blob: 05a358f1ba11a3035e37e8e064a2983f6276ba87 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * Definitions for the 'struct sk_buff' memory handlers.
3 *
4 * Authors:
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14#ifndef _LINUX_SKBUFF_H
15#define _LINUX_SKBUFF_H
16
Linus Torvalds1da177e2005-04-16 15:20:36 -070017#include <linux/kernel.h>
Vegard Nossumfe55f6d2008-08-30 12:16:35 +020018#include <linux/kmemcheck.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070019#include <linux/compiler.h>
20#include <linux/time.h>
21#include <linux/cache.h>
22
23#include <asm/atomic.h>
24#include <asm/types.h>
25#include <linux/spinlock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070026#include <linux/net.h>
Thomas Graf3fc7e8a2005-06-23 21:00:17 -070027#include <linux/textsearch.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070028#include <net/checksum.h>
Al Viroa80958f2006-12-04 20:41:19 +000029#include <linux/rcupdate.h>
Chris Leech97fc2f02006-05-23 17:55:33 -070030#include <linux/dmaengine.h>
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -070031#include <linux/hrtimer.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070032
Herbert Xu60476372007-04-09 11:59:39 -070033/* Don't change this without changing skb_csum_unnecessary! */
Linus Torvalds1da177e2005-04-16 15:20:36 -070034#define CHECKSUM_NONE 0
Herbert Xu60476372007-04-09 11:59:39 -070035#define CHECKSUM_UNNECESSARY 1
36#define CHECKSUM_COMPLETE 2
37#define CHECKSUM_PARTIAL 3
Linus Torvalds1da177e2005-04-16 15:20:36 -070038
39#define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
40 ~(SMP_CACHE_BYTES - 1))
David S. Millerfc910a22007-03-25 20:27:59 -070041#define SKB_WITH_OVERHEAD(X) \
Herbert Xudeea84b2007-10-21 16:27:46 -070042 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
David S. Millerfc910a22007-03-25 20:27:59 -070043#define SKB_MAX_ORDER(X, ORDER) \
44 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
Linus Torvalds1da177e2005-04-16 15:20:36 -070045#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
46#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
47
48/* A. Checksumming of received packets by device.
49 *
50 * NONE: device failed to checksum this packet.
51 * skb->csum is undefined.
52 *
53 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
54 * skb->csum is undefined.
55 * It is bad option, but, unfortunately, many of vendors do this.
56 * Apparently with secret goal to sell you new device, when you
57 * will add new protocol to your host. F.e. IPv6. 8)
58 *
Patrick McHardy84fa7932006-08-29 16:44:56 -070059 * COMPLETE: the most generic way. Device supplied checksum of _all_
Linus Torvalds1da177e2005-04-16 15:20:36 -070060 * the packet as seen by netif_rx in skb->csum.
61 * NOTE: Even if device supports only some protocols, but
Patrick McHardy84fa7932006-08-29 16:44:56 -070062 * is able to produce some skb->csum, it MUST use COMPLETE,
Linus Torvalds1da177e2005-04-16 15:20:36 -070063 * not UNNECESSARY.
64 *
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070065 * PARTIAL: identical to the case for output below. This may occur
66 * on a packet received directly from another Linux OS, e.g.,
67 * a virtualised Linux kernel on the same host. The packet can
68 * be treated in the same way as UNNECESSARY except that on
69 * output (i.e., forwarding) the checksum must be filled in
70 * by the OS or the hardware.
71 *
Linus Torvalds1da177e2005-04-16 15:20:36 -070072 * B. Checksumming on output.
73 *
74 * NONE: skb is checksummed by protocol or csum is not required.
75 *
Patrick McHardy84fa7932006-08-29 16:44:56 -070076 * PARTIAL: device is required to csum packet as seen by hard_start_xmit
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070077 * from skb->csum_start to the end and to record the checksum
78 * at skb->csum_start + skb->csum_offset.
Linus Torvalds1da177e2005-04-16 15:20:36 -070079 *
80 * Device must show its capabilities in dev->features, set
81 * at device setup time.
82 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
83 * everything.
84 * NETIF_F_NO_CSUM - loopback or reliable single hop media.
85 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
86 * TCP/UDP over IPv4. Sigh. Vendors like this
87 * way by an unknown reason. Though, see comment above
88 * about CHECKSUM_UNNECESSARY. 8)
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070089 * NETIF_F_IPV6_CSUM about as dumb as the last one but does IPv6 instead.
Linus Torvalds1da177e2005-04-16 15:20:36 -070090 *
91 * Any questions? No questions, good. --ANK
92 */
93
Linus Torvalds1da177e2005-04-16 15:20:36 -070094struct net_device;
David Howells716ea3a2007-04-02 20:19:53 -070095struct scatterlist;
Jens Axboe9c55e012007-11-06 23:30:13 -080096struct pipe_inode_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -070097
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -070098#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Linus Torvalds1da177e2005-04-16 15:20:36 -070099struct nf_conntrack {
100 atomic_t use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101};
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700102#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700103
104#ifdef CONFIG_BRIDGE_NETFILTER
105struct nf_bridge_info {
106 atomic_t use;
107 struct net_device *physindev;
108 struct net_device *physoutdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700109 unsigned int mask;
110 unsigned long data[32 / sizeof(unsigned long)];
111};
112#endif
113
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114struct sk_buff_head {
115 /* These two members must be first. */
116 struct sk_buff *next;
117 struct sk_buff *prev;
118
119 __u32 qlen;
120 spinlock_t lock;
121};
122
123struct sk_buff;
124
125/* To allow 64K frame to be packed as single skb without frag_list */
126#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
127
128typedef struct skb_frag_struct skb_frag_t;
129
130struct skb_frag_struct {
131 struct page *page;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000132#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
David S. Millera309bb02007-07-30 18:47:03 -0700133 __u32 page_offset;
134 __u32 size;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000135#else
136 __u16 page_offset;
137 __u16 size;
138#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700139};
140
Patrick Ohlyac45f602009-02-12 05:03:37 +0000141#define HAVE_HW_TIME_STAMP
142
143/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800144 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000145 * @hwtstamp: hardware time stamp transformed into duration
146 * since arbitrary point in time
147 * @syststamp: hwtstamp transformed to system time base
148 *
149 * Software time stamps generated by ktime_get_real() are stored in
150 * skb->tstamp. The relation between the different kinds of time
151 * stamps is as follows:
152 *
153 * syststamp and tstamp can be compared against each other in
154 * arbitrary combinations. The accuracy of a
155 * syststamp/tstamp/"syststamp from other device" comparison is
156 * limited by the accuracy of the transformation into system time
157 * base. This depends on the device driver and its underlying
158 * hardware.
159 *
160 * hwtstamps can only be compared against other hwtstamps from
161 * the same device.
162 *
163 * This structure is attached to packets as part of the
164 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
165 */
166struct skb_shared_hwtstamps {
167 ktime_t hwtstamp;
168 ktime_t syststamp;
169};
170
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000171/* Definitions for tx_flags in struct skb_shared_info */
172enum {
173 /* generate hardware time stamp */
174 SKBTX_HW_TSTAMP = 1 << 0,
175
176 /* generate software time stamp */
177 SKBTX_SW_TSTAMP = 1 << 1,
178
179 /* device driver is going to provide hardware time stamp */
180 SKBTX_IN_PROGRESS = 1 << 2,
181
182 /* ensure the originating sk reference is available on driver level */
183 SKBTX_DRV_NEEDS_SK_REF = 1 << 3,
Patrick Ohlyac45f602009-02-12 05:03:37 +0000184};
185
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186/* This data is invariant across clones and lives at
187 * the end of the header data, ie. at skb->end.
188 */
189struct skb_shared_info {
Benjamin LaHaise4947d3e2006-01-03 14:06:50 -0800190 unsigned short nr_frags;
Herbert Xu79671682006-06-22 02:40:14 -0700191 unsigned short gso_size;
192 /* Warning: this field is not always filled in (UFO)! */
193 unsigned short gso_segs;
194 unsigned short gso_type;
Al Viroae08e1f2006-11-08 00:27:11 -0800195 __be32 ip6_frag_id;
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000196 __u8 tx_flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700197 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000198 struct skb_shared_hwtstamps hwtstamps;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700199
200 /*
201 * Warning : all fields before dataref are cleared in __alloc_skb()
202 */
203 atomic_t dataref;
204
Johann Baudy69e3c752009-05-18 22:11:22 -0700205 /* Intermediate layers must ensure that destructor_arg
206 * remains valid until skb destructor */
207 void * destructor_arg;
Eric Dumazetfed66382010-07-22 19:09:08 +0000208 /* must be last field, see pskb_expand_head() */
209 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700210};
211
212/* We divide dataref into two halves. The higher 16 bits hold references
213 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700214 * the entire skb->data. A clone of a headerless skb holds the length of
215 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700216 *
217 * All users must obey the rule that the skb->data reference count must be
218 * greater than or equal to the payload reference count.
219 *
220 * Holding a reference to the payload part means that the user does not
221 * care about modifications to the header part of skb->data.
222 */
223#define SKB_DATAREF_SHIFT 16
224#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
225
David S. Millerd179cd12005-08-17 14:57:30 -0700226
227enum {
228 SKB_FCLONE_UNAVAILABLE,
229 SKB_FCLONE_ORIG,
230 SKB_FCLONE_CLONE,
231};
232
Herbert Xu79671682006-06-22 02:40:14 -0700233enum {
234 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700235 SKB_GSO_UDP = 1 << 1,
Herbert Xu576a30e2006-06-27 13:22:38 -0700236
237 /* This indicates the skb is from an untrusted source. */
238 SKB_GSO_DODGY = 1 << 2,
Michael Chanb0da85372006-06-29 12:30:00 -0700239
240 /* This indicates the tcp segment has CWR set. */
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700241 SKB_GSO_TCP_ECN = 1 << 3,
242
243 SKB_GSO_TCPV6 = 1 << 4,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800244
245 SKB_GSO_FCOE = 1 << 5,
Herbert Xu79671682006-06-22 02:40:14 -0700246};
247
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700248#if BITS_PER_LONG > 32
249#define NET_SKBUFF_DATA_USES_OFFSET 1
250#endif
251
252#ifdef NET_SKBUFF_DATA_USES_OFFSET
253typedef unsigned int sk_buff_data_t;
254#else
255typedef unsigned char *sk_buff_data_t;
256#endif
257
Linus Torvalds1da177e2005-04-16 15:20:36 -0700258/**
259 * struct sk_buff - socket buffer
260 * @next: Next buffer in list
261 * @prev: Previous buffer in list
Linus Torvalds1da177e2005-04-16 15:20:36 -0700262 * @sk: Socket we are owned by
Herbert Xu325ed822005-10-03 13:57:23 -0700263 * @tstamp: Time we arrived
Linus Torvalds1da177e2005-04-16 15:20:36 -0700264 * @dev: Device we arrived on/are leaving by
Randy Dunlapbe52178b2007-05-03 03:16:20 -0700265 * @transport_header: Transport layer header
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -0700266 * @network_header: Network layer header
267 * @mac_header: Link layer header
Eric Dumazet7fee2262010-05-11 23:19:48 +0000268 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700269 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700270 * @cb: Control buffer. Free for use by every layer. Put private vars here
271 * @len: Length of actual data
272 * @data_len: Data length
273 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700274 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700275 * @csum: Checksum (must include start/offset pair)
276 * @csum_start: Offset from skb->head where checksumming should start
277 * @csum_offset: Offset from csum_start where checksum should be stored
Martin Waitz67be2dd2005-05-01 08:59:26 -0700278 * @local_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279 * @cloned: Head may be cloned (check refcnt to be sure)
280 * @nohdr: Payload reference only, must not modify header
281 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700282 * @fclone: skbuff clone status
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283 * @ip_summed: Driver fed us an IP checksum
284 * @priority: Packet queueing priority
285 * @users: User count - see {datagram,tcp}.c
286 * @protocol: Packet protocol from driver
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287 * @truesize: Buffer size
288 * @head: Head of buffer
289 * @data: Data head pointer
290 * @tail: Tail pointer
291 * @end: End pointer
292 * @destructor: Destruct function
Thomas Graf82e91ff2006-11-09 15:19:14 -0800293 * @mark: Generic packet mark
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294 * @nfct: Associated connection, if any
Randy Dunlapc83c2482005-10-18 22:07:41 -0700295 * @ipvs_property: skbuff is owned by ipvs
Randy Dunlap31729362008-02-18 20:52:13 -0800296 * @peeked: this packet has been seen already, so stats have been
297 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700298 * @nf_trace: netfilter packet trace flag
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299 * @nfctinfo: Relationship of this skb to the connection
Randy Dunlap461ddf32005-11-20 21:25:15 -0800300 * @nfct_reasm: netfilter conntrack re-assembly pointer
Linus Torvalds1da177e2005-04-16 15:20:36 -0700301 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800302 * @skb_iif: ifindex of device we arrived on
Tom Herbert0a9627f2010-03-16 08:03:29 +0000303 * @rxhash: the packet hash computed on receive
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700304 * @queue_mapping: Queue mapping for multiqueue devices
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305 * @tc_index: Traffic control index
306 * @tc_verd: traffic control verdict
Randy Dunlap553a5672008-04-20 10:51:01 -0700307 * @ndisc_nodetype: router type (from link layer)
Randy Dunlapf4b8ea72006-06-22 16:00:11 -0700308 * @dma_cookie: a cookie to one of several possible DMA operations
309 * done by skb DMA functions
James Morris984bc162006-06-09 00:29:17 -0700310 * @secmark: security marking
Patrick McHardy6aa895b02008-07-14 22:49:06 -0700311 * @vlan_tci: vlan tag control information
Linus Torvalds1da177e2005-04-16 15:20:36 -0700312 */
313
314struct sk_buff {
315 /* These two members must be first. */
316 struct sk_buff *next;
317 struct sk_buff *prev;
318
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700319 ktime_t tstamp;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000320
321 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324 /*
325 * This is the control buffer. It is free to use for every
326 * layer. Please put your private variables there. If you
327 * want to keep them across layers you have to do a skb_clone()
328 * first. This is owned by whoever has the skb queued ATM.
329 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000330 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331
Eric Dumazet7fee2262010-05-11 23:19:48 +0000332 unsigned long _skb_refdst;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000333#ifdef CONFIG_XFRM
334 struct sec_path *sp;
335#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700337 data_len;
338 __u16 mac_len,
339 hdr_len;
Al Viroff1dcad2006-11-20 18:07:29 -0800340 union {
341 __wsum csum;
Herbert Xu663ead32007-04-09 11:59:07 -0700342 struct {
343 __u16 csum_start;
344 __u16 csum_offset;
345 };
Al Viroff1dcad2006-11-20 18:07:29 -0800346 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347 __u32 priority;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200348 kmemcheck_bitfield_begin(flags1);
Thomas Graf1cbb3382005-07-05 14:13:41 -0700349 __u8 local_df:1,
350 cloned:1,
351 ip_summed:2,
Harald Welte6869c4d2005-08-09 19:24:19 -0700352 nohdr:1,
353 nfctinfo:3;
David S. Millerd179cd12005-08-17 14:57:30 -0700354 __u8 pkt_type:3,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800355 fclone:2,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700356 ipvs_property:1,
Herbert Xua59322b2007-12-05 01:53:40 -0800357 peeked:1,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700358 nf_trace:1;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200359 kmemcheck_bitfield_end(flags1);
Eric Dumazet4ab408d2010-03-01 03:09:26 +0000360 __be16 protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361
362 void (*destructor)(struct sk_buff *skb);
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800363#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700364 struct nf_conntrack *nfct;
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800365 struct sk_buff *nfct_reasm;
366#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367#ifdef CONFIG_BRIDGE_NETFILTER
368 struct nf_bridge_info *nf_bridge;
369#endif
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700370
Eric Dumazet8964be42009-11-20 15:35:04 -0800371 int skb_iif;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700372#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700373 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700374#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700375 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700376#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700377#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200378
Tom Herbert0a9627f2010-03-16 08:03:29 +0000379 __u32 rxhash;
380
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200381 kmemcheck_bitfield_begin(flags2);
Eric Dumazet14d18a82009-10-29 00:10:37 +0000382 __u16 queue_mapping:16;
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400383#ifdef CONFIG_IPV6_NDISC_NODETYPE
John Fastabend597a2642010-06-03 09:30:11 +0000384 __u8 ndisc_nodetype:2,
385 deliver_no_wcard:1;
386#else
387 __u8 deliver_no_wcard:1;
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400388#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200389 kmemcheck_bitfield_end(flags2);
390
Johannes Berg72bce622009-06-17 17:45:28 +0200391 /* 0/14 bit hole */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700392
Chris Leech97fc2f02006-05-23 17:55:33 -0700393#ifdef CONFIG_NET_DMA
394 dma_cookie_t dma_cookie;
395#endif
James Morris984bc162006-06-09 00:29:17 -0700396#ifdef CONFIG_NETWORK_SECMARK
397 __u32 secmark;
398#endif
Neil Horman3b885782009-10-12 13:26:31 -0700399 union {
400 __u32 mark;
401 __u32 dropcount;
402 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700403
Patrick McHardy6aa895b02008-07-14 22:49:06 -0700404 __u16 vlan_tci;
405
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700406 sk_buff_data_t transport_header;
407 sk_buff_data_t network_header;
408 sk_buff_data_t mac_header;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700410 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700411 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700413 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700414 unsigned int truesize;
415 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416};
417
418#ifdef __KERNEL__
419/*
420 * Handling routines are only of interest to the kernel
421 */
422#include <linux/slab.h>
423
424#include <asm/system.h>
425
Eric Dumazet7fee2262010-05-11 23:19:48 +0000426/*
427 * skb might have a dst pointer attached, refcounted or not.
428 * _skb_refdst low order bit is set if refcount was _not_ taken
429 */
430#define SKB_DST_NOREF 1UL
431#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
432
433/**
434 * skb_dst - returns skb dst_entry
435 * @skb: buffer
436 *
437 * Returns skb dst_entry, regardless of reference taken or not.
438 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000439static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
440{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000441 /* If refdst was not refcounted, check we still are in a
442 * rcu_read_lock section
443 */
444 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
445 !rcu_read_lock_held() &&
446 !rcu_read_lock_bh_held());
447 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000448}
449
Eric Dumazet7fee2262010-05-11 23:19:48 +0000450/**
451 * skb_dst_set - sets skb dst
452 * @skb: buffer
453 * @dst: dst entry
454 *
455 * Sets skb dst, assuming a reference was taken on dst and should
456 * be released by skb_dst_drop()
457 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000458static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
459{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000460 skb->_skb_refdst = (unsigned long)dst;
461}
462
463/**
464 * skb_dst_set_noref - sets skb dst, without a reference
465 * @skb: buffer
466 * @dst: dst entry
467 *
468 * Sets skb dst, assuming a reference was not taken on dst
469 * skb_dst_drop() should not dst_release() this dst
470 */
471static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
472{
473 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
474 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
475}
476
477/**
478 * skb_dst_is_noref - Test if skb dst isnt refcounted
479 * @skb: buffer
480 */
481static inline bool skb_dst_is_noref(const struct sk_buff *skb)
482{
483 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000484}
485
Eric Dumazet511c3f92009-06-02 05:14:27 +0000486static inline struct rtable *skb_rtable(const struct sk_buff *skb)
487{
Eric Dumazetadf30902009-06-02 05:19:30 +0000488 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000489}
490
Jörn Engel231d06a2006-03-20 21:28:35 -0800491extern void kfree_skb(struct sk_buff *skb);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000492extern void consume_skb(struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493extern void __kfree_skb(struct sk_buff *skb);
David S. Millerd179cd12005-08-17 14:57:30 -0700494extern struct sk_buff *__alloc_skb(unsigned int size,
Christoph Hellwigb30973f2006-12-06 20:32:36 -0800495 gfp_t priority, int fclone, int node);
David S. Millerd179cd12005-08-17 14:57:30 -0700496static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100497 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700498{
Eric Dumazet564824b2010-10-11 19:05:25 +0000499 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700500}
501
502static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100503 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700504{
Eric Dumazet564824b2010-10-11 19:05:25 +0000505 return __alloc_skb(size, priority, 1, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700506}
507
Changli Gao5b0daa32010-05-29 00:12:13 -0700508extern bool skb_recycle_check(struct sk_buff *skb, int skb_size);
Lennert Buytenhek04a4bb52008-10-01 02:33:12 -0700509
Herbert Xue0053ec2007-10-14 00:37:52 -0700510extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700511extern struct sk_buff *skb_clone(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100512 gfp_t priority);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700513extern struct sk_buff *skb_copy(const struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100514 gfp_t priority);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700515extern struct sk_buff *pskb_copy(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100516 gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517extern int pskb_expand_head(struct sk_buff *skb,
Victor Fusco86a76ca2005-07-08 14:57:47 -0700518 int nhead, int ntail,
Al Virodd0fc662005-10-07 07:46:04 +0100519 gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
521 unsigned int headroom);
522extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
523 int newheadroom, int newtailroom,
Al Virodd0fc662005-10-07 07:46:04 +0100524 gfp_t priority);
David Howells716ea3a2007-04-02 20:19:53 -0700525extern int skb_to_sgvec(struct sk_buff *skb,
526 struct scatterlist *sg, int offset,
527 int len);
528extern int skb_cow_data(struct sk_buff *skb, int tailbits,
529 struct sk_buff **trailer);
Herbert Xu5b057c62006-06-23 02:06:41 -0700530extern int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000531#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700533extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
534 int getfrag(void *from, char *to, int offset,
535 int len,int odd, struct sk_buff *skb),
536 void *from, int length);
537
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800538struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700539 __u32 lower_offset;
540 __u32 upper_offset;
541 __u32 frag_idx;
542 __u32 stepped_offset;
543 struct sk_buff *root_skb;
544 struct sk_buff *cur_skb;
545 __u8 *frag_data;
546};
547
548extern void skb_prepare_seq_read(struct sk_buff *skb,
549 unsigned int from, unsigned int to,
550 struct skb_seq_state *st);
551extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
552 struct skb_seq_state *st);
553extern void skb_abort_seq_read(struct skb_seq_state *st);
554
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700555extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
556 unsigned int to, struct ts_config *config,
557 struct ts_state *state);
558
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000559extern __u32 __skb_get_rxhash(struct sk_buff *skb);
560static inline __u32 skb_get_rxhash(struct sk_buff *skb)
561{
562 if (!skb->rxhash)
563 skb->rxhash = __skb_get_rxhash(skb);
564
565 return skb->rxhash;
566}
567
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700568#ifdef NET_SKBUFF_DATA_USES_OFFSET
569static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
570{
571 return skb->head + skb->end;
572}
573#else
574static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
575{
576 return skb->end;
577}
578#endif
579
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700581#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582
Patrick Ohlyac45f602009-02-12 05:03:37 +0000583static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
584{
585 return &skb_shinfo(skb)->hwtstamps;
586}
587
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588/**
589 * skb_queue_empty - check if a queue is empty
590 * @list: queue head
591 *
592 * Returns true if the queue is empty, false otherwise.
593 */
594static inline int skb_queue_empty(const struct sk_buff_head *list)
595{
596 return list->next == (struct sk_buff *)list;
597}
598
599/**
David S. Millerfc7ebb22008-09-23 00:34:07 -0700600 * skb_queue_is_last - check if skb is the last entry in the queue
601 * @list: queue head
602 * @skb: buffer
603 *
604 * Returns true if @skb is the last buffer on the list.
605 */
606static inline bool skb_queue_is_last(const struct sk_buff_head *list,
607 const struct sk_buff *skb)
608{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000609 return skb->next == (struct sk_buff *)list;
David S. Millerfc7ebb22008-09-23 00:34:07 -0700610}
611
612/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800613 * skb_queue_is_first - check if skb is the first entry in the queue
614 * @list: queue head
615 * @skb: buffer
616 *
617 * Returns true if @skb is the first buffer on the list.
618 */
619static inline bool skb_queue_is_first(const struct sk_buff_head *list,
620 const struct sk_buff *skb)
621{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000622 return skb->prev == (struct sk_buff *)list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800623}
624
625/**
David S. Miller249c8b42008-09-23 00:44:42 -0700626 * skb_queue_next - return the next packet in the queue
627 * @list: queue head
628 * @skb: current buffer
629 *
630 * Return the next packet in @list after @skb. It is only valid to
631 * call this if skb_queue_is_last() evaluates to false.
632 */
633static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
634 const struct sk_buff *skb)
635{
636 /* This BUG_ON may seem severe, but if we just return then we
637 * are going to dereference garbage.
638 */
639 BUG_ON(skb_queue_is_last(list, skb));
640 return skb->next;
641}
642
643/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800644 * skb_queue_prev - return the prev packet in the queue
645 * @list: queue head
646 * @skb: current buffer
647 *
648 * Return the prev packet in @list before @skb. It is only valid to
649 * call this if skb_queue_is_first() evaluates to false.
650 */
651static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
652 const struct sk_buff *skb)
653{
654 /* This BUG_ON may seem severe, but if we just return then we
655 * are going to dereference garbage.
656 */
657 BUG_ON(skb_queue_is_first(list, skb));
658 return skb->prev;
659}
660
661/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700662 * skb_get - reference buffer
663 * @skb: buffer to reference
664 *
665 * Makes another reference to a socket buffer and returns a pointer
666 * to the buffer.
667 */
668static inline struct sk_buff *skb_get(struct sk_buff *skb)
669{
670 atomic_inc(&skb->users);
671 return skb;
672}
673
674/*
675 * If users == 1, we are the only owner and are can avoid redundant
676 * atomic change.
677 */
678
679/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680 * skb_cloned - is the buffer a clone
681 * @skb: buffer to check
682 *
683 * Returns true if the buffer was generated with skb_clone() and is
684 * one of multiple shared copies of the buffer. Cloned buffers are
685 * shared data so must not be written to under normal circumstances.
686 */
687static inline int skb_cloned(const struct sk_buff *skb)
688{
689 return skb->cloned &&
690 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
691}
692
693/**
694 * skb_header_cloned - is the header a clone
695 * @skb: buffer to check
696 *
697 * Returns true if modifying the header part of the buffer requires
698 * the data to be copied.
699 */
700static inline int skb_header_cloned(const struct sk_buff *skb)
701{
702 int dataref;
703
704 if (!skb->cloned)
705 return 0;
706
707 dataref = atomic_read(&skb_shinfo(skb)->dataref);
708 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
709 return dataref != 1;
710}
711
712/**
713 * skb_header_release - release reference to header
714 * @skb: buffer to operate on
715 *
716 * Drop a reference to the header part of the buffer. This is done
717 * by acquiring a payload reference. You must not read from the header
718 * part of skb->data after this.
719 */
720static inline void skb_header_release(struct sk_buff *skb)
721{
722 BUG_ON(skb->nohdr);
723 skb->nohdr = 1;
724 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
725}
726
727/**
728 * skb_shared - is the buffer shared
729 * @skb: buffer to check
730 *
731 * Returns true if more than one person has a reference to this
732 * buffer.
733 */
734static inline int skb_shared(const struct sk_buff *skb)
735{
736 return atomic_read(&skb->users) != 1;
737}
738
739/**
740 * skb_share_check - check if buffer is shared and if so clone it
741 * @skb: buffer to check
742 * @pri: priority for memory allocation
743 *
744 * If the buffer is shared the buffer is cloned and the old copy
745 * drops a reference. A new clone with a single reference is returned.
746 * If the buffer is not shared the original buffer is returned. When
747 * being called from interrupt status or with spinlocks held pri must
748 * be GFP_ATOMIC.
749 *
750 * NULL is returned on a memory allocation failure.
751 */
Victor Fusco86a76ca2005-07-08 14:57:47 -0700752static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100753 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754{
755 might_sleep_if(pri & __GFP_WAIT);
756 if (skb_shared(skb)) {
757 struct sk_buff *nskb = skb_clone(skb, pri);
758 kfree_skb(skb);
759 skb = nskb;
760 }
761 return skb;
762}
763
764/*
765 * Copy shared buffers into a new sk_buff. We effectively do COW on
766 * packets to handle cases where we have a local reader and forward
767 * and a couple of other messy ones. The normal one is tcpdumping
768 * a packet thats being forwarded.
769 */
770
771/**
772 * skb_unshare - make a copy of a shared buffer
773 * @skb: buffer to check
774 * @pri: priority for memory allocation
775 *
776 * If the socket buffer is a clone then this function creates a new
777 * copy of the data, drops a reference count on the old copy and returns
778 * the new copy with the reference count at 1. If the buffer is not a clone
779 * the original buffer is returned. When called with a spinlock held or
780 * from interrupt state @pri must be %GFP_ATOMIC
781 *
782 * %NULL is returned on a memory allocation failure.
783 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -0700784static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100785 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700786{
787 might_sleep_if(pri & __GFP_WAIT);
788 if (skb_cloned(skb)) {
789 struct sk_buff *nskb = skb_copy(skb, pri);
790 kfree_skb(skb); /* Free our shared copy */
791 skb = nskb;
792 }
793 return skb;
794}
795
796/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -0800797 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -0700798 * @list_: list to peek at
799 *
800 * Peek an &sk_buff. Unlike most other operations you _MUST_
801 * be careful with this one. A peek leaves the buffer on the
802 * list and someone else may run off with it. You must hold
803 * the appropriate locks or have a private queue to do this.
804 *
805 * Returns %NULL for an empty list or a pointer to the head element.
806 * The reference count is not incremented and the reference is therefore
807 * volatile. Use with caution.
808 */
809static inline struct sk_buff *skb_peek(struct sk_buff_head *list_)
810{
811 struct sk_buff *list = ((struct sk_buff *)list_)->next;
812 if (list == (struct sk_buff *)list_)
813 list = NULL;
814 return list;
815}
816
817/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -0800818 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -0700819 * @list_: list to peek at
820 *
821 * Peek an &sk_buff. Unlike most other operations you _MUST_
822 * be careful with this one. A peek leaves the buffer on the
823 * list and someone else may run off with it. You must hold
824 * the appropriate locks or have a private queue to do this.
825 *
826 * Returns %NULL for an empty list or a pointer to the tail element.
827 * The reference count is not incremented and the reference is therefore
828 * volatile. Use with caution.
829 */
830static inline struct sk_buff *skb_peek_tail(struct sk_buff_head *list_)
831{
832 struct sk_buff *list = ((struct sk_buff *)list_)->prev;
833 if (list == (struct sk_buff *)list_)
834 list = NULL;
835 return list;
836}
837
838/**
839 * skb_queue_len - get queue length
840 * @list_: list to measure
841 *
842 * Return the length of an &sk_buff queue.
843 */
844static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
845{
846 return list_->qlen;
847}
848
David S. Miller67fed452008-09-21 22:36:24 -0700849/**
850 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
851 * @list: queue to initialize
852 *
853 * This initializes only the list and queue length aspects of
854 * an sk_buff_head object. This allows to initialize the list
855 * aspects of an sk_buff_head without reinitializing things like
856 * the spinlock. It can also be used for on-stack sk_buff_head
857 * objects where the spinlock is known to not be used.
858 */
859static inline void __skb_queue_head_init(struct sk_buff_head *list)
860{
861 list->prev = list->next = (struct sk_buff *)list;
862 list->qlen = 0;
863}
864
Arjan van de Ven76f10ad2006-08-02 14:06:55 -0700865/*
866 * This function creates a split out lock class for each invocation;
867 * this is needed for now since a whole lot of users of the skb-queue
868 * infrastructure in drivers have different locking usage (in hardirq)
869 * than the networking core (in softirq only). In the long run either the
870 * network layer or drivers should need annotation to consolidate the
871 * main types of usage into 3 classes.
872 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700873static inline void skb_queue_head_init(struct sk_buff_head *list)
874{
875 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -0700876 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877}
878
Pavel Emelianovc2ecba72007-04-17 12:45:31 -0700879static inline void skb_queue_head_init_class(struct sk_buff_head *list,
880 struct lock_class_key *class)
881{
882 skb_queue_head_init(list);
883 lockdep_set_class(&list->lock, class);
884}
885
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886/*
Gerrit Renkerbf299272008-04-14 00:04:51 -0700887 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888 *
889 * The "__skb_xxxx()" functions are the non-atomic ones that
890 * can only be called with interrupts disabled.
891 */
Gerrit Renkerbf299272008-04-14 00:04:51 -0700892extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
893static inline void __skb_insert(struct sk_buff *newsk,
894 struct sk_buff *prev, struct sk_buff *next,
895 struct sk_buff_head *list)
896{
897 newsk->next = next;
898 newsk->prev = prev;
899 next->prev = prev->next = newsk;
900 list->qlen++;
901}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902
David S. Miller67fed452008-09-21 22:36:24 -0700903static inline void __skb_queue_splice(const struct sk_buff_head *list,
904 struct sk_buff *prev,
905 struct sk_buff *next)
906{
907 struct sk_buff *first = list->next;
908 struct sk_buff *last = list->prev;
909
910 first->prev = prev;
911 prev->next = first;
912
913 last->next = next;
914 next->prev = last;
915}
916
917/**
918 * skb_queue_splice - join two skb lists, this is designed for stacks
919 * @list: the new list to add
920 * @head: the place to add it in the first list
921 */
922static inline void skb_queue_splice(const struct sk_buff_head *list,
923 struct sk_buff_head *head)
924{
925 if (!skb_queue_empty(list)) {
926 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -0700927 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -0700928 }
929}
930
931/**
932 * skb_queue_splice - join two skb lists and reinitialise the emptied list
933 * @list: the new list to add
934 * @head: the place to add it in the first list
935 *
936 * The list at @list is reinitialised
937 */
938static inline void skb_queue_splice_init(struct sk_buff_head *list,
939 struct sk_buff_head *head)
940{
941 if (!skb_queue_empty(list)) {
942 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -0700943 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -0700944 __skb_queue_head_init(list);
945 }
946}
947
948/**
949 * skb_queue_splice_tail - join two skb lists, each list being a queue
950 * @list: the new list to add
951 * @head: the place to add it in the first list
952 */
953static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
954 struct sk_buff_head *head)
955{
956 if (!skb_queue_empty(list)) {
957 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -0700958 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -0700959 }
960}
961
962/**
963 * skb_queue_splice_tail - join two skb lists and reinitialise the emptied list
964 * @list: the new list to add
965 * @head: the place to add it in the first list
966 *
967 * Each of the lists is a queue.
968 * The list at @list is reinitialised
969 */
970static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
971 struct sk_buff_head *head)
972{
973 if (!skb_queue_empty(list)) {
974 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -0700975 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -0700976 __skb_queue_head_init(list);
977 }
978}
979
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980/**
Stephen Hemminger300ce172005-10-30 13:47:34 -0800981 * __skb_queue_after - queue a buffer at the list head
982 * @list: list to use
983 * @prev: place after this buffer
984 * @newsk: buffer to queue
985 *
986 * Queue a buffer int the middle of a list. This function takes no locks
987 * and you must therefore hold required locks before calling it.
988 *
989 * A buffer cannot be placed on two lists at the same time.
990 */
991static inline void __skb_queue_after(struct sk_buff_head *list,
992 struct sk_buff *prev,
993 struct sk_buff *newsk)
994{
Gerrit Renkerbf299272008-04-14 00:04:51 -0700995 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -0800996}
997
Gerrit Renker7de6c032008-04-14 00:05:09 -0700998extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
999 struct sk_buff_head *list);
1000
Gerrit Renkerf5572852008-04-14 00:05:28 -07001001static inline void __skb_queue_before(struct sk_buff_head *list,
1002 struct sk_buff *next,
1003 struct sk_buff *newsk)
1004{
1005 __skb_insert(newsk, next->prev, next, list);
1006}
1007
Stephen Hemminger300ce172005-10-30 13:47:34 -08001008/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009 * __skb_queue_head - queue a buffer at the list head
1010 * @list: list to use
1011 * @newsk: buffer to queue
1012 *
1013 * Queue a buffer at the start of a list. This function takes no locks
1014 * and you must therefore hold required locks before calling it.
1015 *
1016 * A buffer cannot be placed on two lists at the same time.
1017 */
1018extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1019static inline void __skb_queue_head(struct sk_buff_head *list,
1020 struct sk_buff *newsk)
1021{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001022 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023}
1024
1025/**
1026 * __skb_queue_tail - queue a buffer at the list tail
1027 * @list: list to use
1028 * @newsk: buffer to queue
1029 *
1030 * Queue a buffer at the end of a list. This function takes no locks
1031 * and you must therefore hold required locks before calling it.
1032 *
1033 * A buffer cannot be placed on two lists at the same time.
1034 */
1035extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1036static inline void __skb_queue_tail(struct sk_buff_head *list,
1037 struct sk_buff *newsk)
1038{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001039 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001040}
1041
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043 * remove sk_buff from list. _Must_ be called atomically, and with
1044 * the list known..
1045 */
David S. Miller8728b832005-08-09 19:25:21 -07001046extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001047static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1048{
1049 struct sk_buff *next, *prev;
1050
1051 list->qlen--;
1052 next = skb->next;
1053 prev = skb->prev;
1054 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001055 next->prev = prev;
1056 prev->next = next;
1057}
1058
Gerrit Renkerf525c062008-04-14 00:04:12 -07001059/**
1060 * __skb_dequeue - remove from the head of the queue
1061 * @list: list to dequeue from
1062 *
1063 * Remove the head of the list. This function does not take any locks
1064 * so must be used with appropriate locks held only. The head item is
1065 * returned or %NULL if the list is empty.
1066 */
1067extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1068static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1069{
1070 struct sk_buff *skb = skb_peek(list);
1071 if (skb)
1072 __skb_unlink(skb, list);
1073 return skb;
1074}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001075
1076/**
1077 * __skb_dequeue_tail - remove from the tail of the queue
1078 * @list: list to dequeue from
1079 *
1080 * Remove the tail of the list. This function does not take any locks
1081 * so must be used with appropriate locks held only. The tail item is
1082 * returned or %NULL if the list is empty.
1083 */
1084extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1085static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1086{
1087 struct sk_buff *skb = skb_peek_tail(list);
1088 if (skb)
1089 __skb_unlink(skb, list);
1090 return skb;
1091}
1092
1093
1094static inline int skb_is_nonlinear(const struct sk_buff *skb)
1095{
1096 return skb->data_len;
1097}
1098
1099static inline unsigned int skb_headlen(const struct sk_buff *skb)
1100{
1101 return skb->len - skb->data_len;
1102}
1103
1104static inline int skb_pagelen(const struct sk_buff *skb)
1105{
1106 int i, len = 0;
1107
1108 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
1109 len += skb_shinfo(skb)->frags[i].size;
1110 return len + skb_headlen(skb);
1111}
1112
1113static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1114 struct page *page, int off, int size)
1115{
1116 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1117
1118 frag->page = page;
1119 frag->page_offset = off;
1120 frag->size = size;
1121 skb_shinfo(skb)->nr_frags = i + 1;
1122}
1123
Peter Zijlstra654bed12008-10-07 14:22:33 -07001124extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
1125 int off, int size);
1126
Linus Torvalds1da177e2005-04-16 15:20:36 -07001127#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001128#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001129#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1130
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001131#ifdef NET_SKBUFF_DATA_USES_OFFSET
1132static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1133{
1134 return skb->head + skb->tail;
1135}
1136
1137static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1138{
1139 skb->tail = skb->data - skb->head;
1140}
1141
1142static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1143{
1144 skb_reset_tail_pointer(skb);
1145 skb->tail += offset;
1146}
1147#else /* NET_SKBUFF_DATA_USES_OFFSET */
1148static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1149{
1150 return skb->tail;
1151}
1152
1153static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1154{
1155 skb->tail = skb->data;
1156}
1157
1158static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1159{
1160 skb->tail = skb->data + offset;
1161}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001162
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001163#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1164
Linus Torvalds1da177e2005-04-16 15:20:36 -07001165/*
1166 * Add data to an sk_buff
1167 */
Ilpo Järvinen0dde3e12008-03-27 17:43:41 -07001168extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001169static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1170{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001171 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172 SKB_LINEAR_ASSERT(skb);
1173 skb->tail += len;
1174 skb->len += len;
1175 return tmp;
1176}
1177
Ilpo Järvinenc2aa2702008-03-27 17:52:40 -07001178extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001179static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1180{
1181 skb->data -= len;
1182 skb->len += len;
1183 return skb->data;
1184}
1185
Ilpo Järvinen6be8ac22008-03-27 17:47:24 -07001186extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001187static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1188{
1189 skb->len -= len;
1190 BUG_ON(skb->len < skb->data_len);
1191 return skb->data += len;
1192}
1193
David S. Miller47d29642010-05-02 02:21:44 -07001194static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1195{
1196 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1197}
1198
Linus Torvalds1da177e2005-04-16 15:20:36 -07001199extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1200
1201static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1202{
1203 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001204 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001205 return NULL;
1206 skb->len -= len;
1207 return skb->data += len;
1208}
1209
1210static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1211{
1212 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1213}
1214
1215static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1216{
1217 if (likely(len <= skb_headlen(skb)))
1218 return 1;
1219 if (unlikely(len > skb->len))
1220 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001221 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001222}
1223
1224/**
1225 * skb_headroom - bytes at buffer head
1226 * @skb: buffer to check
1227 *
1228 * Return the number of bytes of free space at the head of an &sk_buff.
1229 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001230static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001231{
1232 return skb->data - skb->head;
1233}
1234
1235/**
1236 * skb_tailroom - bytes at buffer end
1237 * @skb: buffer to check
1238 *
1239 * Return the number of bytes of free space at the tail of an sk_buff
1240 */
1241static inline int skb_tailroom(const struct sk_buff *skb)
1242{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001243 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001244}
1245
1246/**
1247 * skb_reserve - adjust headroom
1248 * @skb: buffer to alter
1249 * @len: bytes to move
1250 *
1251 * Increase the headroom of an empty &sk_buff by reducing the tail
1252 * room. This is only allowed for an empty buffer.
1253 */
David S. Miller8243126c2006-01-17 02:54:21 -08001254static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255{
1256 skb->data += len;
1257 skb->tail += len;
1258}
1259
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001260#ifdef NET_SKBUFF_DATA_USES_OFFSET
1261static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1262{
1263 return skb->head + skb->transport_header;
1264}
1265
1266static inline void skb_reset_transport_header(struct sk_buff *skb)
1267{
1268 skb->transport_header = skb->data - skb->head;
1269}
1270
1271static inline void skb_set_transport_header(struct sk_buff *skb,
1272 const int offset)
1273{
1274 skb_reset_transport_header(skb);
1275 skb->transport_header += offset;
1276}
1277
1278static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1279{
1280 return skb->head + skb->network_header;
1281}
1282
1283static inline void skb_reset_network_header(struct sk_buff *skb)
1284{
1285 skb->network_header = skb->data - skb->head;
1286}
1287
1288static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1289{
1290 skb_reset_network_header(skb);
1291 skb->network_header += offset;
1292}
1293
1294static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1295{
1296 return skb->head + skb->mac_header;
1297}
1298
1299static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1300{
1301 return skb->mac_header != ~0U;
1302}
1303
1304static inline void skb_reset_mac_header(struct sk_buff *skb)
1305{
1306 skb->mac_header = skb->data - skb->head;
1307}
1308
1309static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1310{
1311 skb_reset_mac_header(skb);
1312 skb->mac_header += offset;
1313}
1314
1315#else /* NET_SKBUFF_DATA_USES_OFFSET */
1316
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001317static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1318{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001319 return skb->transport_header;
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001320}
1321
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001322static inline void skb_reset_transport_header(struct sk_buff *skb)
1323{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001324 skb->transport_header = skb->data;
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001325}
1326
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001327static inline void skb_set_transport_header(struct sk_buff *skb,
1328 const int offset)
1329{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001330 skb->transport_header = skb->data + offset;
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001331}
1332
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001333static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1334{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001335 return skb->network_header;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001336}
1337
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001338static inline void skb_reset_network_header(struct sk_buff *skb)
1339{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001340 skb->network_header = skb->data;
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001341}
1342
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001343static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1344{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001345 skb->network_header = skb->data + offset;
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001346}
1347
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001348static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1349{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001350 return skb->mac_header;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001351}
1352
1353static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1354{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001355 return skb->mac_header != NULL;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001356}
1357
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001358static inline void skb_reset_mac_header(struct sk_buff *skb)
1359{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001360 skb->mac_header = skb->data;
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001361}
1362
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001363static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1364{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001365 skb->mac_header = skb->data + offset;
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001366}
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001367#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1368
1369static inline int skb_transport_offset(const struct sk_buff *skb)
1370{
1371 return skb_transport_header(skb) - skb->data;
1372}
1373
1374static inline u32 skb_network_header_len(const struct sk_buff *skb)
1375{
1376 return skb->transport_header - skb->network_header;
1377}
1378
1379static inline int skb_network_offset(const struct sk_buff *skb)
1380{
1381 return skb_network_header(skb) - skb->data;
1382}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001383
Changli Gaof9599ce2010-08-04 04:43:44 +00001384static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1385{
1386 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1387}
1388
Linus Torvalds1da177e2005-04-16 15:20:36 -07001389/*
1390 * CPUs often take a performance hit when accessing unaligned memory
1391 * locations. The actual performance hit varies, it can be small if the
1392 * hardware handles it or large if we have to take an exception and fix it
1393 * in software.
1394 *
1395 * Since an ethernet header is 14 bytes network drivers often end up with
1396 * the IP header at an unaligned offset. The IP header can be aligned by
1397 * shifting the start of the packet by 2 bytes. Drivers should do this
1398 * with:
1399 *
Tobias Klauser8660c122009-07-13 22:48:16 +00001400 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001401 *
1402 * The downside to this alignment of the IP header is that the DMA is now
1403 * unaligned. On some architectures the cost of an unaligned DMA is high
1404 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00001405 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001406 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1407 * to be overridden.
1408 */
1409#ifndef NET_IP_ALIGN
1410#define NET_IP_ALIGN 2
1411#endif
1412
Anton Blanchard025be812006-03-31 02:27:06 -08001413/*
1414 * The networking layer reserves some headroom in skb data (via
1415 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1416 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08001417 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08001418 *
1419 * Unfortunately this headroom changes the DMA alignment of the resulting
1420 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1421 * on some architectures. An architecture can override this value,
1422 * perhaps setting it to a cacheline in size (since that will maintain
1423 * cacheline alignment of the DMA). It must be a power of 2.
1424 *
David S. Millerd6301d32009-02-08 19:24:13 -08001425 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08001426 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07001427 *
1428 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1429 * to reduce average number of cache lines per packet.
1430 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07001431 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08001432 */
1433#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07001434#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08001435#endif
1436
Herbert Xu3cc0e872006-06-09 16:13:38 -07001437extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001438
1439static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1440{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001441 if (unlikely(skb->data_len)) {
1442 WARN_ON(1);
1443 return;
1444 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001445 skb->len = len;
1446 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001447}
1448
Ilpo Järvinen419ae742008-03-27 17:54:01 -07001449extern void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001450
1451static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1452{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001453 if (skb->data_len)
1454 return ___pskb_trim(skb, len);
1455 __skb_trim(skb, len);
1456 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001457}
1458
1459static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1460{
1461 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1462}
1463
1464/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07001465 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1466 * @skb: buffer to alter
1467 * @len: new length
1468 *
1469 * This is identical to pskb_trim except that the caller knows that
1470 * the skb is not cloned so we should never get an error due to out-
1471 * of-memory.
1472 */
1473static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1474{
1475 int err = pskb_trim(skb, len);
1476 BUG_ON(err);
1477}
1478
1479/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001480 * skb_orphan - orphan a buffer
1481 * @skb: buffer to orphan
1482 *
1483 * If a buffer currently has an owner then we call the owner's
1484 * destructor function and make the @skb unowned. The buffer continues
1485 * to exist but is no longer charged to its former owner.
1486 */
1487static inline void skb_orphan(struct sk_buff *skb)
1488{
1489 if (skb->destructor)
1490 skb->destructor(skb);
1491 skb->destructor = NULL;
1492 skb->sk = NULL;
1493}
1494
1495/**
1496 * __skb_queue_purge - empty a list
1497 * @list: list to empty
1498 *
1499 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1500 * the list and one reference dropped. This function does not take the
1501 * list lock and the caller must hold the relevant locks to use it.
1502 */
1503extern void skb_queue_purge(struct sk_buff_head *list);
1504static inline void __skb_queue_purge(struct sk_buff_head *list)
1505{
1506 struct sk_buff *skb;
1507 while ((skb = __skb_dequeue(list)) != NULL)
1508 kfree_skb(skb);
1509}
1510
1511/**
Christoph Hellwigb4e54de2006-07-24 15:31:14 -07001512 * __dev_alloc_skb - allocate an skbuff for receiving
Linus Torvalds1da177e2005-04-16 15:20:36 -07001513 * @length: length to allocate
1514 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1515 *
1516 * Allocate a new &sk_buff and assign it a usage count of one. The
1517 * buffer has unspecified headroom built in. Users should allocate
1518 * the headroom they think they need without accounting for the
1519 * built in space. The built in space is used for optimisations.
1520 *
Christoph Hellwig766ea8c2006-08-07 15:49:53 -07001521 * %NULL is returned if there is no free memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001522 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
Al Virodd0fc662005-10-07 07:46:04 +01001524 gfp_t gfp_mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001525{
Anton Blanchard025be812006-03-31 02:27:06 -08001526 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001527 if (likely(skb))
Anton Blanchard025be812006-03-31 02:27:06 -08001528 skb_reserve(skb, NET_SKB_PAD);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001529 return skb;
1530}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531
Ilpo Järvinenf58518e2008-03-27 17:51:31 -07001532extern struct sk_buff *dev_alloc_skb(unsigned int length);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001533
Christoph Hellwig8af27452006-07-31 22:35:23 -07001534extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1535 unsigned int length, gfp_t gfp_mask);
1536
1537/**
1538 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1539 * @dev: network device to receive on
1540 * @length: length to allocate
1541 *
1542 * Allocate a new &sk_buff and assign it a usage count of one. The
1543 * buffer has unspecified headroom built in. Users should allocate
1544 * the headroom they think they need without accounting for the
1545 * built in space. The built in space is used for optimisations.
1546 *
1547 * %NULL is returned if there is no free memory. Although this function
1548 * allocates memory it can be called from an interrupt.
1549 */
1550static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1551 unsigned int length)
1552{
1553 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1554}
1555
Eric Dumazet61321bb2009-10-07 17:11:23 +00001556static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1557 unsigned int length)
1558{
1559 struct sk_buff *skb = netdev_alloc_skb(dev, length + NET_IP_ALIGN);
1560
1561 if (NET_IP_ALIGN && skb)
1562 skb_reserve(skb, NET_IP_ALIGN);
1563 return skb;
1564}
1565
Eric Dumazet564824b2010-10-11 19:05:25 +00001566/**
1567 * __netdev_alloc_page - allocate a page for ps-rx on a specific device
1568 * @dev: network device to receive on
1569 * @gfp_mask: alloc_pages_node mask
1570 *
1571 * Allocate a new page. dev currently unused.
1572 *
1573 * %NULL is returned if there is no free memory.
1574 */
1575static inline struct page *__netdev_alloc_page(struct net_device *dev, gfp_t gfp_mask)
1576{
1577 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, 0);
1578}
Peter Zijlstra654bed12008-10-07 14:22:33 -07001579
1580/**
1581 * netdev_alloc_page - allocate a page for ps-rx on a specific device
1582 * @dev: network device to receive on
1583 *
Eric Dumazet564824b2010-10-11 19:05:25 +00001584 * Allocate a new page. dev currently unused.
Peter Zijlstra654bed12008-10-07 14:22:33 -07001585 *
1586 * %NULL is returned if there is no free memory.
1587 */
1588static inline struct page *netdev_alloc_page(struct net_device *dev)
1589{
1590 return __netdev_alloc_page(dev, GFP_ATOMIC);
1591}
1592
1593static inline void netdev_free_page(struct net_device *dev, struct page *page)
1594{
1595 __free_page(page);
1596}
1597
Linus Torvalds1da177e2005-04-16 15:20:36 -07001598/**
Patrick McHardy334a8132007-06-25 04:35:20 -07001599 * skb_clone_writable - is the header of a clone writable
1600 * @skb: buffer to check
1601 * @len: length up to which to write
1602 *
1603 * Returns true if modifying the header part of the cloned buffer
1604 * does not requires the data to be copied.
1605 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001606static inline int skb_clone_writable(struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07001607{
1608 return !skb_header_cloned(skb) &&
1609 skb_headroom(skb) + len <= skb->hdr_len;
1610}
1611
Herbert Xud9cc2042007-09-16 16:21:16 -07001612static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
1613 int cloned)
1614{
1615 int delta = 0;
1616
1617 if (headroom < NET_SKB_PAD)
1618 headroom = NET_SKB_PAD;
1619 if (headroom > skb_headroom(skb))
1620 delta = headroom - skb_headroom(skb);
1621
1622 if (delta || cloned)
1623 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
1624 GFP_ATOMIC);
1625 return 0;
1626}
1627
Patrick McHardy334a8132007-06-25 04:35:20 -07001628/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001629 * skb_cow - copy header of skb when it is required
1630 * @skb: buffer to cow
1631 * @headroom: needed headroom
1632 *
1633 * If the skb passed lacks sufficient headroom or its data part
1634 * is shared, data is reallocated. If reallocation fails, an error
1635 * is returned and original skb is not changed.
1636 *
1637 * The result is skb with writable area skb->head...skb->tail
1638 * and at least @headroom of space at head.
1639 */
1640static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1641{
Herbert Xud9cc2042007-09-16 16:21:16 -07001642 return __skb_cow(skb, headroom, skb_cloned(skb));
1643}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001644
Herbert Xud9cc2042007-09-16 16:21:16 -07001645/**
1646 * skb_cow_head - skb_cow but only making the head writable
1647 * @skb: buffer to cow
1648 * @headroom: needed headroom
1649 *
1650 * This function is identical to skb_cow except that we replace the
1651 * skb_cloned check by skb_header_cloned. It should be used when
1652 * you only need to push on some header and do not need to modify
1653 * the data.
1654 */
1655static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
1656{
1657 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001658}
1659
1660/**
1661 * skb_padto - pad an skbuff up to a minimal size
1662 * @skb: buffer to pad
1663 * @len: minimal length
1664 *
1665 * Pads up a buffer to ensure the trailing bytes exist and are
1666 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07001667 * is untouched. Otherwise it is extended. Returns zero on
1668 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669 */
1670
Herbert Xu5b057c62006-06-23 02:06:41 -07001671static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001672{
1673 unsigned int size = skb->len;
1674 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07001675 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001676 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677}
1678
1679static inline int skb_add_data(struct sk_buff *skb,
1680 char __user *from, int copy)
1681{
1682 const int off = skb->len;
1683
1684 if (skb->ip_summed == CHECKSUM_NONE) {
1685 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08001686 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687 copy, 0, &err);
1688 if (!err) {
1689 skb->csum = csum_block_add(skb->csum, csum, off);
1690 return 0;
1691 }
1692 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1693 return 0;
1694
1695 __skb_trim(skb, off);
1696 return -EFAULT;
1697}
1698
1699static inline int skb_can_coalesce(struct sk_buff *skb, int i,
1700 struct page *page, int off)
1701{
1702 if (i) {
1703 struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
1704
1705 return page == frag->page &&
1706 off == frag->page_offset + frag->size;
1707 }
1708 return 0;
1709}
1710
Herbert Xu364c6ba2006-06-09 16:10:40 -07001711static inline int __skb_linearize(struct sk_buff *skb)
1712{
1713 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1714}
1715
Linus Torvalds1da177e2005-04-16 15:20:36 -07001716/**
1717 * skb_linearize - convert paged skb to linear one
1718 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07001719 *
1720 * If there is no free memory -ENOMEM is returned, otherwise zero
1721 * is returned and the old skb data released.
1722 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07001723static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724{
Herbert Xu364c6ba2006-06-09 16:10:40 -07001725 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1726}
1727
1728/**
1729 * skb_linearize_cow - make sure skb is linear and writable
1730 * @skb: buffer to process
1731 *
1732 * If there is no free memory -ENOMEM is returned, otherwise zero
1733 * is returned and the old skb data released.
1734 */
1735static inline int skb_linearize_cow(struct sk_buff *skb)
1736{
1737 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1738 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001739}
1740
1741/**
1742 * skb_postpull_rcsum - update checksum for received skb after pull
1743 * @skb: buffer to update
1744 * @start: start of data before pull
1745 * @len: length of data pulled
1746 *
1747 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07001748 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1749 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001750 */
1751
1752static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08001753 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001754{
Patrick McHardy84fa7932006-08-29 16:44:56 -07001755 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001756 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1757}
1758
Herbert Xucbb042f2006-03-20 22:43:56 -08001759unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1760
Linus Torvalds1da177e2005-04-16 15:20:36 -07001761/**
1762 * pskb_trim_rcsum - trim received skb and update checksum
1763 * @skb: buffer to trim
1764 * @len: new length
1765 *
1766 * This is exactly the same as pskb_trim except that it ensures the
1767 * checksum of received packets are still valid after the operation.
1768 */
1769
1770static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
1771{
Stephen Hemminger0e4e4222005-09-08 12:32:03 -07001772 if (likely(len >= skb->len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001773 return 0;
Patrick McHardy84fa7932006-08-29 16:44:56 -07001774 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001775 skb->ip_summed = CHECKSUM_NONE;
1776 return __pskb_trim(skb, len);
1777}
1778
Linus Torvalds1da177e2005-04-16 15:20:36 -07001779#define skb_queue_walk(queue, skb) \
1780 for (skb = (queue)->next; \
1781 prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1782 skb = skb->next)
1783
James Chapman46f89142007-04-30 00:07:31 -07001784#define skb_queue_walk_safe(queue, skb, tmp) \
1785 for (skb = (queue)->next, tmp = skb->next; \
1786 skb != (struct sk_buff *)(queue); \
1787 skb = tmp, tmp = skb->next)
1788
David S. Miller1164f522008-09-23 00:49:44 -07001789#define skb_queue_walk_from(queue, skb) \
1790 for (; prefetch(skb->next), (skb != (struct sk_buff *)(queue)); \
1791 skb = skb->next)
1792
1793#define skb_queue_walk_from_safe(queue, skb, tmp) \
1794 for (tmp = skb->next; \
1795 skb != (struct sk_buff *)(queue); \
1796 skb = tmp, tmp = skb->next)
1797
Stephen Hemminger300ce172005-10-30 13:47:34 -08001798#define skb_queue_reverse_walk(queue, skb) \
1799 for (skb = (queue)->prev; \
1800 prefetch(skb->prev), (skb != (struct sk_buff *)(queue)); \
1801 skb = skb->prev)
1802
Linus Torvalds1da177e2005-04-16 15:20:36 -07001803
David S. Miller21dc3302010-08-23 00:13:46 -07001804static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07001805{
1806 return skb_shinfo(skb)->frag_list != NULL;
1807}
1808
1809static inline void skb_frag_list_init(struct sk_buff *skb)
1810{
1811 skb_shinfo(skb)->frag_list = NULL;
1812}
1813
1814static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
1815{
1816 frag->next = skb_shinfo(skb)->frag_list;
1817 skb_shinfo(skb)->frag_list = frag;
1818}
1819
1820#define skb_walk_frags(skb, iter) \
1821 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
1822
Herbert Xua59322b2007-12-05 01:53:40 -08001823extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
1824 int *peeked, int *err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001825extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
1826 int noblock, int *err);
1827extern unsigned int datagram_poll(struct file *file, struct socket *sock,
1828 struct poll_table_struct *wait);
1829extern int skb_copy_datagram_iovec(const struct sk_buff *from,
1830 int offset, struct iovec *to,
1831 int size);
Herbert Xufb286bb2005-11-10 13:01:24 -08001832extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001833 int hlen,
1834 struct iovec *iov);
Rusty Russelldb543c12008-08-15 15:13:53 -07001835extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
1836 int offset,
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +00001837 const struct iovec *from,
1838 int from_offset,
Rusty Russelldb543c12008-08-15 15:13:53 -07001839 int len);
Michael S. Tsirkin0a1ec072009-04-20 01:25:46 +00001840extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
1841 int offset,
1842 const struct iovec *to,
1843 int to_offset,
1844 int size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001845extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
Eric Dumazet9d410c72009-10-30 05:03:53 +00001846extern void skb_free_datagram_locked(struct sock *sk,
1847 struct sk_buff *skb);
Herbert Xu27ab2562007-12-05 01:51:58 -08001848extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
Herbert Xu3305b802005-12-13 23:16:37 -08001849 unsigned int flags);
Al Viro2bbbc862006-11-14 21:37:14 -08001850extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
1851 int len, __wsum csum);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001852extern int skb_copy_bits(const struct sk_buff *skb, int offset,
1853 void *to, int len);
Stephen Hemminger0c6fcc82007-04-20 16:40:01 -07001854extern int skb_store_bits(struct sk_buff *skb, int offset,
1855 const void *from, int len);
Al Viro81d77662006-11-14 21:37:33 -08001856extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07001857 int offset, u8 *to, int len,
Al Viro81d77662006-11-14 21:37:33 -08001858 __wsum csum);
Jens Axboe9c55e012007-11-06 23:30:13 -08001859extern int skb_splice_bits(struct sk_buff *skb,
1860 unsigned int offset,
1861 struct pipe_inode_info *pipe,
1862 unsigned int len,
1863 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001864extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
1865extern void skb_split(struct sk_buff *skb,
1866 struct sk_buff *skb1, const u32 len);
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001867extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
1868 int shiftlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001869
Herbert Xu576a30e2006-06-27 13:22:38 -07001870extern struct sk_buff *skb_segment(struct sk_buff *skb, int features);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03001871
Linus Torvalds1da177e2005-04-16 15:20:36 -07001872static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
1873 int len, void *buffer)
1874{
1875 int hlen = skb_headlen(skb);
1876
Patrick McHardy55820ee2005-07-05 14:08:10 -07001877 if (hlen - offset >= len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001878 return skb->data + offset;
1879
1880 if (skb_copy_bits(skb, offset, buffer, len) < 0)
1881 return NULL;
1882
1883 return buffer;
1884}
1885
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03001886static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
1887 void *to,
1888 const unsigned int len)
1889{
1890 memcpy(to, skb->data, len);
1891}
1892
1893static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
1894 const int offset, void *to,
1895 const unsigned int len)
1896{
1897 memcpy(to, skb->data + offset, len);
1898}
1899
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03001900static inline void skb_copy_to_linear_data(struct sk_buff *skb,
1901 const void *from,
1902 const unsigned int len)
1903{
1904 memcpy(skb->data, from, len);
1905}
1906
1907static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
1908 const int offset,
1909 const void *from,
1910 const unsigned int len)
1911{
1912 memcpy(skb->data + offset, from, len);
1913}
1914
Linus Torvalds1da177e2005-04-16 15:20:36 -07001915extern void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001916
Patrick Ohlyac45f602009-02-12 05:03:37 +00001917static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
1918{
1919 return skb->tstamp;
1920}
1921
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001922/**
1923 * skb_get_timestamp - get timestamp from a skb
1924 * @skb: skb to get stamp from
1925 * @stamp: pointer to struct timeval to store stamp in
1926 *
1927 * Timestamps are stored in the skb as offsets to a base timestamp.
1928 * This function converts the offset back to a struct timeval and stores
1929 * it in stamp.
1930 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00001931static inline void skb_get_timestamp(const struct sk_buff *skb,
1932 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001933{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001934 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001935}
1936
Patrick Ohlyac45f602009-02-12 05:03:37 +00001937static inline void skb_get_timestampns(const struct sk_buff *skb,
1938 struct timespec *stamp)
1939{
1940 *stamp = ktime_to_timespec(skb->tstamp);
1941}
1942
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001943static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001944{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07001945 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001946}
1947
Stephen Hemminger164891a2007-04-23 22:26:16 -07001948static inline ktime_t net_timedelta(ktime_t t)
1949{
1950 return ktime_sub(ktime_get_real(), t);
1951}
1952
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07001953static inline ktime_t net_invalid_timestamp(void)
1954{
1955 return ktime_set(0, 0);
1956}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07001957
Richard Cochranc1f19b52010-07-17 08:49:36 +00001958extern void skb_timestamping_init(void);
1959
1960#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
1961
1962extern void skb_clone_tx_timestamp(struct sk_buff *skb);
1963extern bool skb_defer_rx_timestamp(struct sk_buff *skb);
1964
1965#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
1966
1967static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
1968{
1969}
1970
1971static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
1972{
1973 return false;
1974}
1975
1976#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
1977
1978/**
1979 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
1980 *
1981 * @skb: clone of the the original outgoing packet
1982 * @hwtstamps: hardware time stamps
1983 *
1984 */
1985void skb_complete_tx_timestamp(struct sk_buff *skb,
1986 struct skb_shared_hwtstamps *hwtstamps);
1987
Patrick Ohlyac45f602009-02-12 05:03:37 +00001988/**
1989 * skb_tstamp_tx - queue clone of skb with send time stamps
1990 * @orig_skb: the original outgoing packet
1991 * @hwtstamps: hardware time stamps, may be NULL if not available
1992 *
1993 * If the skb has a socket associated, then this function clones the
1994 * skb (thus sharing the actual data and optional structures), stores
1995 * the optional hardware time stamping information (if non NULL) or
1996 * generates a software time stamp (otherwise), then queues the clone
1997 * to the error queue of the socket. Errors are silently ignored.
1998 */
1999extern void skb_tstamp_tx(struct sk_buff *orig_skb,
2000 struct skb_shared_hwtstamps *hwtstamps);
2001
Richard Cochran4507a712010-07-17 08:48:28 +00002002static inline void sw_tx_timestamp(struct sk_buff *skb)
2003{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002004 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2005 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00002006 skb_tstamp_tx(skb, NULL);
2007}
2008
2009/**
2010 * skb_tx_timestamp() - Driver hook for transmit timestamping
2011 *
2012 * Ethernet MAC Drivers should call this function in their hard_xmit()
2013 * function as soon as possible after giving the sk_buff to the MAC
2014 * hardware, but before freeing the sk_buff.
2015 *
2016 * @skb: A socket buffer.
2017 */
2018static inline void skb_tx_timestamp(struct sk_buff *skb)
2019{
Richard Cochranc1f19b52010-07-17 08:49:36 +00002020 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002021 sw_tx_timestamp(skb);
2022}
2023
Herbert Xu759e5d02007-03-25 20:10:56 -07002024extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
Al Virob51655b2006-11-14 21:40:42 -08002025extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002026
Herbert Xu60476372007-04-09 11:59:39 -07002027static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2028{
2029 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2030}
2031
Herbert Xufb286bb2005-11-10 13:01:24 -08002032/**
2033 * skb_checksum_complete - Calculate checksum of an entire packet
2034 * @skb: packet to process
2035 *
2036 * This function calculates the checksum over the entire packet plus
2037 * the value of skb->csum. The latter can be used to supply the
2038 * checksum of a pseudo header as used by TCP/UDP. It returns the
2039 * checksum.
2040 *
2041 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2042 * this function can be used to verify that checksum on received
2043 * packets. In that case the function should return zero if the
2044 * checksum is correct. In particular, this function will return zero
2045 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2046 * hardware has already verified the correctness of the checksum.
2047 */
Al Viro4381ca32007-07-15 21:00:11 +01002048static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002049{
Herbert Xu60476372007-04-09 11:59:39 -07002050 return skb_csum_unnecessary(skb) ?
2051 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002052}
2053
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002054#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002055extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002056static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2057{
2058 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002059 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002060}
2061static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2062{
2063 if (nfct)
2064 atomic_inc(&nfct->use);
2065}
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -08002066static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
2067{
2068 if (skb)
2069 atomic_inc(&skb->users);
2070}
2071static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
2072{
2073 if (skb)
2074 kfree_skb(skb);
2075}
2076#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002077#ifdef CONFIG_BRIDGE_NETFILTER
2078static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2079{
2080 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2081 kfree(nf_bridge);
2082}
2083static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2084{
2085 if (nf_bridge)
2086 atomic_inc(&nf_bridge->use);
2087}
2088#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08002089static inline void nf_reset(struct sk_buff *skb)
2090{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002091#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08002092 nf_conntrack_put(skb->nfct);
2093 skb->nfct = NULL;
Patrick McHardya193a4a2006-03-20 19:23:05 -08002094 nf_conntrack_put_reasm(skb->nfct_reasm);
2095 skb->nfct_reasm = NULL;
2096#endif
2097#ifdef CONFIG_BRIDGE_NETFILTER
2098 nf_bridge_put(skb->nf_bridge);
2099 skb->nf_bridge = NULL;
2100#endif
2101}
2102
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002103/* Note: This doesn't put any conntrack and bridge info in dst. */
2104static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2105{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002106#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002107 dst->nfct = src->nfct;
2108 nf_conntrack_get(src->nfct);
2109 dst->nfctinfo = src->nfctinfo;
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002110 dst->nfct_reasm = src->nfct_reasm;
2111 nf_conntrack_get_reasm(src->nfct_reasm);
2112#endif
2113#ifdef CONFIG_BRIDGE_NETFILTER
2114 dst->nf_bridge = src->nf_bridge;
2115 nf_bridge_get(src->nf_bridge);
2116#endif
2117}
2118
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002119static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2120{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002121#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002122 nf_conntrack_put(dst->nfct);
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002123 nf_conntrack_put_reasm(dst->nfct_reasm);
2124#endif
2125#ifdef CONFIG_BRIDGE_NETFILTER
2126 nf_bridge_put(dst->nf_bridge);
2127#endif
2128 __nf_copy(dst, src);
2129}
2130
James Morris984bc162006-06-09 00:29:17 -07002131#ifdef CONFIG_NETWORK_SECMARK
2132static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2133{
2134 to->secmark = from->secmark;
2135}
2136
2137static inline void skb_init_secmark(struct sk_buff *skb)
2138{
2139 skb->secmark = 0;
2140}
2141#else
2142static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2143{ }
2144
2145static inline void skb_init_secmark(struct sk_buff *skb)
2146{ }
2147#endif
2148
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002149static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2150{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002151 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002152}
2153
Stephen Hemminger92477442009-03-21 13:39:26 -07002154static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002155{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002156 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002157}
2158
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002159static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2160{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002161 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002162}
2163
David S. Millerd5a9e242009-01-27 16:22:11 -08002164static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2165{
2166 skb->queue_mapping = rx_queue + 1;
2167}
2168
Stephen Hemminger92477442009-03-21 13:39:26 -07002169static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002170{
2171 return skb->queue_mapping - 1;
2172}
2173
Stephen Hemminger92477442009-03-21 13:39:26 -07002174static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002175{
Eric Dumazeta02cec22010-09-22 20:43:57 +00002176 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08002177}
2178
Stephen Hemminger92477442009-03-21 13:39:26 -07002179extern u16 skb_tx_hash(const struct net_device *dev,
2180 const struct sk_buff *skb);
2181
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002182#ifdef CONFIG_XFRM
2183static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2184{
2185 return skb->sp;
2186}
2187#else
2188static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2189{
2190 return NULL;
2191}
2192#endif
2193
Herbert Xu89114af2006-07-08 13:34:32 -07002194static inline int skb_is_gso(const struct sk_buff *skb)
2195{
2196 return skb_shinfo(skb)->gso_size;
2197}
2198
Brice Goglineabd7e32007-10-13 12:33:32 +02002199static inline int skb_is_gso_v6(const struct sk_buff *skb)
2200{
2201 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2202}
2203
Ben Hutchings4497b072008-06-19 16:22:28 -07002204extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2205
2206static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2207{
2208 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2209 * wanted then gso_type will be set. */
2210 struct skb_shared_info *shinfo = skb_shinfo(skb);
Alexander Duyckb78462e2010-06-02 12:24:37 +00002211 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2212 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07002213 __skb_warn_lro_forwarding(skb);
2214 return true;
2215 }
2216 return false;
2217}
2218
Herbert Xu35fc92a2007-03-26 23:22:20 -07002219static inline void skb_forward_csum(struct sk_buff *skb)
2220{
2221 /* Unfortunately we don't support this one. Any brave souls? */
2222 if (skb->ip_summed == CHECKSUM_COMPLETE)
2223 skb->ip_summed = CHECKSUM_NONE;
2224}
2225
Eric Dumazetbc8acf22010-09-02 13:07:41 -07002226/**
2227 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2228 * @skb: skb to check
2229 *
2230 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2231 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2232 * use this helper, to document places where we make this assertion.
2233 */
2234static inline void skb_checksum_none_assert(struct sk_buff *skb)
2235{
2236#ifdef DEBUG
2237 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2238#endif
2239}
2240
Rusty Russellf35d9d82008-02-04 23:49:54 -05002241bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002242#endif /* __KERNEL__ */
2243#endif /* _LINUX_SKBUFF_H */