blob: 2725f4e5a9bf837f691ddc88361fcc280af4fb96 [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
Lennert Buytenhek04a4bb52008-10-01 02:33:12 -0700386extern int skb_recycle_check(struct sk_buff *skb, int skb_size);
387
Herbert Xue0053ec2007-10-14 00:37:52 -0700388extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700389extern struct sk_buff *skb_clone(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 *skb_copy(const struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100392 gfp_t priority);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700393extern struct sk_buff *pskb_copy(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100394 gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395extern int pskb_expand_head(struct sk_buff *skb,
Victor Fusco86a76ca2005-07-08 14:57:47 -0700396 int nhead, int ntail,
Al Virodd0fc662005-10-07 07:46:04 +0100397 gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
399 unsigned int headroom);
400extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
401 int newheadroom, int newtailroom,
Al Virodd0fc662005-10-07 07:46:04 +0100402 gfp_t priority);
David Howells716ea3a2007-04-02 20:19:53 -0700403extern int skb_to_sgvec(struct sk_buff *skb,
404 struct scatterlist *sg, int offset,
405 int len);
406extern int skb_cow_data(struct sk_buff *skb, int tailbits,
407 struct sk_buff **trailer);
Herbert Xu5b057c62006-06-23 02:06:41 -0700408extern int skb_pad(struct sk_buff *skb, int pad);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409#define dev_kfree_skb(a) kfree_skb(a)
410extern void skb_over_panic(struct sk_buff *skb, int len,
411 void *here);
412extern void skb_under_panic(struct sk_buff *skb, int len,
413 void *here);
David S. Millerdc6de332006-04-20 00:10:50 -0700414extern void skb_truesize_bug(struct sk_buff *skb);
415
416static inline void skb_truesize_check(struct sk_buff *skb)
417{
Chuck Lever78608ba2007-11-10 21:53:30 -0800418 int len = sizeof(struct sk_buff) + skb->len;
419
420 if (unlikely((int)skb->truesize < len))
David S. Millerdc6de332006-04-20 00:10:50 -0700421 skb_truesize_bug(skb);
422}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700424extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
425 int getfrag(void *from, char *to, int offset,
426 int len,int odd, struct sk_buff *skb),
427 void *from, int length);
428
Thomas Graf677e90e2005-06-23 20:59:51 -0700429struct skb_seq_state
430{
431 __u32 lower_offset;
432 __u32 upper_offset;
433 __u32 frag_idx;
434 __u32 stepped_offset;
435 struct sk_buff *root_skb;
436 struct sk_buff *cur_skb;
437 __u8 *frag_data;
438};
439
440extern void skb_prepare_seq_read(struct sk_buff *skb,
441 unsigned int from, unsigned int to,
442 struct skb_seq_state *st);
443extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
444 struct skb_seq_state *st);
445extern void skb_abort_seq_read(struct skb_seq_state *st);
446
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700447extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
448 unsigned int to, struct ts_config *config,
449 struct ts_state *state);
450
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700451#ifdef NET_SKBUFF_DATA_USES_OFFSET
452static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
453{
454 return skb->head + skb->end;
455}
456#else
457static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
458{
459 return skb->end;
460}
461#endif
462
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700464#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700465
466/**
467 * skb_queue_empty - check if a queue is empty
468 * @list: queue head
469 *
470 * Returns true if the queue is empty, false otherwise.
471 */
472static inline int skb_queue_empty(const struct sk_buff_head *list)
473{
474 return list->next == (struct sk_buff *)list;
475}
476
477/**
David S. Millerfc7ebb22008-09-23 00:34:07 -0700478 * skb_queue_is_last - check if skb is the last entry in the queue
479 * @list: queue head
480 * @skb: buffer
481 *
482 * Returns true if @skb is the last buffer on the list.
483 */
484static inline bool skb_queue_is_last(const struct sk_buff_head *list,
485 const struct sk_buff *skb)
486{
487 return (skb->next == (struct sk_buff *) list);
488}
489
490/**
David S. Miller249c8b42008-09-23 00:44:42 -0700491 * skb_queue_next - return the next packet in the queue
492 * @list: queue head
493 * @skb: current buffer
494 *
495 * Return the next packet in @list after @skb. It is only valid to
496 * call this if skb_queue_is_last() evaluates to false.
497 */
498static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
499 const struct sk_buff *skb)
500{
501 /* This BUG_ON may seem severe, but if we just return then we
502 * are going to dereference garbage.
503 */
504 BUG_ON(skb_queue_is_last(list, skb));
505 return skb->next;
506}
507
508/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700509 * skb_get - reference buffer
510 * @skb: buffer to reference
511 *
512 * Makes another reference to a socket buffer and returns a pointer
513 * to the buffer.
514 */
515static inline struct sk_buff *skb_get(struct sk_buff *skb)
516{
517 atomic_inc(&skb->users);
518 return skb;
519}
520
521/*
522 * If users == 1, we are the only owner and are can avoid redundant
523 * atomic change.
524 */
525
526/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 * skb_cloned - is the buffer a clone
528 * @skb: buffer to check
529 *
530 * Returns true if the buffer was generated with skb_clone() and is
531 * one of multiple shared copies of the buffer. Cloned buffers are
532 * shared data so must not be written to under normal circumstances.
533 */
534static inline int skb_cloned(const struct sk_buff *skb)
535{
536 return skb->cloned &&
537 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
538}
539
540/**
541 * skb_header_cloned - is the header a clone
542 * @skb: buffer to check
543 *
544 * Returns true if modifying the header part of the buffer requires
545 * the data to be copied.
546 */
547static inline int skb_header_cloned(const struct sk_buff *skb)
548{
549 int dataref;
550
551 if (!skb->cloned)
552 return 0;
553
554 dataref = atomic_read(&skb_shinfo(skb)->dataref);
555 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
556 return dataref != 1;
557}
558
559/**
560 * skb_header_release - release reference to header
561 * @skb: buffer to operate on
562 *
563 * Drop a reference to the header part of the buffer. This is done
564 * by acquiring a payload reference. You must not read from the header
565 * part of skb->data after this.
566 */
567static inline void skb_header_release(struct sk_buff *skb)
568{
569 BUG_ON(skb->nohdr);
570 skb->nohdr = 1;
571 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
572}
573
574/**
575 * skb_shared - is the buffer shared
576 * @skb: buffer to check
577 *
578 * Returns true if more than one person has a reference to this
579 * buffer.
580 */
581static inline int skb_shared(const struct sk_buff *skb)
582{
583 return atomic_read(&skb->users) != 1;
584}
585
586/**
587 * skb_share_check - check if buffer is shared and if so clone it
588 * @skb: buffer to check
589 * @pri: priority for memory allocation
590 *
591 * If the buffer is shared the buffer is cloned and the old copy
592 * drops a reference. A new clone with a single reference is returned.
593 * If the buffer is not shared the original buffer is returned. When
594 * being called from interrupt status or with spinlocks held pri must
595 * be GFP_ATOMIC.
596 *
597 * NULL is returned on a memory allocation failure.
598 */
Victor Fusco86a76ca2005-07-08 14:57:47 -0700599static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100600 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601{
602 might_sleep_if(pri & __GFP_WAIT);
603 if (skb_shared(skb)) {
604 struct sk_buff *nskb = skb_clone(skb, pri);
605 kfree_skb(skb);
606 skb = nskb;
607 }
608 return skb;
609}
610
611/*
612 * Copy shared buffers into a new sk_buff. We effectively do COW on
613 * packets to handle cases where we have a local reader and forward
614 * and a couple of other messy ones. The normal one is tcpdumping
615 * a packet thats being forwarded.
616 */
617
618/**
619 * skb_unshare - make a copy of a shared buffer
620 * @skb: buffer to check
621 * @pri: priority for memory allocation
622 *
623 * If the socket buffer is a clone then this function creates a new
624 * copy of the data, drops a reference count on the old copy and returns
625 * the new copy with the reference count at 1. If the buffer is not a clone
626 * the original buffer is returned. When called with a spinlock held or
627 * from interrupt state @pri must be %GFP_ATOMIC
628 *
629 * %NULL is returned on a memory allocation failure.
630 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -0700631static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100632 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633{
634 might_sleep_if(pri & __GFP_WAIT);
635 if (skb_cloned(skb)) {
636 struct sk_buff *nskb = skb_copy(skb, pri);
637 kfree_skb(skb); /* Free our shared copy */
638 skb = nskb;
639 }
640 return skb;
641}
642
643/**
644 * skb_peek
645 * @list_: list to peek at
646 *
647 * Peek an &sk_buff. Unlike most other operations you _MUST_
648 * be careful with this one. A peek leaves the buffer on the
649 * list and someone else may run off with it. You must hold
650 * the appropriate locks or have a private queue to do this.
651 *
652 * Returns %NULL for an empty list or a pointer to the head element.
653 * The reference count is not incremented and the reference is therefore
654 * volatile. Use with caution.
655 */
656static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
657{
658 struct sk_buff *list = ((struct sk_buff *)list_)->next;
659 if (list == (struct sk_buff *)list_)
660 list = NULL;
661 return list;
662}
663
664/**
665 * skb_peek_tail
666 * @list_: list to peek at
667 *
668 * Peek an &sk_buff. Unlike most other operations you _MUST_
669 * be careful with this one. A peek leaves the buffer on the
670 * list and someone else may run off with it. You must hold
671 * the appropriate locks or have a private queue to do this.
672 *
673 * Returns %NULL for an empty list or a pointer to the tail element.
674 * The reference count is not incremented and the reference is therefore
675 * volatile. Use with caution.
676 */
677static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
678{
679 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
680 if (list == (struct sk_buff *)list_)
681 list = NULL;
682 return list;
683}
684
685/**
686 * skb_queue_len - get queue length
687 * @list_: list to measure
688 *
689 * Return the length of an &sk_buff queue.
690 */
691static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
692{
693 return list_->qlen;
694}
695
David S. Miller67fed452008-09-21 22:36:24 -0700696/**
697 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
698 * @list: queue to initialize
699 *
700 * This initializes only the list and queue length aspects of
701 * an sk_buff_head object. This allows to initialize the list
702 * aspects of an sk_buff_head without reinitializing things like
703 * the spinlock. It can also be used for on-stack sk_buff_head
704 * objects where the spinlock is known to not be used.
705 */
706static inline void __skb_queue_head_init(struct sk_buff_head *list)
707{
708 list->prev = list->next = (struct sk_buff *)list;
709 list->qlen = 0;
710}
711
Arjan van de Ven76f10ad2006-08-02 14:06:55 -0700712/*
713 * This function creates a split out lock class for each invocation;
714 * this is needed for now since a whole lot of users of the skb-queue
715 * infrastructure in drivers have different locking usage (in hardirq)
716 * than the networking core (in softirq only). In the long run either the
717 * network layer or drivers should need annotation to consolidate the
718 * main types of usage into 3 classes.
719 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700720static inline void skb_queue_head_init(struct sk_buff_head *list)
721{
722 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -0700723 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700724}
725
Pavel Emelianovc2ecba72007-04-17 12:45:31 -0700726static inline void skb_queue_head_init_class(struct sk_buff_head *list,
727 struct lock_class_key *class)
728{
729 skb_queue_head_init(list);
730 lockdep_set_class(&list->lock, class);
731}
732
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733/*
Gerrit Renkerbf299272008-04-14 00:04:51 -0700734 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735 *
736 * The "__skb_xxxx()" functions are the non-atomic ones that
737 * can only be called with interrupts disabled.
738 */
Gerrit Renkerbf299272008-04-14 00:04:51 -0700739extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
740static inline void __skb_insert(struct sk_buff *newsk,
741 struct sk_buff *prev, struct sk_buff *next,
742 struct sk_buff_head *list)
743{
744 newsk->next = next;
745 newsk->prev = prev;
746 next->prev = prev->next = newsk;
747 list->qlen++;
748}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749
David S. Miller67fed452008-09-21 22:36:24 -0700750static inline void __skb_queue_splice(const struct sk_buff_head *list,
751 struct sk_buff *prev,
752 struct sk_buff *next)
753{
754 struct sk_buff *first = list->next;
755 struct sk_buff *last = list->prev;
756
757 first->prev = prev;
758 prev->next = first;
759
760 last->next = next;
761 next->prev = last;
762}
763
764/**
765 * skb_queue_splice - join two skb lists, this is designed for stacks
766 * @list: the new list to add
767 * @head: the place to add it in the first list
768 */
769static inline void skb_queue_splice(const struct sk_buff_head *list,
770 struct sk_buff_head *head)
771{
772 if (!skb_queue_empty(list)) {
773 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -0700774 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -0700775 }
776}
777
778/**
779 * skb_queue_splice - join two skb lists and reinitialise the emptied list
780 * @list: the new list to add
781 * @head: the place to add it in the first list
782 *
783 * The list at @list is reinitialised
784 */
785static inline void skb_queue_splice_init(struct sk_buff_head *list,
786 struct sk_buff_head *head)
787{
788 if (!skb_queue_empty(list)) {
789 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -0700790 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -0700791 __skb_queue_head_init(list);
792 }
793}
794
795/**
796 * skb_queue_splice_tail - join two skb lists, each list being a queue
797 * @list: the new list to add
798 * @head: the place to add it in the first list
799 */
800static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
801 struct sk_buff_head *head)
802{
803 if (!skb_queue_empty(list)) {
804 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -0700805 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -0700806 }
807}
808
809/**
810 * skb_queue_splice_tail - join two skb lists and reinitialise the emptied list
811 * @list: the new list to add
812 * @head: the place to add it in the first list
813 *
814 * Each of the lists is a queue.
815 * The list at @list is reinitialised
816 */
817static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
818 struct sk_buff_head *head)
819{
820 if (!skb_queue_empty(list)) {
821 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -0700822 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -0700823 __skb_queue_head_init(list);
824 }
825}
826
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827/**
Stephen Hemminger300ce172005-10-30 13:47:34 -0800828 * __skb_queue_after - queue a buffer at the list head
829 * @list: list to use
830 * @prev: place after this buffer
831 * @newsk: buffer to queue
832 *
833 * Queue a buffer int the middle of a list. This function takes no locks
834 * and you must therefore hold required locks before calling it.
835 *
836 * A buffer cannot be placed on two lists at the same time.
837 */
838static inline void __skb_queue_after(struct sk_buff_head *list,
839 struct sk_buff *prev,
840 struct sk_buff *newsk)
841{
Gerrit Renkerbf299272008-04-14 00:04:51 -0700842 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -0800843}
844
Gerrit Renker7de6c032008-04-14 00:05:09 -0700845extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
846 struct sk_buff_head *list);
847
Gerrit Renkerf5572852008-04-14 00:05:28 -0700848static inline void __skb_queue_before(struct sk_buff_head *list,
849 struct sk_buff *next,
850 struct sk_buff *newsk)
851{
852 __skb_insert(newsk, next->prev, next, list);
853}
854
Stephen Hemminger300ce172005-10-30 13:47:34 -0800855/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856 * __skb_queue_head - queue a buffer at the list head
857 * @list: list to use
858 * @newsk: buffer to queue
859 *
860 * Queue a buffer at the start of a list. This function takes no locks
861 * and you must therefore hold required locks before calling it.
862 *
863 * A buffer cannot be placed on two lists at the same time.
864 */
865extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
866static inline void __skb_queue_head(struct sk_buff_head *list,
867 struct sk_buff *newsk)
868{
Stephen Hemminger300ce172005-10-30 13:47:34 -0800869 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870}
871
872/**
873 * __skb_queue_tail - queue a buffer at the list tail
874 * @list: list to use
875 * @newsk: buffer to queue
876 *
877 * Queue a buffer at the end of a list. This function takes no locks
878 * and you must therefore hold required locks before calling it.
879 *
880 * A buffer cannot be placed on two lists at the same time.
881 */
882extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
883static inline void __skb_queue_tail(struct sk_buff_head *list,
884 struct sk_buff *newsk)
885{
Gerrit Renkerf5572852008-04-14 00:05:28 -0700886 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887}
888
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890 * remove sk_buff from list. _Must_ be called atomically, and with
891 * the list known..
892 */
David S. Miller8728b832005-08-09 19:25:21 -0700893extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
895{
896 struct sk_buff *next, *prev;
897
898 list->qlen--;
899 next = skb->next;
900 prev = skb->prev;
901 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902 next->prev = prev;
903 prev->next = next;
904}
905
Gerrit Renkerf525c062008-04-14 00:04:12 -0700906/**
907 * __skb_dequeue - remove from the head of the queue
908 * @list: list to dequeue from
909 *
910 * Remove the head of the list. This function does not take any locks
911 * so must be used with appropriate locks held only. The head item is
912 * returned or %NULL if the list is empty.
913 */
914extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
915static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
916{
917 struct sk_buff *skb = skb_peek(list);
918 if (skb)
919 __skb_unlink(skb, list);
920 return skb;
921}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922
923/**
924 * __skb_dequeue_tail - remove from the tail of the queue
925 * @list: list to dequeue from
926 *
927 * Remove the tail of the list. This function does not take any locks
928 * so must be used with appropriate locks held only. The tail item is
929 * returned or %NULL if the list is empty.
930 */
931extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
932static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
933{
934 struct sk_buff *skb = skb_peek_tail(list);
935 if (skb)
936 __skb_unlink(skb, list);
937 return skb;
938}
939
940
941static inline int skb_is_nonlinear(const struct sk_buff *skb)
942{
943 return skb->data_len;
944}
945
946static inline unsigned int skb_headlen(const struct sk_buff *skb)
947{
948 return skb->len - skb->data_len;
949}
950
951static inline int skb_pagelen(const struct sk_buff *skb)
952{
953 int i, len = 0;
954
955 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
956 len += skb_shinfo(skb)->frags[i].size;
957 return len + skb_headlen(skb);
958}
959
960static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
961 struct page *page, int off, int size)
962{
963 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
964
965 frag->page = page;
966 frag->page_offset = off;
967 frag->size = size;
968 skb_shinfo(skb)->nr_frags = i + 1;
969}
970
Peter Zijlstra654bed12008-10-07 14:22:33 -0700971extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
972 int off, int size);
973
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
975#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
976#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
977
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700978#ifdef NET_SKBUFF_DATA_USES_OFFSET
979static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
980{
981 return skb->head + skb->tail;
982}
983
984static inline void skb_reset_tail_pointer(struct sk_buff *skb)
985{
986 skb->tail = skb->data - skb->head;
987}
988
989static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
990{
991 skb_reset_tail_pointer(skb);
992 skb->tail += offset;
993}
994#else /* NET_SKBUFF_DATA_USES_OFFSET */
995static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
996{
997 return skb->tail;
998}
999
1000static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1001{
1002 skb->tail = skb->data;
1003}
1004
1005static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1006{
1007 skb->tail = skb->data + offset;
1008}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001009
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001010#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1011
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012/*
1013 * Add data to an sk_buff
1014 */
Ilpo Järvinen0dde3e12008-03-27 17:43:41 -07001015extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001016static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1017{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001018 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019 SKB_LINEAR_ASSERT(skb);
1020 skb->tail += len;
1021 skb->len += len;
1022 return tmp;
1023}
1024
Ilpo Järvinenc2aa2702008-03-27 17:52:40 -07001025extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1027{
1028 skb->data -= len;
1029 skb->len += len;
1030 return skb->data;
1031}
1032
Ilpo Järvinen6be8ac22008-03-27 17:47:24 -07001033extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001034static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1035{
1036 skb->len -= len;
1037 BUG_ON(skb->len < skb->data_len);
1038 return skb->data += len;
1039}
1040
Linus Torvalds1da177e2005-04-16 15:20:36 -07001041extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1042
1043static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1044{
1045 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001046 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047 return NULL;
1048 skb->len -= len;
1049 return skb->data += len;
1050}
1051
1052static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1053{
1054 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1055}
1056
1057static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1058{
1059 if (likely(len <= skb_headlen(skb)))
1060 return 1;
1061 if (unlikely(len > skb->len))
1062 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001063 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001064}
1065
1066/**
1067 * skb_headroom - bytes at buffer head
1068 * @skb: buffer to check
1069 *
1070 * Return the number of bytes of free space at the head of an &sk_buff.
1071 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001072static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001073{
1074 return skb->data - skb->head;
1075}
1076
1077/**
1078 * skb_tailroom - bytes at buffer end
1079 * @skb: buffer to check
1080 *
1081 * Return the number of bytes of free space at the tail of an sk_buff
1082 */
1083static inline int skb_tailroom(const struct sk_buff *skb)
1084{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001085 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001086}
1087
1088/**
1089 * skb_reserve - adjust headroom
1090 * @skb: buffer to alter
1091 * @len: bytes to move
1092 *
1093 * Increase the headroom of an empty &sk_buff by reducing the tail
1094 * room. This is only allowed for an empty buffer.
1095 */
David S. Miller8243126c2006-01-17 02:54:21 -08001096static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001097{
1098 skb->data += len;
1099 skb->tail += len;
1100}
1101
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001102#ifdef NET_SKBUFF_DATA_USES_OFFSET
1103static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1104{
1105 return skb->head + skb->transport_header;
1106}
1107
1108static inline void skb_reset_transport_header(struct sk_buff *skb)
1109{
1110 skb->transport_header = skb->data - skb->head;
1111}
1112
1113static inline void skb_set_transport_header(struct sk_buff *skb,
1114 const int offset)
1115{
1116 skb_reset_transport_header(skb);
1117 skb->transport_header += offset;
1118}
1119
1120static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1121{
1122 return skb->head + skb->network_header;
1123}
1124
1125static inline void skb_reset_network_header(struct sk_buff *skb)
1126{
1127 skb->network_header = skb->data - skb->head;
1128}
1129
1130static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1131{
1132 skb_reset_network_header(skb);
1133 skb->network_header += offset;
1134}
1135
1136static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1137{
1138 return skb->head + skb->mac_header;
1139}
1140
1141static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1142{
1143 return skb->mac_header != ~0U;
1144}
1145
1146static inline void skb_reset_mac_header(struct sk_buff *skb)
1147{
1148 skb->mac_header = skb->data - skb->head;
1149}
1150
1151static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1152{
1153 skb_reset_mac_header(skb);
1154 skb->mac_header += offset;
1155}
1156
1157#else /* NET_SKBUFF_DATA_USES_OFFSET */
1158
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001159static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1160{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001161 return skb->transport_header;
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001162}
1163
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001164static inline void skb_reset_transport_header(struct sk_buff *skb)
1165{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001166 skb->transport_header = skb->data;
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001167}
1168
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001169static inline void skb_set_transport_header(struct sk_buff *skb,
1170 const int offset)
1171{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001172 skb->transport_header = skb->data + offset;
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001173}
1174
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001175static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1176{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001177 return skb->network_header;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001178}
1179
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001180static inline void skb_reset_network_header(struct sk_buff *skb)
1181{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001182 skb->network_header = skb->data;
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001183}
1184
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001185static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1186{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001187 skb->network_header = skb->data + offset;
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001188}
1189
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001190static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1191{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001192 return skb->mac_header;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001193}
1194
1195static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1196{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001197 return skb->mac_header != NULL;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001198}
1199
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001200static inline void skb_reset_mac_header(struct sk_buff *skb)
1201{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001202 skb->mac_header = skb->data;
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001203}
1204
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001205static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1206{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001207 skb->mac_header = skb->data + offset;
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001208}
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001209#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1210
1211static inline int skb_transport_offset(const struct sk_buff *skb)
1212{
1213 return skb_transport_header(skb) - skb->data;
1214}
1215
1216static inline u32 skb_network_header_len(const struct sk_buff *skb)
1217{
1218 return skb->transport_header - skb->network_header;
1219}
1220
1221static inline int skb_network_offset(const struct sk_buff *skb)
1222{
1223 return skb_network_header(skb) - skb->data;
1224}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001225
Linus Torvalds1da177e2005-04-16 15:20:36 -07001226/*
1227 * CPUs often take a performance hit when accessing unaligned memory
1228 * locations. The actual performance hit varies, it can be small if the
1229 * hardware handles it or large if we have to take an exception and fix it
1230 * in software.
1231 *
1232 * Since an ethernet header is 14 bytes network drivers often end up with
1233 * the IP header at an unaligned offset. The IP header can be aligned by
1234 * shifting the start of the packet by 2 bytes. Drivers should do this
1235 * with:
1236 *
1237 * skb_reserve(NET_IP_ALIGN);
1238 *
1239 * The downside to this alignment of the IP header is that the DMA is now
1240 * unaligned. On some architectures the cost of an unaligned DMA is high
1241 * and this cost outweighs the gains made by aligning the IP header.
1242 *
1243 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1244 * to be overridden.
1245 */
1246#ifndef NET_IP_ALIGN
1247#define NET_IP_ALIGN 2
1248#endif
1249
Anton Blanchard025be812006-03-31 02:27:06 -08001250/*
1251 * The networking layer reserves some headroom in skb data (via
1252 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1253 * the header has to grow. In the default case, if the header has to grow
1254 * 16 bytes or less we avoid the reallocation.
1255 *
1256 * Unfortunately this headroom changes the DMA alignment of the resulting
1257 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1258 * on some architectures. An architecture can override this value,
1259 * perhaps setting it to a cacheline in size (since that will maintain
1260 * cacheline alignment of the DMA). It must be a power of 2.
1261 *
1262 * Various parts of the networking layer expect at least 16 bytes of
1263 * headroom, you should not reduce this.
1264 */
1265#ifndef NET_SKB_PAD
1266#define NET_SKB_PAD 16
1267#endif
1268
Herbert Xu3cc0e872006-06-09 16:13:38 -07001269extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001270
1271static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1272{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001273 if (unlikely(skb->data_len)) {
1274 WARN_ON(1);
1275 return;
1276 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001277 skb->len = len;
1278 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001279}
1280
Ilpo Järvinen419ae742008-03-27 17:54:01 -07001281extern void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001282
1283static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1284{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001285 if (skb->data_len)
1286 return ___pskb_trim(skb, len);
1287 __skb_trim(skb, len);
1288 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289}
1290
1291static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1292{
1293 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1294}
1295
1296/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07001297 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1298 * @skb: buffer to alter
1299 * @len: new length
1300 *
1301 * This is identical to pskb_trim except that the caller knows that
1302 * the skb is not cloned so we should never get an error due to out-
1303 * of-memory.
1304 */
1305static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1306{
1307 int err = pskb_trim(skb, len);
1308 BUG_ON(err);
1309}
1310
1311/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001312 * skb_orphan - orphan a buffer
1313 * @skb: buffer to orphan
1314 *
1315 * If a buffer currently has an owner then we call the owner's
1316 * destructor function and make the @skb unowned. The buffer continues
1317 * to exist but is no longer charged to its former owner.
1318 */
1319static inline void skb_orphan(struct sk_buff *skb)
1320{
1321 if (skb->destructor)
1322 skb->destructor(skb);
1323 skb->destructor = NULL;
1324 skb->sk = NULL;
1325}
1326
1327/**
1328 * __skb_queue_purge - empty a list
1329 * @list: list to empty
1330 *
1331 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1332 * the list and one reference dropped. This function does not take the
1333 * list lock and the caller must hold the relevant locks to use it.
1334 */
1335extern void skb_queue_purge(struct sk_buff_head *list);
1336static inline void __skb_queue_purge(struct sk_buff_head *list)
1337{
1338 struct sk_buff *skb;
1339 while ((skb = __skb_dequeue(list)) != NULL)
1340 kfree_skb(skb);
1341}
1342
1343/**
Christoph Hellwigb4e54de2006-07-24 15:31:14 -07001344 * __dev_alloc_skb - allocate an skbuff for receiving
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345 * @length: length to allocate
1346 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1347 *
1348 * Allocate a new &sk_buff and assign it a usage count of one. The
1349 * buffer has unspecified headroom built in. Users should allocate
1350 * the headroom they think they need without accounting for the
1351 * built in space. The built in space is used for optimisations.
1352 *
Christoph Hellwig766ea8c2006-08-07 15:49:53 -07001353 * %NULL is returned if there is no free memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001354 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001355static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
Al Virodd0fc662005-10-07 07:46:04 +01001356 gfp_t gfp_mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001357{
Anton Blanchard025be812006-03-31 02:27:06 -08001358 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001359 if (likely(skb))
Anton Blanchard025be812006-03-31 02:27:06 -08001360 skb_reserve(skb, NET_SKB_PAD);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001361 return skb;
1362}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363
Ilpo Järvinenf58518e2008-03-27 17:51:31 -07001364extern struct sk_buff *dev_alloc_skb(unsigned int length);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365
Christoph Hellwig8af27452006-07-31 22:35:23 -07001366extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1367 unsigned int length, gfp_t gfp_mask);
1368
1369/**
1370 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1371 * @dev: network device to receive on
1372 * @length: length to allocate
1373 *
1374 * Allocate a new &sk_buff and assign it a usage count of one. The
1375 * buffer has unspecified headroom built in. Users should allocate
1376 * the headroom they think they need without accounting for the
1377 * built in space. The built in space is used for optimisations.
1378 *
1379 * %NULL is returned if there is no free memory. Although this function
1380 * allocates memory it can be called from an interrupt.
1381 */
1382static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1383 unsigned int length)
1384{
1385 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1386}
1387
Peter Zijlstra654bed12008-10-07 14:22:33 -07001388extern struct page *__netdev_alloc_page(struct net_device *dev, gfp_t gfp_mask);
1389
1390/**
1391 * netdev_alloc_page - allocate a page for ps-rx on a specific device
1392 * @dev: network device to receive on
1393 *
1394 * Allocate a new page node local to the specified device.
1395 *
1396 * %NULL is returned if there is no free memory.
1397 */
1398static inline struct page *netdev_alloc_page(struct net_device *dev)
1399{
1400 return __netdev_alloc_page(dev, GFP_ATOMIC);
1401}
1402
1403static inline void netdev_free_page(struct net_device *dev, struct page *page)
1404{
1405 __free_page(page);
1406}
1407
Linus Torvalds1da177e2005-04-16 15:20:36 -07001408/**
Patrick McHardy334a8132007-06-25 04:35:20 -07001409 * skb_clone_writable - is the header of a clone writable
1410 * @skb: buffer to check
1411 * @len: length up to which to write
1412 *
1413 * Returns true if modifying the header part of the cloned buffer
1414 * does not requires the data to be copied.
1415 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001416static inline int skb_clone_writable(struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07001417{
1418 return !skb_header_cloned(skb) &&
1419 skb_headroom(skb) + len <= skb->hdr_len;
1420}
1421
Herbert Xud9cc2042007-09-16 16:21:16 -07001422static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
1423 int cloned)
1424{
1425 int delta = 0;
1426
1427 if (headroom < NET_SKB_PAD)
1428 headroom = NET_SKB_PAD;
1429 if (headroom > skb_headroom(skb))
1430 delta = headroom - skb_headroom(skb);
1431
1432 if (delta || cloned)
1433 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
1434 GFP_ATOMIC);
1435 return 0;
1436}
1437
Patrick McHardy334a8132007-06-25 04:35:20 -07001438/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001439 * skb_cow - copy header of skb when it is required
1440 * @skb: buffer to cow
1441 * @headroom: needed headroom
1442 *
1443 * If the skb passed lacks sufficient headroom or its data part
1444 * is shared, data is reallocated. If reallocation fails, an error
1445 * is returned and original skb is not changed.
1446 *
1447 * The result is skb with writable area skb->head...skb->tail
1448 * and at least @headroom of space at head.
1449 */
1450static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1451{
Herbert Xud9cc2042007-09-16 16:21:16 -07001452 return __skb_cow(skb, headroom, skb_cloned(skb));
1453}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001454
Herbert Xud9cc2042007-09-16 16:21:16 -07001455/**
1456 * skb_cow_head - skb_cow but only making the head writable
1457 * @skb: buffer to cow
1458 * @headroom: needed headroom
1459 *
1460 * This function is identical to skb_cow except that we replace the
1461 * skb_cloned check by skb_header_cloned. It should be used when
1462 * you only need to push on some header and do not need to modify
1463 * the data.
1464 */
1465static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
1466{
1467 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001468}
1469
1470/**
1471 * skb_padto - pad an skbuff up to a minimal size
1472 * @skb: buffer to pad
1473 * @len: minimal length
1474 *
1475 * Pads up a buffer to ensure the trailing bytes exist and are
1476 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07001477 * is untouched. Otherwise it is extended. Returns zero on
1478 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001479 */
1480
Herbert Xu5b057c62006-06-23 02:06:41 -07001481static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001482{
1483 unsigned int size = skb->len;
1484 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07001485 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001486 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001487}
1488
1489static inline int skb_add_data(struct sk_buff *skb,
1490 char __user *from, int copy)
1491{
1492 const int off = skb->len;
1493
1494 if (skb->ip_summed == CHECKSUM_NONE) {
1495 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08001496 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
Linus Torvalds1da177e2005-04-16 15:20:36 -07001497 copy, 0, &err);
1498 if (!err) {
1499 skb->csum = csum_block_add(skb->csum, csum, off);
1500 return 0;
1501 }
1502 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1503 return 0;
1504
1505 __skb_trim(skb, off);
1506 return -EFAULT;
1507}
1508
1509static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1510 struct page *page, int off)
1511{
1512 if (i) {
1513 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1514
1515 return page == frag->page &&
1516 off == frag->page_offset + frag->size;
1517 }
1518 return 0;
1519}
1520
Herbert Xu364c6ba2006-06-09 16:10:40 -07001521static inline int __skb_linearize(struct sk_buff *skb)
1522{
1523 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1524}
1525
Linus Torvalds1da177e2005-04-16 15:20:36 -07001526/**
1527 * skb_linearize - convert paged skb to linear one
1528 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07001529 *
1530 * If there is no free memory -ENOMEM is returned, otherwise zero
1531 * is returned and the old skb data released.
1532 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07001533static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001534{
Herbert Xu364c6ba2006-06-09 16:10:40 -07001535 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1536}
1537
1538/**
1539 * skb_linearize_cow - make sure skb is linear and writable
1540 * @skb: buffer to process
1541 *
1542 * If there is no free memory -ENOMEM is returned, otherwise zero
1543 * is returned and the old skb data released.
1544 */
1545static inline int skb_linearize_cow(struct sk_buff *skb)
1546{
1547 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1548 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549}
1550
1551/**
1552 * skb_postpull_rcsum - update checksum for received skb after pull
1553 * @skb: buffer to update
1554 * @start: start of data before pull
1555 * @len: length of data pulled
1556 *
1557 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07001558 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1559 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001560 */
1561
1562static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08001563 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001564{
Patrick McHardy84fa7932006-08-29 16:44:56 -07001565 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001566 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1567}
1568
Herbert Xucbb042f2006-03-20 22:43:56 -08001569unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1570
Linus Torvalds1da177e2005-04-16 15:20:36 -07001571/**
1572 * pskb_trim_rcsum - trim received skb and update checksum
1573 * @skb: buffer to trim
1574 * @len: new length
1575 *
1576 * This is exactly the same as pskb_trim except that it ensures the
1577 * checksum of received packets are still valid after the operation.
1578 */
1579
1580static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1581{
Stephen Hemminger0e4e4222005-09-08 12:32:03 -07001582 if (likely(len >= skb->len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583 return 0;
Patrick McHardy84fa7932006-08-29 16:44:56 -07001584 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001585 skb->ip_summed = CHECKSUM_NONE;
1586 return __pskb_trim(skb, len);
1587}
1588
Linus Torvalds1da177e2005-04-16 15:20:36 -07001589#define skb_queue_walk(queue, skb) \
1590 for (skb = (queue)->next; \
1591 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1592 skb = skb->next)
1593
James Chapman46f89142007-04-30 00:07:31 -07001594#define skb_queue_walk_safe(queue, skb, tmp) \
1595 for (skb = (queue)->next, tmp = skb->next; \
1596 skb != (struct sk_buff *)(queue); \
1597 skb = tmp, tmp = skb->next)
1598
David S. Miller1164f522008-09-23 00:49:44 -07001599#define skb_queue_walk_from(queue, skb) \
1600 for (; prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1601 skb = skb->next)
1602
1603#define skb_queue_walk_from_safe(queue, skb, tmp) \
1604 for (tmp = skb->next; \
1605 skb != (struct sk_buff *)(queue); \
1606 skb = tmp, tmp = skb->next)
1607
Stephen Hemminger300ce172005-10-30 13:47:34 -08001608#define skb_queue_reverse_walk(queue, skb) \
1609 for (skb = (queue)->prev; \
1610 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1611 skb = skb->prev)
1612
Linus Torvalds1da177e2005-04-16 15:20:36 -07001613
Herbert Xua59322b2007-12-05 01:53:40 -08001614extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
1615 int *peeked, int *err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001616extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1617 int noblock, int *err);
1618extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1619 struct poll_table_struct *wait);
1620extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1621 int offset, struct iovec *to,
1622 int size);
Herbert Xufb286bb2005-11-10 13:01:24 -08001623extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001624 int hlen,
1625 struct iovec *iov);
Rusty Russelldb543c12008-08-15 15:13:53 -07001626extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
1627 int offset,
1628 struct iovec *from,
1629 int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
Herbert Xu27ab2562007-12-05 01:51:58 -08001631extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
Herbert Xu3305b802005-12-13 23:16:37 -08001632 unsigned int flags);
Al Viro2bbbc862006-11-14 21:37:14 -08001633extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
1634 int len, __wsum csum);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001635extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1636 void *to, int len);
Stephen Hemminger0c6fcc82007-04-20 16:40:01 -07001637extern int skb_store_bits(struct sk_buff *skb, int offset,
1638 const void *from, int len);
Al Viro81d77662006-11-14 21:37:33 -08001639extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001640 int offset, u8 *to, int len,
Al Viro81d77662006-11-14 21:37:33 -08001641 __wsum csum);
Jens Axboe9c55e012007-11-06 23:30:13 -08001642extern int skb_splice_bits(struct sk_buff *skb,
1643 unsigned int offset,
1644 struct pipe_inode_info *pipe,
1645 unsigned int len,
1646 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1648extern void skb_split(struct sk_buff *skb,
1649 struct sk_buff *skb1, const u32 len);
1650
Herbert Xu576a30e2006-06-27 13:22:38 -07001651extern struct sk_buff *skb_segment(struct sk_buff *skb, int features);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001652
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1654 int len, void *buffer)
1655{
1656 int hlen = skb_headlen(skb);
1657
Patrick McHardy55820ee2005-07-05 14:08:10 -07001658 if (hlen - offset >= len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001659 return skb->data + offset;
1660
1661 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1662 return NULL;
1663
1664 return buffer;
1665}
1666
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03001667static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
1668 void *to,
1669 const unsigned int len)
1670{
1671 memcpy(to, skb->data, len);
1672}
1673
1674static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
1675 const int offset, void *to,
1676 const unsigned int len)
1677{
1678 memcpy(to, skb->data + offset, len);
1679}
1680
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03001681static inline void skb_copy_to_linear_data(struct sk_buff *skb,
1682 const void *from,
1683 const unsigned int len)
1684{
1685 memcpy(skb->data, from, len);
1686}
1687
1688static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
1689 const int offset,
1690 const void *from,
1691 const unsigned int len)
1692{
1693 memcpy(skb->data + offset, from, len);
1694}
1695
Linus Torvalds1da177e2005-04-16 15:20:36 -07001696extern void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001697
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001698/**
1699 * skb_get_timestamp - get timestamp from a skb
1700 * @skb: skb to get stamp from
1701 * @stamp: pointer to struct timeval to store stamp in
1702 *
1703 * Timestamps are stored in the skb as offsets to a base timestamp.
1704 * This function converts the offset back to a struct timeval and stores
1705 * it in stamp.
1706 */
Stephen Hemmingerf2c38392005-09-06 15:48:03 -07001707static inline void skb_get_timestamp(const struct sk_buff *skb, struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001708{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001709 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001710}
1711
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001712static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001713{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001714 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001715}
1716
Stephen Hemminger164891a2007-04-23 22:26:16 -07001717static inline ktime_t net_timedelta(ktime_t t)
1718{
1719 return ktime_sub(ktime_get_real(), t);
1720}
1721
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07001722static inline ktime_t net_invalid_timestamp(void)
1723{
1724 return ktime_set(0, 0);
1725}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001726
Herbert Xu759e5d02007-03-25 20:10:56 -07001727extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
Al Virob51655b2006-11-14 21:40:42 -08001728extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08001729
Herbert Xu60476372007-04-09 11:59:39 -07001730static inline int skb_csum_unnecessary(const struct sk_buff *skb)
1731{
1732 return skb->ip_summed & CHECKSUM_UNNECESSARY;
1733}
1734
Herbert Xufb286bb2005-11-10 13:01:24 -08001735/**
1736 * skb_checksum_complete - Calculate checksum of an entire packet
1737 * @skb: packet to process
1738 *
1739 * This function calculates the checksum over the entire packet plus
1740 * the value of skb->csum. The latter can be used to supply the
1741 * checksum of a pseudo header as used by TCP/UDP. It returns the
1742 * checksum.
1743 *
1744 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1745 * this function can be used to verify that checksum on received
1746 * packets. In that case the function should return zero if the
1747 * checksum is correct. In particular, this function will return zero
1748 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1749 * hardware has already verified the correctness of the checksum.
1750 */
Al Viro4381ca32007-07-15 21:00:11 +01001751static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08001752{
Herbert Xu60476372007-04-09 11:59:39 -07001753 return skb_csum_unnecessary(skb) ?
1754 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08001755}
1756
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07001757#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07001758extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001759static inline void nf_conntrack_put(struct nf_conntrack *nfct)
1760{
1761 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07001762 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001763}
1764static inline void nf_conntrack_get(struct nf_conntrack *nfct)
1765{
1766 if (nfct)
1767 atomic_inc(&nfct->use);
1768}
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -08001769static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
1770{
1771 if (skb)
1772 atomic_inc(&skb->users);
1773}
1774static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
1775{
1776 if (skb)
1777 kfree_skb(skb);
1778}
1779#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001780#ifdef CONFIG_BRIDGE_NETFILTER
1781static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
1782{
1783 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
1784 kfree(nf_bridge);
1785}
1786static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
1787{
1788 if (nf_bridge)
1789 atomic_inc(&nf_bridge->use);
1790}
1791#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08001792static inline void nf_reset(struct sk_buff *skb)
1793{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07001794#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08001795 nf_conntrack_put(skb->nfct);
1796 skb->nfct = NULL;
Patrick McHardya193a4a2006-03-20 19:23:05 -08001797 nf_conntrack_put_reasm(skb->nfct_reasm);
1798 skb->nfct_reasm = NULL;
1799#endif
1800#ifdef CONFIG_BRIDGE_NETFILTER
1801 nf_bridge_put(skb->nf_bridge);
1802 skb->nf_bridge = NULL;
1803#endif
1804}
1805
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07001806/* Note: This doesn't put any conntrack and bridge info in dst. */
1807static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
1808{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07001809#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07001810 dst->nfct = src->nfct;
1811 nf_conntrack_get(src->nfct);
1812 dst->nfctinfo = src->nfctinfo;
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07001813 dst->nfct_reasm = src->nfct_reasm;
1814 nf_conntrack_get_reasm(src->nfct_reasm);
1815#endif
1816#ifdef CONFIG_BRIDGE_NETFILTER
1817 dst->nf_bridge = src->nf_bridge;
1818 nf_bridge_get(src->nf_bridge);
1819#endif
1820}
1821
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07001822static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
1823{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07001824#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07001825 nf_conntrack_put(dst->nfct);
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07001826 nf_conntrack_put_reasm(dst->nfct_reasm);
1827#endif
1828#ifdef CONFIG_BRIDGE_NETFILTER
1829 nf_bridge_put(dst->nf_bridge);
1830#endif
1831 __nf_copy(dst, src);
1832}
1833
James Morris984bc162006-06-09 00:29:17 -07001834#ifdef CONFIG_NETWORK_SECMARK
1835static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1836{
1837 to->secmark = from->secmark;
1838}
1839
1840static inline void skb_init_secmark(struct sk_buff *skb)
1841{
1842 skb->secmark = 0;
1843}
1844#else
1845static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1846{ }
1847
1848static inline void skb_init_secmark(struct sk_buff *skb)
1849{ }
1850#endif
1851
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001852static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
1853{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001854 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001855}
1856
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07001857static inline u16 skb_get_queue_mapping(struct sk_buff *skb)
1858{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07001859 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07001860}
1861
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001862static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
1863{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001864 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001865}
1866
Herbert Xu89114af2006-07-08 13:34:32 -07001867static inline int skb_is_gso(const struct sk_buff *skb)
1868{
1869 return skb_shinfo(skb)->gso_size;
1870}
1871
Brice Goglineabd7e32007-10-13 12:33:32 +02001872static inline int skb_is_gso_v6(const struct sk_buff *skb)
1873{
1874 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
1875}
1876
Ben Hutchings4497b072008-06-19 16:22:28 -07001877extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
1878
1879static inline bool skb_warn_if_lro(const struct sk_buff *skb)
1880{
1881 /* LRO sets gso_size but not gso_type, whereas if GSO is really
1882 * wanted then gso_type will be set. */
1883 struct skb_shared_info *shinfo = skb_shinfo(skb);
1884 if (shinfo->gso_size != 0 && unlikely(shinfo->gso_type == 0)) {
1885 __skb_warn_lro_forwarding(skb);
1886 return true;
1887 }
1888 return false;
1889}
1890
Herbert Xu35fc92a2007-03-26 23:22:20 -07001891static inline void skb_forward_csum(struct sk_buff *skb)
1892{
1893 /* Unfortunately we don't support this one. Any brave souls? */
1894 if (skb->ip_summed == CHECKSUM_COMPLETE)
1895 skb->ip_summed = CHECKSUM_NONE;
1896}
1897
Rusty Russellf35d9d82008-02-04 23:49:54 -05001898bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001899#endif /* __KERNEL__ */
1900#endif /* _LINUX_SKBUFF_H */