blob: 0f966460a345d8ff0ad82ce9e7b7fa6b84251bc8 [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
Randy Dunlapf4b8ea72006-06-22 16:00:11 -0700325 * @dma_cookie: a cookie to one of several possible DMA operations
326 * done by skb DMA functions
James Morris984bc162006-06-09 00:29:17 -0700327 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700328 * @mark: Generic packet mark
329 * @dropcount: total number of sk_receive_queue overflows
Patrick McHardy6aa895b02008-07-14 22:49:06 -0700330 * @vlan_tci: vlan tag control information
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700331 * @transport_header: Transport layer header
332 * @network_header: Network layer header
333 * @mac_header: Link layer header
334 * @tail: Tail pointer
335 * @end: End pointer
336 * @head: Head of buffer
337 * @data: Data head pointer
338 * @truesize: Buffer size
339 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340 */
341
342struct sk_buff {
343 /* These two members must be first. */
344 struct sk_buff *next;
345 struct sk_buff *prev;
346
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700347 ktime_t tstamp;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000348
349 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352 /*
353 * This is the control buffer. It is free to use for every
354 * layer. Please put your private variables there. If you
355 * want to keep them across layers you have to do a skb_clone()
356 * first. This is owned by whoever has the skb queued ATM.
357 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000358 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359
Eric Dumazet7fee2262010-05-11 23:19:48 +0000360 unsigned long _skb_refdst;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000361#ifdef CONFIG_XFRM
362 struct sec_path *sp;
363#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700365 data_len;
366 __u16 mac_len,
367 hdr_len;
Al Viroff1dcad2006-11-20 18:07:29 -0800368 union {
369 __wsum csum;
Herbert Xu663ead32007-04-09 11:59:07 -0700370 struct {
371 __u16 csum_start;
372 __u16 csum_offset;
373 };
Al Viroff1dcad2006-11-20 18:07:29 -0800374 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700375 __u32 priority;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200376 kmemcheck_bitfield_begin(flags1);
Thomas Graf1cbb3382005-07-05 14:13:41 -0700377 __u8 local_df:1,
378 cloned:1,
379 ip_summed:2,
Harald Welte6869c4d2005-08-09 19:24:19 -0700380 nohdr:1,
381 nfctinfo:3;
David S. Millerd179cd12005-08-17 14:57:30 -0700382 __u8 pkt_type:3,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800383 fclone:2,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700384 ipvs_property:1,
Herbert Xua59322b2007-12-05 01:53:40 -0800385 peeked:1,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700386 nf_trace:1;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200387 kmemcheck_bitfield_end(flags1);
Eric Dumazet4ab408d2010-03-01 03:09:26 +0000388 __be16 protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389
390 void (*destructor)(struct sk_buff *skb);
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800391#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700392 struct nf_conntrack *nfct;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100393#endif
394#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800395 struct sk_buff *nfct_reasm;
396#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397#ifdef CONFIG_BRIDGE_NETFILTER
398 struct nf_bridge_info *nf_bridge;
399#endif
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700400
Eric Dumazet8964be42009-11-20 15:35:04 -0800401 int skb_iif;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700403 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700404#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700405 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200408
Tom Herbert0a9627f2010-03-16 08:03:29 +0000409 __u32 rxhash;
410
Eric Dumazet0a148422011-04-20 09:27:32 +0000411 __u16 queue_mapping;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200412 kmemcheck_bitfield_begin(flags2);
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400413#ifdef CONFIG_IPV6_NDISC_NODETYPE
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000414 __u8 ndisc_nodetype:2;
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400415#endif
Tom Herbert3853b582010-11-21 13:17:29 +0000416 __u8 ooo_okay:1;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200417 kmemcheck_bitfield_end(flags2);
418
Tom Herbert3853b582010-11-21 13:17:29 +0000419 /* 0/13 bit hole */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700420
Chris Leech97fc2f02006-05-23 17:55:33 -0700421#ifdef CONFIG_NET_DMA
422 dma_cookie_t dma_cookie;
423#endif
James Morris984bc162006-06-09 00:29:17 -0700424#ifdef CONFIG_NETWORK_SECMARK
425 __u32 secmark;
426#endif
Neil Horman3b885782009-10-12 13:26:31 -0700427 union {
428 __u32 mark;
429 __u32 dropcount;
430 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431
Patrick McHardy6aa895b02008-07-14 22:49:06 -0700432 __u16 vlan_tci;
433
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700434 sk_buff_data_t transport_header;
435 sk_buff_data_t network_header;
436 sk_buff_data_t mac_header;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700438 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700439 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700441 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700442 unsigned int truesize;
443 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444};
445
446#ifdef __KERNEL__
447/*
448 * Handling routines are only of interest to the kernel
449 */
450#include <linux/slab.h>
451
452#include <asm/system.h>
453
Eric Dumazet7fee2262010-05-11 23:19:48 +0000454/*
455 * skb might have a dst pointer attached, refcounted or not.
456 * _skb_refdst low order bit is set if refcount was _not_ taken
457 */
458#define SKB_DST_NOREF 1UL
459#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
460
461/**
462 * skb_dst - returns skb dst_entry
463 * @skb: buffer
464 *
465 * Returns skb dst_entry, regardless of reference taken or not.
466 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000467static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
468{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000469 /* If refdst was not refcounted, check we still are in a
470 * rcu_read_lock section
471 */
472 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
473 !rcu_read_lock_held() &&
474 !rcu_read_lock_bh_held());
475 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000476}
477
Eric Dumazet7fee2262010-05-11 23:19:48 +0000478/**
479 * skb_dst_set - sets skb dst
480 * @skb: buffer
481 * @dst: dst entry
482 *
483 * Sets skb dst, assuming a reference was taken on dst and should
484 * be released by skb_dst_drop()
485 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000486static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
487{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000488 skb->_skb_refdst = (unsigned long)dst;
489}
490
Eric Dumazet27b75c92010-10-15 05:44:11 +0000491extern void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst);
Eric Dumazet7fee2262010-05-11 23:19:48 +0000492
493/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300494 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000495 * @skb: buffer
496 */
497static inline bool skb_dst_is_noref(const struct sk_buff *skb)
498{
499 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000500}
501
Eric Dumazet511c3f92009-06-02 05:14:27 +0000502static inline struct rtable *skb_rtable(const struct sk_buff *skb)
503{
Eric Dumazetadf30902009-06-02 05:19:30 +0000504 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000505}
506
Jörn Engel231d06a2006-03-20 21:28:35 -0800507extern void kfree_skb(struct sk_buff *skb);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000508extern void consume_skb(struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700509extern void __kfree_skb(struct sk_buff *skb);
David S. Millerd179cd12005-08-17 14:57:30 -0700510extern struct sk_buff *__alloc_skb(unsigned int size,
Christoph Hellwigb30973f2006-12-06 20:32:36 -0800511 gfp_t priority, int fclone, int node);
David S. Millerd179cd12005-08-17 14:57:30 -0700512static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100513 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700514{
Eric Dumazet564824b2010-10-11 19:05:25 +0000515 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700516}
517
518static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100519 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700520{
Eric Dumazet564824b2010-10-11 19:05:25 +0000521 return __alloc_skb(size, priority, 1, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700522}
523
Changli Gao5b0daa32010-05-29 00:12:13 -0700524extern bool skb_recycle_check(struct sk_buff *skb, int skb_size);
Lennert Buytenhek04a4bb52008-10-01 02:33:12 -0700525
Herbert Xue0053ec2007-10-14 00:37:52 -0700526extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
Michael S. Tsirkin48c83012011-08-31 08:03:29 +0000527extern int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700528extern struct sk_buff *skb_clone(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100529 gfp_t priority);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700530extern struct sk_buff *skb_copy(const 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 *pskb_copy(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100533 gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534extern int pskb_expand_head(struct sk_buff *skb,
Victor Fusco86a76ca2005-07-08 14:57:47 -0700535 int nhead, int ntail,
Al Virodd0fc662005-10-07 07:46:04 +0100536 gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
538 unsigned int headroom);
539extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
540 int newheadroom, int newtailroom,
Al Virodd0fc662005-10-07 07:46:04 +0100541 gfp_t priority);
David Howells716ea3a2007-04-02 20:19:53 -0700542extern int skb_to_sgvec(struct sk_buff *skb,
543 struct scatterlist *sg, int offset,
544 int len);
545extern int skb_cow_data(struct sk_buff *skb, int tailbits,
546 struct sk_buff **trailer);
Herbert Xu5b057c62006-06-23 02:06:41 -0700547extern int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000548#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700550extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
551 int getfrag(void *from, char *to, int offset,
552 int len,int odd, struct sk_buff *skb),
553 void *from, int length);
554
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800555struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700556 __u32 lower_offset;
557 __u32 upper_offset;
558 __u32 frag_idx;
559 __u32 stepped_offset;
560 struct sk_buff *root_skb;
561 struct sk_buff *cur_skb;
562 __u8 *frag_data;
563};
564
565extern void skb_prepare_seq_read(struct sk_buff *skb,
566 unsigned int from, unsigned int to,
567 struct skb_seq_state *st);
568extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
569 struct skb_seq_state *st);
570extern void skb_abort_seq_read(struct skb_seq_state *st);
571
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700572extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
573 unsigned int to, struct ts_config *config,
574 struct ts_state *state);
575
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000576extern __u32 __skb_get_rxhash(struct sk_buff *skb);
577static inline __u32 skb_get_rxhash(struct sk_buff *skb)
578{
579 if (!skb->rxhash)
580 skb->rxhash = __skb_get_rxhash(skb);
581
582 return skb->rxhash;
583}
584
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700585#ifdef NET_SKBUFF_DATA_USES_OFFSET
586static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
587{
588 return skb->head + skb->end;
589}
590#else
591static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
592{
593 return skb->end;
594}
595#endif
596
Linus Torvalds1da177e2005-04-16 15:20:36 -0700597/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700598#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700599
Patrick Ohlyac45f602009-02-12 05:03:37 +0000600static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
601{
602 return &skb_shinfo(skb)->hwtstamps;
603}
604
Linus Torvalds1da177e2005-04-16 15:20:36 -0700605/**
606 * skb_queue_empty - check if a queue is empty
607 * @list: queue head
608 *
609 * Returns true if the queue is empty, false otherwise.
610 */
611static inline int skb_queue_empty(const struct sk_buff_head *list)
612{
613 return list->next == (struct sk_buff *)list;
614}
615
616/**
David S. Millerfc7ebb22008-09-23 00:34:07 -0700617 * skb_queue_is_last - check if skb is the last entry in the queue
618 * @list: queue head
619 * @skb: buffer
620 *
621 * Returns true if @skb is the last buffer on the list.
622 */
623static inline bool skb_queue_is_last(const struct sk_buff_head *list,
624 const struct sk_buff *skb)
625{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000626 return skb->next == (struct sk_buff *)list;
David S. Millerfc7ebb22008-09-23 00:34:07 -0700627}
628
629/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800630 * skb_queue_is_first - check if skb is the first entry in the queue
631 * @list: queue head
632 * @skb: buffer
633 *
634 * Returns true if @skb is the first buffer on the list.
635 */
636static inline bool skb_queue_is_first(const struct sk_buff_head *list,
637 const struct sk_buff *skb)
638{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000639 return skb->prev == (struct sk_buff *)list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800640}
641
642/**
David S. Miller249c8b42008-09-23 00:44:42 -0700643 * skb_queue_next - return the next packet in the queue
644 * @list: queue head
645 * @skb: current buffer
646 *
647 * Return the next packet in @list after @skb. It is only valid to
648 * call this if skb_queue_is_last() evaluates to false.
649 */
650static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
651 const struct sk_buff *skb)
652{
653 /* This BUG_ON may seem severe, but if we just return then we
654 * are going to dereference garbage.
655 */
656 BUG_ON(skb_queue_is_last(list, skb));
657 return skb->next;
658}
659
660/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800661 * skb_queue_prev - return the prev packet in the queue
662 * @list: queue head
663 * @skb: current buffer
664 *
665 * Return the prev packet in @list before @skb. It is only valid to
666 * call this if skb_queue_is_first() evaluates to false.
667 */
668static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
669 const struct sk_buff *skb)
670{
671 /* This BUG_ON may seem severe, but if we just return then we
672 * are going to dereference garbage.
673 */
674 BUG_ON(skb_queue_is_first(list, skb));
675 return skb->prev;
676}
677
678/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679 * skb_get - reference buffer
680 * @skb: buffer to reference
681 *
682 * Makes another reference to a socket buffer and returns a pointer
683 * to the buffer.
684 */
685static inline struct sk_buff *skb_get(struct sk_buff *skb)
686{
687 atomic_inc(&skb->users);
688 return skb;
689}
690
691/*
692 * If users == 1, we are the only owner and are can avoid redundant
693 * atomic change.
694 */
695
696/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697 * skb_cloned - is the buffer a clone
698 * @skb: buffer to check
699 *
700 * Returns true if the buffer was generated with skb_clone() and is
701 * one of multiple shared copies of the buffer. Cloned buffers are
702 * shared data so must not be written to under normal circumstances.
703 */
704static inline int skb_cloned(const struct sk_buff *skb)
705{
706 return skb->cloned &&
707 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
708}
709
710/**
711 * skb_header_cloned - is the header a clone
712 * @skb: buffer to check
713 *
714 * Returns true if modifying the header part of the buffer requires
715 * the data to be copied.
716 */
717static inline int skb_header_cloned(const struct sk_buff *skb)
718{
719 int dataref;
720
721 if (!skb->cloned)
722 return 0;
723
724 dataref = atomic_read(&skb_shinfo(skb)->dataref);
725 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
726 return dataref != 1;
727}
728
729/**
730 * skb_header_release - release reference to header
731 * @skb: buffer to operate on
732 *
733 * Drop a reference to the header part of the buffer. This is done
734 * by acquiring a payload reference. You must not read from the header
735 * part of skb->data after this.
736 */
737static inline void skb_header_release(struct sk_buff *skb)
738{
739 BUG_ON(skb->nohdr);
740 skb->nohdr = 1;
741 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
742}
743
744/**
745 * skb_shared - is the buffer shared
746 * @skb: buffer to check
747 *
748 * Returns true if more than one person has a reference to this
749 * buffer.
750 */
751static inline int skb_shared(const struct sk_buff *skb)
752{
753 return atomic_read(&skb->users) != 1;
754}
755
756/**
757 * skb_share_check - check if buffer is shared and if so clone it
758 * @skb: buffer to check
759 * @pri: priority for memory allocation
760 *
761 * If the buffer is shared the buffer is cloned and the old copy
762 * drops a reference. A new clone with a single reference is returned.
763 * If the buffer is not shared the original buffer is returned. When
764 * being called from interrupt status or with spinlocks held pri must
765 * be GFP_ATOMIC.
766 *
767 * NULL is returned on a memory allocation failure.
768 */
Victor Fusco86a76ca2005-07-08 14:57:47 -0700769static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100770 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700771{
772 might_sleep_if(pri & __GFP_WAIT);
773 if (skb_shared(skb)) {
774 struct sk_buff *nskb = skb_clone(skb, pri);
775 kfree_skb(skb);
776 skb = nskb;
777 }
778 return skb;
779}
780
781/*
782 * Copy shared buffers into a new sk_buff. We effectively do COW on
783 * packets to handle cases where we have a local reader and forward
784 * and a couple of other messy ones. The normal one is tcpdumping
785 * a packet thats being forwarded.
786 */
787
788/**
789 * skb_unshare - make a copy of a shared buffer
790 * @skb: buffer to check
791 * @pri: priority for memory allocation
792 *
793 * If the socket buffer is a clone then this function creates a new
794 * copy of the data, drops a reference count on the old copy and returns
795 * the new copy with the reference count at 1. If the buffer is not a clone
796 * the original buffer is returned. When called with a spinlock held or
797 * from interrupt state @pri must be %GFP_ATOMIC
798 *
799 * %NULL is returned on a memory allocation failure.
800 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -0700801static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100802 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700803{
804 might_sleep_if(pri & __GFP_WAIT);
805 if (skb_cloned(skb)) {
806 struct sk_buff *nskb = skb_copy(skb, pri);
807 kfree_skb(skb); /* Free our shared copy */
808 skb = nskb;
809 }
810 return skb;
811}
812
813/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -0800814 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815 * @list_: list to peek at
816 *
817 * Peek an &sk_buff. Unlike most other operations you _MUST_
818 * be careful with this one. A peek leaves the buffer on the
819 * list and someone else may run off with it. You must hold
820 * the appropriate locks or have a private queue to do this.
821 *
822 * Returns %NULL for an empty list or a pointer to the head element.
823 * The reference count is not incremented and the reference is therefore
824 * volatile. Use with caution.
825 */
826static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
827{
828 struct sk_buff *list = ((struct sk_buff *)list_)->next;
829 if (list == (struct sk_buff *)list_)
830 list = NULL;
831 return list;
832}
833
834/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -0800835 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -0700836 * @list_: list to peek at
837 *
838 * Peek an &sk_buff. Unlike most other operations you _MUST_
839 * be careful with this one. A peek leaves the buffer on the
840 * list and someone else may run off with it. You must hold
841 * the appropriate locks or have a private queue to do this.
842 *
843 * Returns %NULL for an empty list or a pointer to the tail element.
844 * The reference count is not incremented and the reference is therefore
845 * volatile. Use with caution.
846 */
847static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
848{
849 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
850 if (list == (struct sk_buff *)list_)
851 list = NULL;
852 return list;
853}
854
855/**
856 * skb_queue_len - get queue length
857 * @list_: list to measure
858 *
859 * Return the length of an &sk_buff queue.
860 */
861static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
862{
863 return list_->qlen;
864}
865
David S. Miller67fed452008-09-21 22:36:24 -0700866/**
867 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
868 * @list: queue to initialize
869 *
870 * This initializes only the list and queue length aspects of
871 * an sk_buff_head object. This allows to initialize the list
872 * aspects of an sk_buff_head without reinitializing things like
873 * the spinlock. It can also be used for on-stack sk_buff_head
874 * objects where the spinlock is known to not be used.
875 */
876static inline void __skb_queue_head_init(struct sk_buff_head *list)
877{
878 list->prev = list->next = (struct sk_buff *)list;
879 list->qlen = 0;
880}
881
Arjan van de Ven76f10ad2006-08-02 14:06:55 -0700882/*
883 * This function creates a split out lock class for each invocation;
884 * this is needed for now since a whole lot of users of the skb-queue
885 * infrastructure in drivers have different locking usage (in hardirq)
886 * than the networking core (in softirq only). In the long run either the
887 * network layer or drivers should need annotation to consolidate the
888 * main types of usage into 3 classes.
889 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890static inline void skb_queue_head_init(struct sk_buff_head *list)
891{
892 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -0700893 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894}
895
Pavel Emelianovc2ecba72007-04-17 12:45:31 -0700896static inline void skb_queue_head_init_class(struct sk_buff_head *list,
897 struct lock_class_key *class)
898{
899 skb_queue_head_init(list);
900 lockdep_set_class(&list->lock, class);
901}
902
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903/*
Gerrit Renkerbf299272008-04-14 00:04:51 -0700904 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905 *
906 * The "__skb_xxxx()" functions are the non-atomic ones that
907 * can only be called with interrupts disabled.
908 */
Gerrit Renkerbf299272008-04-14 00:04:51 -0700909extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
910static inline void __skb_insert(struct sk_buff *newsk,
911 struct sk_buff *prev, struct sk_buff *next,
912 struct sk_buff_head *list)
913{
914 newsk->next = next;
915 newsk->prev = prev;
916 next->prev = prev->next = newsk;
917 list->qlen++;
918}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700919
David S. Miller67fed452008-09-21 22:36:24 -0700920static inline void __skb_queue_splice(const struct sk_buff_head *list,
921 struct sk_buff *prev,
922 struct sk_buff *next)
923{
924 struct sk_buff *first = list->next;
925 struct sk_buff *last = list->prev;
926
927 first->prev = prev;
928 prev->next = first;
929
930 last->next = next;
931 next->prev = last;
932}
933
934/**
935 * skb_queue_splice - join two skb lists, this is designed for stacks
936 * @list: the new list to add
937 * @head: the place to add it in the first list
938 */
939static inline void skb_queue_splice(const struct sk_buff_head *list,
940 struct sk_buff_head *head)
941{
942 if (!skb_queue_empty(list)) {
943 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -0700944 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -0700945 }
946}
947
948/**
949 * skb_queue_splice - join two skb lists and reinitialise the emptied list
950 * @list: the new list to add
951 * @head: the place to add it in the first list
952 *
953 * The list at @list is reinitialised
954 */
955static inline void skb_queue_splice_init(struct sk_buff_head *list,
956 struct sk_buff_head *head)
957{
958 if (!skb_queue_empty(list)) {
959 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -0700960 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -0700961 __skb_queue_head_init(list);
962 }
963}
964
965/**
966 * skb_queue_splice_tail - join two skb lists, each list being a queue
967 * @list: the new list to add
968 * @head: the place to add it in the first list
969 */
970static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
971 struct sk_buff_head *head)
972{
973 if (!skb_queue_empty(list)) {
974 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -0700975 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -0700976 }
977}
978
979/**
980 * skb_queue_splice_tail - join two skb lists and reinitialise the emptied list
981 * @list: the new list to add
982 * @head: the place to add it in the first list
983 *
984 * Each of the lists is a queue.
985 * The list at @list is reinitialised
986 */
987static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
988 struct sk_buff_head *head)
989{
990 if (!skb_queue_empty(list)) {
991 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -0700992 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -0700993 __skb_queue_head_init(list);
994 }
995}
996
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997/**
Stephen Hemminger300ce172005-10-30 13:47:34 -0800998 * __skb_queue_after - queue a buffer at the list head
999 * @list: list to use
1000 * @prev: place after this buffer
1001 * @newsk: buffer to queue
1002 *
1003 * Queue a buffer int the middle of a list. This function takes no locks
1004 * and you must therefore hold required locks before calling it.
1005 *
1006 * A buffer cannot be placed on two lists at the same time.
1007 */
1008static inline void __skb_queue_after(struct sk_buff_head *list,
1009 struct sk_buff *prev,
1010 struct sk_buff *newsk)
1011{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001012 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001013}
1014
Gerrit Renker7de6c032008-04-14 00:05:09 -07001015extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1016 struct sk_buff_head *list);
1017
Gerrit Renkerf5572852008-04-14 00:05:28 -07001018static inline void __skb_queue_before(struct sk_buff_head *list,
1019 struct sk_buff *next,
1020 struct sk_buff *newsk)
1021{
1022 __skb_insert(newsk, next->prev, next, list);
1023}
1024
Stephen Hemminger300ce172005-10-30 13:47:34 -08001025/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026 * __skb_queue_head - queue a buffer at the list head
1027 * @list: list to use
1028 * @newsk: buffer to queue
1029 *
1030 * Queue a buffer at the start of a list. This function takes no locks
1031 * and you must therefore hold required locks before calling it.
1032 *
1033 * A buffer cannot be placed on two lists at the same time.
1034 */
1035extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1036static inline void __skb_queue_head(struct sk_buff_head *list,
1037 struct sk_buff *newsk)
1038{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001039 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001040}
1041
1042/**
1043 * __skb_queue_tail - queue a buffer at the list tail
1044 * @list: list to use
1045 * @newsk: buffer to queue
1046 *
1047 * Queue a buffer at the end of a list. This function takes no locks
1048 * and you must therefore hold required locks before calling it.
1049 *
1050 * A buffer cannot be placed on two lists at the same time.
1051 */
1052extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1053static inline void __skb_queue_tail(struct sk_buff_head *list,
1054 struct sk_buff *newsk)
1055{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001056 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057}
1058
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060 * remove sk_buff from list. _Must_ be called atomically, and with
1061 * the list known..
1062 */
David S. Miller8728b832005-08-09 19:25:21 -07001063extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001064static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1065{
1066 struct sk_buff *next, *prev;
1067
1068 list->qlen--;
1069 next = skb->next;
1070 prev = skb->prev;
1071 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001072 next->prev = prev;
1073 prev->next = next;
1074}
1075
Gerrit Renkerf525c062008-04-14 00:04:12 -07001076/**
1077 * __skb_dequeue - remove from the head of the queue
1078 * @list: list to dequeue from
1079 *
1080 * Remove the head of the list. This function does not take any locks
1081 * so must be used with appropriate locks held only. The head item is
1082 * returned or %NULL if the list is empty.
1083 */
1084extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1085static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1086{
1087 struct sk_buff *skb = skb_peek(list);
1088 if (skb)
1089 __skb_unlink(skb, list);
1090 return skb;
1091}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001092
1093/**
1094 * __skb_dequeue_tail - remove from the tail of the queue
1095 * @list: list to dequeue from
1096 *
1097 * Remove the tail of the list. This function does not take any locks
1098 * so must be used with appropriate locks held only. The tail item is
1099 * returned or %NULL if the list is empty.
1100 */
1101extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1102static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1103{
1104 struct sk_buff *skb = skb_peek_tail(list);
1105 if (skb)
1106 __skb_unlink(skb, list);
1107 return skb;
1108}
1109
1110
1111static inline int skb_is_nonlinear(const struct sk_buff *skb)
1112{
1113 return skb->data_len;
1114}
1115
1116static inline unsigned int skb_headlen(const struct sk_buff *skb)
1117{
1118 return skb->len - skb->data_len;
1119}
1120
1121static inline int skb_pagelen(const struct sk_buff *skb)
1122{
1123 int i, len = 0;
1124
1125 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1126 len += skb_shinfo(skb)->frags[i].size;
1127 return len + skb_headlen(skb);
1128}
1129
1130static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1131 struct page *page, int off, int size)
1132{
1133 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1134
1135 frag->page = page;
1136 frag->page_offset = off;
1137 frag->size = size;
1138 skb_shinfo(skb)->nr_frags = i + 1;
1139}
1140
Peter Zijlstra654bed12008-10-07 14:22:33 -07001141extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
1142 int off, int size);
1143
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001145#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1147
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001148#ifdef NET_SKBUFF_DATA_USES_OFFSET
1149static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1150{
1151 return skb->head + skb->tail;
1152}
1153
1154static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1155{
1156 skb->tail = skb->data - skb->head;
1157}
1158
1159static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1160{
1161 skb_reset_tail_pointer(skb);
1162 skb->tail += offset;
1163}
1164#else /* NET_SKBUFF_DATA_USES_OFFSET */
1165static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1166{
1167 return skb->tail;
1168}
1169
1170static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1171{
1172 skb->tail = skb->data;
1173}
1174
1175static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1176{
1177 skb->tail = skb->data + offset;
1178}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001179
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001180#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1181
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182/*
1183 * Add data to an sk_buff
1184 */
Ilpo Järvinen0dde3e12008-03-27 17:43:41 -07001185extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001186static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1187{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001188 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001189 SKB_LINEAR_ASSERT(skb);
1190 skb->tail += len;
1191 skb->len += len;
1192 return tmp;
1193}
1194
Ilpo Järvinenc2aa2702008-03-27 17:52:40 -07001195extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1197{
1198 skb->data -= len;
1199 skb->len += len;
1200 return skb->data;
1201}
1202
Ilpo Järvinen6be8ac22008-03-27 17:47:24 -07001203extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001204static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1205{
1206 skb->len -= len;
1207 BUG_ON(skb->len < skb->data_len);
1208 return skb->data += len;
1209}
1210
David S. Miller47d29642010-05-02 02:21:44 -07001211static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1212{
1213 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1214}
1215
Linus Torvalds1da177e2005-04-16 15:20:36 -07001216extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1217
1218static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1219{
1220 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001221 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001222 return NULL;
1223 skb->len -= len;
1224 return skb->data += len;
1225}
1226
1227static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1228{
1229 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1230}
1231
1232static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1233{
1234 if (likely(len <= skb_headlen(skb)))
1235 return 1;
1236 if (unlikely(len > skb->len))
1237 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001238 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001239}
1240
1241/**
1242 * skb_headroom - bytes at buffer head
1243 * @skb: buffer to check
1244 *
1245 * Return the number of bytes of free space at the head of an &sk_buff.
1246 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001247static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248{
1249 return skb->data - skb->head;
1250}
1251
1252/**
1253 * skb_tailroom - bytes at buffer end
1254 * @skb: buffer to check
1255 *
1256 * Return the number of bytes of free space at the tail of an sk_buff
1257 */
1258static inline int skb_tailroom(const struct sk_buff *skb)
1259{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001260 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001261}
1262
1263/**
1264 * skb_reserve - adjust headroom
1265 * @skb: buffer to alter
1266 * @len: bytes to move
1267 *
1268 * Increase the headroom of an empty &sk_buff by reducing the tail
1269 * room. This is only allowed for an empty buffer.
1270 */
David S. Miller8243126c2006-01-17 02:54:21 -08001271static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001272{
1273 skb->data += len;
1274 skb->tail += len;
1275}
1276
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00001277static inline void skb_reset_mac_len(struct sk_buff *skb)
1278{
1279 skb->mac_len = skb->network_header - skb->mac_header;
1280}
1281
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001282#ifdef NET_SKBUFF_DATA_USES_OFFSET
1283static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1284{
1285 return skb->head + skb->transport_header;
1286}
1287
1288static inline void skb_reset_transport_header(struct sk_buff *skb)
1289{
1290 skb->transport_header = skb->data - skb->head;
1291}
1292
1293static inline void skb_set_transport_header(struct sk_buff *skb,
1294 const int offset)
1295{
1296 skb_reset_transport_header(skb);
1297 skb->transport_header += offset;
1298}
1299
1300static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1301{
1302 return skb->head + skb->network_header;
1303}
1304
1305static inline void skb_reset_network_header(struct sk_buff *skb)
1306{
1307 skb->network_header = skb->data - skb->head;
1308}
1309
1310static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1311{
1312 skb_reset_network_header(skb);
1313 skb->network_header += offset;
1314}
1315
1316static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1317{
1318 return skb->head + skb->mac_header;
1319}
1320
1321static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1322{
1323 return skb->mac_header != ~0U;
1324}
1325
1326static inline void skb_reset_mac_header(struct sk_buff *skb)
1327{
1328 skb->mac_header = skb->data - skb->head;
1329}
1330
1331static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1332{
1333 skb_reset_mac_header(skb);
1334 skb->mac_header += offset;
1335}
1336
1337#else /* NET_SKBUFF_DATA_USES_OFFSET */
1338
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001339static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1340{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001341 return skb->transport_header;
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001342}
1343
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001344static inline void skb_reset_transport_header(struct sk_buff *skb)
1345{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001346 skb->transport_header = skb->data;
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001347}
1348
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001349static inline void skb_set_transport_header(struct sk_buff *skb,
1350 const int offset)
1351{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001352 skb->transport_header = skb->data + offset;
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001353}
1354
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001355static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1356{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001357 return skb->network_header;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001358}
1359
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001360static inline void skb_reset_network_header(struct sk_buff *skb)
1361{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001362 skb->network_header = skb->data;
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001363}
1364
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001365static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1366{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001367 skb->network_header = skb->data + offset;
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001368}
1369
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001370static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1371{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001372 return skb->mac_header;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001373}
1374
1375static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1376{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001377 return skb->mac_header != NULL;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001378}
1379
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001380static inline void skb_reset_mac_header(struct sk_buff *skb)
1381{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001382 skb->mac_header = skb->data;
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001383}
1384
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001385static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1386{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001387 skb->mac_header = skb->data + offset;
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001388}
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001389#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1390
Michał Mirosław04fb4512010-12-14 15:24:08 +00001391static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1392{
1393 return skb->csum_start - skb_headroom(skb);
1394}
1395
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001396static inline int skb_transport_offset(const struct sk_buff *skb)
1397{
1398 return skb_transport_header(skb) - skb->data;
1399}
1400
1401static inline u32 skb_network_header_len(const struct sk_buff *skb)
1402{
1403 return skb->transport_header - skb->network_header;
1404}
1405
1406static inline int skb_network_offset(const struct sk_buff *skb)
1407{
1408 return skb_network_header(skb) - skb->data;
1409}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001410
Changli Gaof9599ce2010-08-04 04:43:44 +00001411static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1412{
1413 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1414}
1415
Linus Torvalds1da177e2005-04-16 15:20:36 -07001416/*
1417 * CPUs often take a performance hit when accessing unaligned memory
1418 * locations. The actual performance hit varies, it can be small if the
1419 * hardware handles it or large if we have to take an exception and fix it
1420 * in software.
1421 *
1422 * Since an ethernet header is 14 bytes network drivers often end up with
1423 * the IP header at an unaligned offset. The IP header can be aligned by
1424 * shifting the start of the packet by 2 bytes. Drivers should do this
1425 * with:
1426 *
Tobias Klauser8660c122009-07-13 22:48:16 +00001427 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428 *
1429 * The downside to this alignment of the IP header is that the DMA is now
1430 * unaligned. On some architectures the cost of an unaligned DMA is high
1431 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00001432 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001433 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1434 * to be overridden.
1435 */
1436#ifndef NET_IP_ALIGN
1437#define NET_IP_ALIGN 2
1438#endif
1439
Anton Blanchard025be812006-03-31 02:27:06 -08001440/*
1441 * The networking layer reserves some headroom in skb data (via
1442 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1443 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08001444 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08001445 *
1446 * Unfortunately this headroom changes the DMA alignment of the resulting
1447 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1448 * on some architectures. An architecture can override this value,
1449 * perhaps setting it to a cacheline in size (since that will maintain
1450 * cacheline alignment of the DMA). It must be a power of 2.
1451 *
David S. Millerd6301d32009-02-08 19:24:13 -08001452 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08001453 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07001454 *
1455 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1456 * to reduce average number of cache lines per packet.
1457 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07001458 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08001459 */
1460#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07001461#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08001462#endif
1463
Herbert Xu3cc0e872006-06-09 16:13:38 -07001464extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001465
1466static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1467{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00001468 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07001469 WARN_ON(1);
1470 return;
1471 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001472 skb->len = len;
1473 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474}
1475
Ilpo Järvinen419ae742008-03-27 17:54:01 -07001476extern void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001477
1478static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1479{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001480 if (skb->data_len)
1481 return ___pskb_trim(skb, len);
1482 __skb_trim(skb, len);
1483 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001484}
1485
1486static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1487{
1488 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1489}
1490
1491/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07001492 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1493 * @skb: buffer to alter
1494 * @len: new length
1495 *
1496 * This is identical to pskb_trim except that the caller knows that
1497 * the skb is not cloned so we should never get an error due to out-
1498 * of-memory.
1499 */
1500static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1501{
1502 int err = pskb_trim(skb, len);
1503 BUG_ON(err);
1504}
1505
1506/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507 * skb_orphan - orphan a buffer
1508 * @skb: buffer to orphan
1509 *
1510 * If a buffer currently has an owner then we call the owner's
1511 * destructor function and make the @skb unowned. The buffer continues
1512 * to exist but is no longer charged to its former owner.
1513 */
1514static inline void skb_orphan(struct sk_buff *skb)
1515{
1516 if (skb->destructor)
1517 skb->destructor(skb);
1518 skb->destructor = NULL;
1519 skb->sk = NULL;
1520}
1521
1522/**
1523 * __skb_queue_purge - empty a list
1524 * @list: list to empty
1525 *
1526 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1527 * the list and one reference dropped. This function does not take the
1528 * list lock and the caller must hold the relevant locks to use it.
1529 */
1530extern void skb_queue_purge(struct sk_buff_head *list);
1531static inline void __skb_queue_purge(struct sk_buff_head *list)
1532{
1533 struct sk_buff *skb;
1534 while ((skb = __skb_dequeue(list)) != NULL)
1535 kfree_skb(skb);
1536}
1537
1538/**
Christoph Hellwigb4e54de2006-07-24 15:31:14 -07001539 * __dev_alloc_skb - allocate an skbuff for receiving
Linus Torvalds1da177e2005-04-16 15:20:36 -07001540 * @length: length to allocate
1541 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1542 *
1543 * Allocate a new &sk_buff and assign it a usage count of one. The
1544 * buffer has unspecified headroom built in. Users should allocate
1545 * the headroom they think they need without accounting for the
1546 * built in space. The built in space is used for optimisations.
1547 *
Christoph Hellwig766ea8c2006-08-07 15:49:53 -07001548 * %NULL is returned if there is no free memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
Al Virodd0fc662005-10-07 07:46:04 +01001551 gfp_t gfp_mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001552{
Anton Blanchard025be812006-03-31 02:27:06 -08001553 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001554 if (likely(skb))
Anton Blanchard025be812006-03-31 02:27:06 -08001555 skb_reserve(skb, NET_SKB_PAD);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001556 return skb;
1557}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558
Ilpo Järvinenf58518e2008-03-27 17:51:31 -07001559extern struct sk_buff *dev_alloc_skb(unsigned int length);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001560
Christoph Hellwig8af27452006-07-31 22:35:23 -07001561extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1562 unsigned int length, gfp_t gfp_mask);
1563
1564/**
1565 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1566 * @dev: network device to receive on
1567 * @length: length to allocate
1568 *
1569 * Allocate a new &sk_buff and assign it a usage count of one. The
1570 * buffer has unspecified headroom built in. Users should allocate
1571 * the headroom they think they need without accounting for the
1572 * built in space. The built in space is used for optimisations.
1573 *
1574 * %NULL is returned if there is no free memory. Although this function
1575 * allocates memory it can be called from an interrupt.
1576 */
1577static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1578 unsigned int length)
1579{
1580 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1581}
1582
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001583static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
1584 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00001585{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001586 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00001587
1588 if (NET_IP_ALIGN && skb)
1589 skb_reserve(skb, NET_IP_ALIGN);
1590 return skb;
1591}
1592
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001593static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1594 unsigned int length)
1595{
1596 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
1597}
1598
Eric Dumazet564824b2010-10-11 19:05:25 +00001599/**
1600 * __netdev_alloc_page - allocate a page for ps-rx on a specific device
1601 * @dev: network device to receive on
1602 * @gfp_mask: alloc_pages_node mask
1603 *
1604 * Allocate a new page. dev currently unused.
1605 *
1606 * %NULL is returned if there is no free memory.
1607 */
1608static inline struct page *__netdev_alloc_page(struct net_device *dev, gfp_t gfp_mask)
1609{
1610 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, 0);
1611}
Peter Zijlstra654bed12008-10-07 14:22:33 -07001612
1613/**
1614 * netdev_alloc_page - allocate a page for ps-rx on a specific device
1615 * @dev: network device to receive on
1616 *
Eric Dumazet564824b2010-10-11 19:05:25 +00001617 * Allocate a new page. dev currently unused.
Peter Zijlstra654bed12008-10-07 14:22:33 -07001618 *
1619 * %NULL is returned if there is no free memory.
1620 */
1621static inline struct page *netdev_alloc_page(struct net_device *dev)
1622{
1623 return __netdev_alloc_page(dev, GFP_ATOMIC);
1624}
1625
1626static inline void netdev_free_page(struct net_device *dev, struct page *page)
1627{
1628 __free_page(page);
1629}
1630
Linus Torvalds1da177e2005-04-16 15:20:36 -07001631/**
Patrick McHardy334a8132007-06-25 04:35:20 -07001632 * skb_clone_writable - is the header of a clone writable
1633 * @skb: buffer to check
1634 * @len: length up to which to write
1635 *
1636 * Returns true if modifying the header part of the cloned buffer
1637 * does not requires the data to be copied.
1638 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001639static inline int skb_clone_writable(struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07001640{
1641 return !skb_header_cloned(skb) &&
1642 skb_headroom(skb) + len <= skb->hdr_len;
1643}
1644
Herbert Xud9cc2042007-09-16 16:21:16 -07001645static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
1646 int cloned)
1647{
1648 int delta = 0;
1649
1650 if (headroom < NET_SKB_PAD)
1651 headroom = NET_SKB_PAD;
1652 if (headroom > skb_headroom(skb))
1653 delta = headroom - skb_headroom(skb);
1654
1655 if (delta || cloned)
1656 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
1657 GFP_ATOMIC);
1658 return 0;
1659}
1660
Patrick McHardy334a8132007-06-25 04:35:20 -07001661/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001662 * skb_cow - copy header of skb when it is required
1663 * @skb: buffer to cow
1664 * @headroom: needed headroom
1665 *
1666 * If the skb passed lacks sufficient headroom or its data part
1667 * is shared, data is reallocated. If reallocation fails, an error
1668 * is returned and original skb is not changed.
1669 *
1670 * The result is skb with writable area skb->head...skb->tail
1671 * and at least @headroom of space at head.
1672 */
1673static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1674{
Herbert Xud9cc2042007-09-16 16:21:16 -07001675 return __skb_cow(skb, headroom, skb_cloned(skb));
1676}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677
Herbert Xud9cc2042007-09-16 16:21:16 -07001678/**
1679 * skb_cow_head - skb_cow but only making the head writable
1680 * @skb: buffer to cow
1681 * @headroom: needed headroom
1682 *
1683 * This function is identical to skb_cow except that we replace the
1684 * skb_cloned check by skb_header_cloned. It should be used when
1685 * you only need to push on some header and do not need to modify
1686 * the data.
1687 */
1688static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
1689{
1690 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001691}
1692
1693/**
1694 * skb_padto - pad an skbuff up to a minimal size
1695 * @skb: buffer to pad
1696 * @len: minimal length
1697 *
1698 * Pads up a buffer to ensure the trailing bytes exist and are
1699 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07001700 * is untouched. Otherwise it is extended. Returns zero on
1701 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001702 */
1703
Herbert Xu5b057c62006-06-23 02:06:41 -07001704static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001705{
1706 unsigned int size = skb->len;
1707 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07001708 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001709 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710}
1711
1712static inline int skb_add_data(struct sk_buff *skb,
1713 char __user *from, int copy)
1714{
1715 const int off = skb->len;
1716
1717 if (skb->ip_summed == CHECKSUM_NONE) {
1718 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08001719 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
Linus Torvalds1da177e2005-04-16 15:20:36 -07001720 copy, 0, &err);
1721 if (!err) {
1722 skb->csum = csum_block_add(skb->csum, csum, off);
1723 return 0;
1724 }
1725 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1726 return 0;
1727
1728 __skb_trim(skb, off);
1729 return -EFAULT;
1730}
1731
1732static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1733 struct page *page, int off)
1734{
1735 if (i) {
1736 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1737
1738 return page == frag->page &&
1739 off == frag->page_offset + frag->size;
1740 }
1741 return 0;
1742}
1743
Herbert Xu364c6ba2006-06-09 16:10:40 -07001744static inline int __skb_linearize(struct sk_buff *skb)
1745{
1746 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1747}
1748
Linus Torvalds1da177e2005-04-16 15:20:36 -07001749/**
1750 * skb_linearize - convert paged skb to linear one
1751 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07001752 *
1753 * If there is no free memory -ENOMEM is returned, otherwise zero
1754 * is returned and the old skb data released.
1755 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07001756static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001757{
Herbert Xu364c6ba2006-06-09 16:10:40 -07001758 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1759}
1760
1761/**
1762 * skb_linearize_cow - make sure skb is linear and writable
1763 * @skb: buffer to process
1764 *
1765 * If there is no free memory -ENOMEM is returned, otherwise zero
1766 * is returned and the old skb data released.
1767 */
1768static inline int skb_linearize_cow(struct sk_buff *skb)
1769{
1770 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1771 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001772}
1773
1774/**
1775 * skb_postpull_rcsum - update checksum for received skb after pull
1776 * @skb: buffer to update
1777 * @start: start of data before pull
1778 * @len: length of data pulled
1779 *
1780 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07001781 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1782 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001783 */
1784
1785static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08001786 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001787{
Patrick McHardy84fa7932006-08-29 16:44:56 -07001788 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001789 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1790}
1791
Herbert Xucbb042f2006-03-20 22:43:56 -08001792unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1793
Linus Torvalds1da177e2005-04-16 15:20:36 -07001794/**
1795 * pskb_trim_rcsum - trim received skb and update checksum
1796 * @skb: buffer to trim
1797 * @len: new length
1798 *
1799 * This is exactly the same as pskb_trim except that it ensures the
1800 * checksum of received packets are still valid after the operation.
1801 */
1802
1803static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1804{
Stephen Hemminger0e4e4222005-09-08 12:32:03 -07001805 if (likely(len >= skb->len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001806 return 0;
Patrick McHardy84fa7932006-08-29 16:44:56 -07001807 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001808 skb->ip_summed = CHECKSUM_NONE;
1809 return __pskb_trim(skb, len);
1810}
1811
Linus Torvalds1da177e2005-04-16 15:20:36 -07001812#define skb_queue_walk(queue, skb) \
1813 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07001814 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07001815 skb = skb->next)
1816
James Chapman46f89142007-04-30 00:07:31 -07001817#define skb_queue_walk_safe(queue, skb, tmp) \
1818 for (skb = (queue)->next, tmp = skb->next; \
1819 skb != (struct sk_buff *)(queue); \
1820 skb = tmp, tmp = skb->next)
1821
David S. Miller1164f522008-09-23 00:49:44 -07001822#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07001823 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07001824 skb = skb->next)
1825
1826#define skb_queue_walk_from_safe(queue, skb, tmp) \
1827 for (tmp = skb->next; \
1828 skb != (struct sk_buff *)(queue); \
1829 skb = tmp, tmp = skb->next)
1830
Stephen Hemminger300ce172005-10-30 13:47:34 -08001831#define skb_queue_reverse_walk(queue, skb) \
1832 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07001833 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08001834 skb = skb->prev)
1835
David S. Miller686a2952011-01-20 22:47:32 -08001836#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
1837 for (skb = (queue)->prev, tmp = skb->prev; \
1838 skb != (struct sk_buff *)(queue); \
1839 skb = tmp, tmp = skb->prev)
1840
1841#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
1842 for (tmp = skb->prev; \
1843 skb != (struct sk_buff *)(queue); \
1844 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001845
David S. Miller21dc3302010-08-23 00:13:46 -07001846static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07001847{
1848 return skb_shinfo(skb)->frag_list != NULL;
1849}
1850
1851static inline void skb_frag_list_init(struct sk_buff *skb)
1852{
1853 skb_shinfo(skb)->frag_list = NULL;
1854}
1855
1856static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
1857{
1858 frag->next = skb_shinfo(skb)->frag_list;
1859 skb_shinfo(skb)->frag_list = frag;
1860}
1861
1862#define skb_walk_frags(skb, iter) \
1863 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
1864
Herbert Xua59322b2007-12-05 01:53:40 -08001865extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
1866 int *peeked, int *err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001867extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1868 int noblock, int *err);
1869extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1870 struct poll_table_struct *wait);
1871extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1872 int offset, struct iovec *to,
1873 int size);
Herbert Xufb286bb2005-11-10 13:01:24 -08001874extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001875 int hlen,
1876 struct iovec *iov);
Rusty Russelldb543c12008-08-15 15:13:53 -07001877extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
1878 int offset,
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +00001879 const struct iovec *from,
1880 int from_offset,
Rusty Russelldb543c12008-08-15 15:13:53 -07001881 int len);
Michael S. Tsirkin0a1ec072009-04-20 01:25:46 +00001882extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
1883 int offset,
1884 const struct iovec *to,
1885 int to_offset,
1886 int size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001887extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
Eric Dumazet9d410c72009-10-30 05:03:53 +00001888extern void skb_free_datagram_locked(struct sock *sk,
1889 struct sk_buff *skb);
Herbert Xu27ab2562007-12-05 01:51:58 -08001890extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
Herbert Xu3305b802005-12-13 23:16:37 -08001891 unsigned int flags);
Al Viro2bbbc862006-11-14 21:37:14 -08001892extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
1893 int len, __wsum csum);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001894extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1895 void *to, int len);
Stephen Hemminger0c6fcc82007-04-20 16:40:01 -07001896extern int skb_store_bits(struct sk_buff *skb, int offset,
1897 const void *from, int len);
Al Viro81d77662006-11-14 21:37:33 -08001898extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001899 int offset, u8 *to, int len,
Al Viro81d77662006-11-14 21:37:33 -08001900 __wsum csum);
Jens Axboe9c55e012007-11-06 23:30:13 -08001901extern int skb_splice_bits(struct sk_buff *skb,
1902 unsigned int offset,
1903 struct pipe_inode_info *pipe,
1904 unsigned int len,
1905 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001906extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1907extern void skb_split(struct sk_buff *skb,
1908 struct sk_buff *skb1, const u32 len);
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001909extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
1910 int shiftlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001911
Michał Mirosław04ed3e72011-01-24 15:32:47 -08001912extern struct sk_buff *skb_segment(struct sk_buff *skb, u32 features);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001913
Linus Torvalds1da177e2005-04-16 15:20:36 -07001914static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1915 int len, void *buffer)
1916{
1917 int hlen = skb_headlen(skb);
1918
Patrick McHardy55820ee2005-07-05 14:08:10 -07001919 if (hlen - offset >= len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001920 return skb->data + offset;
1921
1922 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1923 return NULL;
1924
1925 return buffer;
1926}
1927
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03001928static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
1929 void *to,
1930 const unsigned int len)
1931{
1932 memcpy(to, skb->data, len);
1933}
1934
1935static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
1936 const int offset, void *to,
1937 const unsigned int len)
1938{
1939 memcpy(to, skb->data + offset, len);
1940}
1941
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03001942static inline void skb_copy_to_linear_data(struct sk_buff *skb,
1943 const void *from,
1944 const unsigned int len)
1945{
1946 memcpy(skb->data, from, len);
1947}
1948
1949static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
1950 const int offset,
1951 const void *from,
1952 const unsigned int len)
1953{
1954 memcpy(skb->data + offset, from, len);
1955}
1956
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957extern void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001958
Patrick Ohlyac45f602009-02-12 05:03:37 +00001959static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
1960{
1961 return skb->tstamp;
1962}
1963
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001964/**
1965 * skb_get_timestamp - get timestamp from a skb
1966 * @skb: skb to get stamp from
1967 * @stamp: pointer to struct timeval to store stamp in
1968 *
1969 * Timestamps are stored in the skb as offsets to a base timestamp.
1970 * This function converts the offset back to a struct timeval and stores
1971 * it in stamp.
1972 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00001973static inline void skb_get_timestamp(const struct sk_buff *skb,
1974 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001975{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001976 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001977}
1978
Patrick Ohlyac45f602009-02-12 05:03:37 +00001979static inline void skb_get_timestampns(const struct sk_buff *skb,
1980 struct timespec *stamp)
1981{
1982 *stamp = ktime_to_timespec(skb->tstamp);
1983}
1984
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001985static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001986{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001987 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001988}
1989
Stephen Hemminger164891a2007-04-23 22:26:16 -07001990static inline ktime_t net_timedelta(ktime_t t)
1991{
1992 return ktime_sub(ktime_get_real(), t);
1993}
1994
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07001995static inline ktime_t net_invalid_timestamp(void)
1996{
1997 return ktime_set(0, 0);
1998}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001999
Richard Cochranc1f19b52010-07-17 08:49:36 +00002000extern void skb_timestamping_init(void);
2001
2002#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2003
2004extern void skb_clone_tx_timestamp(struct sk_buff *skb);
2005extern bool skb_defer_rx_timestamp(struct sk_buff *skb);
2006
2007#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2008
2009static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2010{
2011}
2012
2013static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2014{
2015 return false;
2016}
2017
2018#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2019
2020/**
2021 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2022 *
Richard Cochranda92b192011-10-21 00:49:15 +00002023 * PHY drivers may accept clones of transmitted packets for
2024 * timestamping via their phy_driver.txtstamp method. These drivers
2025 * must call this function to return the skb back to the stack, with
2026 * or without a timestamp.
2027 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00002028 * @skb: clone of the the original outgoing packet
Richard Cochranda92b192011-10-21 00:49:15 +00002029 * @hwtstamps: hardware time stamps, may be NULL if not available
Richard Cochranc1f19b52010-07-17 08:49:36 +00002030 *
2031 */
2032void skb_complete_tx_timestamp(struct sk_buff *skb,
2033 struct skb_shared_hwtstamps *hwtstamps);
2034
Patrick Ohlyac45f602009-02-12 05:03:37 +00002035/**
2036 * skb_tstamp_tx - queue clone of skb with send time stamps
2037 * @orig_skb: the original outgoing packet
2038 * @hwtstamps: hardware time stamps, may be NULL if not available
2039 *
2040 * If the skb has a socket associated, then this function clones the
2041 * skb (thus sharing the actual data and optional structures), stores
2042 * the optional hardware time stamping information (if non NULL) or
2043 * generates a software time stamp (otherwise), then queues the clone
2044 * to the error queue of the socket. Errors are silently ignored.
2045 */
2046extern void skb_tstamp_tx(struct sk_buff *orig_skb,
2047 struct skb_shared_hwtstamps *hwtstamps);
2048
Richard Cochran4507a712010-07-17 08:48:28 +00002049static inline void sw_tx_timestamp(struct sk_buff *skb)
2050{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002051 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2052 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00002053 skb_tstamp_tx(skb, NULL);
2054}
2055
2056/**
2057 * skb_tx_timestamp() - Driver hook for transmit timestamping
2058 *
2059 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00002060 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00002061 *
2062 * @skb: A socket buffer.
2063 */
2064static inline void skb_tx_timestamp(struct sk_buff *skb)
2065{
Richard Cochranc1f19b52010-07-17 08:49:36 +00002066 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002067 sw_tx_timestamp(skb);
2068}
2069
Herbert Xu759e5d02007-03-25 20:10:56 -07002070extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
Al Virob51655b2006-11-14 21:40:42 -08002071extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002072
Herbert Xu60476372007-04-09 11:59:39 -07002073static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2074{
2075 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2076}
2077
Herbert Xufb286bb2005-11-10 13:01:24 -08002078/**
2079 * skb_checksum_complete - Calculate checksum of an entire packet
2080 * @skb: packet to process
2081 *
2082 * This function calculates the checksum over the entire packet plus
2083 * the value of skb->csum. The latter can be used to supply the
2084 * checksum of a pseudo header as used by TCP/UDP. It returns the
2085 * checksum.
2086 *
2087 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2088 * this function can be used to verify that checksum on received
2089 * packets. In that case the function should return zero if the
2090 * checksum is correct. In particular, this function will return zero
2091 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2092 * hardware has already verified the correctness of the checksum.
2093 */
Al Viro4381ca32007-07-15 21:00:11 +01002094static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002095{
Herbert Xu60476372007-04-09 11:59:39 -07002096 return skb_csum_unnecessary(skb) ?
2097 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002098}
2099
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002100#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002101extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002102static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2103{
2104 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002105 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002106}
2107static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2108{
2109 if (nfct)
2110 atomic_inc(&nfct->use);
2111}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002112#endif
2113#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -08002114static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
2115{
2116 if (skb)
2117 atomic_inc(&skb->users);
2118}
2119static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
2120{
2121 if (skb)
2122 kfree_skb(skb);
2123}
2124#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002125#ifdef CONFIG_BRIDGE_NETFILTER
2126static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2127{
2128 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2129 kfree(nf_bridge);
2130}
2131static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2132{
2133 if (nf_bridge)
2134 atomic_inc(&nf_bridge->use);
2135}
2136#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08002137static inline void nf_reset(struct sk_buff *skb)
2138{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002139#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08002140 nf_conntrack_put(skb->nfct);
2141 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002142#endif
2143#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Patrick McHardya193a4a2006-03-20 19:23:05 -08002144 nf_conntrack_put_reasm(skb->nfct_reasm);
2145 skb->nfct_reasm = NULL;
2146#endif
2147#ifdef CONFIG_BRIDGE_NETFILTER
2148 nf_bridge_put(skb->nf_bridge);
2149 skb->nf_bridge = NULL;
2150#endif
2151}
2152
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002153/* Note: This doesn't put any conntrack and bridge info in dst. */
2154static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2155{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002156#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002157 dst->nfct = src->nfct;
2158 nf_conntrack_get(src->nfct);
2159 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002160#endif
2161#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002162 dst->nfct_reasm = src->nfct_reasm;
2163 nf_conntrack_get_reasm(src->nfct_reasm);
2164#endif
2165#ifdef CONFIG_BRIDGE_NETFILTER
2166 dst->nf_bridge = src->nf_bridge;
2167 nf_bridge_get(src->nf_bridge);
2168#endif
2169}
2170
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002171static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2172{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002173#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002174 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002175#endif
2176#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002177 nf_conntrack_put_reasm(dst->nfct_reasm);
2178#endif
2179#ifdef CONFIG_BRIDGE_NETFILTER
2180 nf_bridge_put(dst->nf_bridge);
2181#endif
2182 __nf_copy(dst, src);
2183}
2184
James Morris984bc162006-06-09 00:29:17 -07002185#ifdef CONFIG_NETWORK_SECMARK
2186static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2187{
2188 to->secmark = from->secmark;
2189}
2190
2191static inline void skb_init_secmark(struct sk_buff *skb)
2192{
2193 skb->secmark = 0;
2194}
2195#else
2196static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2197{ }
2198
2199static inline void skb_init_secmark(struct sk_buff *skb)
2200{ }
2201#endif
2202
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002203static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2204{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002205 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002206}
2207
Stephen Hemminger92477442009-03-21 13:39:26 -07002208static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002209{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002210 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002211}
2212
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002213static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2214{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002215 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002216}
2217
David S. Millerd5a9e242009-01-27 16:22:11 -08002218static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2219{
2220 skb->queue_mapping = rx_queue + 1;
2221}
2222
Stephen Hemminger92477442009-03-21 13:39:26 -07002223static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002224{
2225 return skb->queue_mapping - 1;
2226}
2227
Stephen Hemminger92477442009-03-21 13:39:26 -07002228static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002229{
Eric Dumazeta02cec22010-09-22 20:43:57 +00002230 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08002231}
2232
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00002233extern u16 __skb_tx_hash(const struct net_device *dev,
2234 const struct sk_buff *skb,
2235 unsigned int num_tx_queues);
Stephen Hemminger92477442009-03-21 13:39:26 -07002236
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002237#ifdef CONFIG_XFRM
2238static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2239{
2240 return skb->sp;
2241}
2242#else
2243static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2244{
2245 return NULL;
2246}
2247#endif
2248
Herbert Xu89114af2006-07-08 13:34:32 -07002249static inline int skb_is_gso(const struct sk_buff *skb)
2250{
2251 return skb_shinfo(skb)->gso_size;
2252}
2253
Brice Goglineabd7e32007-10-13 12:33:32 +02002254static inline int skb_is_gso_v6(const struct sk_buff *skb)
2255{
2256 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2257}
2258
Ben Hutchings4497b072008-06-19 16:22:28 -07002259extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2260
2261static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2262{
2263 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2264 * wanted then gso_type will be set. */
2265 struct skb_shared_info *shinfo = skb_shinfo(skb);
Alexander Duyckb78462e2010-06-02 12:24:37 +00002266 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2267 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07002268 __skb_warn_lro_forwarding(skb);
2269 return true;
2270 }
2271 return false;
2272}
2273
Herbert Xu35fc92a2007-03-26 23:22:20 -07002274static inline void skb_forward_csum(struct sk_buff *skb)
2275{
2276 /* Unfortunately we don't support this one. Any brave souls? */
2277 if (skb->ip_summed == CHECKSUM_COMPLETE)
2278 skb->ip_summed = CHECKSUM_NONE;
2279}
2280
Eric Dumazetbc8acf22010-09-02 13:07:41 -07002281/**
2282 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2283 * @skb: skb to check
2284 *
2285 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2286 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2287 * use this helper, to document places where we make this assertion.
2288 */
2289static inline void skb_checksum_none_assert(struct sk_buff *skb)
2290{
2291#ifdef DEBUG
2292 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2293#endif
2294}
2295
Rusty Russellf35d9d82008-02-04 23:49:54 -05002296bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00002297
Linus Torvalds1da177e2005-04-16 15:20:36 -07002298#endif /* __KERNEL__ */
2299#endif /* _LINUX_SKBUFF_H */