksum_complete - Calculate checksum of an entire packet
2013 * @skb: packet to process
2014 *
2015 * This function calculates the checksum over the entire packet plus
2016 * the value of skb->csum. The latter can be used to supply the
2017 * checksum of a pseudo header as used by TCP/UDP. It returns the
2018 * checksum.
2019 *
2020 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2021 * this function can be used to verify that checksum on received
2022 * packets. In that case the function should return zero if the
2023 * checksum is correct. In particular, this function will return zero
2024 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2025 * hardware has already verified the correctness of the checksum.
2026 */
Al Viro4381ca32007-07-15 21:00:11 +01002027static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002028{
Herbert Xu60476372007-04-09 11:59:39 -07002029 return skb_csum_unnecessary(skb) ?
2030 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002031}
2032
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002033#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002034extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002035static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2036{
2037 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002038 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002039}
2040static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2041{
2042 if (nfct)
2043 atomic_inc(&nfct->use);
2044}
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -08002045static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
2046{
2047 if (skb)
2048 atomic_inc(&skb->users);
2049}
2050static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
2051{
2052 if (skb)
2053 kfree_skb(skb);
2054}
2055#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002056#ifdef CONFIG_BRIDGE_NETFILTER
2057static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2058{
2059 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2060 kfree(nf_bridge);
2061}
2062static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2063{
2064 if (nf_bridge)
2065 atomic_inc(&nf_bridge->use);
2066}
2067#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08002068static inline void nf_reset(struct sk_buff *skb)
2069{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002070#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08002071 nf_conntrack_put(skb->nfct);
2072 skb->nfct = NULL;
Patrick McHardya193a4a2006-03-20 19:23:05 -08002073 nf_conntrack_put_reasm(skb->nfct_reasm);
2074 skb->nfct_reasm = NULL;
2075#endif
2076#ifdef CONFIG_BRIDGE_NETFILTER
2077 nf_bridge_put(skb->nf_bridge);
2078 skb->nf_bridge = NULL;
2079#endif
2080}
2081
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002082/* Note: This doesn't put any conntrack and bridge info in dst. */
2083static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2084{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002085#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002086 dst->nfct = src->nfct;
2087 nf_conntrack_get(src->nfct);
2088 dst->nfctinfo = src->nfctinfo;
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002089 dst->nfct_reasm = src->nfct_reasm;
2090 nf_conntrack_get_reasm(src->nfct_reasm);
2091#endif
2092#ifdef CONFIG_BRIDGE_NETFILTER
2093 dst->nf_bridge = src->nf_bridge;
2094 nf_bridge_get(src->nf_bridge);
2095#endif
2096}
2097
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002098static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2099{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002100#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002101 nf_conntrack_put(dst->nfct);
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002102 nf_conntrack_put_reasm(dst->nfct_reasm);
2103#endif
2104#ifdef CONFIG_BRIDGE_NETFILTER
2105 nf_bridge_put(dst->nf_bridge);
2106#endif
2107 __nf_copy(dst, src);
2108}
2109
James Morris984bc162006-06-09 00:29:17 -07002110#ifdef CONFIG_NETWORK_SECMARK
2111static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2112{
2113 to->secmark = from->secmark;
2114}
2115
2116static inline void skb_init_secmark(struct sk_buff *skb)
2117{
2118 skb->secmark = 0;
2119}
2120#else
2121static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2122{ }
2123
2124static inline void skb_init_secmark(struct sk_buff *skb)
2125{ }
2126#endif
2127
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002128static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2129{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002130 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002131}
2132
Stephen Hemminger92477442009-03-21 13:39:26 -07002133static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002134{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002135 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002136}
2137
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002138static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2139{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002140 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002141}
2142
David S. Millerd5a9e242009-01-27 16:22:11 -08002143static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2144{
2145 skb->queue_mapping = rx_queue + 1;
2146}
2147
Stephen Hemminger92477442009-03-21 13:39:26 -07002148static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002149{
2150 return skb->queue_mapping - 1;
2151}
2152
Stephen Hemminger92477442009-03-21 13:39:26 -07002153static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002154{
2155 return (skb->queue_mapping != 0);
2156}
2157
Stephen Hemminger92477442009-03-21 13:39:26 -07002158extern u16 skb_tx_hash(const struct net_device *dev,
2159 const struct sk_buff *skb);
2160
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002161#ifdef CONFIG_XFRM
2162static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2163{
2164 return skb->sp;
2165}
2166#else
2167static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2168{
2169 return NULL;
2170}
2171#endif
2172
Herbert Xu89114af2006-07-08 13:34:32 -07002173static inline int skb_is_gso(const struct sk_buff *skb)
2174{
2175 return skb_shinfo(skb)->gso_size;
2176}
2177
Brice Goglineabd7e32007-10-13 12:33:32 +02002178static inline int skb_is_gso_v6(const struct sk_buff *skb)
2179{
2180 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2181}
2182
Ben Hutchings4497b072008-06-19 16:22:28 -07002183extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2184
2185static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2186{
2187 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2188 * wanted then gso_type will be set. */
2189 struct skb_shared_info *shinfo = skb_shinfo(skb);
Alexander Duyckb78462e2010-06-02 12:24:37 +00002190 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2191 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07002192 __skb_warn_lro_forwarding(skb);
2193 return true;
2194 }
2195 return false;
2196}
2197
Herbert Xu35fc92a2007-03-26 23:22:20 -07002198static inline void skb_forward_csum(struct sk_buff *skb)
2199{
2200 /* Unfortunately we don't support this one. Any brave souls? */
2201 if (skb->ip_summed == CHECKSUM_COMPLETE)
2202 skb->ip_summed = CHECKSUM_NONE;
2203}
2204
Rusty Russellf35d9d82008-02-04 23:49:54 -05002205bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002206#endif /* __KERNEL__ */
2207#endif /* _LINUX_SKBUFF_H */