blob: ea0b3746386030e444fd92e3813501e7c84d6a5a [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
Arun Sharma600634972011-07-26 16:09:06 -070023#include <linux/atomic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070024#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
Anton Blancharda715dea2011-03-27 14:57:26 +0000125/* To allow 64K frame to be packed as single skb without frag_list. Since
126 * GRO uses frags we allocate at least 16 regardless of page size.
127 */
128#if (65536/PAGE_SIZE + 2) < 16
David S. Millereec00952011-03-29 23:34:08 -0700129#define MAX_SKB_FRAGS 16UL
Anton Blancharda715dea2011-03-27 14:57:26 +0000130#else
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
Anton Blancharda715dea2011-03-27 14:57:26 +0000132#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133
134typedef struct skb_frag_struct skb_frag_t;
135
136struct skb_frag_struct {
137 struct page *page;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000138#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
David S. Millera309bb02007-07-30 18:47:03 -0700139 __u32 page_offset;
140 __u32 size;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000141#else
142 __u16 page_offset;
143 __u16 size;
144#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145};
146
Patrick Ohlyac45f602009-02-12 05:03:37 +0000147#define HAVE_HW_TIME_STAMP
148
149/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800150 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000151 * @hwtstamp: hardware time stamp transformed into duration
152 * since arbitrary point in time
153 * @syststamp: hwtstamp transformed to system time base
154 *
155 * Software time stamps generated by ktime_get_real() are stored in
156 * skb->tstamp. The relation between the different kinds of time
157 * stamps is as follows:
158 *
159 * syststamp and tstamp can be compared against each other in
160 * arbitrary combinations. The accuracy of a
161 * syststamp/tstamp/"syststamp from other device" comparison is
162 * limited by the accuracy of the transformation into system time
163 * base. This depends on the device driver and its underlying
164 * hardware.
165 *
166 * hwtstamps can only be compared against other hwtstamps from
167 * the same device.
168 *
169 * This structure is attached to packets as part of the
170 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
171 */
172struct skb_shared_hwtstamps {
173 ktime_t hwtstamp;
174 ktime_t syststamp;
175};
176
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000177/* Definitions for tx_flags in struct skb_shared_info */
178enum {
179 /* generate hardware time stamp */
180 SKBTX_HW_TSTAMP = 1 << 0,
181
182 /* generate software time stamp */
183 SKBTX_SW_TSTAMP = 1 << 1,
184
185 /* device driver is going to provide hardware time stamp */
186 SKBTX_IN_PROGRESS = 1 << 2,
187
188 /* ensure the originating sk reference is available on driver level */
189 SKBTX_DRV_NEEDS_SK_REF = 1 << 3,
Shirley Maa6686f22011-07-06 12:22:12 +0000190
191 /* device driver supports TX zero-copy buffers */
192 SKBTX_DEV_ZEROCOPY = 1 << 4,
193};
194
195/*
196 * The callback notifies userspace to release buffers when skb DMA is done in
197 * lower device, the skb last reference should be 0 when calling this.
198 * The desc is used to track userspace buffer index.
199 */
200struct ubuf_info {
201 void (*callback)(void *);
202 void *arg;
203 unsigned long desc;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000204};
205
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206/* This data is invariant across clones and lives at
207 * the end of the header data, ie. at skb->end.
208 */
209struct skb_shared_info {
Benjamin LaHaise4947d3e2006-01-03 14:06:50 -0800210 unsigned short nr_frags;
Herbert Xu79671682006-06-22 02:40:14 -0700211 unsigned short gso_size;
212 /* Warning: this field is not always filled in (UFO)! */
213 unsigned short gso_segs;
214 unsigned short gso_type;
Al Viroae08e1f2006-11-08 00:27:11 -0800215 __be32 ip6_frag_id;
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000216 __u8 tx_flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700217 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000218 struct skb_shared_hwtstamps hwtstamps;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700219
220 /*
221 * Warning : all fields before dataref are cleared in __alloc_skb()
222 */
223 atomic_t dataref;
224
Johann Baudy69e3c752009-05-18 22:11:22 -0700225 /* Intermediate layers must ensure that destructor_arg
226 * remains valid until skb destructor */
227 void * destructor_arg;
Shirley Maa6686f22011-07-06 12:22:12 +0000228
Eric Dumazetfed66382010-07-22 19:09:08 +0000229 /* must be last field, see pskb_expand_head() */
230 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231};
232
233/* We divide dataref into two halves. The higher 16 bits hold references
234 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700235 * the entire skb->data. A clone of a headerless skb holds the length of
236 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700237 *
238 * All users must obey the rule that the skb->data reference count must be
239 * greater than or equal to the payload reference count.
240 *
241 * Holding a reference to the payload part means that the user does not
242 * care about modifications to the header part of skb->data.
243 */
244#define SKB_DATAREF_SHIFT 16
245#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
246
David S. Millerd179cd12005-08-17 14:57:30 -0700247
248enum {
249 SKB_FCLONE_UNAVAILABLE,
250 SKB_FCLONE_ORIG,
251 SKB_FCLONE_CLONE,
252};
253
Herbert Xu79671682006-06-22 02:40:14 -0700254enum {
255 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700256 SKB_GSO_UDP = 1 << 1,
Herbert Xu576a30e2006-06-27 13:22:38 -0700257
258 /* This indicates the skb is from an untrusted source. */
259 SKB_GSO_DODGY = 1 << 2,
Michael Chanb0da85372006-06-29 12:30:00 -0700260
261 /* This indicates the tcp segment has CWR set. */
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700262 SKB_GSO_TCP_ECN = 1 << 3,
263
264 SKB_GSO_TCPV6 = 1 << 4,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800265
266 SKB_GSO_FCOE = 1 << 5,
Herbert Xu79671682006-06-22 02:40:14 -0700267};
268
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700269#if BITS_PER_LONG > 32
270#define NET_SKBUFF_DATA_USES_OFFSET 1
271#endif
272
273#ifdef NET_SKBUFF_DATA_USES_OFFSET
274typedef unsigned int sk_buff_data_t;
275#else
276typedef unsigned char *sk_buff_data_t;
277#endif
278
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100279#if defined(CONFIG_NF_DEFRAG_IPV4) || defined(CONFIG_NF_DEFRAG_IPV4_MODULE) || \
280 defined(CONFIG_NF_DEFRAG_IPV6) || defined(CONFIG_NF_DEFRAG_IPV6_MODULE)
281#define NET_SKBUFF_NF_DEFRAG_NEEDED 1
282#endif
283
Linus Torvalds1da177e2005-04-16 15:20:36 -0700284/**
285 * struct sk_buff - socket buffer
286 * @next: Next buffer in list
287 * @prev: Previous buffer in list
Herbert Xu325ed822005-10-03 13:57:23 -0700288 * @tstamp: Time we arrived
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700289 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700290 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700291 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000292 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700293 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294 * @len: Length of actual data
295 * @data_len: Data length
296 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700297 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700298 * @csum: Checksum (must include start/offset pair)
299 * @csum_start: Offset from skb->head where checksumming should start
300 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700301 * @priority: Packet queueing priority
Martin Waitz67be2dd2005-05-01 08:59:26 -0700302 * @local_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700304 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305 * @nohdr: Payload reference only, must not modify header
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700306 * @nfctinfo: Relationship of this skb to the connection
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700308 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700309 * @ipvs_property: skbuff is owned by ipvs
Randy Dunlap31729362008-02-18 20:52:13 -0800310 * @peeked: this packet has been seen already, so stats have been
311 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700312 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700313 * @protocol: Packet protocol from driver
314 * @destructor: Destruct function
315 * @nfct: Associated connection, if any
Randy Dunlap461ddf32005-11-20 21:25:15 -0800316 * @nfct_reasm: netfilter conntrack re-assembly pointer
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800318 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319 * @tc_index: Traffic control index
320 * @tc_verd: traffic control verdict
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700321 * @rxhash: the packet hash computed on receive
322 * @queue_mapping: Queue mapping for multiqueue devices
Randy Dunlap553a5672008-04-20 10:51:01 -0700323 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700324 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Changli Gao4ca24622011-08-19 07:26:44 -0700325 * @l4_rxhash: indicate rxhash is a canonical 4-tuple hash over transport
326 * ports.
Randy Dunlapf4b8ea72006-06-22 16:00:11 -0700327 * @dma_cookie: a cookie to one of several possible DMA operations
328 * done by skb DMA functions
James Morris984bc162006-06-09 00:29:17 -0700329 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700330 * @mark: Generic packet mark
331 * @dropcount: total number of sk_receive_queue overflows
Patrick McHardy6aa895b02008-07-14 22:49:06 -0700332 * @vlan_tci: vlan tag control information
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700333 * @transport_header: Transport layer header
334 * @network_header: Network layer header
335 * @mac_header: Link layer header
336 * @tail: Tail pointer
337 * @end: End pointer
338 * @head: Head of buffer
339 * @data: Data head pointer
340 * @truesize: Buffer size
341 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700342 */
343
344struct sk_buff {
345 /* These two members must be first. */
346 struct sk_buff *next;
347 struct sk_buff *prev;
348
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700349 ktime_t tstamp;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000350
351 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353
Linus Torvalds1da177e2005-04-16 15:20:36 -0700354 /*
355 * This is the control buffer. It is free to use for every
356 * layer. Please put your private variables there. If you
357 * want to keep them across layers you have to do a skb_clone()
358 * first. This is owned by whoever has the skb queued ATM.
359 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000360 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361
Eric Dumazet7fee2262010-05-11 23:19:48 +0000362 unsigned long _skb_refdst;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000363#ifdef CONFIG_XFRM
364 struct sec_path *sp;
365#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700367 data_len;
368 __u16 mac_len,
369 hdr_len;
Al Viroff1dcad2006-11-20 18:07:29 -0800370 union {
371 __wsum csum;
Herbert Xu663ead32007-04-09 11:59:07 -0700372 struct {
373 __u16 csum_start;
374 __u16 csum_offset;
375 };
Al Viroff1dcad2006-11-20 18:07:29 -0800376 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700377 __u32 priority;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200378 kmemcheck_bitfield_begin(flags1);
Thomas Graf1cbb3382005-07-05 14:13:41 -0700379 __u8 local_df:1,
380 cloned:1,
381 ip_summed:2,
Harald Welte6869c4d2005-08-09 19:24:19 -0700382 nohdr:1,
383 nfctinfo:3;
David S. Millerd179cd12005-08-17 14:57:30 -0700384 __u8 pkt_type:3,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800385 fclone:2,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700386 ipvs_property:1,
Herbert Xua59322b2007-12-05 01:53:40 -0800387 peeked:1,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700388 nf_trace:1;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200389 kmemcheck_bitfield_end(flags1);
Eric Dumazet4ab408d2010-03-01 03:09:26 +0000390 __be16 protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391
392 void (*destructor)(struct sk_buff *skb);
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800393#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700394 struct nf_conntrack *nfct;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100395#endif
396#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800397 struct sk_buff *nfct_reasm;
398#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399#ifdef CONFIG_BRIDGE_NETFILTER
400 struct nf_bridge_info *nf_bridge;
401#endif
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700402
Eric Dumazet8964be42009-11-20 15:35:04 -0800403 int skb_iif;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700405 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700407 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200410
Tom Herbert0a9627f2010-03-16 08:03:29 +0000411 __u32 rxhash;
412
Eric Dumazet0a148422011-04-20 09:27:32 +0000413 __u16 queue_mapping;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200414 kmemcheck_bitfield_begin(flags2);
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400415#ifdef CONFIG_IPV6_NDISC_NODETYPE
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000416 __u8 ndisc_nodetype:2;
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400417#endif
Tom Herbert3853b582010-11-21 13:17:29 +0000418 __u8 ooo_okay:1;
Tom Herbertbdeab992011-08-14 19:45:55 +0000419 __u8 l4_rxhash:1;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200420 kmemcheck_bitfield_end(flags2);
421
Tom Herbert3853b582010-11-21 13:17:29 +0000422 /* 0/13 bit hole */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700423
Chris Leech97fc2f02006-05-23 17:55:33 -0700424#ifdef CONFIG_NET_DMA
425 dma_cookie_t dma_cookie;
426#endif
James Morris984bc162006-06-09 00:29:17 -0700427#ifdef CONFIG_NETWORK_SECMARK
428 __u32 secmark;
429#endif
Neil Horman3b885782009-10-12 13:26:31 -0700430 union {
431 __u32 mark;
432 __u32 dropcount;
433 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434
Patrick McHardy6aa895b02008-07-14 22:49:06 -0700435 __u16 vlan_tci;
436
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700437 sk_buff_data_t transport_header;
438 sk_buff_data_t network_header;
439 sk_buff_data_t mac_header;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700441 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700442 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700444 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700445 unsigned int truesize;
446 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447};
448
449#ifdef __KERNEL__
450/*
451 * Handling routines are only of interest to the kernel
452 */
453#include <linux/slab.h>
454
455#include <asm/system.h>
456
Eric Dumazet7fee2262010-05-11 23:19:48 +0000457/*
458 * skb might have a dst pointer attached, refcounted or not.
459 * _skb_refdst low order bit is set if refcount was _not_ taken
460 */
461#define SKB_DST_NOREF 1UL
462#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
463
464/**
465 * skb_dst - returns skb dst_entry
466 * @skb: buffer
467 *
468 * Returns skb dst_entry, regardless of reference taken or not.
469 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000470static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
471{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000472 /* If refdst was not refcounted, check we still are in a
473 * rcu_read_lock section
474 */
475 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
476 !rcu_read_lock_held() &&
477 !rcu_read_lock_bh_held());
478 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000479}
480
Eric Dumazet7fee2262010-05-11 23:19:48 +0000481/**
482 * skb_dst_set - sets skb dst
483 * @skb: buffer
484 * @dst: dst entry
485 *
486 * Sets skb dst, assuming a reference was taken on dst and should
487 * be released by skb_dst_drop()
488 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000489static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
490{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000491 skb->_skb_refdst = (unsigned long)dst;
492}
493
Eric Dumazet27b75c92010-10-15 05:44:11 +0000494extern void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst);
Eric Dumazet7fee2262010-05-11 23:19:48 +0000495
496/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300497 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000498 * @skb: buffer
499 */
500static inline bool skb_dst_is_noref(const struct sk_buff *skb)
501{
502 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000503}
504
Eric Dumazet511c3f92009-06-02 05:14:27 +0000505static inline struct rtable *skb_rtable(const struct sk_buff *skb)
506{
Eric Dumazetadf30902009-06-02 05:19:30 +0000507 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000508}
509
Jörn Engel231d06a2006-03-20 21:28:35 -0800510extern void kfree_skb(struct sk_buff *skb);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000511extern void consume_skb(struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512extern void __kfree_skb(struct sk_buff *skb);
David S. Millerd179cd12005-08-17 14:57:30 -0700513extern struct sk_buff *__alloc_skb(unsigned int size,
Christoph Hellwigb30973f2006-12-06 20:32:36 -0800514 gfp_t priority, int fclone, int node);
David S. Millerd179cd12005-08-17 14:57:30 -0700515static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100516 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700517{
Eric Dumazet564824b2010-10-11 19:05:25 +0000518 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700519}
520
521static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100522 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700523{
Eric Dumazet564824b2010-10-11 19:05:25 +0000524 return __alloc_skb(size, priority, 1, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700525}
526
Changli Gao5b0daa32010-05-29 00:12:13 -0700527extern bool skb_recycle_check(struct sk_buff *skb, int skb_size);
Lennert Buytenhek04a4bb52008-10-01 02:33:12 -0700528
Herbert Xue0053ec2007-10-14 00:37:52 -0700529extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700530extern struct sk_buff *skb_clone(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100531 gfp_t priority);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700532extern struct sk_buff *skb_copy(const struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100533 gfp_t priority);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700534extern struct sk_buff *pskb_copy(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100535 gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536extern int pskb_expand_head(struct sk_buff *skb,
Victor Fusco86a76ca2005-07-08 14:57:47 -0700537 int nhead, int ntail,
Al Virodd0fc662005-10-07 07:46:04 +0100538 gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
540 unsigned int headroom);
541extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
542 int newheadroom, int newtailroom,
Al Virodd0fc662005-10-07 07:46:04 +0100543 gfp_t priority);
David Howells716ea3a2007-04-02 20:19:53 -0700544extern int skb_to_sgvec(struct sk_buff *skb,
545 struct scatterlist *sg, int offset,
546 int len);
547extern int skb_cow_data(struct sk_buff *skb, int tailbits,
548 struct sk_buff **trailer);
Herbert Xu5b057c62006-06-23 02:06:41 -0700549extern int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000550#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700552extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
553 int getfrag(void *from, char *to, int offset,
554 int len,int odd, struct sk_buff *skb),
555 void *from, int length);
556
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800557struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700558 __u32 lower_offset;
559 __u32 upper_offset;
560 __u32 frag_idx;
561 __u32 stepped_offset;
562 struct sk_buff *root_skb;
563 struct sk_buff *cur_skb;
564 __u8 *frag_data;
565};
566
567extern void skb_prepare_seq_read(struct sk_buff *skb,
568 unsigned int from, unsigned int to,
569 struct skb_seq_state *st);
570extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
571 struct skb_seq_state *st);
572extern void skb_abort_seq_read(struct skb_seq_state *st);
573
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700574extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
575 unsigned int to, struct ts_config *config,
576 struct ts_state *state);
577
Tom Herbertbdeab992011-08-14 19:45:55 +0000578extern void __skb_get_rxhash(struct sk_buff *skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000579static inline __u32 skb_get_rxhash(struct sk_buff *skb)
580{
581 if (!skb->rxhash)
Tom Herbertbdeab992011-08-14 19:45:55 +0000582 __skb_get_rxhash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000583
584 return skb->rxhash;
585}
586
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700587#ifdef NET_SKBUFF_DATA_USES_OFFSET
588static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
589{
590 return skb->head + skb->end;
591}
592#else
593static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
594{
595 return skb->end;
596}
597#endif
598
Linus Torvalds1da177e2005-04-16 15:20:36 -0700599/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700600#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601
Patrick Ohlyac45f602009-02-12 05:03:37 +0000602static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
603{
604 return &skb_shinfo(skb)->hwtstamps;
605}
606
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607/**
608 * skb_queue_empty - check if a queue is empty
609 * @list: queue head
610 *
611 * Returns true if the queue is empty, false otherwise.
612 */
613static inline int skb_queue_empty(const struct sk_buff_head *list)
614{
615 return list->next == (struct sk_buff *)list;
616}
617
618/**
David S. Millerfc7ebb22008-09-23 00:34:07 -0700619 * skb_queue_is_last - check if skb is the last entry in the queue
620 * @list: queue head
621 * @skb: buffer
622 *
623 * Returns true if @skb is the last buffer on the list.
624 */
625static inline bool skb_queue_is_last(const struct sk_buff_head *list,
626 const struct sk_buff *skb)
627{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000628 return skb->next == (struct sk_buff *)list;
David S. Millerfc7ebb22008-09-23 00:34:07 -0700629}
630
631/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800632 * skb_queue_is_first - check if skb is the first entry in the queue
633 * @list: queue head
634 * @skb: buffer
635 *
636 * Returns true if @skb is the first buffer on the list.
637 */
638static inline bool skb_queue_is_first(const struct sk_buff_head *list,
639 const struct sk_buff *skb)
640{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000641 return skb->prev == (struct sk_buff *)list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800642}
643
644/**
David S. Miller249c8b42008-09-23 00:44:42 -0700645 * skb_queue_next - return the next packet in the queue
646 * @list: queue head
647 * @skb: current buffer
648 *
649 * Return the next packet in @list after @skb. It is only valid to
650 * call this if skb_queue_is_last() evaluates to false.
651 */
652static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
653 const struct sk_buff *skb)
654{
655 /* This BUG_ON may seem severe, but if we just return then we
656 * are going to dereference garbage.
657 */
658 BUG_ON(skb_queue_is_last(list, skb));
659 return skb->next;
660}
661
662/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800663 * skb_queue_prev - return the prev packet in the queue
664 * @list: queue head
665 * @skb: current buffer
666 *
667 * Return the prev packet in @list before @skb. It is only valid to
668 * call this if skb_queue_is_first() evaluates to false.
669 */
670static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
671 const struct sk_buff *skb)
672{
673 /* This BUG_ON may seem severe, but if we just return then we
674 * are going to dereference garbage.
675 */
676 BUG_ON(skb_queue_is_first(list, skb));
677 return skb->prev;
678}
679
680/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700681 * skb_get - reference buffer
682 * @skb: buffer to reference
683 *
684 * Makes another reference to a socket buffer and returns a pointer
685 * to the buffer.
686 */
687static inline struct sk_buff *skb_get(struct sk_buff *skb)
688{
689 atomic_inc(&skb->users);
690 return skb;
691}
692
693/*
694 * If users == 1, we are the only owner and are can avoid redundant
695 * atomic change.
696 */
697
698/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699 * skb_cloned - is the buffer a clone
700 * @skb: buffer to check
701 *
702 * Returns true if the buffer was generated with skb_clone() and is
703 * one of multiple shared copies of the buffer. Cloned buffers are
704 * shared data so must not be written to under normal circumstances.
705 */
706static inline int skb_cloned(const struct sk_buff *skb)
707{
708 return skb->cloned &&
709 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
710}
711
712/**
713 * skb_header_cloned - is the header a clone
714 * @skb: buffer to check
715 *
716 * Returns true if modifying the header part of the buffer requires
717 * the data to be copied.
718 */
719static inline int skb_header_cloned(const struct sk_buff *skb)
720{
721 int dataref;
722
723 if (!skb->cloned)
724 return 0;
725
726 dataref = atomic_read(&skb_shinfo(skb)->dataref);
727 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
728 return dataref != 1;
729}
730
731/**
732 * skb_header_release - release reference to header
733 * @skb: buffer to operate on
734 *
735 * Drop a reference to the header part of the buffer. This is done
736 * by acquiring a payload reference. You must not read from the header
737 * part of skb->data after this.
738 */
739static inline void skb_header_release(struct sk_buff *skb)
740{
741 BUG_ON(skb->nohdr);
742 skb->nohdr = 1;
743 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
744}
745
746/**
747 * skb_shared - is the buffer shared
748 * @skb: buffer to check
749 *
750 * Returns true if more than one person has a reference to this
751 * buffer.
752 */
753static inline int skb_shared(const struct sk_buff *skb)
754{
755 return atomic_read(&skb->users) != 1;
756}
757
758/**
759 * skb_share_check - check if buffer is shared and if so clone it
760 * @skb: buffer to check
761 * @pri: priority for memory allocation
762 *
763 * If the buffer is shared the buffer is cloned and the old copy
764 * drops a reference. A new clone with a single reference is returned.
765 * If the buffer is not shared the original buffer is returned. When
766 * being called from interrupt status or with spinlocks held pri must
767 * be GFP_ATOMIC.
768 *
769 * NULL is returned on a memory allocation failure.
770 */
Victor Fusco86a76ca2005-07-08 14:57:47 -0700771static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100772 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773{
774 might_sleep_if(pri & __GFP_WAIT);
775 if (skb_shared(skb)) {
776 struct sk_buff *nskb = skb_clone(skb, pri);
777 kfree_skb(skb);
778 skb = nskb;
779 }
780 return skb;
781}
782
783/*
784 * Copy shared buffers into a new sk_buff. We effectively do COW on
785 * packets to handle cases where we have a local reader and forward
786 * and a couple of other messy ones. The normal one is tcpdumping
787 * a packet thats being forwarded.
788 */
789
790/**
791 * skb_unshare - make a copy of a shared buffer
792 * @skb: buffer to check
793 * @pri: priority for memory allocation
794 *
795 * If the socket buffer is a clone then this function creates a new
796 * copy of the data, drops a reference count on the old copy and returns
797 * the new copy with the reference count at 1. If the buffer is not a clone
798 * the original buffer is returned. When called with a spinlock held or
799 * from interrupt state @pri must be %GFP_ATOMIC
800 *
801 * %NULL is returned on a memory allocation failure.
802 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -0700803static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100804 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805{
806 might_sleep_if(pri & __GFP_WAIT);
807 if (skb_cloned(skb)) {
808 struct sk_buff *nskb = skb_copy(skb, pri);
809 kfree_skb(skb); /* Free our shared copy */
810 skb = nskb;
811 }
812 return skb;
813}
814
815/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -0800816 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817 * @list_: list to peek at
818 *
819 * Peek an &sk_buff. Unlike most other operations you _MUST_
820 * be careful with this one. A peek leaves the buffer on the
821 * list and someone else may run off with it. You must hold
822 * the appropriate locks or have a private queue to do this.
823 *
824 * Returns %NULL for an empty list or a pointer to the head element.
825 * The reference count is not incremented and the reference is therefore
826 * volatile. Use with caution.
827 */
828static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
829{
830 struct sk_buff *list = ((struct sk_buff *)list_)->next;
831 if (list == (struct sk_buff *)list_)
832 list = NULL;
833 return list;
834}
835
836/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -0800837 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -0700838 * @list_: list to peek at
839 *
840 * Peek an &sk_buff. Unlike most other operations you _MUST_
841 * be careful with this one. A peek leaves the buffer on the
842 * list and someone else may run off with it. You must hold
843 * the appropriate locks or have a private queue to do this.
844 *
845 * Returns %NULL for an empty list or a pointer to the tail element.
846 * The reference count is not incremented and the reference is therefore
847 * volatile. Use with caution.
848 */
849static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
850{
851 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
852 if (list == (struct sk_buff *)list_)
853 list = NULL;
854 return list;
855}
856
857/**
858 * skb_queue_len - get queue length
859 * @list_: list to measure
860 *
861 * Return the length of an &sk_buff queue.
862 */
863static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
864{
865 return list_->qlen;
866}
867
David S. Miller67fed452008-09-21 22:36:24 -0700868/**
869 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
870 * @list: queue to initialize
871 *
872 * This initializes only the list and queue length aspects of
873 * an sk_buff_head object. This allows to initialize the list
874 * aspects of an sk_buff_head without reinitializing things like
875 * the spinlock. It can also be used for on-stack sk_buff_head
876 * objects where the spinlock is known to not be used.
877 */
878static inline void __skb_queue_head_init(struct sk_buff_head *list)
879{
880 list->prev = list->next = (struct sk_buff *)list;
881 list->qlen = 0;
882}
883
Arjan van de Ven76f10ad2006-08-02 14:06:55 -0700884/*
885 * This function creates a split out lock class for each invocation;
886 * this is needed for now since a whole lot of users of the skb-queue
887 * infrastructure in drivers have different locking usage (in hardirq)
888 * than the networking core (in softirq only). In the long run either the
889 * network layer or drivers should need annotation to consolidate the
890 * main types of usage into 3 classes.
891 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700892static inline void skb_queue_head_init(struct sk_buff_head *list)
893{
894 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -0700895 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896}
897
Pavel Emelianovc2ecba72007-04-17 12:45:31 -0700898static inline void skb_queue_head_init_class(struct sk_buff_head *list,
899 struct lock_class_key *class)
900{
901 skb_queue_head_init(list);
902 lockdep_set_class(&list->lock, class);
903}
904
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905/*
Gerrit Renkerbf299272008-04-14 00:04:51 -0700906 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907 *
908 * The "__skb_xxxx()" functions are the non-atomic ones that
909 * can only be called with interrupts disabled.
910 */
Gerrit Renkerbf299272008-04-14 00:04:51 -0700911extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
912static inline void __skb_insert(struct sk_buff *newsk,
913 struct sk_buff *prev, struct sk_buff *next,
914 struct sk_buff_head *list)
915{
916 newsk->next = next;
917 newsk->prev = prev;
918 next->prev = prev->next = newsk;
919 list->qlen++;
920}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700921
David S. Miller67fed452008-09-21 22:36:24 -0700922static inline void __skb_queue_splice(const struct sk_buff_head *list,
923 struct sk_buff *prev,
924 struct sk_buff *next)
925{
926 struct sk_buff *first = list->next;
927 struct sk_buff *last = list->prev;
928
929 first->prev = prev;
930 prev->next = first;
931
932 last->next = next;
933 next->prev = last;
934}
935
936/**
937 * skb_queue_splice - join two skb lists, this is designed for stacks
938 * @list: the new list to add
939 * @head: the place to add it in the first list
940 */
941static inline void skb_queue_splice(const struct sk_buff_head *list,
942 struct sk_buff_head *head)
943{
944 if (!skb_queue_empty(list)) {
945 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -0700946 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -0700947 }
948}
949
950/**
951 * skb_queue_splice - join two skb lists and reinitialise the emptied list
952 * @list: the new list to add
953 * @head: the place to add it in the first list
954 *
955 * The list at @list is reinitialised
956 */
957static inline void skb_queue_splice_init(struct sk_buff_head *list,
958 struct sk_buff_head *head)
959{
960 if (!skb_queue_empty(list)) {
961 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -0700962 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -0700963 __skb_queue_head_init(list);
964 }
965}
966
967/**
968 * skb_queue_splice_tail - join two skb lists, each list being a queue
969 * @list: the new list to add
970 * @head: the place to add it in the first list
971 */
972static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
973 struct sk_buff_head *head)
974{
975 if (!skb_queue_empty(list)) {
976 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -0700977 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -0700978 }
979}
980
981/**
982 * skb_queue_splice_tail - join two skb lists and reinitialise the emptied list
983 * @list: the new list to add
984 * @head: the place to add it in the first list
985 *
986 * Each of the lists is a queue.
987 * The list at @list is reinitialised
988 */
989static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
990 struct sk_buff_head *head)
991{
992 if (!skb_queue_empty(list)) {
993 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -0700994 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -0700995 __skb_queue_head_init(list);
996 }
997}
998
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001000 * __skb_queue_after - queue a buffer at the list head
1001 * @list: list to use
1002 * @prev: place after this buffer
1003 * @newsk: buffer to queue
1004 *
1005 * Queue a buffer int the middle of a list. This function takes no locks
1006 * and you must therefore hold required locks before calling it.
1007 *
1008 * A buffer cannot be placed on two lists at the same time.
1009 */
1010static inline void __skb_queue_after(struct sk_buff_head *list,
1011 struct sk_buff *prev,
1012 struct sk_buff *newsk)
1013{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001014 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001015}
1016
Gerrit Renker7de6c032008-04-14 00:05:09 -07001017extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1018 struct sk_buff_head *list);
1019
Gerrit Renkerf5572852008-04-14 00:05:28 -07001020static inline void __skb_queue_before(struct sk_buff_head *list,
1021 struct sk_buff *next,
1022 struct sk_buff *newsk)
1023{
1024 __skb_insert(newsk, next->prev, next, list);
1025}
1026
Stephen Hemminger300ce172005-10-30 13:47:34 -08001027/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001028 * __skb_queue_head - queue a buffer at the list head
1029 * @list: list to use
1030 * @newsk: buffer to queue
1031 *
1032 * Queue a buffer at the start of a list. This function takes no locks
1033 * and you must therefore hold required locks before calling it.
1034 *
1035 * A buffer cannot be placed on two lists at the same time.
1036 */
1037extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1038static inline void __skb_queue_head(struct sk_buff_head *list,
1039 struct sk_buff *newsk)
1040{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001041 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042}
1043
1044/**
1045 * __skb_queue_tail - queue a buffer at the list tail
1046 * @list: list to use
1047 * @newsk: buffer to queue
1048 *
1049 * Queue a buffer at the end of a list. This function takes no locks
1050 * and you must therefore hold required locks before calling it.
1051 *
1052 * A buffer cannot be placed on two lists at the same time.
1053 */
1054extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1055static inline void __skb_queue_tail(struct sk_buff_head *list,
1056 struct sk_buff *newsk)
1057{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001058 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059}
1060
Linus Torvalds1da177e2005-04-16 15:20:36 -07001061/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001062 * remove sk_buff from list. _Must_ be called atomically, and with
1063 * the list known..
1064 */
David S. Miller8728b832005-08-09 19:25:21 -07001065extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001066static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1067{
1068 struct sk_buff *next, *prev;
1069
1070 list->qlen--;
1071 next = skb->next;
1072 prev = skb->prev;
1073 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074 next->prev = prev;
1075 prev->next = next;
1076}
1077
Gerrit Renkerf525c062008-04-14 00:04:12 -07001078/**
1079 * __skb_dequeue - remove from the head of the queue
1080 * @list: list to dequeue from
1081 *
1082 * Remove the head of the list. This function does not take any locks
1083 * so must be used with appropriate locks held only. The head item is
1084 * returned or %NULL if the list is empty.
1085 */
1086extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1087static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1088{
1089 struct sk_buff *skb = skb_peek(list);
1090 if (skb)
1091 __skb_unlink(skb, list);
1092 return skb;
1093}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001094
1095/**
1096 * __skb_dequeue_tail - remove from the tail of the queue
1097 * @list: list to dequeue from
1098 *
1099 * Remove the tail of the list. This function does not take any locks
1100 * so must be used with appropriate locks held only. The tail item is
1101 * returned or %NULL if the list is empty.
1102 */
1103extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1104static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1105{
1106 struct sk_buff *skb = skb_peek_tail(list);
1107 if (skb)
1108 __skb_unlink(skb, list);
1109 return skb;
1110}
1111
1112
1113static inline int skb_is_nonlinear(const struct sk_buff *skb)
1114{
1115 return skb->data_len;
1116}
1117
1118static inline unsigned int skb_headlen(const struct sk_buff *skb)
1119{
1120 return skb->len - skb->data_len;
1121}
1122
1123static inline int skb_pagelen(const struct sk_buff *skb)
1124{
1125 int i, len = 0;
1126
1127 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1128 len += skb_shinfo(skb)->frags[i].size;
1129 return len + skb_headlen(skb);
1130}
1131
1132static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1133 struct page *page, int off, int size)
1134{
1135 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1136
1137 frag->page = page;
1138 frag->page_offset = off;
1139 frag->size = size;
1140 skb_shinfo(skb)->nr_frags = i + 1;
1141}
1142
Peter Zijlstra654bed12008-10-07 14:22:33 -07001143extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
1144 int off, int size);
1145
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001147#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001148#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1149
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001150#ifdef NET_SKBUFF_DATA_USES_OFFSET
1151static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1152{
1153 return skb->head + skb->tail;
1154}
1155
1156static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1157{
1158 skb->tail = skb->data - skb->head;
1159}
1160
1161static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1162{
1163 skb_reset_tail_pointer(skb);
1164 skb->tail += offset;
1165}
1166#else /* NET_SKBUFF_DATA_USES_OFFSET */
1167static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1168{
1169 return skb->tail;
1170}
1171
1172static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1173{
1174 skb->tail = skb->data;
1175}
1176
1177static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1178{
1179 skb->tail = skb->data + offset;
1180}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001181
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001182#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1183
Linus Torvalds1da177e2005-04-16 15:20:36 -07001184/*
1185 * Add data to an sk_buff
1186 */
Ilpo Järvinen0dde3e12008-03-27 17:43:41 -07001187extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001188static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1189{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001190 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001191 SKB_LINEAR_ASSERT(skb);
1192 skb->tail += len;
1193 skb->len += len;
1194 return tmp;
1195}
1196
Ilpo Järvinenc2aa2702008-03-27 17:52:40 -07001197extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001198static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1199{
1200 skb->data -= len;
1201 skb->len += len;
1202 return skb->data;
1203}
1204
Ilpo Järvinen6be8ac22008-03-27 17:47:24 -07001205extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001206static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1207{
1208 skb->len -= len;
1209 BUG_ON(skb->len < skb->data_len);
1210 return skb->data += len;
1211}
1212
David S. Miller47d29642010-05-02 02:21:44 -07001213static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1214{
1215 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1216}
1217
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1219
1220static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1221{
1222 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001223 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001224 return NULL;
1225 skb->len -= len;
1226 return skb->data += len;
1227}
1228
1229static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1230{
1231 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1232}
1233
1234static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1235{
1236 if (likely(len <= skb_headlen(skb)))
1237 return 1;
1238 if (unlikely(len > skb->len))
1239 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001240 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001241}
1242
1243/**
1244 * skb_headroom - bytes at buffer head
1245 * @skb: buffer to check
1246 *
1247 * Return the number of bytes of free space at the head of an &sk_buff.
1248 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001249static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250{
1251 return skb->data - skb->head;
1252}
1253
1254/**
1255 * skb_tailroom - bytes at buffer end
1256 * @skb: buffer to check
1257 *
1258 * Return the number of bytes of free space at the tail of an sk_buff
1259 */
1260static inline int skb_tailroom(const struct sk_buff *skb)
1261{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001262 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001263}
1264
1265/**
1266 * skb_reserve - adjust headroom
1267 * @skb: buffer to alter
1268 * @len: bytes to move
1269 *
1270 * Increase the headroom of an empty &sk_buff by reducing the tail
1271 * room. This is only allowed for an empty buffer.
1272 */
David S. Miller8243126c2006-01-17 02:54:21 -08001273static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274{
1275 skb->data += len;
1276 skb->tail += len;
1277}
1278
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00001279static inline void skb_reset_mac_len(struct sk_buff *skb)
1280{
1281 skb->mac_len = skb->network_header - skb->mac_header;
1282}
1283
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001284#ifdef NET_SKBUFF_DATA_USES_OFFSET
1285static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1286{
1287 return skb->head + skb->transport_header;
1288}
1289
1290static inline void skb_reset_transport_header(struct sk_buff *skb)
1291{
1292 skb->transport_header = skb->data - skb->head;
1293}
1294
1295static inline void skb_set_transport_header(struct sk_buff *skb,
1296 const int offset)
1297{
1298 skb_reset_transport_header(skb);
1299 skb->transport_header += offset;
1300}
1301
1302static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1303{
1304 return skb->head + skb->network_header;
1305}
1306
1307static inline void skb_reset_network_header(struct sk_buff *skb)
1308{
1309 skb->network_header = skb->data - skb->head;
1310}
1311
1312static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1313{
1314 skb_reset_network_header(skb);
1315 skb->network_header += offset;
1316}
1317
1318static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1319{
1320 return skb->head + skb->mac_header;
1321}
1322
1323static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1324{
1325 return skb->mac_header != ~0U;
1326}
1327
1328static inline void skb_reset_mac_header(struct sk_buff *skb)
1329{
1330 skb->mac_header = skb->data - skb->head;
1331}
1332
1333static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1334{
1335 skb_reset_mac_header(skb);
1336 skb->mac_header += offset;
1337}
1338
1339#else /* NET_SKBUFF_DATA_USES_OFFSET */
1340
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001341static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1342{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001343 return skb->transport_header;
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001344}
1345
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001346static inline void skb_reset_transport_header(struct sk_buff *skb)
1347{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001348 skb->transport_header = skb->data;
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001349}
1350
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001351static inline void skb_set_transport_header(struct sk_buff *skb,
1352 const int offset)
1353{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001354 skb->transport_header = skb->data + offset;
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001355}
1356
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001357static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1358{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001359 return skb->network_header;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001360}
1361
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001362static inline void skb_reset_network_header(struct sk_buff *skb)
1363{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001364 skb->network_header = skb->data;
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001365}
1366
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001367static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1368{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001369 skb->network_header = skb->data + offset;
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001370}
1371
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001372static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1373{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001374 return skb->mac_header;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001375}
1376
1377static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1378{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001379 return skb->mac_header != NULL;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001380}
1381
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001382static inline void skb_reset_mac_header(struct sk_buff *skb)
1383{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001384 skb->mac_header = skb->data;
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001385}
1386
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001387static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1388{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001389 skb->mac_header = skb->data + offset;
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001390}
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001391#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1392
Michał Mirosław04fb4512010-12-14 15:24:08 +00001393static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1394{
1395 return skb->csum_start - skb_headroom(skb);
1396}
1397
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001398static inline int skb_transport_offset(const struct sk_buff *skb)
1399{
1400 return skb_transport_header(skb) - skb->data;
1401}
1402
1403static inline u32 skb_network_header_len(const struct sk_buff *skb)
1404{
1405 return skb->transport_header - skb->network_header;
1406}
1407
1408static inline int skb_network_offset(const struct sk_buff *skb)
1409{
1410 return skb_network_header(skb) - skb->data;
1411}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001412
Changli Gaof9599ce2010-08-04 04:43:44 +00001413static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1414{
1415 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1416}
1417
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418/*
1419 * CPUs often take a performance hit when accessing unaligned memory
1420 * locations. The actual performance hit varies, it can be small if the
1421 * hardware handles it or large if we have to take an exception and fix it
1422 * in software.
1423 *
1424 * Since an ethernet header is 14 bytes network drivers often end up with
1425 * the IP header at an unaligned offset. The IP header can be aligned by
1426 * shifting the start of the packet by 2 bytes. Drivers should do this
1427 * with:
1428 *
Tobias Klauser8660c122009-07-13 22:48:16 +00001429 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001430 *
1431 * The downside to this alignment of the IP header is that the DMA is now
1432 * unaligned. On some architectures the cost of an unaligned DMA is high
1433 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00001434 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001435 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1436 * to be overridden.
1437 */
1438#ifndef NET_IP_ALIGN
1439#define NET_IP_ALIGN 2
1440#endif
1441
Anton Blanchard025be812006-03-31 02:27:06 -08001442/*
1443 * The networking layer reserves some headroom in skb data (via
1444 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1445 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08001446 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08001447 *
1448 * Unfortunately this headroom changes the DMA alignment of the resulting
1449 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1450 * on some architectures. An architecture can override this value,
1451 * perhaps setting it to a cacheline in size (since that will maintain
1452 * cacheline alignment of the DMA). It must be a power of 2.
1453 *
David S. Millerd6301d32009-02-08 19:24:13 -08001454 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08001455 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07001456 *
1457 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1458 * to reduce average number of cache lines per packet.
1459 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07001460 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08001461 */
1462#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07001463#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08001464#endif
1465
Herbert Xu3cc0e872006-06-09 16:13:38 -07001466extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001467
1468static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1469{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00001470 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07001471 WARN_ON(1);
1472 return;
1473 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001474 skb->len = len;
1475 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476}
1477
Ilpo Järvinen419ae742008-03-27 17:54:01 -07001478extern void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001479
1480static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1481{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001482 if (skb->data_len)
1483 return ___pskb_trim(skb, len);
1484 __skb_trim(skb, len);
1485 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001486}
1487
1488static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1489{
1490 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1491}
1492
1493/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07001494 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1495 * @skb: buffer to alter
1496 * @len: new length
1497 *
1498 * This is identical to pskb_trim except that the caller knows that
1499 * the skb is not cloned so we should never get an error due to out-
1500 * of-memory.
1501 */
1502static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1503{
1504 int err = pskb_trim(skb, len);
1505 BUG_ON(err);
1506}
1507
1508/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001509 * skb_orphan - orphan a buffer
1510 * @skb: buffer to orphan
1511 *
1512 * If a buffer currently has an owner then we call the owner's
1513 * destructor function and make the @skb unowned. The buffer continues
1514 * to exist but is no longer charged to its former owner.
1515 */
1516static inline void skb_orphan(struct sk_buff *skb)
1517{
1518 if (skb->destructor)
1519 skb->destructor(skb);
1520 skb->destructor = NULL;
1521 skb->sk = NULL;
1522}
1523
1524/**
1525 * __skb_queue_purge - empty a list
1526 * @list: list to empty
1527 *
1528 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1529 * the list and one reference dropped. This function does not take the
1530 * list lock and the caller must hold the relevant locks to use it.
1531 */
1532extern void skb_queue_purge(struct sk_buff_head *list);
1533static inline void __skb_queue_purge(struct sk_buff_head *list)
1534{
1535 struct sk_buff *skb;
1536 while ((skb = __skb_dequeue(list)) != NULL)
1537 kfree_skb(skb);
1538}
1539
1540/**
Christoph Hellwigb4e54de2006-07-24 15:31:14 -07001541 * __dev_alloc_skb - allocate an skbuff for receiving
Linus Torvalds1da177e2005-04-16 15:20:36 -07001542 * @length: length to allocate
1543 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1544 *
1545 * Allocate a new &sk_buff and assign it a usage count of one. The
1546 * buffer has unspecified headroom built in. Users should allocate
1547 * the headroom they think they need without accounting for the
1548 * built in space. The built in space is used for optimisations.
1549 *
Christoph Hellwig766ea8c2006-08-07 15:49:53 -07001550 * %NULL is returned if there is no free memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001551 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001552static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
Al Virodd0fc662005-10-07 07:46:04 +01001553 gfp_t gfp_mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001554{
Anton Blanchard025be812006-03-31 02:27:06 -08001555 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001556 if (likely(skb))
Anton Blanchard025be812006-03-31 02:27:06 -08001557 skb_reserve(skb, NET_SKB_PAD);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558 return skb;
1559}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001560
Ilpo Järvinenf58518e2008-03-27 17:51:31 -07001561extern struct sk_buff *dev_alloc_skb(unsigned int length);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001562
Christoph Hellwig8af27452006-07-31 22:35:23 -07001563extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1564 unsigned int length, gfp_t gfp_mask);
1565
1566/**
1567 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1568 * @dev: network device to receive on
1569 * @length: length to allocate
1570 *
1571 * Allocate a new &sk_buff and assign it a usage count of one. The
1572 * buffer has unspecified headroom built in. Users should allocate
1573 * the headroom they think they need without accounting for the
1574 * built in space. The built in space is used for optimisations.
1575 *
1576 * %NULL is returned if there is no free memory. Although this function
1577 * allocates memory it can be called from an interrupt.
1578 */
1579static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1580 unsigned int length)
1581{
1582 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1583}
1584
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001585static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
1586 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00001587{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001588 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00001589
1590 if (NET_IP_ALIGN && skb)
1591 skb_reserve(skb, NET_IP_ALIGN);
1592 return skb;
1593}
1594
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001595static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1596 unsigned int length)
1597{
1598 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
1599}
1600
Eric Dumazet564824b2010-10-11 19:05:25 +00001601/**
1602 * __netdev_alloc_page - allocate a page for ps-rx on a specific device
1603 * @dev: network device to receive on
1604 * @gfp_mask: alloc_pages_node mask
1605 *
1606 * Allocate a new page. dev currently unused.
1607 *
1608 * %NULL is returned if there is no free memory.
1609 */
1610static inline struct page *__netdev_alloc_page(struct net_device *dev, gfp_t gfp_mask)
1611{
1612 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, 0);
1613}
Peter Zijlstra654bed12008-10-07 14:22:33 -07001614
1615/**
1616 * netdev_alloc_page - allocate a page for ps-rx on a specific device
1617 * @dev: network device to receive on
1618 *
Eric Dumazet564824b2010-10-11 19:05:25 +00001619 * Allocate a new page. dev currently unused.
Peter Zijlstra654bed12008-10-07 14:22:33 -07001620 *
1621 * %NULL is returned if there is no free memory.
1622 */
1623static inline struct page *netdev_alloc_page(struct net_device *dev)
1624{
1625 return __netdev_alloc_page(dev, GFP_ATOMIC);
1626}
1627
1628static inline void netdev_free_page(struct net_device *dev, struct page *page)
1629{
1630 __free_page(page);
1631}
1632
Linus Torvalds1da177e2005-04-16 15:20:36 -07001633/**
Patrick McHardy334a8132007-06-25 04:35:20 -07001634 * skb_clone_writable - is the header of a clone writable
1635 * @skb: buffer to check
1636 * @len: length up to which to write
1637 *
1638 * Returns true if modifying the header part of the cloned buffer
1639 * does not requires the data to be copied.
1640 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001641static inline int skb_clone_writable(struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07001642{
1643 return !skb_header_cloned(skb) &&
1644 skb_headroom(skb) + len <= skb->hdr_len;
1645}
1646
Herbert Xud9cc2042007-09-16 16:21:16 -07001647static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
1648 int cloned)
1649{
1650 int delta = 0;
1651
1652 if (headroom < NET_SKB_PAD)
1653 headroom = NET_SKB_PAD;
1654 if (headroom > skb_headroom(skb))
1655 delta = headroom - skb_headroom(skb);
1656
1657 if (delta || cloned)
1658 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
1659 GFP_ATOMIC);
1660 return 0;
1661}
1662
Patrick McHardy334a8132007-06-25 04:35:20 -07001663/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001664 * skb_cow - copy header of skb when it is required
1665 * @skb: buffer to cow
1666 * @headroom: needed headroom
1667 *
1668 * If the skb passed lacks sufficient headroom or its data part
1669 * is shared, data is reallocated. If reallocation fails, an error
1670 * is returned and original skb is not changed.
1671 *
1672 * The result is skb with writable area skb->head...skb->tail
1673 * and at least @headroom of space at head.
1674 */
1675static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1676{
Herbert Xud9cc2042007-09-16 16:21:16 -07001677 return __skb_cow(skb, headroom, skb_cloned(skb));
1678}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001679
Herbert Xud9cc2042007-09-16 16:21:16 -07001680/**
1681 * skb_cow_head - skb_cow but only making the head writable
1682 * @skb: buffer to cow
1683 * @headroom: needed headroom
1684 *
1685 * This function is identical to skb_cow except that we replace the
1686 * skb_cloned check by skb_header_cloned. It should be used when
1687 * you only need to push on some header and do not need to modify
1688 * the data.
1689 */
1690static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
1691{
1692 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001693}
1694
1695/**
1696 * skb_padto - pad an skbuff up to a minimal size
1697 * @skb: buffer to pad
1698 * @len: minimal length
1699 *
1700 * Pads up a buffer to ensure the trailing bytes exist and are
1701 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07001702 * is untouched. Otherwise it is extended. Returns zero on
1703 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001704 */
1705
Herbert Xu5b057c62006-06-23 02:06:41 -07001706static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001707{
1708 unsigned int size = skb->len;
1709 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07001710 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001711 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001712}
1713
1714static inline int skb_add_data(struct sk_buff *skb,
1715 char __user *from, int copy)
1716{
1717 const int off = skb->len;
1718
1719 if (skb->ip_summed == CHECKSUM_NONE) {
1720 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08001721 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
Linus Torvalds1da177e2005-04-16 15:20:36 -07001722 copy, 0, &err);
1723 if (!err) {
1724 skb->csum = csum_block_add(skb->csum, csum, off);
1725 return 0;
1726 }
1727 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1728 return 0;
1729
1730 __skb_trim(skb, off);
1731 return -EFAULT;
1732}
1733
1734static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1735 struct page *page, int off)
1736{
1737 if (i) {
1738 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1739
1740 return page == frag->page &&
1741 off == frag->page_offset + frag->size;
1742 }
1743 return 0;
1744}
1745
Herbert Xu364c6ba2006-06-09 16:10:40 -07001746static inline int __skb_linearize(struct sk_buff *skb)
1747{
1748 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1749}
1750
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751/**
1752 * skb_linearize - convert paged skb to linear one
1753 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07001754 *
1755 * If there is no free memory -ENOMEM is returned, otherwise zero
1756 * is returned and the old skb data released.
1757 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07001758static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001759{
Herbert Xu364c6ba2006-06-09 16:10:40 -07001760 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1761}
1762
1763/**
1764 * skb_linearize_cow - make sure skb is linear and writable
1765 * @skb: buffer to process
1766 *
1767 * If there is no free memory -ENOMEM is returned, otherwise zero
1768 * is returned and the old skb data released.
1769 */
1770static inline int skb_linearize_cow(struct sk_buff *skb)
1771{
1772 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1773 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001774}
1775
1776/**
1777 * skb_postpull_rcsum - update checksum for received skb after pull
1778 * @skb: buffer to update
1779 * @start: start of data before pull
1780 * @len: length of data pulled
1781 *
1782 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07001783 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1784 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001785 */
1786
1787static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08001788 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001789{
Patrick McHardy84fa7932006-08-29 16:44:56 -07001790 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001791 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1792}
1793
Herbert Xucbb042f2006-03-20 22:43:56 -08001794unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1795
Linus Torvalds1da177e2005-04-16 15:20:36 -07001796/**
1797 * pskb_trim_rcsum - trim received skb and update checksum
1798 * @skb: buffer to trim
1799 * @len: new length
1800 *
1801 * This is exactly the same as pskb_trim except that it ensures the
1802 * checksum of received packets are still valid after the operation.
1803 */
1804
1805static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1806{
Stephen Hemminger0e4e4222005-09-08 12:32:03 -07001807 if (likely(len >= skb->len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001808 return 0;
Patrick McHardy84fa7932006-08-29 16:44:56 -07001809 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001810 skb->ip_summed = CHECKSUM_NONE;
1811 return __pskb_trim(skb, len);
1812}
1813
Linus Torvalds1da177e2005-04-16 15:20:36 -07001814#define skb_queue_walk(queue, skb) \
1815 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07001816 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07001817 skb = skb->next)
1818
James Chapman46f89142007-04-30 00:07:31 -07001819#define skb_queue_walk_safe(queue, skb, tmp) \
1820 for (skb = (queue)->next, tmp = skb->next; \
1821 skb != (struct sk_buff *)(queue); \
1822 skb = tmp, tmp = skb->next)
1823
David S. Miller1164f522008-09-23 00:49:44 -07001824#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07001825 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07001826 skb = skb->next)
1827
1828#define skb_queue_walk_from_safe(queue, skb, tmp) \
1829 for (tmp = skb->next; \
1830 skb != (struct sk_buff *)(queue); \
1831 skb = tmp, tmp = skb->next)
1832
Stephen Hemminger300ce172005-10-30 13:47:34 -08001833#define skb_queue_reverse_walk(queue, skb) \
1834 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07001835 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08001836 skb = skb->prev)
1837
David S. Miller686a2952011-01-20 22:47:32 -08001838#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
1839 for (skb = (queue)->prev, tmp = skb->prev; \
1840 skb != (struct sk_buff *)(queue); \
1841 skb = tmp, tmp = skb->prev)
1842
1843#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
1844 for (tmp = skb->prev; \
1845 skb != (struct sk_buff *)(queue); \
1846 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001847
David S. Miller21dc3302010-08-23 00:13:46 -07001848static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07001849{
1850 return skb_shinfo(skb)->frag_list != NULL;
1851}
1852
1853static inline void skb_frag_list_init(struct sk_buff *skb)
1854{
1855 skb_shinfo(skb)->frag_list = NULL;
1856}
1857
1858static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
1859{
1860 frag->next = skb_shinfo(skb)->frag_list;
1861 skb_shinfo(skb)->frag_list = frag;
1862}
1863
1864#define skb_walk_frags(skb, iter) \
1865 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
1866
Herbert Xua59322b2007-12-05 01:53:40 -08001867extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
1868 int *peeked, int *err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001869extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1870 int noblock, int *err);
1871extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1872 struct poll_table_struct *wait);
1873extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1874 int offset, struct iovec *to,
1875 int size);
Herbert Xufb286bb2005-11-10 13:01:24 -08001876extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001877 int hlen,
1878 struct iovec *iov);
Rusty Russelldb543c12008-08-15 15:13:53 -07001879extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
1880 int offset,
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +00001881 const struct iovec *from,
1882 int from_offset,
Rusty Russelldb543c12008-08-15 15:13:53 -07001883 int len);
Michael S. Tsirkin0a1ec072009-04-20 01:25:46 +00001884extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
1885 int offset,
1886 const struct iovec *to,
1887 int to_offset,
1888 int size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001889extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
Eric Dumazet9d410c72009-10-30 05:03:53 +00001890extern void skb_free_datagram_locked(struct sock *sk,
1891 struct sk_buff *skb);
Herbert Xu27ab2562007-12-05 01:51:58 -08001892extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
Herbert Xu3305b802005-12-13 23:16:37 -08001893 unsigned int flags);
Al Viro2bbbc862006-11-14 21:37:14 -08001894extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
1895 int len, __wsum csum);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001896extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1897 void *to, int len);
Stephen Hemminger0c6fcc82007-04-20 16:40:01 -07001898extern int skb_store_bits(struct sk_buff *skb, int offset,
1899 const void *from, int len);
Al Viro81d77662006-11-14 21:37:33 -08001900extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001901 int offset, u8 *to, int len,
Al Viro81d77662006-11-14 21:37:33 -08001902 __wsum csum);
Jens Axboe9c55e012007-11-06 23:30:13 -08001903extern int skb_splice_bits(struct sk_buff *skb,
1904 unsigned int offset,
1905 struct pipe_inode_info *pipe,
1906 unsigned int len,
1907 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001908extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1909extern void skb_split(struct sk_buff *skb,
1910 struct sk_buff *skb1, const u32 len);
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001911extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
1912 int shiftlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001913
Michał Mirosław04ed3e72011-01-24 15:32:47 -08001914extern struct sk_buff *skb_segment(struct sk_buff *skb, u32 features);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001915
Linus Torvalds1da177e2005-04-16 15:20:36 -07001916static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1917 int len, void *buffer)
1918{
1919 int hlen = skb_headlen(skb);
1920
Patrick McHardy55820ee2005-07-05 14:08:10 -07001921 if (hlen - offset >= len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001922 return skb->data + offset;
1923
1924 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1925 return NULL;
1926
1927 return buffer;
1928}
1929
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03001930static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
1931 void *to,
1932 const unsigned int len)
1933{
1934 memcpy(to, skb->data, len);
1935}
1936
1937static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
1938 const int offset, void *to,
1939 const unsigned int len)
1940{
1941 memcpy(to, skb->data + offset, len);
1942}
1943
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03001944static inline void skb_copy_to_linear_data(struct sk_buff *skb,
1945 const void *from,
1946 const unsigned int len)
1947{
1948 memcpy(skb->data, from, len);
1949}
1950
1951static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
1952 const int offset,
1953 const void *from,
1954 const unsigned int len)
1955{
1956 memcpy(skb->data + offset, from, len);
1957}
1958
Linus Torvalds1da177e2005-04-16 15:20:36 -07001959extern void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001960
Patrick Ohlyac45f602009-02-12 05:03:37 +00001961static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
1962{
1963 return skb->tstamp;
1964}
1965
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001966/**
1967 * skb_get_timestamp - get timestamp from a skb
1968 * @skb: skb to get stamp from
1969 * @stamp: pointer to struct timeval to store stamp in
1970 *
1971 * Timestamps are stored in the skb as offsets to a base timestamp.
1972 * This function converts the offset back to a struct timeval and stores
1973 * it in stamp.
1974 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00001975static inline void skb_get_timestamp(const struct sk_buff *skb,
1976 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001977{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001978 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001979}
1980
Patrick Ohlyac45f602009-02-12 05:03:37 +00001981static inline void skb_get_timestampns(const struct sk_buff *skb,
1982 struct timespec *stamp)
1983{
1984 *stamp = ktime_to_timespec(skb->tstamp);
1985}
1986
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001987static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001988{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001989 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001990}
1991
Stephen Hemminger164891a2007-04-23 22:26:16 -07001992static inline ktime_t net_timedelta(ktime_t t)
1993{
1994 return ktime_sub(ktime_get_real(), t);
1995}
1996
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07001997static inline ktime_t net_invalid_timestamp(void)
1998{
1999 return ktime_set(0, 0);
2000}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002001
Richard Cochranc1f19b52010-07-17 08:49:36 +00002002extern void skb_timestamping_init(void);
2003
2004#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2005
2006extern void skb_clone_tx_timestamp(struct sk_buff *skb);
2007extern bool skb_defer_rx_timestamp(struct sk_buff *skb);
2008
2009#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2010
2011static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2012{
2013}
2014
2015static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2016{
2017 return false;
2018}
2019
2020#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2021
2022/**
2023 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2024 *
2025 * @skb: clone of the the original outgoing packet
2026 * @hwtstamps: hardware time stamps
2027 *
2028 */
2029void skb_complete_tx_timestamp(struct sk_buff *skb,
2030 struct skb_shared_hwtstamps *hwtstamps);
2031
Patrick Ohlyac45f602009-02-12 05:03:37 +00002032/**
2033 * skb_tstamp_tx - queue clone of skb with send time stamps
2034 * @orig_skb: the original outgoing packet
2035 * @hwtstamps: hardware time stamps, may be NULL if not available
2036 *
2037 * If the skb has a socket associated, then this function clones the
2038 * skb (thus sharing the actual data and optional structures), stores
2039 * the optional hardware time stamping information (if non NULL) or
2040 * generates a software time stamp (otherwise), then queues the clone
2041 * to the error queue of the socket. Errors are silently ignored.
2042 */
2043extern void skb_tstamp_tx(struct sk_buff *orig_skb,
2044 struct skb_shared_hwtstamps *hwtstamps);
2045
Richard Cochran4507a712010-07-17 08:48:28 +00002046static inline void sw_tx_timestamp(struct sk_buff *skb)
2047{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002048 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2049 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00002050 skb_tstamp_tx(skb, NULL);
2051}
2052
2053/**
2054 * skb_tx_timestamp() - Driver hook for transmit timestamping
2055 *
2056 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00002057 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00002058 *
2059 * @skb: A socket buffer.
2060 */
2061static inline void skb_tx_timestamp(struct sk_buff *skb)
2062{
Richard Cochranc1f19b52010-07-17 08:49:36 +00002063 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002064 sw_tx_timestamp(skb);
2065}
2066
Herbert Xu759e5d02007-03-25 20:10:56 -07002067extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
Al Virob51655b2006-11-14 21:40:42 -08002068extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002069
Herbert Xu60476372007-04-09 11:59:39 -07002070static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2071{
2072 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2073}
2074
Herbert Xufb286bb2005-11-10 13:01:24 -08002075/**
2076 * skb_checksum_complete - Calculate checksum of an entire packet
2077 * @skb: packet to process
2078 *
2079 * This function calculates the checksum over the entire packet plus
2080 * the value of skb->csum. The latter can be used to supply the
2081 * checksum of a pseudo header as used by TCP/UDP. It returns the
2082 * checksum.
2083 *
2084 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2085 * this function can be used to verify that checksum on received
2086 * packets. In that case the function should return zero if the
2087 * checksum is correct. In particular, this function will return zero
2088 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2089 * hardware has already verified the correctness of the checksum.
2090 */
Al Viro4381ca32007-07-15 21:00:11 +01002091static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002092{
Herbert Xu60476372007-04-09 11:59:39 -07002093 return skb_csum_unnecessary(skb) ?
2094 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002095}
2096
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002097#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002098extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002099static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2100{
2101 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002102 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002103}
2104static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2105{
2106 if (nfct)
2107 atomic_inc(&nfct->use);
2108}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002109#endif
2110#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -08002111static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
2112{
2113 if (skb)
2114 atomic_inc(&skb->users);
2115}
2116static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
2117{
2118 if (skb)
2119 kfree_skb(skb);
2120}
2121#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002122#ifdef CONFIG_BRIDGE_NETFILTER
2123static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2124{
2125 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2126 kfree(nf_bridge);
2127}
2128static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2129{
2130 if (nf_bridge)
2131 atomic_inc(&nf_bridge->use);
2132}
2133#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08002134static inline void nf_reset(struct sk_buff *skb)
2135{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002136#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08002137 nf_conntrack_put(skb->nfct);
2138 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002139#endif
2140#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Patrick McHardya193a4a2006-03-20 19:23:05 -08002141 nf_conntrack_put_reasm(skb->nfct_reasm);
2142 skb->nfct_reasm = NULL;
2143#endif
2144#ifdef CONFIG_BRIDGE_NETFILTER
2145 nf_bridge_put(skb->nf_bridge);
2146 skb->nf_bridge = NULL;
2147#endif
2148}
2149
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002150/* Note: This doesn't put any conntrack and bridge info in dst. */
2151static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2152{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002153#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002154 dst->nfct = src->nfct;
2155 nf_conntrack_get(src->nfct);
2156 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002157#endif
2158#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002159 dst->nfct_reasm = src->nfct_reasm;
2160 nf_conntrack_get_reasm(src->nfct_reasm);
2161#endif
2162#ifdef CONFIG_BRIDGE_NETFILTER
2163 dst->nf_bridge = src->nf_bridge;
2164 nf_bridge_get(src->nf_bridge);
2165#endif
2166}
2167
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002168static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2169{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002170#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002171 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002172#endif
2173#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002174 nf_conntrack_put_reasm(dst->nfct_reasm);
2175#endif
2176#ifdef CONFIG_BRIDGE_NETFILTER
2177 nf_bridge_put(dst->nf_bridge);
2178#endif
2179 __nf_copy(dst, src);
2180}
2181
James Morris984bc162006-06-09 00:29:17 -07002182#ifdef CONFIG_NETWORK_SECMARK
2183static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2184{
2185 to->secmark = from->secmark;
2186}
2187
2188static inline void skb_init_secmark(struct sk_buff *skb)
2189{
2190 skb->secmark = 0;
2191}
2192#else
2193static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2194{ }
2195
2196static inline void skb_init_secmark(struct sk_buff *skb)
2197{ }
2198#endif
2199
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002200static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2201{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002202 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002203}
2204
Stephen Hemminger92477442009-03-21 13:39:26 -07002205static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002206{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002207 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002208}
2209
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002210static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2211{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002212 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002213}
2214
David S. Millerd5a9e242009-01-27 16:22:11 -08002215static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2216{
2217 skb->queue_mapping = rx_queue + 1;
2218}
2219
Stephen Hemminger92477442009-03-21 13:39:26 -07002220static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002221{
2222 return skb->queue_mapping - 1;
2223}
2224
Stephen Hemminger92477442009-03-21 13:39:26 -07002225static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002226{
Eric Dumazeta02cec22010-09-22 20:43:57 +00002227 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08002228}
2229
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00002230extern u16 __skb_tx_hash(const struct net_device *dev,
2231 const struct sk_buff *skb,
2232 unsigned int num_tx_queues);
Stephen Hemminger92477442009-03-21 13:39:26 -07002233
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002234#ifdef CONFIG_XFRM
2235static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2236{
2237 return skb->sp;
2238}
2239#else
2240static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2241{
2242 return NULL;
2243}
2244#endif
2245
Herbert Xu89114af2006-07-08 13:34:32 -07002246static inline int skb_is_gso(const struct sk_buff *skb)
2247{
2248 return skb_shinfo(skb)->gso_size;
2249}
2250
Brice Goglineabd7e32007-10-13 12:33:32 +02002251static inline int skb_is_gso_v6(const struct sk_buff *skb)
2252{
2253 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2254}
2255
Ben Hutchings4497b072008-06-19 16:22:28 -07002256extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2257
2258static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2259{
2260 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2261 * wanted then gso_type will be set. */
2262 struct skb_shared_info *shinfo = skb_shinfo(skb);
Alexander Duyckb78462e2010-06-02 12:24:37 +00002263 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2264 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07002265 __skb_warn_lro_forwarding(skb);
2266 return true;
2267 }
2268 return false;
2269}
2270
Herbert Xu35fc92a2007-03-26 23:22:20 -07002271static inline void skb_forward_csum(struct sk_buff *skb)
2272{
2273 /* Unfortunately we don't support this one. Any brave souls? */
2274 if (skb->ip_summed == CHECKSUM_COMPLETE)
2275 skb->ip_summed = CHECKSUM_NONE;
2276}
2277
Eric Dumazetbc8acf22010-09-02 13:07:41 -07002278/**
2279 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2280 * @skb: skb to check
2281 *
2282 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2283 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2284 * use this helper, to document places where we make this assertion.
2285 */
2286static inline void skb_checksum_none_assert(struct sk_buff *skb)
2287{
2288#ifdef DEBUG
2289 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2290#endif
2291}
2292
Rusty Russellf35d9d82008-02-04 23:49:54 -05002293bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00002294
Linus Torvalds1da177e2005-04-16 15:20:36 -07002295#endif /* __KERNEL__ */
2296#endif /* _LINUX_SKBUFF_H */