blob: 487e34507b4198ea1dbccab63b0580b1360c2151 [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 };
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -0700272#ifdef CONFIG_XFRM
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273 struct sec_path *sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -0700274#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275 /*
276 * This is the control buffer. It is free to use for every
277 * layer. Please put your private variables there. If you
278 * want to keep them across layers you have to do a skb_clone()
279 * first. This is owned by whoever has the skb queued ATM.
280 */
Patrick McHardy3e3850e2006-01-06 23:04:54 -0800281 char cb[48];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282
283 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700284 data_len;
285 __u16 mac_len,
286 hdr_len;
Al Viroff1dcad2006-11-20 18:07:29 -0800287 union {
288 __wsum csum;
Herbert Xu663ead32007-04-09 11:59:07 -0700289 struct {
290 __u16 csum_start;
291 __u16 csum_offset;
292 };
Al Viroff1dcad2006-11-20 18:07:29 -0800293 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294 __u32 priority;
Thomas Graf1cbb3382005-07-05 14:13:41 -0700295 __u8 local_df:1,
296 cloned:1,
297 ip_summed:2,
Harald Welte6869c4d2005-08-09 19:24:19 -0700298 nohdr:1,
299 nfctinfo:3;
David S. Millerd179cd12005-08-17 14:57:30 -0700300 __u8 pkt_type:3,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800301 fclone:2,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700302 ipvs_property:1,
Herbert Xua59322b2007-12-05 01:53:40 -0800303 peeked:1,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700304 nf_trace:1;
Alexey Dobriyana0d3bea2005-08-11 16:05:50 -0700305 __be16 protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700306
307 void (*destructor)(struct sk_buff *skb);
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800308#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700309 struct nf_conntrack *nfct;
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800310 struct sk_buff *nfct_reasm;
311#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312#ifdef CONFIG_BRIDGE_NETFILTER
313 struct nf_bridge_info *nf_bridge;
314#endif
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700315
316 int iif;
317 __u16 queue_mapping;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700318#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700319 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700321 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323#endif
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400324#ifdef CONFIG_IPV6_NDISC_NODETYPE
Templin, Fred Lfadf6bf2008-03-11 18:35:59 -0400325 __u8 ndisc_nodetype:2;
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400326#endif
Johannes Bergd0f09802008-07-29 11:32:07 +0200327#if defined(CONFIG_MAC80211) || defined(CONFIG_MAC80211_MODULE)
328 __u8 do_not_encrypt:1;
329#endif
330 /* 0/13/14 bit hole */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700331
Chris Leech97fc2f02006-05-23 17:55:33 -0700332#ifdef CONFIG_NET_DMA
333 dma_cookie_t dma_cookie;
334#endif
James Morris984bc162006-06-09 00:29:17 -0700335#ifdef CONFIG_NETWORK_SECMARK
336 __u32 secmark;
337#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338
Thomas Graf82e91ff2006-11-09 15:19:14 -0800339 __u32 mark;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340
Patrick McHardy6aa895b02008-07-14 22:49:06 -0700341 __u16 vlan_tci;
342
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700343 sk_buff_data_t transport_header;
344 sk_buff_data_t network_header;
345 sk_buff_data_t mac_header;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700347 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700348 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700350 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700351 unsigned int truesize;
352 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353};
354
355#ifdef __KERNEL__
356/*
357 * Handling routines are only of interest to the kernel
358 */
359#include <linux/slab.h>
360
361#include <asm/system.h>
362
David S. Millera40c24a2008-09-11 04:51:14 -0700363#ifdef CONFIG_HAS_DMA
364#include <linux/dma-mapping.h>
365extern int skb_dma_map(struct device *dev, struct sk_buff *skb,
366 enum dma_data_direction dir);
367extern void skb_dma_unmap(struct device *dev, struct sk_buff *skb,
368 enum dma_data_direction dir);
369#endif
370
Jörn Engel231d06a2006-03-20 21:28:35 -0800371extern void kfree_skb(struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372extern void __kfree_skb(struct sk_buff *skb);
David S. Millerd179cd12005-08-17 14:57:30 -0700373extern struct sk_buff *__alloc_skb(unsigned int size,
Christoph Hellwigb30973f2006-12-06 20:32:36 -0800374 gfp_t priority, int fclone, int node);
David S. Millerd179cd12005-08-17 14:57:30 -0700375static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100376 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700377{
Christoph Hellwigb30973f2006-12-06 20:32:36 -0800378 return __alloc_skb(size, priority, 0, -1);
David S. Millerd179cd12005-08-17 14:57:30 -0700379}
380
381static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100382 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700383{
Christoph Hellwigb30973f2006-12-06 20:32:36 -0800384 return __alloc_skb(size, priority, 1, -1);
David S. Millerd179cd12005-08-17 14:57:30 -0700385}
386
Lennert Buytenhek04a4bb52008-10-01 02:33:12 -0700387extern int skb_recycle_check(struct sk_buff *skb, int skb_size);
388
Herbert Xue0053ec2007-10-14 00:37:52 -0700389extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700390extern struct sk_buff *skb_clone(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100391 gfp_t priority);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700392extern struct sk_buff *skb_copy(const struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100393 gfp_t priority);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700394extern struct sk_buff *pskb_copy(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100395 gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396extern int pskb_expand_head(struct sk_buff *skb,
Victor Fusco86a76ca2005-07-08 14:57:47 -0700397 int nhead, int ntail,
Al Virodd0fc662005-10-07 07:46:04 +0100398 gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
400 unsigned int headroom);
401extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
402 int newheadroom, int newtailroom,
Al Virodd0fc662005-10-07 07:46:04 +0100403 gfp_t priority);
David Howells716ea3a2007-04-02 20:19:53 -0700404extern int skb_to_sgvec(struct sk_buff *skb,
405 struct scatterlist *sg, int offset,
406 int len);
407extern int skb_cow_data(struct sk_buff *skb, int tailbits,
408 struct sk_buff **trailer);
Herbert Xu5b057c62006-06-23 02:06:41 -0700409extern int skb_pad(struct sk_buff *skb, int pad);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410#define dev_kfree_skb(a) kfree_skb(a)
411extern void skb_over_panic(struct sk_buff *skb, int len,
412 void *here);
413extern void skb_under_panic(struct sk_buff *skb, int len,
414 void *here);
David S. Millerdc6de332006-04-20 00:10:50 -0700415extern void skb_truesize_bug(struct sk_buff *skb);
416
417static inline void skb_truesize_check(struct sk_buff *skb)
418{
Chuck Lever78608ba2007-11-10 21:53:30 -0800419 int len = sizeof(struct sk_buff) + skb->len;
420
421 if (unlikely((int)skb->truesize < len))
David S. Millerdc6de332006-04-20 00:10:50 -0700422 skb_truesize_bug(skb);
423}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700425extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
426 int getfrag(void *from, char *to, int offset,
427 int len,int odd, struct sk_buff *skb),
428 void *from, int length);
429
Thomas Graf677e90e2005-06-23 20:59:51 -0700430struct skb_seq_state
431{
432 __u32 lower_offset;
433 __u32 upper_offset;
434 __u32 frag_idx;
435 __u32 stepped_offset;
436 struct sk_buff *root_skb;
437 struct sk_buff *cur_skb;
438 __u8 *frag_data;
439};
440
441extern void skb_prepare_seq_read(struct sk_buff *skb,
442 unsigned int from, unsigned int to,
443 struct skb_seq_state *st);
444extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
445 struct skb_seq_state *st);
446extern void skb_abort_seq_read(struct skb_seq_state *st);
447
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700448extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
449 unsigned int to, struct ts_config *config,
450 struct ts_state *state);
451
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700452#ifdef NET_SKBUFF_DATA_USES_OFFSET
453static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
454{
455 return skb->head + skb->end;
456}
457#else
458static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
459{
460 return skb->end;
461}
462#endif
463
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700465#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466
467/**
468 * skb_queue_empty - check if a queue is empty
469 * @list: queue head
470 *
471 * Returns true if the queue is empty, false otherwise.
472 */
473static inline int skb_queue_empty(const struct sk_buff_head *list)
474{
475 return list->next == (struct sk_buff *)list;
476}
477
478/**
David S. Millerfc7ebb22008-09-23 00:34:07 -0700479 * skb_queue_is_last - check if skb is the last entry in the queue
480 * @list: queue head
481 * @skb: buffer
482 *
483 * Returns true if @skb is the last buffer on the list.
484 */
485static inline bool skb_queue_is_last(const struct sk_buff_head *list,
486 const struct sk_buff *skb)
487{
488 return (skb->next == (struct sk_buff *) list);
489}
490
491/**
David S. Miller249c8b42008-09-23 00:44:42 -0700492 * skb_queue_next - return the next packet in the queue
493 * @list: queue head
494 * @skb: current buffer
495 *
496 * Return the next packet in @list after @skb. It is only valid to
497 * call this if skb_queue_is_last() evaluates to false.
498 */
499static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
500 const struct sk_buff *skb)
501{
502 /* This BUG_ON may seem severe, but if we just return then we
503 * are going to dereference garbage.
504 */
505 BUG_ON(skb_queue_is_last(list, skb));
506 return skb->next;
507}
508
509/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700510 * skb_get - reference buffer
511 * @skb: buffer to reference
512 *
513 * Makes another reference to a socket buffer and returns a pointer
514 * to the buffer.
515 */
516static inline struct sk_buff *skb_get(struct sk_buff *skb)
517{
518 atomic_inc(&skb->users);
519 return skb;
520}
521
522/*
523 * If users == 1, we are the only owner and are can avoid redundant
524 * atomic change.
525 */
526
527/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528 * skb_cloned - is the buffer a clone
529 * @skb: buffer to check
530 *
531 * Returns true if the buffer was generated with skb_clone() and is
532 * one of multiple shared copies of the buffer. Cloned buffers are
533 * shared data so must not be written to under normal circumstances.
534 */
535static inline int skb_cloned(const struct sk_buff *skb)
536{
537 return skb->cloned &&
538 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
539}
540
541/**
542 * skb_header_cloned - is the header a clone
543 * @skb: buffer to check
544 *
545 * Returns true if modifying the header part of the buffer requires
546 * the data to be copied.
547 */
548static inline int skb_header_cloned(const struct sk_buff *skb)
549{
550 int dataref;
551
552 if (!skb->cloned)
553 return 0;
554
555 dataref = atomic_read(&skb_shinfo(skb)->dataref);
556 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
557 return dataref != 1;
558}
559
560/**
561 * skb_header_release - release reference to header
562 * @skb: buffer to operate on
563 *
564 * Drop a reference to the header part of the buffer. This is done
565 * by acquiring a payload reference. You must not read from the header
566 * part of skb->data after this.
567 */
568static inline void skb_header_release(struct sk_buff *skb)
569{
570 BUG_ON(skb->nohdr);
571 skb->nohdr = 1;
572 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
573}
574
575/**
576 * skb_shared - is the buffer shared
577 * @skb: buffer to check
578 *
579 * Returns true if more than one person has a reference to this
580 * buffer.
581 */
582static inline int skb_shared(const struct sk_buff *skb)
583{
584 return atomic_read(&skb->users) != 1;
585}
586
587/**
588 * skb_share_check - check if buffer is shared and if so clone it
589 * @skb: buffer to check
590 * @pri: priority for memory allocation
591 *
592 * If the buffer is shared the buffer is cloned and the old copy
593 * drops a reference. A new clone with a single reference is returned.
594 * If the buffer is not shared the original buffer is returned. When
595 * being called from interrupt status or with spinlocks held pri must
596 * be GFP_ATOMIC.
597 *
598 * NULL is returned on a memory allocation failure.
599 */
Victor Fusco86a76ca2005-07-08 14:57:47 -0700600static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100601 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602{
603 might_sleep_if(pri & __GFP_WAIT);
604 if (skb_shared(skb)) {
605 struct sk_buff *nskb = skb_clone(skb, pri);
606 kfree_skb(skb);
607 skb = nskb;
608 }
609 return skb;
610}
611
612/*
613 * Copy shared buffers into a new sk_buff. We effectively do COW on
614 * packets to handle cases where we have a local reader and forward
615 * and a couple of other messy ones. The normal one is tcpdumping
616 * a packet thats being forwarded.
617 */
618
619/**
620 * skb_unshare - make a copy of a shared buffer
621 * @skb: buffer to check
622 * @pri: priority for memory allocation
623 *
624 * If the socket buffer is a clone then this function creates a new
625 * copy of the data, drops a reference count on the old copy and returns
626 * the new copy with the reference count at 1. If the buffer is not a clone
627 * the original buffer is returned. When called with a spinlock held or
628 * from interrupt state @pri must be %GFP_ATOMIC
629 *
630 * %NULL is returned on a memory allocation failure.
631 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -0700632static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100633 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634{
635 might_sleep_if(pri & __GFP_WAIT);
636 if (skb_cloned(skb)) {
637 struct sk_buff *nskb = skb_copy(skb, pri);
638 kfree_skb(skb); /* Free our shared copy */
639 skb = nskb;
640 }
641 return skb;
642}
643
644/**
645 * skb_peek
646 * @list_: list to peek at
647 *
648 * Peek an &sk_buff. Unlike most other operations you _MUST_
649 * be careful with this one. A peek leaves the buffer on the
650 * list and someone else may run off with it. You must hold
651 * the appropriate locks or have a private queue to do this.
652 *
653 * Returns %NULL for an empty list or a pointer to the head element.
654 * The reference count is not incremented and the reference is therefore
655 * volatile. Use with caution.
656 */
657static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
658{
659 struct sk_buff *list = ((struct sk_buff *)list_)->next;
660 if (list == (struct sk_buff *)list_)
661 list = NULL;
662 return list;
663}
664
665/**
666 * skb_peek_tail
667 * @list_: list to peek at
668 *
669 * Peek an &sk_buff. Unlike most other operations you _MUST_
670 * be careful with this one. A peek leaves the buffer on the
671 * list and someone else may run off with it. You must hold
672 * the appropriate locks or have a private queue to do this.
673 *
674 * Returns %NULL for an empty list or a pointer to the tail element.
675 * The reference count is not incremented and the reference is therefore
676 * volatile. Use with caution.
677 */
678static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
679{
680 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
681 if (list == (struct sk_buff *)list_)
682 list = NULL;
683 return list;
684}
685
686/**
687 * skb_queue_len - get queue length
688 * @list_: list to measure
689 *
690 * Return the length of an &sk_buff queue.
691 */
692static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
693{
694 return list_->qlen;
695}
696
David S. Miller67fed452008-09-21 22:36:24 -0700697/**
698 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
699 * @list: queue to initialize
700 *
701 * This initializes only the list and queue length aspects of
702 * an sk_buff_head object. This allows to initialize the list
703 * aspects of an sk_buff_head without reinitializing things like
704 * the spinlock. It can also be used for on-stack sk_buff_head
705 * objects where the spinlock is known to not be used.
706 */
707static inline void __skb_queue_head_init(struct sk_buff_head *list)
708{
709 list->prev = list->next = (struct sk_buff *)list;
710 list->qlen = 0;
711}
712
Arjan van de Ven76f10ad2006-08-02 14:06:55 -0700713/*
714 * This function creates a split out lock class for each invocation;
715 * this is needed for now since a whole lot of users of the skb-queue
716 * infrastructure in drivers have different locking usage (in hardirq)
717 * than the networking core (in softirq only). In the long run either the
718 * network layer or drivers should need annotation to consolidate the
719 * main types of usage into 3 classes.
720 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700721static inline void skb_queue_head_init(struct sk_buff_head *list)
722{
723 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -0700724 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700725}
726
Pavel Emelianovc2ecba72007-04-17 12:45:31 -0700727static inline void skb_queue_head_init_class(struct sk_buff_head *list,
728 struct lock_class_key *class)
729{
730 skb_queue_head_init(list);
731 lockdep_set_class(&list->lock, class);
732}
733
Linus Torvalds1da177e2005-04-16 15:20:36 -0700734/*
Gerrit Renkerbf299272008-04-14 00:04:51 -0700735 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736 *
737 * The "__skb_xxxx()" functions are the non-atomic ones that
738 * can only be called with interrupts disabled.
739 */
Gerrit Renkerbf299272008-04-14 00:04:51 -0700740extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
741static inline void __skb_insert(struct sk_buff *newsk,
742 struct sk_buff *prev, struct sk_buff *next,
743 struct sk_buff_head *list)
744{
745 newsk->next = next;
746 newsk->prev = prev;
747 next->prev = prev->next = newsk;
748 list->qlen++;
749}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750
David S. Miller67fed452008-09-21 22:36:24 -0700751static inline void __skb_queue_splice(const struct sk_buff_head *list,
752 struct sk_buff *prev,
753 struct sk_buff *next)
754{
755 struct sk_buff *first = list->next;
756 struct sk_buff *last = list->prev;
757
758 first->prev = prev;
759 prev->next = first;
760
761 last->next = next;
762 next->prev = last;
763}
764
765/**
766 * skb_queue_splice - join two skb lists, this is designed for stacks
767 * @list: the new list to add
768 * @head: the place to add it in the first list
769 */
770static inline void skb_queue_splice(const struct sk_buff_head *list,
771 struct sk_buff_head *head)
772{
773 if (!skb_queue_empty(list)) {
774 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -0700775 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -0700776 }
777}
778
779/**
780 * skb_queue_splice - join two skb lists and reinitialise the emptied list
781 * @list: the new list to add
782 * @head: the place to add it in the first list
783 *
784 * The list at @list is reinitialised
785 */
786static inline void skb_queue_splice_init(struct sk_buff_head *list,
787 struct sk_buff_head *head)
788{
789 if (!skb_queue_empty(list)) {
790 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -0700791 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -0700792 __skb_queue_head_init(list);
793 }
794}
795
796/**
797 * skb_queue_splice_tail - join two skb lists, each list being a queue
798 * @list: the new list to add
799 * @head: the place to add it in the first list
800 */
801static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
802 struct sk_buff_head *head)
803{
804 if (!skb_queue_empty(list)) {
805 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -0700806 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -0700807 }
808}
809
810/**
811 * skb_queue_splice_tail - join two skb lists and reinitialise the emptied list
812 * @list: the new list to add
813 * @head: the place to add it in the first list
814 *
815 * Each of the lists is a queue.
816 * The list at @list is reinitialised
817 */
818static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
819 struct sk_buff_head *head)
820{
821 if (!skb_queue_empty(list)) {
822 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -0700823 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -0700824 __skb_queue_head_init(list);
825 }
826}
827
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828/**
Stephen Hemminger300ce172005-10-30 13:47:34 -0800829 * __skb_queue_after - queue a buffer at the list head
830 * @list: list to use
831 * @prev: place after this buffer
832 * @newsk: buffer to queue
833 *
834 * Queue a buffer int the middle of a list. This function takes no locks
835 * and you must therefore hold required locks before calling it.
836 *
837 * A buffer cannot be placed on two lists at the same time.
838 */
839static inline void __skb_queue_after(struct sk_buff_head *list,
840 struct sk_buff *prev,
841 struct sk_buff *newsk)
842{
Gerrit Renkerbf299272008-04-14 00:04:51 -0700843 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -0800844}
845
Gerrit Renker7de6c032008-04-14 00:05:09 -0700846extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
847 struct sk_buff_head *list);
848
Gerrit Renkerf5572852008-04-14 00:05:28 -0700849static inline void __skb_queue_before(struct sk_buff_head *list,
850 struct sk_buff *next,
851 struct sk_buff *newsk)
852{
853 __skb_insert(newsk, next->prev, next, list);
854}
855
Stephen Hemminger300ce172005-10-30 13:47:34 -0800856/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857 * __skb_queue_head - queue a buffer at the list head
858 * @list: list to use
859 * @newsk: buffer to queue
860 *
861 * Queue a buffer at the start of a list. This function takes no locks
862 * and you must therefore hold required locks before calling it.
863 *
864 * A buffer cannot be placed on two lists at the same time.
865 */
866extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
867static inline void __skb_queue_head(struct sk_buff_head *list,
868 struct sk_buff *newsk)
869{
Stephen Hemminger300ce172005-10-30 13:47:34 -0800870 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871}
872
873/**
874 * __skb_queue_tail - queue a buffer at the list tail
875 * @list: list to use
876 * @newsk: buffer to queue
877 *
878 * Queue a buffer at the end of a list. This function takes no locks
879 * and you must therefore hold required locks before calling it.
880 *
881 * A buffer cannot be placed on two lists at the same time.
882 */
883extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
884static inline void __skb_queue_tail(struct sk_buff_head *list,
885 struct sk_buff *newsk)
886{
Gerrit Renkerf5572852008-04-14 00:05:28 -0700887 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888}
889
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890/*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891 * remove sk_buff from list. _Must_ be called atomically, and with
892 * the list known..
893 */
David S. Miller8728b832005-08-09 19:25:21 -0700894extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
896{
897 struct sk_buff *next, *prev;
898
899 list->qlen--;
900 next = skb->next;
901 prev = skb->prev;
902 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700903 next->prev = prev;
904 prev->next = next;
905}
906
Gerrit Renkerf525c062008-04-14 00:04:12 -0700907/**
908 * __skb_dequeue - remove from the head of the queue
909 * @list: list to dequeue from
910 *
911 * Remove the head of the list. This function does not take any locks
912 * so must be used with appropriate locks held only. The head item is
913 * returned or %NULL if the list is empty.
914 */
915extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
916static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
917{
918 struct sk_buff *skb = skb_peek(list);
919 if (skb)
920 __skb_unlink(skb, list);
921 return skb;
922}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700923
924/**
925 * __skb_dequeue_tail - remove from the tail of the queue
926 * @list: list to dequeue from
927 *
928 * Remove the tail of the list. This function does not take any locks
929 * so must be used with appropriate locks held only. The tail item is
930 * returned or %NULL if the list is empty.
931 */
932extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
933static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
934{
935 struct sk_buff *skb = skb_peek_tail(list);
936 if (skb)
937 __skb_unlink(skb, list);
938 return skb;
939}
940
941
942static inline int skb_is_nonlinear(const struct sk_buff *skb)
943{
944 return skb->data_len;
945}
946
947static inline unsigned int skb_headlen(const struct sk_buff *skb)
948{
949 return skb->len - skb->data_len;
950}
951
952static inline int skb_pagelen(const struct sk_buff *skb)
953{
954 int i, len = 0;
955
956 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
957 len += skb_shinfo(skb)->frags[i].size;
958 return len + skb_headlen(skb);
959}
960
961static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
962 struct page *page, int off, int size)
963{
964 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
965
966 frag->page = page;
967 frag->page_offset = off;
968 frag->size = size;
969 skb_shinfo(skb)->nr_frags = i + 1;
970}
971
Peter Zijlstra654bed12008-10-07 14:22:33 -0700972extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
973 int off, int size);
974
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
976#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_shinfo(skb)->frag_list)
977#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
978
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700979#ifdef NET_SKBUFF_DATA_USES_OFFSET
980static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
981{
982 return skb->head + skb->tail;
983}
984
985static inline void skb_reset_tail_pointer(struct sk_buff *skb)
986{
987 skb->tail = skb->data - skb->head;
988}
989
990static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
991{
992 skb_reset_tail_pointer(skb);
993 skb->tail += offset;
994}
995#else /* NET_SKBUFF_DATA_USES_OFFSET */
996static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
997{
998 return skb->tail;
999}
1000
1001static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1002{
1003 skb->tail = skb->data;
1004}
1005
1006static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1007{
1008 skb->tail = skb->data + offset;
1009}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001010
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001011#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1012
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013/*
1014 * Add data to an sk_buff
1015 */
Ilpo Järvinen0dde3e12008-03-27 17:43:41 -07001016extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001017static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1018{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001019 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020 SKB_LINEAR_ASSERT(skb);
1021 skb->tail += len;
1022 skb->len += len;
1023 return tmp;
1024}
1025
Ilpo Järvinenc2aa2702008-03-27 17:52:40 -07001026extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1028{
1029 skb->data -= len;
1030 skb->len += len;
1031 return skb->data;
1032}
1033
Ilpo Järvinen6be8ac22008-03-27 17:47:24 -07001034extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1036{
1037 skb->len -= len;
1038 BUG_ON(skb->len < skb->data_len);
1039 return skb->data += len;
1040}
1041
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1043
1044static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1045{
1046 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001047 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001048 return NULL;
1049 skb->len -= len;
1050 return skb->data += len;
1051}
1052
1053static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1054{
1055 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1056}
1057
1058static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1059{
1060 if (likely(len <= skb_headlen(skb)))
1061 return 1;
1062 if (unlikely(len > skb->len))
1063 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001064 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001065}
1066
1067/**
1068 * skb_headroom - bytes at buffer head
1069 * @skb: buffer to check
1070 *
1071 * Return the number of bytes of free space at the head of an &sk_buff.
1072 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001073static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074{
1075 return skb->data - skb->head;
1076}
1077
1078/**
1079 * skb_tailroom - bytes at buffer end
1080 * @skb: buffer to check
1081 *
1082 * Return the number of bytes of free space at the tail of an sk_buff
1083 */
1084static inline int skb_tailroom(const struct sk_buff *skb)
1085{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001086 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087}
1088
1089/**
1090 * skb_reserve - adjust headroom
1091 * @skb: buffer to alter
1092 * @len: bytes to move
1093 *
1094 * Increase the headroom of an empty &sk_buff by reducing the tail
1095 * room. This is only allowed for an empty buffer.
1096 */
David S. Miller8243126c2006-01-17 02:54:21 -08001097static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098{
1099 skb->data += len;
1100 skb->tail += len;
1101}
1102
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001103#ifdef NET_SKBUFF_DATA_USES_OFFSET
1104static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1105{
1106 return skb->head + skb->transport_header;
1107}
1108
1109static inline void skb_reset_transport_header(struct sk_buff *skb)
1110{
1111 skb->transport_header = skb->data - skb->head;
1112}
1113
1114static inline void skb_set_transport_header(struct sk_buff *skb,
1115 const int offset)
1116{
1117 skb_reset_transport_header(skb);
1118 skb->transport_header += offset;
1119}
1120
1121static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1122{
1123 return skb->head + skb->network_header;
1124}
1125
1126static inline void skb_reset_network_header(struct sk_buff *skb)
1127{
1128 skb->network_header = skb->data - skb->head;
1129}
1130
1131static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1132{
1133 skb_reset_network_header(skb);
1134 skb->network_header += offset;
1135}
1136
1137static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1138{
1139 return skb->head + skb->mac_header;
1140}
1141
1142static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1143{
1144 return skb->mac_header != ~0U;
1145}
1146
1147static inline void skb_reset_mac_header(struct sk_buff *skb)
1148{
1149 skb->mac_header = skb->data - skb->head;
1150}
1151
1152static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1153{
1154 skb_reset_mac_header(skb);
1155 skb->mac_header += offset;
1156}
1157
1158#else /* NET_SKBUFF_DATA_USES_OFFSET */
1159
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001160static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1161{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001162 return skb->transport_header;
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001163}
1164
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001165static inline void skb_reset_transport_header(struct sk_buff *skb)
1166{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001167 skb->transport_header = skb->data;
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001168}
1169
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001170static inline void skb_set_transport_header(struct sk_buff *skb,
1171 const int offset)
1172{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001173 skb->transport_header = skb->data + offset;
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001174}
1175
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001176static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1177{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001178 return skb->network_header;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001179}
1180
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001181static inline void skb_reset_network_header(struct sk_buff *skb)
1182{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001183 skb->network_header = skb->data;
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001184}
1185
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001186static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1187{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001188 skb->network_header = skb->data + offset;
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001189}
1190
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001191static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1192{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001193 return skb->mac_header;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001194}
1195
1196static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1197{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001198 return skb->mac_header != NULL;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001199}
1200
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001201static inline void skb_reset_mac_header(struct sk_buff *skb)
1202{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001203 skb->mac_header = skb->data;
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001204}
1205
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001206static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1207{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001208 skb->mac_header = skb->data + offset;
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001209}
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001210#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1211
1212static inline int skb_transport_offset(const struct sk_buff *skb)
1213{
1214 return skb_transport_header(skb) - skb->data;
1215}
1216
1217static inline u32 skb_network_header_len(const struct sk_buff *skb)
1218{
1219 return skb->transport_header - skb->network_header;
1220}
1221
1222static inline int skb_network_offset(const struct sk_buff *skb)
1223{
1224 return skb_network_header(skb) - skb->data;
1225}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001226
Linus Torvalds1da177e2005-04-16 15:20:36 -07001227/*
1228 * CPUs often take a performance hit when accessing unaligned memory
1229 * locations. The actual performance hit varies, it can be small if the
1230 * hardware handles it or large if we have to take an exception and fix it
1231 * in software.
1232 *
1233 * Since an ethernet header is 14 bytes network drivers often end up with
1234 * the IP header at an unaligned offset. The IP header can be aligned by
1235 * shifting the start of the packet by 2 bytes. Drivers should do this
1236 * with:
1237 *
1238 * skb_reserve(NET_IP_ALIGN);
1239 *
1240 * The downside to this alignment of the IP header is that the DMA is now
1241 * unaligned. On some architectures the cost of an unaligned DMA is high
1242 * and this cost outweighs the gains made by aligning the IP header.
1243 *
1244 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1245 * to be overridden.
1246 */
1247#ifndef NET_IP_ALIGN
1248#define NET_IP_ALIGN 2
1249#endif
1250
Anton Blanchard025be812006-03-31 02:27:06 -08001251/*
1252 * The networking layer reserves some headroom in skb data (via
1253 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1254 * the header has to grow. In the default case, if the header has to grow
1255 * 16 bytes or less we avoid the reallocation.
1256 *
1257 * Unfortunately this headroom changes the DMA alignment of the resulting
1258 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1259 * on some architectures. An architecture can override this value,
1260 * perhaps setting it to a cacheline in size (since that will maintain
1261 * cacheline alignment of the DMA). It must be a power of 2.
1262 *
1263 * Various parts of the networking layer expect at least 16 bytes of
1264 * headroom, you should not reduce this.
1265 */
1266#ifndef NET_SKB_PAD
1267#define NET_SKB_PAD 16
1268#endif
1269
Herbert Xu3cc0e872006-06-09 16:13:38 -07001270extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001271
1272static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1273{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001274 if (unlikely(skb->data_len)) {
1275 WARN_ON(1);
1276 return;
1277 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001278 skb->len = len;
1279 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001280}
1281
Ilpo Järvinen419ae742008-03-27 17:54:01 -07001282extern void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283
1284static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1285{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001286 if (skb->data_len)
1287 return ___pskb_trim(skb, len);
1288 __skb_trim(skb, len);
1289 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001290}
1291
1292static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1293{
1294 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1295}
1296
1297/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07001298 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1299 * @skb: buffer to alter
1300 * @len: new length
1301 *
1302 * This is identical to pskb_trim except that the caller knows that
1303 * the skb is not cloned so we should never get an error due to out-
1304 * of-memory.
1305 */
1306static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1307{
1308 int err = pskb_trim(skb, len);
1309 BUG_ON(err);
1310}
1311
1312/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001313 * skb_orphan - orphan a buffer
1314 * @skb: buffer to orphan
1315 *
1316 * If a buffer currently has an owner then we call the owner's
1317 * destructor function and make the @skb unowned. The buffer continues
1318 * to exist but is no longer charged to its former owner.
1319 */
1320static inline void skb_orphan(struct sk_buff *skb)
1321{
1322 if (skb->destructor)
1323 skb->destructor(skb);
1324 skb->destructor = NULL;
1325 skb->sk = NULL;
1326}
1327
1328/**
1329 * __skb_queue_purge - empty a list
1330 * @list: list to empty
1331 *
1332 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1333 * the list and one reference dropped. This function does not take the
1334 * list lock and the caller must hold the relevant locks to use it.
1335 */
1336extern void skb_queue_purge(struct sk_buff_head *list);
1337static inline void __skb_queue_purge(struct sk_buff_head *list)
1338{
1339 struct sk_buff *skb;
1340 while ((skb = __skb_dequeue(list)) != NULL)
1341 kfree_skb(skb);
1342}
1343
1344/**
Christoph Hellwigb4e54de2006-07-24 15:31:14 -07001345 * __dev_alloc_skb - allocate an skbuff for receiving
Linus Torvalds1da177e2005-04-16 15:20:36 -07001346 * @length: length to allocate
1347 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1348 *
1349 * Allocate a new &sk_buff and assign it a usage count of one. The
1350 * buffer has unspecified headroom built in. Users should allocate
1351 * the headroom they think they need without accounting for the
1352 * built in space. The built in space is used for optimisations.
1353 *
Christoph Hellwig766ea8c2006-08-07 15:49:53 -07001354 * %NULL is returned if there is no free memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001355 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
Al Virodd0fc662005-10-07 07:46:04 +01001357 gfp_t gfp_mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358{
Anton Blanchard025be812006-03-31 02:27:06 -08001359 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360 if (likely(skb))
Anton Blanchard025be812006-03-31 02:27:06 -08001361 skb_reserve(skb, NET_SKB_PAD);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362 return skb;
1363}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001364
Ilpo Järvinenf58518e2008-03-27 17:51:31 -07001365extern struct sk_buff *dev_alloc_skb(unsigned int length);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001366
Christoph Hellwig8af27452006-07-31 22:35:23 -07001367extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1368 unsigned int length, gfp_t gfp_mask);
1369
1370/**
1371 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1372 * @dev: network device to receive on
1373 * @length: length to allocate
1374 *
1375 * Allocate a new &sk_buff and assign it a usage count of one. The
1376 * buffer has unspecified headroom built in. Users should allocate
1377 * the headroom they think they need without accounting for the
1378 * built in space. The built in space is used for optimisations.
1379 *
1380 * %NULL is returned if there is no free memory. Although this function
1381 * allocates memory it can be called from an interrupt.
1382 */
1383static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1384 unsigned int length)
1385{
1386 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1387}
1388
Peter Zijlstra654bed12008-10-07 14:22:33 -07001389extern struct page *__netdev_alloc_page(struct net_device *dev, gfp_t gfp_mask);
1390
1391/**
1392 * netdev_alloc_page - allocate a page for ps-rx on a specific device
1393 * @dev: network device to receive on
1394 *
1395 * Allocate a new page node local to the specified device.
1396 *
1397 * %NULL is returned if there is no free memory.
1398 */
1399static inline struct page *netdev_alloc_page(struct net_device *dev)
1400{
1401 return __netdev_alloc_page(dev, GFP_ATOMIC);
1402}
1403
1404static inline void netdev_free_page(struct net_device *dev, struct page *page)
1405{
1406 __free_page(page);
1407}
1408
Linus Torvalds1da177e2005-04-16 15:20:36 -07001409/**
Patrick McHardy334a8132007-06-25 04:35:20 -07001410 * skb_clone_writable - is the header of a clone writable
1411 * @skb: buffer to check
1412 * @len: length up to which to write
1413 *
1414 * Returns true if modifying the header part of the cloned buffer
1415 * does not requires the data to be copied.
1416 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001417static inline int skb_clone_writable(struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07001418{
1419 return !skb_header_cloned(skb) &&
1420 skb_headroom(skb) + len <= skb->hdr_len;
1421}
1422
Herbert Xud9cc2042007-09-16 16:21:16 -07001423static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
1424 int cloned)
1425{
1426 int delta = 0;
1427
1428 if (headroom < NET_SKB_PAD)
1429 headroom = NET_SKB_PAD;
1430 if (headroom > skb_headroom(skb))
1431 delta = headroom - skb_headroom(skb);
1432
1433 if (delta || cloned)
1434 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
1435 GFP_ATOMIC);
1436 return 0;
1437}
1438
Patrick McHardy334a8132007-06-25 04:35:20 -07001439/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001440 * skb_cow - copy header of skb when it is required
1441 * @skb: buffer to cow
1442 * @headroom: needed headroom
1443 *
1444 * If the skb passed lacks sufficient headroom or its data part
1445 * is shared, data is reallocated. If reallocation fails, an error
1446 * is returned and original skb is not changed.
1447 *
1448 * The result is skb with writable area skb->head...skb->tail
1449 * and at least @headroom of space at head.
1450 */
1451static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1452{
Herbert Xud9cc2042007-09-16 16:21:16 -07001453 return __skb_cow(skb, headroom, skb_cloned(skb));
1454}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001455
Herbert Xud9cc2042007-09-16 16:21:16 -07001456/**
1457 * skb_cow_head - skb_cow but only making the head writable
1458 * @skb: buffer to cow
1459 * @headroom: needed headroom
1460 *
1461 * This function is identical to skb_cow except that we replace the
1462 * skb_cloned check by skb_header_cloned. It should be used when
1463 * you only need to push on some header and do not need to modify
1464 * the data.
1465 */
1466static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
1467{
1468 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001469}
1470
1471/**
1472 * skb_padto - pad an skbuff up to a minimal size
1473 * @skb: buffer to pad
1474 * @len: minimal length
1475 *
1476 * Pads up a buffer to ensure the trailing bytes exist and are
1477 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07001478 * is untouched. Otherwise it is extended. Returns zero on
1479 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001480 */
1481
Herbert Xu5b057c62006-06-23 02:06:41 -07001482static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001483{
1484 unsigned int size = skb->len;
1485 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07001486 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001487 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001488}
1489
1490static inline int skb_add_data(struct sk_buff *skb,
1491 char __user *from, int copy)
1492{
1493 const int off = skb->len;
1494
1495 if (skb->ip_summed == CHECKSUM_NONE) {
1496 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08001497 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
Linus Torvalds1da177e2005-04-16 15:20:36 -07001498 copy, 0, &err);
1499 if (!err) {
1500 skb->csum = csum_block_add(skb->csum, csum, off);
1501 return 0;
1502 }
1503 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1504 return 0;
1505
1506 __skb_trim(skb, off);
1507 return -EFAULT;
1508}
1509
1510static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1511 struct page *page, int off)
1512{
1513 if (i) {
1514 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1515
1516 return page == frag->page &&
1517 off == frag->page_offset + frag->size;
1518 }
1519 return 0;
1520}
1521
Herbert Xu364c6ba2006-06-09 16:10:40 -07001522static inline int __skb_linearize(struct sk_buff *skb)
1523{
1524 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1525}
1526
Linus Torvalds1da177e2005-04-16 15:20:36 -07001527/**
1528 * skb_linearize - convert paged skb to linear one
1529 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07001530 *
1531 * If there is no free memory -ENOMEM is returned, otherwise zero
1532 * is returned and the old skb data released.
1533 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07001534static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001535{
Herbert Xu364c6ba2006-06-09 16:10:40 -07001536 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1537}
1538
1539/**
1540 * skb_linearize_cow - make sure skb is linear and writable
1541 * @skb: buffer to process
1542 *
1543 * If there is no free memory -ENOMEM is returned, otherwise zero
1544 * is returned and the old skb data released.
1545 */
1546static inline int skb_linearize_cow(struct sk_buff *skb)
1547{
1548 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1549 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550}
1551
1552/**
1553 * skb_postpull_rcsum - update checksum for received skb after pull
1554 * @skb: buffer to update
1555 * @start: start of data before pull
1556 * @len: length of data pulled
1557 *
1558 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07001559 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1560 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001561 */
1562
1563static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08001564 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565{
Patrick McHardy84fa7932006-08-29 16:44:56 -07001566 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1568}
1569
Herbert Xucbb042f2006-03-20 22:43:56 -08001570unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1571
Linus Torvalds1da177e2005-04-16 15:20:36 -07001572/**
1573 * pskb_trim_rcsum - trim received skb and update checksum
1574 * @skb: buffer to trim
1575 * @len: new length
1576 *
1577 * This is exactly the same as pskb_trim except that it ensures the
1578 * checksum of received packets are still valid after the operation.
1579 */
1580
1581static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1582{
Stephen Hemminger0e4e4222005-09-08 12:32:03 -07001583 if (likely(len >= skb->len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001584 return 0;
Patrick McHardy84fa7932006-08-29 16:44:56 -07001585 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001586 skb->ip_summed = CHECKSUM_NONE;
1587 return __pskb_trim(skb, len);
1588}
1589
Linus Torvalds1da177e2005-04-16 15:20:36 -07001590#define skb_queue_walk(queue, skb) \
1591 for (skb = (queue)->next; \
1592 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1593 skb = skb->next)
1594
James Chapman46f89142007-04-30 00:07:31 -07001595#define skb_queue_walk_safe(queue, skb, tmp) \
1596 for (skb = (queue)->next, tmp = skb->next; \
1597 skb != (struct sk_buff *)(queue); \
1598 skb = tmp, tmp = skb->next)
1599
David S. Miller1164f522008-09-23 00:49:44 -07001600#define skb_queue_walk_from(queue, skb) \
1601 for (; prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1602 skb = skb->next)
1603
1604#define skb_queue_walk_from_safe(queue, skb, tmp) \
1605 for (tmp = skb->next; \
1606 skb != (struct sk_buff *)(queue); \
1607 skb = tmp, tmp = skb->next)
1608
Stephen Hemminger300ce172005-10-30 13:47:34 -08001609#define skb_queue_reverse_walk(queue, skb) \
1610 for (skb = (queue)->prev; \
1611 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1612 skb = skb->prev)
1613
Linus Torvalds1da177e2005-04-16 15:20:36 -07001614
Herbert Xua59322b2007-12-05 01:53:40 -08001615extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
1616 int *peeked, int *err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001617extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1618 int noblock, int *err);
1619extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1620 struct poll_table_struct *wait);
1621extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1622 int offset, struct iovec *to,
1623 int size);
Herbert Xufb286bb2005-11-10 13:01:24 -08001624extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001625 int hlen,
1626 struct iovec *iov);
Rusty Russelldb543c12008-08-15 15:13:53 -07001627extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
1628 int offset,
1629 struct iovec *from,
1630 int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001631extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
Herbert Xu27ab2562007-12-05 01:51:58 -08001632extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
Herbert Xu3305b802005-12-13 23:16:37 -08001633 unsigned int flags);
Al Viro2bbbc862006-11-14 21:37:14 -08001634extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
1635 int len, __wsum csum);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001636extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1637 void *to, int len);
Stephen Hemminger0c6fcc82007-04-20 16:40:01 -07001638extern int skb_store_bits(struct sk_buff *skb, int offset,
1639 const void *from, int len);
Al Viro81d77662006-11-14 21:37:33 -08001640extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001641 int offset, u8 *to, int len,
Al Viro81d77662006-11-14 21:37:33 -08001642 __wsum csum);
Jens Axboe9c55e012007-11-06 23:30:13 -08001643extern int skb_splice_bits(struct sk_buff *skb,
1644 unsigned int offset,
1645 struct pipe_inode_info *pipe,
1646 unsigned int len,
1647 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001648extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1649extern void skb_split(struct sk_buff *skb,
1650 struct sk_buff *skb1, const u32 len);
1651
Herbert Xu576a30e2006-06-27 13:22:38 -07001652extern struct sk_buff *skb_segment(struct sk_buff *skb, int features);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001653
Linus Torvalds1da177e2005-04-16 15:20:36 -07001654static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1655 int len, void *buffer)
1656{
1657 int hlen = skb_headlen(skb);
1658
Patrick McHardy55820ee2005-07-05 14:08:10 -07001659 if (hlen - offset >= len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001660 return skb->data + offset;
1661
1662 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1663 return NULL;
1664
1665 return buffer;
1666}
1667
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03001668static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
1669 void *to,
1670 const unsigned int len)
1671{
1672 memcpy(to, skb->data, len);
1673}
1674
1675static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
1676 const int offset, void *to,
1677 const unsigned int len)
1678{
1679 memcpy(to, skb->data + offset, len);
1680}
1681
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03001682static inline void skb_copy_to_linear_data(struct sk_buff *skb,
1683 const void *from,
1684 const unsigned int len)
1685{
1686 memcpy(skb->data, from, len);
1687}
1688
1689static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
1690 const int offset,
1691 const void *from,
1692 const unsigned int len)
1693{
1694 memcpy(skb->data + offset, from, len);
1695}
1696
Linus Torvalds1da177e2005-04-16 15:20:36 -07001697extern void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001698
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001699/**
1700 * skb_get_timestamp - get timestamp from a skb
1701 * @skb: skb to get stamp from
1702 * @stamp: pointer to struct timeval to store stamp in
1703 *
1704 * Timestamps are stored in the skb as offsets to a base timestamp.
1705 * This function converts the offset back to a struct timeval and stores
1706 * it in stamp.
1707 */
Stephen Hemmingerf2c38392005-09-06 15:48:03 -07001708static inline void skb_get_timestamp(const struct sk_buff *skb, struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001709{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001710 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001711}
1712
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001713static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001714{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001715 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001716}
1717
Stephen Hemminger164891a2007-04-23 22:26:16 -07001718static inline ktime_t net_timedelta(ktime_t t)
1719{
1720 return ktime_sub(ktime_get_real(), t);
1721}
1722
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07001723static inline ktime_t net_invalid_timestamp(void)
1724{
1725 return ktime_set(0, 0);
1726}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001727
Herbert Xu759e5d02007-03-25 20:10:56 -07001728extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
Al Virob51655b2006-11-14 21:40:42 -08001729extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08001730
Herbert Xu60476372007-04-09 11:59:39 -07001731static inline int skb_csum_unnecessary(const struct sk_buff *skb)
1732{
1733 return skb->ip_summed & CHECKSUM_UNNECESSARY;
1734}
1735
Herbert Xufb286bb2005-11-10 13:01:24 -08001736/**
1737 * skb_checksum_complete - Calculate checksum of an entire packet
1738 * @skb: packet to process
1739 *
1740 * This function calculates the checksum over the entire packet plus
1741 * the value of skb->csum. The latter can be used to supply the
1742 * checksum of a pseudo header as used by TCP/UDP. It returns the
1743 * checksum.
1744 *
1745 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
1746 * this function can be used to verify that checksum on received
1747 * packets. In that case the function should return zero if the
1748 * checksum is correct. In particular, this function will return zero
1749 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
1750 * hardware has already verified the correctness of the checksum.
1751 */
Al Viro4381ca32007-07-15 21:00:11 +01001752static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08001753{
Herbert Xu60476372007-04-09 11:59:39 -07001754 return skb_csum_unnecessary(skb) ?
1755 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08001756}
1757
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07001758#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07001759extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001760static inline void nf_conntrack_put(struct nf_conntrack *nfct)
1761{
1762 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07001763 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001764}
1765static inline void nf_conntrack_get(struct nf_conntrack *nfct)
1766{
1767 if (nfct)
1768 atomic_inc(&nfct->use);
1769}
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -08001770static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
1771{
1772 if (skb)
1773 atomic_inc(&skb->users);
1774}
1775static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
1776{
1777 if (skb)
1778 kfree_skb(skb);
1779}
1780#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07001781#ifdef CONFIG_BRIDGE_NETFILTER
1782static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
1783{
1784 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
1785 kfree(nf_bridge);
1786}
1787static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
1788{
1789 if (nf_bridge)
1790 atomic_inc(&nf_bridge->use);
1791}
1792#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08001793static inline void nf_reset(struct sk_buff *skb)
1794{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07001795#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08001796 nf_conntrack_put(skb->nfct);
1797 skb->nfct = NULL;
Patrick McHardya193a4a2006-03-20 19:23:05 -08001798 nf_conntrack_put_reasm(skb->nfct_reasm);
1799 skb->nfct_reasm = NULL;
1800#endif
1801#ifdef CONFIG_BRIDGE_NETFILTER
1802 nf_bridge_put(skb->nf_bridge);
1803 skb->nf_bridge = NULL;
1804#endif
1805}
1806
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07001807/* Note: This doesn't put any conntrack and bridge info in dst. */
1808static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
1809{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07001810#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07001811 dst->nfct = src->nfct;
1812 nf_conntrack_get(src->nfct);
1813 dst->nfctinfo = src->nfctinfo;
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07001814 dst->nfct_reasm = src->nfct_reasm;
1815 nf_conntrack_get_reasm(src->nfct_reasm);
1816#endif
1817#ifdef CONFIG_BRIDGE_NETFILTER
1818 dst->nf_bridge = src->nf_bridge;
1819 nf_bridge_get(src->nf_bridge);
1820#endif
1821}
1822
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07001823static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
1824{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07001825#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07001826 nf_conntrack_put(dst->nfct);
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07001827 nf_conntrack_put_reasm(dst->nfct_reasm);
1828#endif
1829#ifdef CONFIG_BRIDGE_NETFILTER
1830 nf_bridge_put(dst->nf_bridge);
1831#endif
1832 __nf_copy(dst, src);
1833}
1834
James Morris984bc162006-06-09 00:29:17 -07001835#ifdef CONFIG_NETWORK_SECMARK
1836static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1837{
1838 to->secmark = from->secmark;
1839}
1840
1841static inline void skb_init_secmark(struct sk_buff *skb)
1842{
1843 skb->secmark = 0;
1844}
1845#else
1846static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
1847{ }
1848
1849static inline void skb_init_secmark(struct sk_buff *skb)
1850{ }
1851#endif
1852
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001853static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
1854{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001855 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001856}
1857
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07001858static inline u16 skb_get_queue_mapping(struct sk_buff *skb)
1859{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07001860 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07001861}
1862
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001863static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
1864{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001865 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07001866}
1867
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07001868#ifdef CONFIG_XFRM
1869static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
1870{
1871 return skb->sp;
1872}
1873#else
1874static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
1875{
1876 return NULL;
1877}
1878#endif
1879
Herbert Xu89114af2006-07-08 13:34:32 -07001880static inline int skb_is_gso(const struct sk_buff *skb)
1881{
1882 return skb_shinfo(skb)->gso_size;
1883}
1884
Brice Goglineabd7e32007-10-13 12:33:32 +02001885static inline int skb_is_gso_v6(const struct sk_buff *skb)
1886{
1887 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
1888}
1889
Ben Hutchings4497b072008-06-19 16:22:28 -07001890extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
1891
1892static inline bool skb_warn_if_lro(const struct sk_buff *skb)
1893{
1894 /* LRO sets gso_size but not gso_type, whereas if GSO is really
1895 * wanted then gso_type will be set. */
1896 struct skb_shared_info *shinfo = skb_shinfo(skb);
1897 if (shinfo->gso_size != 0 && unlikely(shinfo->gso_type == 0)) {
1898 __skb_warn_lro_forwarding(skb);
1899 return true;
1900 }
1901 return false;
1902}
1903
Herbert Xu35fc92a2007-03-26 23:22:20 -07001904static inline void skb_forward_csum(struct sk_buff *skb)
1905{
1906 /* Unfortunately we don't support this one. Any brave souls? */
1907 if (skb->ip_summed == CHECKSUM_COMPLETE)
1908 skb->ip_summed = CHECKSUM_NONE;
1909}
1910
Rusty Russellf35d9d82008-02-04 23:49:54 -05001911bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001912#endif /* __KERNEL__ */
1913#endif /* _LINUX_SKBUFF_H */