blob: d2f1778877d74066bce0b06cb163a5f97997177a [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>
18#include <linux/compiler.h>
19#include <linux/time.h>
20#include <linux/cache.h>
21
22#include <asm/atomic.h>
23#include <asm/types.h>
24#include <linux/spinlock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <linux/net.h>
Thomas Graf3fc7e8a2005-06-23 21:00:17 -070026#include <linux/textsearch.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <net/checksum.h>
Al Viroa80958f2006-12-04 20:41:19 +000028#include <linux/rcupdate.h>
Chris Leech97fc2f02006-05-23 17:55:33 -070029#include <linux/dmaengine.h>
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -070030#include <linux/hrtimer.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070031
32#define HAVE_ALLOC_SKB /* For the drivers to know */
33#define HAVE_ALIGNABLE_SKB /* Ditto 8) */
Linus Torvalds1da177e2005-04-16 15:20:36 -070034
Herbert Xu60476372007-04-09 11:59:39 -070035/* Don't change this without changing skb_csum_unnecessary! */
Linus Torvalds1da177e2005-04-16 15:20:36 -070036#define CHECKSUM_NONE 0
Herbert Xu60476372007-04-09 11:59:39 -070037#define CHECKSUM_UNNECESSARY 1
38#define CHECKSUM_COMPLETE 2
39#define CHECKSUM_PARTIAL 3
Linus Torvalds1da177e2005-04-16 15:20:36 -070040
41#define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
42 ~(SMP_CACHE_BYTES - 1))
David S. Millerfc910a22007-03-25 20:27:59 -070043#define SKB_WITH_OVERHEAD(X) \
Herbert Xudeea84b2007-10-21 16:27:46 -070044 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
David S. Millerfc910a22007-03-25 20:27:59 -070045#define SKB_MAX_ORDER(X, ORDER) \
46 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
Linus Torvalds1da177e2005-04-16 15:20:36 -070047#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
48#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
49
50/* A. Checksumming of received packets by device.
51 *
52 * NONE: device failed to checksum this packet.
53 * skb->csum is undefined.
54 *
55 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
56 * skb->csum is undefined.
57 * It is bad option, but, unfortunately, many of vendors do this.
58 * Apparently with secret goal to sell you new device, when you
59 * will add new protocol to your host. F.e. IPv6. 8)
60 *
Patrick McHardy84fa7932006-08-29 16:44:56 -070061 * COMPLETE: the most generic way. Device supplied checksum of _all_
Linus Torvalds1da177e2005-04-16 15:20:36 -070062 * the packet as seen by netif_rx in skb->csum.
63 * NOTE: Even if device supports only some protocols, but
Patrick McHardy84fa7932006-08-29 16:44:56 -070064 * is able to produce some skb->csum, it MUST use COMPLETE,
Linus Torvalds1da177e2005-04-16 15:20:36 -070065 * not UNNECESSARY.
66 *
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070067 * PARTIAL: identical to the case for output below. This may occur
68 * on a packet received directly from another Linux OS, e.g.,
69 * a virtualised Linux kernel on the same host. The packet can
70 * be treated in the same way as UNNECESSARY except that on
71 * output (i.e., forwarding) the checksum must be filled in
72 * by the OS or the hardware.
73 *
Linus Torvalds1da177e2005-04-16 15:20:36 -070074 * B. Checksumming on output.
75 *
76 * NONE: skb is checksummed by protocol or csum is not required.
77 *
Patrick McHardy84fa7932006-08-29 16:44:56 -070078 * PARTIAL: device is required to csum packet as seen by hard_start_xmit
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070079 * from skb->csum_start to the end and to record the checksum
80 * at skb->csum_start + skb->csum_offset.
Linus Torvalds1da177e2005-04-16 15:20:36 -070081 *
82 * Device must show its capabilities in dev->features, set
83 * at device setup time.
84 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
85 * everything.
86 * NETIF_F_NO_CSUM - loopback or reliable single hop media.
87 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
88 * TCP/UDP over IPv4. Sigh. Vendors like this
89 * way by an unknown reason. Though, see comment above
90 * about CHECKSUM_UNNECESSARY. 8)
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070091 * NETIF_F_IPV6_CSUM about as dumb as the last one but does IPv6 instead.
Linus Torvalds1da177e2005-04-16 15:20:36 -070092 *
93 * Any questions? No questions, good. --ANK
94 */
95
Linus Torvalds1da177e2005-04-16 15:20:36 -070096struct net_device;
David Howells716ea3a2007-04-02 20:19:53 -070097struct scatterlist;
Jens Axboe9c55e012007-11-06 23:30:13 -080098struct pipe_inode_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -070099
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700100#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101struct nf_conntrack {
102 atomic_t use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103};
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700104#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105
106#ifdef CONFIG_BRIDGE_NETFILTER
107struct nf_bridge_info {
108 atomic_t use;
109 struct net_device *physindev;
110 struct net_device *physoutdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111 unsigned int mask;
112 unsigned long data[32 / sizeof(unsigned long)];
113};
114#endif
115
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116struct sk_buff_head {
117 /* These two members must be first. */
118 struct sk_buff *next;
119 struct sk_buff *prev;
120
121 __u32 qlen;
122 spinlock_t lock;
123};
124
125struct sk_buff;
126
127/* To allow 64K frame to be packed as single skb without frag_list */
128#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
129
130typedef struct skb_frag_struct skb_frag_t;
131
132struct skb_frag_struct {
133 struct page *page;
David S. Millera309bb02007-07-30 18:47:03 -0700134 __u32 page_offset;
135 __u32 size;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136};
137
138/* This data is invariant across clones and lives at
139 * the end of the header data, ie. at skb->end.
140 */
141struct skb_shared_info {
142 atomic_t dataref;
Benjamin LaHaise4947d3e2006-01-03 14:06:50 -0800143 unsigned short nr_frags;
Herbert Xu79671682006-06-22 02:40:14 -0700144 unsigned short gso_size;
145 /* Warning: this field is not always filled in (UFO)! */
146 unsigned short gso_segs;
147 unsigned short gso_type;
Al Viroae08e1f2006-11-08 00:27:11 -0800148 __be32 ip6_frag_id;
David S. Miller271bff72008-09-11 04:48:58 -0700149#ifdef CONFIG_HAS_DMA
150 unsigned int num_dma_maps;
151#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152 struct sk_buff *frag_list;
153 skb_frag_t frags[MAX_SKB_FRAGS];
David S. Miller271bff72008-09-11 04:48:58 -0700154#ifdef CONFIG_HAS_DMA
155 dma_addr_t dma_maps[MAX_SKB_FRAGS + 1];
156#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157};
158
159/* We divide dataref into two halves. The higher 16 bits hold references
160 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700161 * the entire skb->data. A clone of a headerless skb holds the length of
162 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700163 *
164 * All users must obey the rule that the skb->data reference count must be
165 * greater than or equal to the payload reference count.
166 *
167 * Holding a reference to the payload part means that the user does not
168 * care about modifications to the header part of skb->data.
169 */
170#define SKB_DATAREF_SHIFT 16
171#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
172
David S. Millerd179cd12005-08-17 14:57:30 -0700173
174enum {
175 SKB_FCLONE_UNAVAILABLE,
176 SKB_FCLONE_ORIG,
177 SKB_FCLONE_CLONE,
178};
179
Herbert Xu79671682006-06-22 02:40:14 -0700180enum {
181 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700182 SKB_GSO_UDP = 1 << 1,
Herbert Xu576a30e2006-06-27 13:22:38 -0700183
184 /* This indicates the skb is from an untrusted source. */
185 SKB_GSO_DODGY = 1 << 2,
Michael Chanb0da85372006-06-29 12:30:00 -0700186
187 /* This indicates the tcp segment has CWR set. */
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700188 SKB_GSO_TCP_ECN = 1 << 3,
189
190 SKB_GSO_TCPV6 = 1 << 4,
Herbert Xu79671682006-06-22 02:40:14 -0700191};
192
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700193#if BITS_PER_LONG > 32
194#define NET_SKBUFF_DATA_USES_OFFSET 1
195#endif
196
197#ifdef NET_SKBUFF_DATA_USES_OFFSET
198typedef unsigned int sk_buff_data_t;
199#else
200typedef unsigned char *sk_buff_data_t;
201#endif
202
Linus Torvalds1da177e2005-04-16 15:20:36 -0700203/**
204 * struct sk_buff - socket buffer
205 * @next: Next buffer in list
206 * @prev: Previous buffer in list
Linus Torvalds1da177e2005-04-16 15:20:36 -0700207 * @sk: Socket we are owned by
Herbert Xu325ed822005-10-03 13:57:23 -0700208 * @tstamp: Time we arrived
Linus Torvalds1da177e2005-04-16 15:20:36 -0700209 * @dev: Device we arrived on/are leaving by
Randy Dunlapbe52178b2007-05-03 03:16:20 -0700210 * @transport_header: Transport layer header
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -0700211 * @network_header: Network layer header
212 * @mac_header: Link layer header
Martin Waitz67be2dd2005-05-01 08:59:26 -0700213 * @dst: destination entry
214 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215 * @cb: Control buffer. Free for use by every layer. Put private vars here
216 * @len: Length of actual data
217 * @data_len: Data length
218 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700219 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700220 * @csum: Checksum (must include start/offset pair)
221 * @csum_start: Offset from skb->head where checksumming should start
222 * @csum_offset: Offset from csum_start where checksum should be stored
Martin Waitz67be2dd2005-05-01 08:59:26 -0700223 * @local_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224 * @cloned: Head may be cloned (check refcnt to be sure)
225 * @nohdr: Payload reference only, must not modify header
226 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700227 * @fclone: skbuff clone status
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228 * @ip_summed: Driver fed us an IP checksum
229 * @priority: Packet queueing priority
230 * @users: User count - see {datagram,tcp}.c
231 * @protocol: Packet protocol from driver
Linus Torvalds1da177e2005-04-16 15:20:36 -0700232 * @truesize: Buffer size
233 * @head: Head of buffer
234 * @data: Data head pointer
235 * @tail: Tail pointer
236 * @end: End pointer
237 * @destructor: Destruct function
Thomas Graf82e91ff2006-11-09 15:19:14 -0800238 * @mark: Generic packet mark
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239 * @nfct: Associated connection, if any
Randy Dunlapc83c2482005-10-18 22:07:41 -0700240 * @ipvs_property: skbuff is owned by ipvs
Randy Dunlap31729362008-02-18 20:52:13 -0800241 * @peeked: this packet has been seen already, so stats have been
242 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700243 * @nf_trace: netfilter packet trace flag
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244 * @nfctinfo: Relationship of this skb to the connection
Randy Dunlap461ddf32005-11-20 21:25:15 -0800245 * @nfct_reasm: netfilter conntrack re-assembly pointer
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700247 * @iif: ifindex of device we arrived on
248 * @queue_mapping: Queue mapping for multiqueue devices
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249 * @tc_index: Traffic control index
250 * @tc_verd: traffic control verdict
Randy Dunlap553a5672008-04-20 10:51:01 -0700251 * @ndisc_nodetype: router type (from link layer)
Randy Dunlap4a7b61d2008-07-31 20:52:08 -0700252 * @do_not_encrypt: set to prevent encryption of this frame
Randy Dunlapf4b8ea72006-06-22 16:00:11 -0700253 * @dma_cookie: a cookie to one of several possible DMA operations
254 * done by skb DMA functions
James Morris984bc162006-06-09 00:29:17 -0700255 * @secmark: security marking
Patrick McHardy6aa895b02008-07-14 22:49:06 -0700256 * @vlan_tci: vlan tag control information
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257 */
258
259struct sk_buff {
260 /* These two members must be first. */
261 struct sk_buff *next;
262 struct sk_buff *prev;
263
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264 struct sock *sk;
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700265 ktime_t tstamp;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267
Eric Dumazetee6b9672008-03-05 18:30:47 -0800268 union {
269 struct dst_entry *dst;
270 struct rtable *rtable;
271 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272 struct sec_path *sp;
273
274 /*
275 * This is the control buffer. It is free to use for every
276 * layer. Please put your private variables there. If you
277 * want to keep them across layers you have to do a skb_clone()
278 * first. This is owned by whoever has the skb queued ATM.
279 */
Patrick McHardy3e3850e2006-01-06 23:04:54 -0800280 char cb[48];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281
282 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700283 data_len;
284 __u16 mac_len,
285 hdr_len;
Al Viroff1dcad2006-11-20 18:07:29 -0800286 union {
287 __wsum csum;
Herbert Xu663ead32007-04-09 11:59:07 -0700288 struct {
289 __u16 csum_start;
290 __u16 csum_offset;
291 };
Al Viroff1dcad2006-11-20 18:07:29 -0800292 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293 __u32 priority;
Thomas Graf1cbb3382005-07-05 14:13:41 -0700294 __u8 local_df:1,
295 cloned:1,
296 ip_summed:2,
Harald Welte6869c4d2005-08-09 19:24:19 -0700297 nohdr:1,
298 nfctinfo:3;
David S. Millerd179cd12005-08-17 14:57:30 -0700299 __u8 pkt_type:3,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800300 fclone:2,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700301 ipvs_property:1,
Herbert Xua59322b2007-12-05 01:53:40 -0800302 peeked:1,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700303 nf_trace:1;
Alexey Dobriyana0d3bea2005-08-11 16:05:50 -0700304 __be16 protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305
306 void (*destructor)(struct sk_buff *skb);
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800307#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700308 struct nf_conntrack *nfct;
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800309 struct sk_buff *nfct_reasm;
310#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700311#ifdef CONFIG_BRIDGE_NETFILTER
312 struct nf_bridge_info *nf_bridge;
313#endif
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700314
315 int iif;
316 __u16 queue_mapping;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700318 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700320 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322#endif
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400323#ifdef CONFIG_IPV6_NDISC_NODETYPE
Templin, Fred Lfadf6bf2008-03-11 18:35:59 -0400324 __u8 ndisc_nodetype:2;
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400325#endif
Johannes Bergd0f09802008-07-29 11:32:07 +0200326#if defined(CONFIG_MAC80211) || defined(CONFIG_MAC80211_MODULE)
327 __u8 do_not_encrypt:1;
328#endif
329 /* 0/13/14 bit hole */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700330
Chris Leech97fc2f02006-05-23 17:55:33 -0700331#ifdef CONFIG_NET_DMA
332 dma_cookie_t dma_cookie;
333#endif
James Morris984bc162006-06-09 00:29:17 -0700334#ifdef CONFIG_NETWORK_SECMARK
335 __u32 secmark;
336#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337
Thomas Graf82e91ff2006-11-09 15:19:14 -0800338 __u32 mark;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700339
Patrick McHardy6aa895b02008-07-14 22:49:06 -0700340 __u16 vlan_tci;
341
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700342 sk_buff_data_t transport_header;
343 sk_buff_data_t network_header;
344 sk_buff_data_t mac_header;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700346 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700347 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700349 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700350 unsigned int truesize;
351 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352};
353
354#ifdef __KERNEL__
355/*
356 * Handling routines are only of interest to the kernel
357 */
358#include <linux/slab.h>
359
360#include <asm/system.h>
361
David S. Millera40c24a2008-09-11 04:51:14 -0700362#ifdef CONFIG_HAS_DMA
363#include <linux/dma-mapping.h>
364extern int skb_dma_map(struct device *dev, struct sk_buff *skb,
365 enum dma_data_direction dir);
366extern void skb_dma_unmap(struct device *dev, struct sk_buff *skb,
367 enum dma_data_direction dir);
368#endif
369
Jörn Engel231d06a2006-03-20 21:28:35 -0800370extern void kfree_skb(struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371extern void __kfree_skb(struct sk_buff *skb);
David S. Millerd179cd12005-08-17 14:57:30 -0700372extern struct sk_buff *__alloc_skb(unsigned int size,
Christoph Hellwigb30973f2006-12-06 20:32:36 -0800373 gfp_t priority, int fclone, int node);
David S. Millerd179cd12005-08-17 14:57:30 -0700374static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100375 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700376{
Christoph Hellwigb30973f2006-12-06 20:32:36 -0800377 return __alloc_skb(size, priority, 0, -1);
David S. Millerd179cd12005-08-17 14:57:30 -0700378}
379
380static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100381 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700382{
Christoph Hellwigb30973f2006-12-06 20:32:36 -0800383 return __alloc_skb(size, priority, 1, -1);
David S. Millerd179cd12005-08-17 14:57:30 -0700384}
385
Herbert Xue0053ec2007-10-14 00:37:52 -0700386extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700387extern struct sk_buff *skb_clone(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100388 gfp_t priority);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700389extern struct sk_buff *skb_copy(const struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100390 gfp_t priority);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700391extern struct sk_buff *pskb_copy(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100392 gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393extern int pskb_expand_head(struct sk_buff *skb,
Victor Fusco86a76ca2005-07-08 14:57:47 -0700394 int nhead, int ntail,
Al Virodd0fc662005-10-07 07:46:04 +0100395 gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
397 unsigned int headroom);
398extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
399 int newheadroom, int newtailroom,
Al Virodd0fc662005-10-07 07:46:04 +0100400 gfp_t priority);
David Howells716ea3a2007-04-02 20:19:53 -0700401extern int skb_to_sgvec(struct sk_buff *skb,
402 struct scatterlist *sg, int offset,
403 int len);
404extern int skb_cow_data(struct sk_buff *skb, int tailbits,
405 struct sk_buff **trailer);
Herbert Xu5b057c62006-06-23 02:06:41 -0700406extern int skb_pad(struct sk_buff *skb, int pad);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407#define dev_kfree_skb(a) kfree_skb(a)
408extern void skb_over_panic(struct sk_buff *skb, int len,
409 void *here);
410extern void skb_under_panic(struct sk_buff *skb, int len,
411 void *here);
David S. Millerdc6de332006-04-20 00:10:50 -0700412extern void skb_truesize_bug(struct sk_buff *skb);
413
414static inline void skb_truesize_check(struct sk_buff *skb)
415{
Chuck Lever78608ba2007-11-10 21:53:30 -0800416 int len = sizeof(struct sk_buff) + skb->len;
417
418 if (unlikely((int)skb->truesize < len))
David S. Millerdc6de332006-04-20 00:10:50 -0700419 skb_truesize_bug(skb);
420}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700422extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
423 int getfrag(void *from, char *to, int offset,
424 int len,int odd, struct sk_buff *skb),
425 void *from, int length);
426
Thomas Graf677e90e2005-06-23 20:59:51 -0700427struct skb_seq_state
428{
429 __u32 lower_offset;
430 __u32 upper_offset;
431 __u32 frag_idx;
432 __u32 stepped_offset;
433 struct sk_buff *root_skb;
434 struct sk_buff *cur_skb;
435 __u8 *frag_data;
436};
437
438extern void skb_prepare_seq_read(struct sk_buff *skb,
439 unsigned int from, unsigned int to,
440 struct skb_seq_state *st);
441extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
442 struct skb_seq_state *st);
443extern void skb_abort_seq_read(struct skb_seq_state *st);
444
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700445extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
446 unsigned int to, struct ts_config *config,
447 struct ts_state *state);
448
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700449#ifdef NET_SKBUFF_DATA_USES_OFFSET
450static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
451{
452 return skb->head + skb->end;
453}
454#else
455static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
456{
457 return skb->end;
458}
459#endif
460
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700462#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463
464/**
465 * skb_queue_empty - check if a queue is empty
466 * @list: queue head
467 *
468 * Returns true if the queue is empty, false otherwise.
469 */
470static inline int skb_queue_empty(const struct sk_buff_head *list)
471{
472 return list->next == (struct sk_buff *)list;
473}
474
475/**
David S. Millerfc7ebb22008-09-23 00:34:07 -0700476 * skb_queue_is_last - check if skb is the last entry in the queue
477 * @list: queue head
478 * @skb: buffer
479 *
480 * Returns true if @skb is the last buffer on the list.
481 */
482static inline bool skb_queue_is_last(const struct sk_buff_head *list,
483 const struct sk_buff *skb)
484{
485 return (skb->next == (struct sk_buff *) list);
486}
487
488/**
David S. Miller249c8b42008-09-23 00:44:42 -0700489 * skb_queue_next - return the next packet in the queue
490 * @list: queue head
491 * @skb: current buffer
492 *
493 * Return the next packet in @list after @skb. It is only valid to
494 * call this if skb_queue_is_last() evaluates to false.
495 */
496static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
497 const struct sk_buff *skb)
498{
499 /* This BUG_ON may seem severe, but if we just return then we
500 * are going to dereference garbage.
501 */
502 BUG_ON(skb_queue_is_last(list, skb));
503 return skb->next;
504}
505
506/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507 * skb_get - reference buffer
508 * @skb: buffer to reference
509 *
510 * Makes another reference to a socket buffer and returns a pointer
511 * to the buffer.
512 */
513static inline struct sk_buff *skb_get(struct sk_buff *skb)
514{
515 atomic_inc(&skb->users);
516 return skb;
517}
518
519/*
520 * If users == 1, we are the only owner and are can avoid redundant
521 * atomic change.
522 */
523
524/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525 * skb_cloned - is the buffer a clone
526 * @skb: buffer to check
527 *
528 * Returns true if the buffer was generated with skb_clone() and is
529 * one of multiple shared copies of the buffer. Cloned buffers are
530 * shared data so must not be written to under normal circumstances.
531 */
532static inline int skb_cloned(const struct sk_buff *skb)
533{
534 return skb->cloned &&
535 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
536}
537
538/**
539 * skb_header_cloned - is the header a clone
540 * @skb: buffer to check
541 *
542 * Returns true if modifying the header part of the buffer requires
543 * the data to be copied.
544 */
545static inline int skb_header_cloned(const struct sk_buff *skb)
546{
547 int dataref;
548
549 if (!skb->cloned)
550 return 0;
551
552 dataref = atomic_read(&skb_shinfo(skb)->dataref);
553 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
554 return dataref != 1;
555}
556
557/**
558 * skb_header_release - release reference to header
559 * @skb: buffer to operate on
560 *
561 * Drop a reference to the header part of the buffer. This is done
562 * by acquiring a payload reference. You must not read from the header
563 * part of skb->data after this.
564 */
565static inline void skb_header_release(struct sk_buff *skb)
566{
567 BUG_ON(skb->nohdr);
568 skb->nohdr = 1;
569 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
570}
571
572/**
573 * skb_shared - is the buffer shared
574 * @skb: buffer to check
575 *
576 * Returns true if more than one person has a reference to this
577 * buffer.
578 */
579static inline int skb_shared(const struct sk_buff *skb)
580{
581 return atomic_read(&skb->users) != 1;
582}
583
584/**
585 * skb_share_check - check if buffer is shared and if so clone it
586 * @skb: buffer to check
587 * @pri: priority for memory allocation
588 *
589 * If the buffer is shared the buffer is cloned and the old copy
590 * drops a reference. A new clone with a single reference is returned.
591 * If the buffer is not shared the original buffer is returned. When
592 * being called from interrupt status or with spinlocks held pri must
593 * be GFP_ATOMIC.
594 *
595 * NULL is returned on a memory allocation failure.
596 */
Victor Fusco86a76ca2005-07-08 14:57:47 -0700597static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100598 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700599{
600 might_sleep_if(pri & __GFP_WAIT);
601 if (skb_shared(skb)) {
602 struct sk_buff *nskb = skb_clone(skb, pri);
603 kfree_skb(skb);
604 skb = nskb;
605 }
606 return skb;
607}
608
609/*
610 * Copy shared buffers into a new sk_buff. We effectively do COW on
611 * packets to handle cases where we have a local reader and forward
612 * and a couple of other messy ones. The normal one is tcpdumping
613 * a packet thats being forwarded.
614 */
615
616/**
617 * skb_unshare - make a copy of a shared buffer
618 * @skb: buffer to check
619 * @pri: priority for memory allocation
620 *
621 * If the socket buffer is a clone then this function creates a new
622 * copy of the data, drops a reference count on the old copy and returns
623 * the new copy with the reference count at 1. If the buffer is not a clone
624 * the original buffer is returned. When called with a spinlock held or
625 * from interrupt state @pri must be %GFP_ATOMIC
626 *
627 * %NULL is returned on a memory allocation failure.
628 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -0700629static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100630 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631{
632 might_sleep_if(pri & __GFP_WAIT);
633 if (skb_cloned(skb)) {
634 struct sk_buff *nskb = skb_copy(skb, pri);
635 kfree_skb(skb); /* Free our shared copy */
636 skb = nskb;
637 }
638 return skb;
639}
640
641/**
642 * skb_peek
643 * @list_: list to peek at
644 *
645 * Peek an &sk_buff. Unlike most other operations you _MUST_
646 * be careful with this one. A peek leaves the buffer on the
647 * list and someone else may run off with it. You must hold
648 * the appropriate locks or have a private queue to do this.
649 *
650 * Returns %NULL for an empty list or a pointer to the head element.
651 * The reference count is not incremented and the reference is therefore
652 * volatile. Use with caution.
653 */
654static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
655{
656 struct sk_buff *list = ((struct sk_buff *)list_)->next;
657 if (list == (struct sk_buff *)list_)
658 list = NULL;
659 return list;
660}
661
662/**
663 * skb_peek_tail
664 * @list_: list to peek at
665 *
666 * Peek an &sk_buff. Unlike most other operations you _MUST_
667 * be careful with this one. A peek leaves the buffer on the
668 * list and someone else may run off with it. You must hold
669 * the appropriate locks or have a private queue to do this.
670 *
671 * Returns %NULL for an empty list or a pointer to the tail element.
672 * The reference count is not incremented and the reference is therefore
673 * volatile. Use with caution.
674 */
675static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
676{
677 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
678 if (list == (struct sk_buff *)list_)
679 list = NULL;
680 return list;
681}
682
683/**
684 * skb_queue_len - get queue length
685 * @list_: list to measure
686 *
687 * Return the length of an &sk_buff queue.
688 */
689static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
690{
691 return list_->qlen;
692}
693
David S. Miller67fed452008-09-21 22:36:24 -0700694/**
695 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
696 * @list: queue to initialize
697 *
698 * This initializes only the list and queue length aspects of
699 * an sk_buff_head object. This allows to initialize the list
700 * aspects of an sk_buff_head without reinitializing things like
701 * the spinlock. It can also be used for on-stack sk_buff_head
702 * objects where the spinlock is known to not be used.
703 */
704static inline void __skb_queue_head_init(struct sk_buff_head *list)
705{
706 list->prev = list->next = (struct sk_buff *)list;
707 list->qlen = 0;
708}
709
Arjan van de Ven76f10ad2006-08-02 14:06:55 -0700710/*
711 * This function creates a split out lock class for each invocation;
712 * this is needed for now since a whole lot of users of the skb-queue
713 * infrastructure in drivers have different locking usage (in hardirq)
714 * than the networking core (in softirq only). In the long run either the
715 * network layer or drivers should need annotation to consolidate the
716 * main types of usage into 3 classes.
717 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718static inline void skb_queue_head_init(struct sk_buff_head *list)
719{
720 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -0700721 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700722}
723
Pavel Emelianovc2ecba72007-04-17 12:45:31 -0700724static inline void skb_queue_head_init_class(struct sk_buff_head *list,
725 struct lock_class_key *class)
726{
727 skb_queue_head_init(list);
728 lockdep_set_class(&list->lock, class);
729}
730
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731/*
Gerrit Renkerbf299272008-04-14 00:04:51 -0700732 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733 *
734 * The "__skb_xxxx()" functions are the non-atomic ones that
735 * can only be called with interrupts disabled.
736 */
Gerrit Renkerbf299272008-04-14 00:04:51 -0700737extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
738static inline void __skb_insert(struct sk_buff *newsk,
739 struct sk_buff *prev, struct sk_buff *next,
740 struct sk_buff_head *list)
741{
742 newsk->next = next;
743 newsk->prev = prev;
744 next->prev = prev->next = newsk;
745 list->qlen++;
746}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747
David S. Miller67fed452008-09-21 22:36:24 -0700748static inline void __skb_queue_splice(const struct sk_buff_head *list,
749 struct sk_buff *prev,
750 struct sk_buff *next)
751{
752 struct sk_buff *first = list->next;
753 struct sk_buff *last = list->prev;
754
755 first->prev = prev;
756 prev->next = first;
757
758 last->next = next;
759 next->prev = last;
760}
761
762/**
763 * skb_queue_splice - join two skb lists, this is designed for stacks
764 * @list: the new list to add
765 * @head: the place to add it in the first list
766 */
767static inline void skb_queue_splice(const struct sk_buff_head *list,
768 struct sk_buff_head *head)
769{
770 if (!skb_queue_empty(list)) {
771 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -0700772 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -0700773 }
774}
775
776/**
777 * skb_queue_splice - join two skb lists and reinitialise the emptied list
778 * @list: the new list to add
779 * @head: the place to add it in the first list
780 *
781 * The list at @list is reinitialised
782 */
783static inline void skb_queue_splice_init(struct sk_buff_head *list,
784 struct sk_buff_head *head)
785{
786 if (!skb_queue_empty(list)) {
787 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -0700788 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -0700789 __skb_queue_head_init(list);
790 }
791}
792
793/**
794 * skb_queue_splice_tail - join two skb lists, each list being a queue
795 * @list: the new list to add
796 * @head: the place to add it in the first list
797 */
798static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
799 struct sk_buff_head *head)
800{
801 if (!skb_queue_empty(list)) {
802 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -0700803 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -0700804 }
805}
806
807/**
808 * skb_queue_splice_tail - join two skb lists and reinitialise the emptied list
809 * @list: the new list to add
810 * @head: the place to add it in the first list
811 *
812 * Each of the lists is a queue.
813 * The list at @list is reinitialised
814 */
815static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
816 struct sk_buff_head *head)
817{
818 if (!skb_queue_empty(list)) {
819 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -0700820 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -0700821 __skb_queue_head_init(list);
822 }
823}
824
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825/**
Stephen Hemminger300ce172005-10-30 13:47:34 -0800826 * __skb_queue_after - queue a buffer at the list head
827 * @list: list to use
828 * @prev: place after this buffer
829 * @newsk: buffer to queue
830 *
831 * Queue a buffer int the middle of a list. This function takes no locks
832 * and you must therefore hold required locks before calling it.
833 *
834 * A buffer cannot be placed on two lists at the same time.
835 */
836static inline void __skb_queue_after(struct sk_buff_head *list,
837 struct sk_buff *prev,
838 struct sk_buff *newsk)
839{
Gerrit Renkerbf299272008-04-14 00:04:51 -0700840 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -0800841}
842
Gerrit Renker7de6c032008-04-14 00:05:09 -0700843extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
844 struct sk_buff_head *list);
845
Gerrit Renkerf5572852008-04-14 00:05:28 -0700846static inline void __skb_queue_before(struct sk_buff_head *list,
847 struct sk_buff *next,
848 struct sk_buff *newsk)
849{
850 __skb_insert(newsk, next->prev, next, list);
851}
852
Stephen Hemminger300ce172005-10-30 13:47:34 -0800853/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700854 * __skb_queue_head - queue a buffer at the list head
855 * @list: list to use
856 * @newsk: buffer to queue
857 *
858 * Queue a buffer at the start of a list. This function takes no locks
859 * and you must therefore hold required locks before calling it.
860 *
861 * A buffer cannot be placed on two lists at the same time.
862 */
863extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
864static inline void __skb_queue_head(struct sk_buff_head *list,
865 struct sk_buff *newsk)
866{
Stephen Hemminger300ce172005-10-30 13:47:34 -0800867 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868}
869
870/**
871 * __skb_queue_tail - queue a buffer at the list tail
872 * @list: list to use
873 * @newsk: buffer to queue
874 *
875 * Queue a buffer at the end of a list. This function takes no locks
876 * and you must therefore hold required locks before calling it.
877 *
878 * A buffer cannot be placed on two lists at the same time.
879 */
880extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
881static inline void __skb_queue_tail(struct sk_buff_head *list,
882 struct sk_buff *newsk)
883{
Gerrit Renkerf5572852008-04-14 00:05:28 -0700884 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885}
886
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888 * remove sk_buff from list. _Must_ be called atomically, and with
889 * the list known..
890 */
David S. Miller8728b832005-08-09 19:25:21 -0700891extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700892static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
893{
894 struct sk_buff *next, *prev;
895
896 list->qlen--;
897 next = skb->next;
898 prev = skb->prev;
899 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700900 next->prev = prev;
901 prev->next = next;
902}
903
Gerrit Renkerf525c062008-04-14 00:04:12 -0700904/**
905 * __skb_dequeue - remove from the head of the queue
906 * @list: list to dequeue from
907 *
908 * Remove the head of the list. This function does not take any locks
909 * so must be used with appropriate locks held only. The head item is
910 * returned or %NULL if the list is empty.
911 */
912extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
913static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
914{
915 struct sk_buff *skb = skb_peek(list);
916 if (skb)
917 __skb_unlink(skb, list);
918 return skb;
919}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920
921/**
922 * __skb_dequeue_tail - remove from the tail of the queue
923 * @list: list to dequeue from
924 *
925 * Remove the tail of the list. This function does not take any locks
926 * so must be used with appropriate locks held only. The tail item is
927 * returned or %NULL if the list is empty.
928 */
929extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
930static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
931{
932 struct sk_buff *skb = skb_peek_tail(list);
933 if (skb)
934 __skb_unlink(skb, list);
935 return skb;
936}
937
938
939static inline int skb_is_nonlinear(const struct sk_buff *skb)
940{
941 return skb->data_len;
942}
943
944static inline unsigned int skb_headlen(const struct sk_buff *skb)
945{
946 return skb->len - skb->data_len;
947}
948
949static inline int skb_pagelen(const struct sk_buff *skb)
950{
951 int i, len = 0;
952
953 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
954 len += skb_shinfo(skb)->frags[i].size;
955 return len + skb_headlen(skb);
956}
957
958static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
959 struct page *page, int off, int size)
960{
961 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
962
963 frag->page = page;
964 frag->page_offset = off;
965 frag->size = size;
966 skb_shinfo(skb)->nr_frags = i + 1;
967}
968
969#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
970#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
971#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
972
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700973#ifdef NET_SKBUFF_DATA_USES_OFFSET
974static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
975{
976 return skb->head + skb->tail;
977}
978
979static inline void skb_reset_tail_pointer(struct sk_buff *skb)
980{
981 skb->tail = skb->data - skb->head;
982}
983
984static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
985{
986 skb_reset_tail_pointer(skb);
987 skb->tail += offset;
988}
989#else /* NET_SKBUFF_DATA_USES_OFFSET */
990static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
991{
992 return skb->tail;
993}
994
995static inline void skb_reset_tail_pointer(struct sk_buff *skb)
996{
997 skb->tail = skb->data;
998}
999
1000static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1001{
1002 skb->tail = skb->data + offset;
1003}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001004
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001005#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1006
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007/*
1008 * Add data to an sk_buff
1009 */
Ilpo Järvinen0dde3e12008-03-27 17:43:41 -07001010extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1012{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001013 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014 SKB_LINEAR_ASSERT(skb);
1015 skb->tail += len;
1016 skb->len += len;
1017 return tmp;
1018}
1019
Ilpo Järvinenc2aa2702008-03-27 17:52:40 -07001020extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1022{
1023 skb->data -= len;
1024 skb->len += len;
1025 return skb->data;
1026}
1027
Ilpo Järvinen6be8ac22008-03-27 17:47:24 -07001028extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001029static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1030{
1031 skb->len -= len;
1032 BUG_ON(skb->len < skb->data_len);
1033 return skb->data += len;
1034}
1035
Linus Torvalds1da177e2005-04-16 15:20:36 -07001036extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1037
1038static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1039{
1040 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001041 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042 return NULL;
1043 skb->len -= len;
1044 return skb->data += len;
1045}
1046
1047static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1048{
1049 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1050}
1051
1052static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1053{
1054 if (likely(len <= skb_headlen(skb)))
1055 return 1;
1056 if (unlikely(len > skb->len))
1057 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001058 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059}
1060
1061/**
1062 * skb_headroom - bytes at buffer head
1063 * @skb: buffer to check
1064 *
1065 * Return the number of bytes of free space at the head of an &sk_buff.
1066 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001067static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068{
1069 return skb->data - skb->head;
1070}
1071
1072/**
1073 * skb_tailroom - bytes at buffer end
1074 * @skb: buffer to check
1075 *
1076 * Return the number of bytes of free space at the tail of an sk_buff
1077 */
1078static inline int skb_tailroom(const struct sk_buff *skb)
1079{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001080 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081}
1082
1083/**
1084 * skb_reserve - adjust headroom
1085 * @skb: buffer to alter
1086 * @len: bytes to move
1087 *
1088 * Increase the headroom of an empty &sk_buff by reducing the tail
1089 * room. This is only allowed for an empty buffer.
1090 */
David S. Miller8243126c2006-01-17 02:54:21 -08001091static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001092{
1093 skb->data += len;
1094 skb->tail += len;
1095}
1096
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001097#ifdef NET_SKBUFF_DATA_USES_OFFSET
1098static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1099{
1100 return skb->head + skb->transport_header;
1101}
1102
1103static inline void skb_reset_transport_header(struct sk_buff *skb)
1104{
1105 skb->transport_header = skb->data - skb->head;
1106}
1107
1108static inline void skb_set_transport_header(struct sk_buff *skb,
1109 const int offset)
1110{
1111 skb_reset_transport_header(skb);
1112 skb->transport_header += offset;
1113}
1114
1115static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1116{
1117 return skb->head + skb->network_header;
1118}
1119
1120static inline void skb_reset_network_header(struct sk_buff *skb)
1121{
1122 skb->network_header = skb->data - skb->head;
1123}
1124
1125static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1126{
1127 skb_reset_network_header(skb);
1128 skb->network_header += offset;
1129}
1130
1131static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1132{
1133 return skb->head + skb->mac_header;
1134}
1135
1136static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1137{
1138 return skb->mac_header != ~0U;
1139}
1140
1141static inline void skb_reset_mac_header(struct sk_buff *skb)
1142{
1143 skb->mac_header = skb->data - skb->head;
1144}
1145
1146static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1147{
1148 skb_reset_mac_header(skb);
1149 skb->mac_header += offset;
1150}
1151
1152#else /* NET_SKBUFF_DATA_USES_OFFSET */
1153
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001154static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1155{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001156 return skb->transport_header;
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001157}
1158
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001159static inline void skb_reset_transport_header(struct sk_buff *skb)
1160{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001161 skb->transport_header = skb->data;
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001162}
1163
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001164static inline void skb_set_transport_header(struct sk_buff *skb,
1165 const int offset)
1166{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001167 skb->transport_header = skb->data + offset;
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001168}
1169
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001170static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1171{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001172 return skb->network_header;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001173}
1174
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001175static inline void skb_reset_network_header(struct sk_buff *skb)
1176{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001177 skb->network_header = skb->data;
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001178}
1179
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001180static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1181{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001182 skb->network_header = skb->data + offset;
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001183}
1184
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001185static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1186{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001187 return skb->mac_header;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001188}
1189
1190static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1191{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001192 return skb->mac_header != NULL;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001193}
1194
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001195static inline void skb_reset_mac_header(struct sk_buff *skb)
1196{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001197 skb->mac_header = skb->data;
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001198}
1199
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001200static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1201{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001202 skb->mac_header = skb->data + offset;
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001203}
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001204#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1205
1206static inline int skb_transport_offset(const struct sk_buff *skb)
1207{
1208 return skb_transport_header(skb) - skb->data;
1209}
1210
1211static inline u32 skb_network_header_len(const struct sk_buff *skb)
1212{
1213 return skb->transport_header - skb->network_header;
1214}
1215
1216static inline int skb_network_offset(const struct sk_buff *skb)
1217{
1218 return skb_network_header(skb) - skb->data;
1219}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001220
Linus Torvalds1da177e2005-04-16 15:20:36 -07001221/*
1222 * CPUs often take a performance hit when accessing unaligned memory
1223 * locations. The actual performance hit varies, it can be small if the
1224 * hardware handles it or large if we have to take an exception and fix it
1225 * in software.
1226 *
1227 * Since an ethernet header is 14 bytes network drivers often end up with
1228 * the IP header at an unaligned offset. The IP header can be aligned by
1229 * shifting the start of the packet by 2 bytes. Drivers should do this
1230 * with:
1231 *
1232 * skb_reserve(NET_IP_ALIGN);
1233 *
1234 * The downside to this alignment of the IP header is that the DMA is now
1235 * unaligned. On some architectures the cost of an unaligned DMA is high
1236 * and this cost outweighs the gains made by aligning the IP header.
1237 *
1238 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1239 * to be overridden.
1240 */
1241#ifndef NET_IP_ALIGN
1242#define NET_IP_ALIGN 2
1243#endif
1244
Anton Blanchard025be812006-03-31 02:27:06 -08001245/*
1246 * The networking layer reserves some headroom in skb data (via
1247 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1248 * the header has to grow. In the default case, if the header has to grow
1249 * 16 bytes or less we avoid the reallocation.
1250 *
1251 * Unfortunately this headroom changes the DMA alignment of the resulting
1252 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1253 * on some architectures. An architecture can override this value,
1254 * perhaps setting it to a cacheline in size (since that will maintain
1255 * cacheline alignment of the DMA). It must be a power of 2.
1256 *
1257 * Various parts of the networking layer expect at least 16 bytes of
1258 * headroom, you should not reduce this.
1259 */
1260#ifndef NET_SKB_PAD
1261#define NET_SKB_PAD 16
1262#endif
1263
Herbert Xu3cc0e872006-06-09 16:13:38 -07001264extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001265
1266static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1267{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001268 if (unlikely(skb->data_len)) {
1269 WARN_ON(1);
1270 return;
1271 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001272 skb->len = len;
1273 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001274}
1275
Ilpo Järvinen419ae742008-03-27 17:54:01 -07001276extern void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001277
1278static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1279{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001280 if (skb->data_len)
1281 return ___pskb_trim(skb, len);
1282 __skb_trim(skb, len);
1283 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284}
1285
1286static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1287{
1288 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1289}
1290
1291/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07001292 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1293 * @skb: buffer to alter
1294 * @len: new length
1295 *
1296 * This is identical to pskb_trim except that the caller knows that
1297 * the skb is not cloned so we should never get an error due to out-
1298 * of-memory.
1299 */
1300static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1301{
1302 int err = pskb_trim(skb, len);
1303 BUG_ON(err);
1304}
1305
1306/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001307 * skb_orphan - orphan a buffer
1308 * @skb: buffer to orphan
1309 *
1310 * If a buffer currently has an owner then we call the owner's
1311 * destructor function and make the @skb unowned. The buffer continues
1312 * to exist but is no longer charged to its former owner.
1313 */
1314static inline void skb_orphan(struct sk_buff *skb)
1315{
1316 if (skb->destructor)
1317 skb->destructor(skb);
1318 skb->destructor = NULL;
1319 skb->sk = NULL;
1320}
1321
1322/**
1323 * __skb_queue_purge - empty a list
1324 * @list: list to empty
1325 *
1326 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1327 * the list and one reference dropped. This function does not take the
1328 * list lock and the caller must hold the relevant locks to use it.
1329 */
1330extern void skb_queue_purge(struct sk_buff_head *list);
1331static inline void __skb_queue_purge(struct sk_buff_head *list)
1332{
1333 struct sk_buff *skb;
1334 while ((skb = __skb_dequeue(list)) != NULL)
1335 kfree_skb(skb);
1336}
1337
1338/**
Christoph Hellwigb4e54de2006-07-24 15:31:14 -07001339 * __dev_alloc_skb - allocate an skbuff for receiving
Linus Torvalds1da177e2005-04-16 15:20:36 -07001340 * @length: length to allocate
1341 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1342 *
1343 * Allocate a new &sk_buff and assign it a usage count of one. The
1344 * buffer has unspecified headroom built in. Users should allocate
1345 * the headroom they think they need without accounting for the
1346 * built in space. The built in space is used for optimisations.
1347 *
Christoph Hellwig766ea8c2006-08-07 15:49:53 -07001348 * %NULL is returned if there is no free memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001349 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
Al Virodd0fc662005-10-07 07:46:04 +01001351 gfp_t gfp_mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352{
Anton Blanchard025be812006-03-31 02:27:06 -08001353 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001354 if (likely(skb))
Anton Blanchard025be812006-03-31 02:27:06 -08001355 skb_reserve(skb, NET_SKB_PAD);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356 return skb;
1357}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358
Ilpo Järvinenf58518e2008-03-27 17:51:31 -07001359extern struct sk_buff *dev_alloc_skb(unsigned int length);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360
Christoph Hellwig8af27452006-07-31 22:35:23 -07001361extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1362 unsigned int length, gfp_t gfp_mask);
1363
1364/**
1365 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1366 * @dev: network device to receive on
1367 * @length: length to allocate
1368 *
1369 * Allocate a new &sk_buff and assign it a usage count of one. The
1370 * buffer has unspecified headroom built in. Users should allocate
1371 * the headroom they think they need without accounting for the
1372 * built in space. The built in space is used for optimisations.
1373 *
1374 * %NULL is returned if there is no free memory. Although this function
1375 * allocates memory it can be called from an interrupt.
1376 */
1377static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1378 unsigned int length)
1379{
1380 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1381}
1382
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383/**
Patrick McHardy334a8132007-06-25 04:35:20 -07001384 * skb_clone_writable - is the header of a clone writable
1385 * @skb: buffer to check
1386 * @len: length up to which to write
1387 *
1388 * Returns true if modifying the header part of the cloned buffer
1389 * does not requires the data to be copied.
1390 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001391static inline int skb_clone_writable(struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07001392{
1393 return !skb_header_cloned(skb) &&
1394 skb_headroom(skb) + len <= skb->hdr_len;
1395}
1396
Herbert Xud9cc2042007-09-16 16:21:16 -07001397static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
1398 int cloned)
1399{
1400 int delta = 0;
1401
1402 if (headroom < NET_SKB_PAD)
1403 headroom = NET_SKB_PAD;
1404 if (headroom > skb_headroom(skb))
1405 delta = headroom - skb_headroom(skb);
1406
1407 if (delta || cloned)
1408 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
1409 GFP_ATOMIC);
1410 return 0;
1411}
1412
Patrick McHardy334a8132007-06-25 04:35:20 -07001413/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001414 * skb_cow - copy header of skb when it is required
1415 * @skb: buffer to cow
1416 * @headroom: needed headroom
1417 *
1418 * If the skb passed lacks sufficient headroom or its data part
1419 * is shared, data is reallocated. If reallocation fails, an error
1420 * is returned and original skb is not changed.
1421 *
1422 * The result is skb with writable area skb->head...skb->tail
1423 * and at least @headroom of space at head.
1424 */
1425static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1426{
Herbert Xud9cc2042007-09-16 16:21:16 -07001427 return __skb_cow(skb, headroom, skb_cloned(skb));
1428}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001429
Herbert Xud9cc2042007-09-16 16:21:16 -07001430/**
1431 * skb_cow_head - skb_cow but only making the head writable
1432 * @skb: buffer to cow
1433 * @headroom: needed headroom
1434 *
1435 * This function is identical to skb_cow except that we replace the
1436 * skb_cloned check by skb_header_cloned. It should be used when
1437 * you only need to push on some header and do not need to modify
1438 * the data.
1439 */
1440static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
1441{
1442 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001443}
1444
1445/**
1446 * skb_padto - pad an skbuff up to a minimal size
1447 * @skb: buffer to pad
1448 * @len: minimal length
1449 *
1450 * Pads up a buffer to ensure the trailing bytes exist and are
1451 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07001452 * is untouched. Otherwise it is extended. Returns zero on
1453 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001454 */
1455
Herbert Xu5b057c62006-06-23 02:06:41 -07001456static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001457{
1458 unsigned int size = skb->len;
1459 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07001460 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001461 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462}
1463
1464static inline int skb_add_data(struct sk_buff *skb,
1465 char __user *from, int copy)
1466{
1467 const int off = skb->len;
1468
1469 if (skb->ip_summed == CHECKSUM_NONE) {
1470 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08001471 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472 copy, 0, &err);
1473 if (!err) {
1474 skb->csum = csum_block_add(skb->csum, csum, off);
1475 return 0;
1476 }
1477 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1478 return 0;
1479
1480 __skb_trim(skb, off);
1481 return -EFAULT;
1482}
1483
1484static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1485 struct page *page, int off)
1486{
1487 if (i) {
1488 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1489
1490 return page == frag->page &&
1491 off == frag->page_offset + frag->size;
1492 }
1493 return 0;
1494}
1495
Herbert Xu364c6ba2006-06-09 16:10:40 -07001496static inline int __skb_linearize(struct sk_buff *skb)
1497{
1498 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1499}
1500
Linus Torvalds1da177e2005-04-16 15:20:36 -07001501/**
1502 * skb_linearize - convert paged skb to linear one
1503 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07001504 *
1505 * If there is no free memory -ENOMEM is returned, otherwise zero
1506 * is returned and the old skb data released.
1507 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07001508static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001509{
Herbert Xu364c6ba2006-06-09 16:10:40 -07001510 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1511}
1512
1513/**
1514 * skb_linearize_cow - make sure skb is linear and writable
1515 * @skb: buffer to process
1516 *
1517 * If there is no free memory -ENOMEM is returned, otherwise zero
1518 * is returned and the old skb data released.
1519 */
1520static inline int skb_linearize_cow(struct sk_buff *skb)
1521{
1522 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1523 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524}
1525
1526/**
1527 * skb_postpull_rcsum - update checksum for received skb after pull
1528 * @skb: buffer to update
1529 * @start: start of data before pull
1530 * @len: length of data pulled
1531 *
1532 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07001533 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1534 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001535 */
1536
1537static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08001538 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539{
Patrick McHardy84fa7932006-08-29 16:44:56 -07001540 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001541 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1542}
1543
Herbert Xucbb042f2006-03-20 22:43:56 -08001544unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1545
Linus Torvalds1da177e2005-04-16 15:20:36 -07001546/**
1547 * pskb_trim_rcsum - trim received skb and update checksum
1548 * @skb: buffer to trim
1549 * @len: new length
1550 *
1551 * This is exactly the same as pskb_trim except that it ensures the
1552 * checksum of received packets are still valid after the operation.
1553 */
1554
1555static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1556{
Stephen Hemminger0e4e4222005-09-08 12:32:03 -07001557 if (likely(len >= skb->len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558 return 0;
Patrick McHardy84fa7932006-08-29 16:44:56 -07001559 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001560 skb->ip_summed = CHECKSUM_NONE;
1561 return __pskb_trim(skb, len);
1562}
1563
Linus Torvalds1da177e2005-04-16 15:20:36 -07001564#define skb_queue_walk(queue, skb) \
1565 for (skb = (queue)->next; \
1566 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1567 skb = skb->next)
1568
James Chapman46f89142007-04-30 00:07:31 -07001569#define skb_queue_walk_safe(queue, skb, tmp) \
1570 for (skb = (queue)->next, tmp = skb->next; \
1571 skb != (struct sk_buff *)(queue); \
1572 skb = tmp, tmp = skb->next)
1573
Stephen Hemminger300ce172005-10-30 13:47:34 -08001574#define skb_queue_reverse_walk(queue, skb) \
1575 for (skb = (queue)->prev; \
1576 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1577 skb = skb->prev)
1578
Linus Torvalds1da177e2005-04-16 15:20:36 -07001579
Herbert Xua59322b2007-12-05 01:53:40 -08001580extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
1581 int *peeked, int *err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1583 int noblock, int *err);
1584extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1585 struct poll_table_struct *wait);
1586extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1587 int offset, struct iovec *to,
1588 int size);
Herbert Xufb286bb2005-11-10 13:01:24 -08001589extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001590 int hlen,
1591 struct iovec *iov);
Rusty Russelldb543c12008-08-15 15:13:53 -07001592extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
1593 int offset,
1594 struct iovec *from,
1595 int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001596extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
Herbert Xu27ab2562007-12-05 01:51:58 -08001597extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
Herbert Xu3305b802005-12-13 23:16:37 -08001598 unsigned int flags);
Al Viro2bbbc862006-11-14 21:37:14 -08001599extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
1600 int len, __wsum csum);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001601extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1602 void *to, int len);
Stephen Hemminger0c6fcc82007-04-20 16:40:01 -07001603extern int skb_store_bits(struct sk_buff *skb, int offset,
1604 const void *from, int len);
Al Viro81d77662006-11-14 21:37:33 -08001605extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001606 int offset, u8 *to, int len,
Al Viro81d77662006-11-14 21:37:33 -08001607 __wsum csum);
Jens Axboe9c55e012007-11-06 23:30:13 -08001608extern int skb_splice_bits(struct sk_buff *skb,
1609 unsigned int offset,
1610 struct pipe_inode_info *pipe,
1611 unsigned int len,
1612 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001613extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1614extern void skb_split(struct sk_buff *skb,
1615 struct sk_buff *skb1, const u32 len);
1616
Herbert Xu576a30e2006-06-27 13:22:38 -07001617extern struct sk_buff *skb_segment(struct sk_buff *skb, int features);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001618
Linus Torvalds1da177e2005-04-16 15:20:36 -07001619static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1620 int len, void *buffer)
1621{
1622 int hlen = skb_headlen(skb);
1623
Patrick McHardy55820ee2005-07-05 14:08:10 -07001624 if (hlen - offset >= len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001625 return skb->data + offset;
1626
1627 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1628 return NULL;
1629
1630 return buffer;
1631}
1632
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03001633static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
1634 void *to,
1635 const unsigned int len)
1636{
1637 memcpy(to, skb->data, len);
1638}
1639
1640static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
1641 const int offset, void *to,
1642 const unsigned int len)
1643{
1644 memcpy(to, skb->data + offset, len);
1645}
1646
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03001647static inline void skb_copy_to_linear_data(struct sk_buff *skb,
1648 const void *from,
1649 const unsigned int len)
1650{
1651 memcpy(skb->data, from, len);
1652}
1653
1654static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
1655 const int offset,
1656 const void *from,
1657 const unsigned int len)
1658{
1659 memcpy(skb->data + offset, from, len);
1660}
1661
Linus Torvalds1da177e2005-04-16 15:20:36 -07001662extern void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001663
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001664/**
1665 * skb_get_timestamp - get timestamp from a skb
1666 * @skb: skb to get stamp from
1667 * @stamp: pointer to struct timeval to store stamp in
1668 *
1669 * Timestamps are stored in the skb as offsets to a base timestamp.
1670 * This function converts the offset back to a struct timeval and stores
1671 * it in stamp.
1672 */
Stephen Hemmingerf2c38392005-09-06 15:48:03 -07001673static inline void skb_get_timestamp(const struct sk_buff *skb, struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001674{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001675 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001676}
1677
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001678static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001679{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001680 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001681}
1682
Stephen Hemminger164891a2007-04-23 22:26:16 -07001683static inline ktime_t net_timedelta(ktime_t t)
1684{
1685 return ktime_sub(ktime_get_real(), t);
1686}
1687
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07001688static inline ktime_t net_invalid_timestamp(void)
1689{
1690 return ktime_set(0, 0);
1691}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001692
Herbert Xu759e5d02007-03-25 20:10:56 -07001693extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
Al Virob51655b2006-11-14 21:40:42 -08001694extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08001695
Herbert Xu60476372007-04-09 11:59:39 -07001696static inline int skb_csum_unnecessary(const struct sk_buff *skb)
1697{
1698 return skb->ip_summed & CHECKSUM_UNNECESSARY;
1699}
1700
Herbert Xufb286bb2005-11-10 13:01:24 -08001701/**
1702 * skb_checksum_complete - Calculate checksum of an entire packet
1703 * @skb: packet to process
1704 *
1705 * This function calculates the checksum over the entire packet plus
1706 * the value of skb->csum. The latter can be used to supply the
1707 * checksum of a pseudo header as used by TCP/UDP. It returns the
1708 * checksum.
1709 *
1710 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1711 * this function can be used to verify that checksum on received
1712 * packets. In that case the function should return zero if the
1713 * checksum is correct. In particular, this function will return zero
1714 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1715 * hardware has already verified the correctness of the checksum.
1716 */
Al Viro4381ca32007-07-15 21:00:11 +01001717static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08001718{
Herbert Xu60476372007-04-09 11:59:39 -07001719 return skb_csum_unnecessary(skb) ?
1720 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08001721}
1722
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07001723#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07001724extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001725static inline void nf_conntrack_put(struct nf_conntrack *nfct)
1726{
1727 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07001728 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001729}
1730static inline void nf_conntrack_get(struct nf_conntrack *nfct)
1731{
1732 if (nfct)
1733 atomic_inc(&nfct->use);
1734}
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -08001735static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
1736{
1737 if (skb)
1738 atomic_inc(&skb->users);
1739}
1740static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
1741{
1742 if (skb)
1743 kfree_skb(skb);
1744}
1745#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746#ifdef CONFIG_BRIDGE_NETFILTER
1747static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
1748{
1749 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
1750 kfree(nf_bridge);
1751}
1752static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
1753{
1754 if (nf_bridge)
1755 atomic_inc(&nf_bridge->use);
1756}
1757#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08001758static inline void nf_reset(struct sk_buff *skb)
1759{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07001760#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08001761 nf_conntrack_put(skb->nfct);
1762 skb->nfct = NULL;
Patrick McHardya193a4a2006-03-20 19:23:05 -08001763 nf_conntrack_put_reasm(skb->nfct_reasm);
1764 skb->nfct_reasm = NULL;
1765#endif
1766#ifdef CONFIG_BRIDGE_NETFILTER
1767 nf_bridge_put(skb->nf_bridge);
1768 skb->nf_bridge = NULL;
1769#endif
1770}
1771
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07001772/* Note: This doesn't put any conntrack and bridge info in dst. */
1773static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
1774{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07001775#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07001776 dst->nfct = src->nfct;
1777 nf_conntrack_get(src->nfct);
1778 dst->nfctinfo = src->nfctinfo;
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07001779 dst->nfct_reasm = src->nfct_reasm;
1780 nf_conntrack_get_reasm(src->nfct_reasm);
1781#endif
1782#ifdef CONFIG_BRIDGE_NETFILTER
1783 dst->nf_bridge = src->nf_bridge;
1784 nf_bridge_get(src->nf_bridge);
1785#endif
1786}
1787
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07001788static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
1789{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07001790#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07001791 nf_conntrack_put(dst->nfct);
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07001792 nf_conntrack_put_reasm(dst->nfct_reasm);
1793#endif
1794#ifdef CONFIG_BRIDGE_NETFILTER
1795 nf_bridge_put(dst->nf_bridge);
1796#endif
1797 __nf_copy(dst, src);
1798}
1799
James Morris984bc162006-06-09 00:29:17 -07001800#ifdef CONFIG_NETWORK_SECMARK
1801static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1802{
1803 to->secmark = from->secmark;
1804}
1805
1806static inline void skb_init_secmark(struct sk_buff *skb)
1807{
1808 skb->secmark = 0;
1809}
1810#else
1811static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1812{ }
1813
1814static inline void skb_init_secmark(struct sk_buff *skb)
1815{ }
1816#endif
1817
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001818static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
1819{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001820 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001821}
1822
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07001823static inline u16 skb_get_queue_mapping(struct sk_buff *skb)
1824{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07001825 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07001826}
1827
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001828static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
1829{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001830 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001831}
1832
Herbert Xu89114af2006-07-08 13:34:32 -07001833static inline int skb_is_gso(const struct sk_buff *skb)
1834{
1835 return skb_shinfo(skb)->gso_size;
1836}
1837
Brice Goglineabd7e32007-10-13 12:33:32 +02001838static inline int skb_is_gso_v6(const struct sk_buff *skb)
1839{
1840 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
1841}
1842
Ben Hutchings4497b072008-06-19 16:22:28 -07001843extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
1844
1845static inline bool skb_warn_if_lro(const struct sk_buff *skb)
1846{
1847 /* LRO sets gso_size but not gso_type, whereas if GSO is really
1848 * wanted then gso_type will be set. */
1849 struct skb_shared_info *shinfo = skb_shinfo(skb);
1850 if (shinfo->gso_size != 0 && unlikely(shinfo->gso_type == 0)) {
1851 __skb_warn_lro_forwarding(skb);
1852 return true;
1853 }
1854 return false;
1855}
1856
Herbert Xu35fc92a2007-03-26 23:22:20 -07001857static inline void skb_forward_csum(struct sk_buff *skb)
1858{
1859 /* Unfortunately we don't support this one. Any brave souls? */
1860 if (skb->ip_summed == CHECKSUM_COMPLETE)
1861 skb->ip_summed = CHECKSUM_NONE;
1862}
1863
Rusty Russellf35d9d82008-02-04 23:49:54 -05001864bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001865#endif /* __KERNEL__ */
1866#endif /* _LINUX_SKBUFF_H */