blob: a10e487c0864eeefc0b200048f070b3de3d1fb1f [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * Definitions for the 'struct sk_buff' memory handlers.
3 *
4 * Authors:
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14#ifndef _LINUX_SKBUFF_H
15#define _LINUX_SKBUFF_H
16
Linus Torvalds1da177e2005-04-16 15:20:36 -070017#include <linux/kernel.h>
Vegard Nossumfe55f6d2008-08-30 12:16:35 +020018#include <linux/kmemcheck.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070019#include <linux/compiler.h>
20#include <linux/time.h>
21#include <linux/cache.h>
22
Arun Sharma600634972011-07-26 16:09:06 -070023#include <linux/atomic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070024#include <asm/types.h>
25#include <linux/spinlock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070026#include <linux/net.h>
Thomas Graf3fc7e8a2005-06-23 21:00:17 -070027#include <linux/textsearch.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070028#include <net/checksum.h>
Al Viroa80958f2006-12-04 20:41:19 +000029#include <linux/rcupdate.h>
Chris Leech97fc2f02006-05-23 17:55:33 -070030#include <linux/dmaengine.h>
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -070031#include <linux/hrtimer.h>
Ian Campbell131ea662011-08-19 06:25:00 +000032#include <linux/dma-mapping.h>
Michał Mirosławc8f44af2011-11-15 15:29:55 +000033#include <linux/netdev_features.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070034
Herbert Xu60476372007-04-09 11:59:39 -070035/* Don't change this without changing skb_csum_unnecessary! */
Linus Torvalds1da177e2005-04-16 15:20:36 -070036#define CHECKSUM_NONE 0
Herbert Xu60476372007-04-09 11:59:39 -070037#define CHECKSUM_UNNECESSARY 1
38#define CHECKSUM_COMPLETE 2
39#define CHECKSUM_PARTIAL 3
Linus Torvalds1da177e2005-04-16 15:20:36 -070040
41#define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
42 ~(SMP_CACHE_BYTES - 1))
David S. Millerfc910a22007-03-25 20:27:59 -070043#define SKB_WITH_OVERHEAD(X) \
Herbert Xudeea84b2007-10-21 16:27:46 -070044 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
David S. Millerfc910a22007-03-25 20:27:59 -070045#define SKB_MAX_ORDER(X, ORDER) \
46 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
Linus Torvalds1da177e2005-04-16 15:20:36 -070047#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
48#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
49
Eric Dumazet87fb4b72011-10-13 07:28:54 +000050/* return minimum truesize of one skb containing X bytes of data */
51#define SKB_TRUESIZE(X) ((X) + \
52 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
53 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
54
Linus Torvalds1da177e2005-04-16 15:20:36 -070055/* A. Checksumming of received packets by device.
56 *
57 * NONE: device failed to checksum this packet.
58 * skb->csum is undefined.
59 *
60 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
61 * skb->csum is undefined.
62 * It is bad option, but, unfortunately, many of vendors do this.
63 * Apparently with secret goal to sell you new device, when you
64 * will add new protocol to your host. F.e. IPv6. 8)
65 *
Patrick McHardy84fa7932006-08-29 16:44:56 -070066 * COMPLETE: the most generic way. Device supplied checksum of _all_
Linus Torvalds1da177e2005-04-16 15:20:36 -070067 * the packet as seen by netif_rx in skb->csum.
68 * NOTE: Even if device supports only some protocols, but
Patrick McHardy84fa7932006-08-29 16:44:56 -070069 * is able to produce some skb->csum, it MUST use COMPLETE,
Linus Torvalds1da177e2005-04-16 15:20:36 -070070 * not UNNECESSARY.
71 *
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070072 * PARTIAL: identical to the case for output below. This may occur
73 * on a packet received directly from another Linux OS, e.g.,
74 * a virtualised Linux kernel on the same host. The packet can
75 * be treated in the same way as UNNECESSARY except that on
76 * output (i.e., forwarding) the checksum must be filled in
77 * by the OS or the hardware.
78 *
Linus Torvalds1da177e2005-04-16 15:20:36 -070079 * B. Checksumming on output.
80 *
81 * NONE: skb is checksummed by protocol or csum is not required.
82 *
Patrick McHardy84fa7932006-08-29 16:44:56 -070083 * PARTIAL: device is required to csum packet as seen by hard_start_xmit
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070084 * from skb->csum_start to the end and to record the checksum
85 * at skb->csum_start + skb->csum_offset.
Linus Torvalds1da177e2005-04-16 15:20:36 -070086 *
87 * Device must show its capabilities in dev->features, set
88 * at device setup time.
89 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
90 * everything.
91 * NETIF_F_NO_CSUM - loopback or reliable single hop media.
92 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
93 * TCP/UDP over IPv4. Sigh. Vendors like this
94 * way by an unknown reason. Though, see comment above
95 * about CHECKSUM_UNNECESSARY. 8)
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070096 * NETIF_F_IPV6_CSUM about as dumb as the last one but does IPv6 instead.
Linus Torvalds1da177e2005-04-16 15:20:36 -070097 *
98 * Any questions? No questions, good. --ANK
99 */
100
Linus Torvalds1da177e2005-04-16 15:20:36 -0700101struct net_device;
David Howells716ea3a2007-04-02 20:19:53 -0700102struct scatterlist;
Jens Axboe9c55e012007-11-06 23:30:13 -0800103struct pipe_inode_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700104
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700105#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106struct nf_conntrack {
107 atomic_t use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108};
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700109#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700110
111#ifdef CONFIG_BRIDGE_NETFILTER
112struct nf_bridge_info {
113 atomic_t use;
114 struct net_device *physindev;
115 struct net_device *physoutdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116 unsigned int mask;
117 unsigned long data[32 / sizeof(unsigned long)];
118};
119#endif
120
Linus Torvalds1da177e2005-04-16 15:20:36 -0700121struct sk_buff_head {
122 /* These two members must be first. */
123 struct sk_buff *next;
124 struct sk_buff *prev;
125
126 __u32 qlen;
127 spinlock_t lock;
128};
129
130struct sk_buff;
131
Anton Blancharda715dea2011-03-27 14:57:26 +0000132/* To allow 64K frame to be packed as single skb without frag_list. Since
133 * GRO uses frags we allocate at least 16 regardless of page size.
134 */
135#if (65536/PAGE_SIZE + 2) < 16
David S. Millereec00952011-03-29 23:34:08 -0700136#define MAX_SKB_FRAGS 16UL
Anton Blancharda715dea2011-03-27 14:57:26 +0000137#else
Linus Torvalds1da177e2005-04-16 15:20:36 -0700138#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 2)
Anton Blancharda715dea2011-03-27 14:57:26 +0000139#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700140
141typedef struct skb_frag_struct skb_frag_t;
142
143struct skb_frag_struct {
Ian Campbella8605c62011-10-19 23:01:49 +0000144 struct {
145 struct page *p;
146 } page;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000147#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
David S. Millera309bb02007-07-30 18:47:03 -0700148 __u32 page_offset;
149 __u32 size;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000150#else
151 __u16 page_offset;
152 __u16 size;
153#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700154};
155
Eric Dumazet9e903e02011-10-18 21:00:24 +0000156static inline unsigned int skb_frag_size(const skb_frag_t *frag)
157{
158 return frag->size;
159}
160
161static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
162{
163 frag->size = size;
164}
165
166static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
167{
168 frag->size += delta;
169}
170
171static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
172{
173 frag->size -= delta;
174}
175
Patrick Ohlyac45f602009-02-12 05:03:37 +0000176#define HAVE_HW_TIME_STAMP
177
178/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800179 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000180 * @hwtstamp: hardware time stamp transformed into duration
181 * since arbitrary point in time
182 * @syststamp: hwtstamp transformed to system time base
183 *
184 * Software time stamps generated by ktime_get_real() are stored in
185 * skb->tstamp. The relation between the different kinds of time
186 * stamps is as follows:
187 *
188 * syststamp and tstamp can be compared against each other in
189 * arbitrary combinations. The accuracy of a
190 * syststamp/tstamp/"syststamp from other device" comparison is
191 * limited by the accuracy of the transformation into system time
192 * base. This depends on the device driver and its underlying
193 * hardware.
194 *
195 * hwtstamps can only be compared against other hwtstamps from
196 * the same device.
197 *
198 * This structure is attached to packets as part of the
199 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
200 */
201struct skb_shared_hwtstamps {
202 ktime_t hwtstamp;
203 ktime_t syststamp;
204};
205
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000206/* Definitions for tx_flags in struct skb_shared_info */
207enum {
208 /* generate hardware time stamp */
209 SKBTX_HW_TSTAMP = 1 << 0,
210
211 /* generate software time stamp */
212 SKBTX_SW_TSTAMP = 1 << 1,
213
214 /* device driver is going to provide hardware time stamp */
215 SKBTX_IN_PROGRESS = 1 << 2,
216
217 /* ensure the originating sk reference is available on driver level */
218 SKBTX_DRV_NEEDS_SK_REF = 1 << 3,
Shirley Maa6686f22011-07-06 12:22:12 +0000219
220 /* device driver supports TX zero-copy buffers */
221 SKBTX_DEV_ZEROCOPY = 1 << 4,
222};
223
224/*
225 * The callback notifies userspace to release buffers when skb DMA is done in
226 * lower device, the skb last reference should be 0 when calling this.
227 * The desc is used to track userspace buffer index.
228 */
229struct ubuf_info {
230 void (*callback)(void *);
231 void *arg;
232 unsigned long desc;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000233};
234
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235/* This data is invariant across clones and lives at
236 * the end of the header data, ie. at skb->end.
237 */
238struct skb_shared_info {
Benjamin LaHaise4947d3e2006-01-03 14:06:50 -0800239 unsigned short nr_frags;
Herbert Xu79671682006-06-22 02:40:14 -0700240 unsigned short gso_size;
241 /* Warning: this field is not always filled in (UFO)! */
242 unsigned short gso_segs;
243 unsigned short gso_type;
Al Viroae08e1f2006-11-08 00:27:11 -0800244 __be32 ip6_frag_id;
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000245 __u8 tx_flags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000247 struct skb_shared_hwtstamps hwtstamps;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700248
249 /*
250 * Warning : all fields before dataref are cleared in __alloc_skb()
251 */
252 atomic_t dataref;
253
Johann Baudy69e3c752009-05-18 22:11:22 -0700254 /* Intermediate layers must ensure that destructor_arg
255 * remains valid until skb destructor */
256 void * destructor_arg;
Shirley Maa6686f22011-07-06 12:22:12 +0000257
Eric Dumazetfed66382010-07-22 19:09:08 +0000258 /* must be last field, see pskb_expand_head() */
259 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260};
261
262/* We divide dataref into two halves. The higher 16 bits hold references
263 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700264 * the entire skb->data. A clone of a headerless skb holds the length of
265 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266 *
267 * All users must obey the rule that the skb->data reference count must be
268 * greater than or equal to the payload reference count.
269 *
270 * Holding a reference to the payload part means that the user does not
271 * care about modifications to the header part of skb->data.
272 */
273#define SKB_DATAREF_SHIFT 16
274#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
275
David S. Millerd179cd12005-08-17 14:57:30 -0700276
277enum {
278 SKB_FCLONE_UNAVAILABLE,
279 SKB_FCLONE_ORIG,
280 SKB_FCLONE_CLONE,
281};
282
Herbert Xu79671682006-06-22 02:40:14 -0700283enum {
284 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700285 SKB_GSO_UDP = 1 << 1,
Herbert Xu576a30e2006-06-27 13:22:38 -0700286
287 /* This indicates the skb is from an untrusted source. */
288 SKB_GSO_DODGY = 1 << 2,
Michael Chanb0da85372006-06-29 12:30:00 -0700289
290 /* This indicates the tcp segment has CWR set. */
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700291 SKB_GSO_TCP_ECN = 1 << 3,
292
293 SKB_GSO_TCPV6 = 1 << 4,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800294
295 SKB_GSO_FCOE = 1 << 5,
Herbert Xu79671682006-06-22 02:40:14 -0700296};
297
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700298#if BITS_PER_LONG > 32
299#define NET_SKBUFF_DATA_USES_OFFSET 1
300#endif
301
302#ifdef NET_SKBUFF_DATA_USES_OFFSET
303typedef unsigned int sk_buff_data_t;
304#else
305typedef unsigned char *sk_buff_data_t;
306#endif
307
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100308#if defined(CONFIG_NF_DEFRAG_IPV4) || defined(CONFIG_NF_DEFRAG_IPV4_MODULE) || \
309 defined(CONFIG_NF_DEFRAG_IPV6) || defined(CONFIG_NF_DEFRAG_IPV6_MODULE)
310#define NET_SKBUFF_NF_DEFRAG_NEEDED 1
311#endif
312
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313/**
314 * struct sk_buff - socket buffer
315 * @next: Next buffer in list
316 * @prev: Previous buffer in list
Herbert Xu325ed822005-10-03 13:57:23 -0700317 * @tstamp: Time we arrived
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700318 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700319 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700320 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000321 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700322 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323 * @len: Length of actual data
324 * @data_len: Data length
325 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700326 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700327 * @csum: Checksum (must include start/offset pair)
328 * @csum_start: Offset from skb->head where checksumming should start
329 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700330 * @priority: Packet queueing priority
Martin Waitz67be2dd2005-05-01 08:59:26 -0700331 * @local_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700332 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700333 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334 * @nohdr: Payload reference only, must not modify header
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700335 * @nfctinfo: Relationship of this skb to the connection
Linus Torvalds1da177e2005-04-16 15:20:36 -0700336 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700337 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700338 * @ipvs_property: skbuff is owned by ipvs
Randy Dunlap31729362008-02-18 20:52:13 -0800339 * @peeked: this packet has been seen already, so stats have been
340 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700341 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700342 * @protocol: Packet protocol from driver
343 * @destructor: Destruct function
344 * @nfct: Associated connection, if any
Randy Dunlap461ddf32005-11-20 21:25:15 -0800345 * @nfct_reasm: netfilter conntrack re-assembly pointer
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800347 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348 * @tc_index: Traffic control index
349 * @tc_verd: traffic control verdict
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700350 * @rxhash: the packet hash computed on receive
351 * @queue_mapping: Queue mapping for multiqueue devices
Randy Dunlap553a5672008-04-20 10:51:01 -0700352 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700353 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Changli Gao4ca24622011-08-19 07:26:44 -0700354 * @l4_rxhash: indicate rxhash is a canonical 4-tuple hash over transport
355 * ports.
Randy Dunlapf4b8ea72006-06-22 16:00:11 -0700356 * @dma_cookie: a cookie to one of several possible DMA operations
357 * done by skb DMA functions
James Morris984bc162006-06-09 00:29:17 -0700358 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700359 * @mark: Generic packet mark
360 * @dropcount: total number of sk_receive_queue overflows
Patrick McHardy6aa895b02008-07-14 22:49:06 -0700361 * @vlan_tci: vlan tag control information
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700362 * @transport_header: Transport layer header
363 * @network_header: Network layer header
364 * @mac_header: Link layer header
365 * @tail: Tail pointer
366 * @end: End pointer
367 * @head: Head of buffer
368 * @data: Data head pointer
369 * @truesize: Buffer size
370 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700371 */
372
373struct sk_buff {
374 /* These two members must be first. */
375 struct sk_buff *next;
376 struct sk_buff *prev;
377
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700378 ktime_t tstamp;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000379
380 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700381 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382
Linus Torvalds1da177e2005-04-16 15:20:36 -0700383 /*
384 * This is the control buffer. It is free to use for every
385 * layer. Please put your private variables there. If you
386 * want to keep them across layers you have to do a skb_clone()
387 * first. This is owned by whoever has the skb queued ATM.
388 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000389 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390
Eric Dumazet7fee2262010-05-11 23:19:48 +0000391 unsigned long _skb_refdst;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000392#ifdef CONFIG_XFRM
393 struct sec_path *sp;
394#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700395 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700396 data_len;
397 __u16 mac_len,
398 hdr_len;
Al Viroff1dcad2006-11-20 18:07:29 -0800399 union {
400 __wsum csum;
Herbert Xu663ead32007-04-09 11:59:07 -0700401 struct {
402 __u16 csum_start;
403 __u16 csum_offset;
404 };
Al Viroff1dcad2006-11-20 18:07:29 -0800405 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700406 __u32 priority;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200407 kmemcheck_bitfield_begin(flags1);
Thomas Graf1cbb3382005-07-05 14:13:41 -0700408 __u8 local_df:1,
409 cloned:1,
410 ip_summed:2,
Harald Welte6869c4d2005-08-09 19:24:19 -0700411 nohdr:1,
412 nfctinfo:3;
David S. Millerd179cd12005-08-17 14:57:30 -0700413 __u8 pkt_type:3,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800414 fclone:2,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700415 ipvs_property:1,
Herbert Xua59322b2007-12-05 01:53:40 -0800416 peeked:1,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700417 nf_trace:1;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200418 kmemcheck_bitfield_end(flags1);
Eric Dumazet4ab408d2010-03-01 03:09:26 +0000419 __be16 protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700420
421 void (*destructor)(struct sk_buff *skb);
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800422#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700423 struct nf_conntrack *nfct;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100424#endif
425#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800426 struct sk_buff *nfct_reasm;
427#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428#ifdef CONFIG_BRIDGE_NETFILTER
429 struct nf_bridge_info *nf_bridge;
430#endif
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700431
Eric Dumazet8964be42009-11-20 15:35:04 -0800432 int skb_iif;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700434 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700436 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200439
Tom Herbert0a9627f2010-03-16 08:03:29 +0000440 __u32 rxhash;
441
Eric Dumazet0a148422011-04-20 09:27:32 +0000442 __u16 queue_mapping;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200443 kmemcheck_bitfield_begin(flags2);
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400444#ifdef CONFIG_IPV6_NDISC_NODETYPE
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000445 __u8 ndisc_nodetype:2;
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400446#endif
Tom Herbert3853b582010-11-21 13:17:29 +0000447 __u8 ooo_okay:1;
Tom Herbertbdeab992011-08-14 19:45:55 +0000448 __u8 l4_rxhash:1;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200449 kmemcheck_bitfield_end(flags2);
450
Tom Herbert3853b582010-11-21 13:17:29 +0000451 /* 0/13 bit hole */
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700452
Chris Leech97fc2f02006-05-23 17:55:33 -0700453#ifdef CONFIG_NET_DMA
454 dma_cookie_t dma_cookie;
455#endif
James Morris984bc162006-06-09 00:29:17 -0700456#ifdef CONFIG_NETWORK_SECMARK
457 __u32 secmark;
458#endif
Neil Horman3b885782009-10-12 13:26:31 -0700459 union {
460 __u32 mark;
461 __u32 dropcount;
462 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463
Patrick McHardy6aa895b02008-07-14 22:49:06 -0700464 __u16 vlan_tci;
465
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700466 sk_buff_data_t transport_header;
467 sk_buff_data_t network_header;
468 sk_buff_data_t mac_header;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700470 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700471 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700473 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700474 unsigned int truesize;
475 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476};
477
478#ifdef __KERNEL__
479/*
480 * Handling routines are only of interest to the kernel
481 */
482#include <linux/slab.h>
483
484#include <asm/system.h>
485
Eric Dumazet7fee2262010-05-11 23:19:48 +0000486/*
487 * skb might have a dst pointer attached, refcounted or not.
488 * _skb_refdst low order bit is set if refcount was _not_ taken
489 */
490#define SKB_DST_NOREF 1UL
491#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
492
493/**
494 * skb_dst - returns skb dst_entry
495 * @skb: buffer
496 *
497 * Returns skb dst_entry, regardless of reference taken or not.
498 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000499static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
500{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000501 /* If refdst was not refcounted, check we still are in a
502 * rcu_read_lock section
503 */
504 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
505 !rcu_read_lock_held() &&
506 !rcu_read_lock_bh_held());
507 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000508}
509
Eric Dumazet7fee2262010-05-11 23:19:48 +0000510/**
511 * skb_dst_set - sets skb dst
512 * @skb: buffer
513 * @dst: dst entry
514 *
515 * Sets skb dst, assuming a reference was taken on dst and should
516 * be released by skb_dst_drop()
517 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000518static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
519{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000520 skb->_skb_refdst = (unsigned long)dst;
521}
522
Eric Dumazet27b75c92010-10-15 05:44:11 +0000523extern void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst);
Eric Dumazet7fee2262010-05-11 23:19:48 +0000524
525/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300526 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000527 * @skb: buffer
528 */
529static inline bool skb_dst_is_noref(const struct sk_buff *skb)
530{
531 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000532}
533
Eric Dumazet511c3f92009-06-02 05:14:27 +0000534static inline struct rtable *skb_rtable(const struct sk_buff *skb)
535{
Eric Dumazetadf30902009-06-02 05:19:30 +0000536 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000537}
538
Jörn Engel231d06a2006-03-20 21:28:35 -0800539extern void kfree_skb(struct sk_buff *skb);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000540extern void consume_skb(struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541extern void __kfree_skb(struct sk_buff *skb);
David S. Millerd179cd12005-08-17 14:57:30 -0700542extern struct sk_buff *__alloc_skb(unsigned int size,
Christoph Hellwigb30973f2006-12-06 20:32:36 -0800543 gfp_t priority, int fclone, int node);
Eric Dumazetb2b5ce92011-11-14 06:03:34 +0000544extern struct sk_buff *build_skb(void *data);
David S. Millerd179cd12005-08-17 14:57:30 -0700545static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100546 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700547{
Eric Dumazet564824b2010-10-11 19:05:25 +0000548 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700549}
550
551static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100552 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700553{
Eric Dumazet564824b2010-10-11 19:05:25 +0000554 return __alloc_skb(size, priority, 1, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700555}
556
Andy Fleming3d153a72011-10-13 04:33:54 +0000557extern void skb_recycle(struct sk_buff *skb);
Changli Gao5b0daa32010-05-29 00:12:13 -0700558extern bool skb_recycle_check(struct sk_buff *skb, int skb_size);
Lennert Buytenhek04a4bb52008-10-01 02:33:12 -0700559
Herbert Xue0053ec2007-10-14 00:37:52 -0700560extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
Michael S. Tsirkin48c83012011-08-31 08:03:29 +0000561extern int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700562extern struct sk_buff *skb_clone(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100563 gfp_t priority);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700564extern struct sk_buff *skb_copy(const struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100565 gfp_t priority);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700566extern struct sk_buff *pskb_copy(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100567 gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568extern int pskb_expand_head(struct sk_buff *skb,
Victor Fusco86a76ca2005-07-08 14:57:47 -0700569 int nhead, int ntail,
Al Virodd0fc662005-10-07 07:46:04 +0100570 gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700571extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
572 unsigned int headroom);
573extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
574 int newheadroom, int newtailroom,
Al Virodd0fc662005-10-07 07:46:04 +0100575 gfp_t priority);
David Howells716ea3a2007-04-02 20:19:53 -0700576extern int skb_to_sgvec(struct sk_buff *skb,
577 struct scatterlist *sg, int offset,
578 int len);
579extern int skb_cow_data(struct sk_buff *skb, int tailbits,
580 struct sk_buff **trailer);
Herbert Xu5b057c62006-06-23 02:06:41 -0700581extern int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000582#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700584extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
585 int getfrag(void *from, char *to, int offset,
586 int len,int odd, struct sk_buff *skb),
587 void *from, int length);
588
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800589struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700590 __u32 lower_offset;
591 __u32 upper_offset;
592 __u32 frag_idx;
593 __u32 stepped_offset;
594 struct sk_buff *root_skb;
595 struct sk_buff *cur_skb;
596 __u8 *frag_data;
597};
598
599extern void skb_prepare_seq_read(struct sk_buff *skb,
600 unsigned int from, unsigned int to,
601 struct skb_seq_state *st);
602extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
603 struct skb_seq_state *st);
604extern void skb_abort_seq_read(struct skb_seq_state *st);
605
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700606extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
607 unsigned int to, struct ts_config *config,
608 struct ts_state *state);
609
Tom Herbertbdeab992011-08-14 19:45:55 +0000610extern void __skb_get_rxhash(struct sk_buff *skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000611static inline __u32 skb_get_rxhash(struct sk_buff *skb)
612{
613 if (!skb->rxhash)
Tom Herbertbdeab992011-08-14 19:45:55 +0000614 __skb_get_rxhash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000615
616 return skb->rxhash;
617}
618
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700619#ifdef NET_SKBUFF_DATA_USES_OFFSET
620static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
621{
622 return skb->head + skb->end;
623}
624#else
625static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
626{
627 return skb->end;
628}
629#endif
630
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700632#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633
Patrick Ohlyac45f602009-02-12 05:03:37 +0000634static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
635{
636 return &skb_shinfo(skb)->hwtstamps;
637}
638
Linus Torvalds1da177e2005-04-16 15:20:36 -0700639/**
640 * skb_queue_empty - check if a queue is empty
641 * @list: queue head
642 *
643 * Returns true if the queue is empty, false otherwise.
644 */
645static inline int skb_queue_empty(const struct sk_buff_head *list)
646{
647 return list->next == (struct sk_buff *)list;
648}
649
650/**
David S. Millerfc7ebb22008-09-23 00:34:07 -0700651 * skb_queue_is_last - check if skb is the last entry in the queue
652 * @list: queue head
653 * @skb: buffer
654 *
655 * Returns true if @skb is the last buffer on the list.
656 */
657static inline bool skb_queue_is_last(const struct sk_buff_head *list,
658 const struct sk_buff *skb)
659{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000660 return skb->next == (struct sk_buff *)list;
David S. Millerfc7ebb22008-09-23 00:34:07 -0700661}
662
663/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800664 * skb_queue_is_first - check if skb is the first entry in the queue
665 * @list: queue head
666 * @skb: buffer
667 *
668 * Returns true if @skb is the first buffer on the list.
669 */
670static inline bool skb_queue_is_first(const struct sk_buff_head *list,
671 const struct sk_buff *skb)
672{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000673 return skb->prev == (struct sk_buff *)list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800674}
675
676/**
David S. Miller249c8b42008-09-23 00:44:42 -0700677 * skb_queue_next - return the next packet in the queue
678 * @list: queue head
679 * @skb: current buffer
680 *
681 * Return the next packet in @list after @skb. It is only valid to
682 * call this if skb_queue_is_last() evaluates to false.
683 */
684static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
685 const struct sk_buff *skb)
686{
687 /* This BUG_ON may seem severe, but if we just return then we
688 * are going to dereference garbage.
689 */
690 BUG_ON(skb_queue_is_last(list, skb));
691 return skb->next;
692}
693
694/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800695 * skb_queue_prev - return the prev packet in the queue
696 * @list: queue head
697 * @skb: current buffer
698 *
699 * Return the prev packet in @list before @skb. It is only valid to
700 * call this if skb_queue_is_first() evaluates to false.
701 */
702static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
703 const struct sk_buff *skb)
704{
705 /* This BUG_ON may seem severe, but if we just return then we
706 * are going to dereference garbage.
707 */
708 BUG_ON(skb_queue_is_first(list, skb));
709 return skb->prev;
710}
711
712/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700713 * skb_get - reference buffer
714 * @skb: buffer to reference
715 *
716 * Makes another reference to a socket buffer and returns a pointer
717 * to the buffer.
718 */
719static inline struct sk_buff *skb_get(struct sk_buff *skb)
720{
721 atomic_inc(&skb->users);
722 return skb;
723}
724
725/*
726 * If users == 1, we are the only owner and are can avoid redundant
727 * atomic change.
728 */
729
730/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731 * skb_cloned - is the buffer a clone
732 * @skb: buffer to check
733 *
734 * Returns true if the buffer was generated with skb_clone() and is
735 * one of multiple shared copies of the buffer. Cloned buffers are
736 * shared data so must not be written to under normal circumstances.
737 */
738static inline int skb_cloned(const struct sk_buff *skb)
739{
740 return skb->cloned &&
741 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
742}
743
744/**
745 * skb_header_cloned - is the header a clone
746 * @skb: buffer to check
747 *
748 * Returns true if modifying the header part of the buffer requires
749 * the data to be copied.
750 */
751static inline int skb_header_cloned(const struct sk_buff *skb)
752{
753 int dataref;
754
755 if (!skb->cloned)
756 return 0;
757
758 dataref = atomic_read(&skb_shinfo(skb)->dataref);
759 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
760 return dataref != 1;
761}
762
763/**
764 * skb_header_release - release reference to header
765 * @skb: buffer to operate on
766 *
767 * Drop a reference to the header part of the buffer. This is done
768 * by acquiring a payload reference. You must not read from the header
769 * part of skb->data after this.
770 */
771static inline void skb_header_release(struct sk_buff *skb)
772{
773 BUG_ON(skb->nohdr);
774 skb->nohdr = 1;
775 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
776}
777
778/**
779 * skb_shared - is the buffer shared
780 * @skb: buffer to check
781 *
782 * Returns true if more than one person has a reference to this
783 * buffer.
784 */
785static inline int skb_shared(const struct sk_buff *skb)
786{
787 return atomic_read(&skb->users) != 1;
788}
789
790/**
791 * skb_share_check - check if buffer is shared and if so clone it
792 * @skb: buffer to check
793 * @pri: priority for memory allocation
794 *
795 * If the buffer is shared the buffer is cloned and the old copy
796 * drops a reference. A new clone with a single reference is returned.
797 * If the buffer is not shared the original buffer is returned. When
798 * being called from interrupt status or with spinlocks held pri must
799 * be GFP_ATOMIC.
800 *
801 * NULL is returned on a memory allocation failure.
802 */
Victor Fusco86a76ca2005-07-08 14:57:47 -0700803static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100804 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700805{
806 might_sleep_if(pri & __GFP_WAIT);
807 if (skb_shared(skb)) {
808 struct sk_buff *nskb = skb_clone(skb, pri);
809 kfree_skb(skb);
810 skb = nskb;
811 }
812 return skb;
813}
814
815/*
816 * Copy shared buffers into a new sk_buff. We effectively do COW on
817 * packets to handle cases where we have a local reader and forward
818 * and a couple of other messy ones. The normal one is tcpdumping
819 * a packet thats being forwarded.
820 */
821
822/**
823 * skb_unshare - make a copy of a shared buffer
824 * @skb: buffer to check
825 * @pri: priority for memory allocation
826 *
827 * If the socket buffer is a clone then this function creates a new
828 * copy of the data, drops a reference count on the old copy and returns
829 * the new copy with the reference count at 1. If the buffer is not a clone
830 * the original buffer is returned. When called with a spinlock held or
831 * from interrupt state @pri must be %GFP_ATOMIC
832 *
833 * %NULL is returned on a memory allocation failure.
834 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -0700835static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100836 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837{
838 might_sleep_if(pri & __GFP_WAIT);
839 if (skb_cloned(skb)) {
840 struct sk_buff *nskb = skb_copy(skb, pri);
841 kfree_skb(skb); /* Free our shared copy */
842 skb = nskb;
843 }
844 return skb;
845}
846
847/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -0800848 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849 * @list_: list to peek at
850 *
851 * Peek an &sk_buff. Unlike most other operations you _MUST_
852 * be careful with this one. A peek leaves the buffer on the
853 * list and someone else may run off with it. You must hold
854 * the appropriate locks or have a private queue to do this.
855 *
856 * Returns %NULL for an empty list or a pointer to the head element.
857 * The reference count is not incremented and the reference is therefore
858 * volatile. Use with caution.
859 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -0400860static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700861{
Eric Dumazet05bdd2f2011-10-20 17:45:43 -0400862 struct sk_buff *list = ((const struct sk_buff *)list_)->next;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700863 if (list == (struct sk_buff *)list_)
864 list = NULL;
865 return list;
866}
867
868/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -0800869 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -0700870 * @list_: list to peek at
871 *
872 * Peek an &sk_buff. Unlike most other operations you _MUST_
873 * be careful with this one. A peek leaves the buffer on the
874 * list and someone else may run off with it. You must hold
875 * the appropriate locks or have a private queue to do this.
876 *
877 * Returns %NULL for an empty list or a pointer to the tail element.
878 * The reference count is not incremented and the reference is therefore
879 * volatile. Use with caution.
880 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -0400881static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700882{
Eric Dumazet05bdd2f2011-10-20 17:45:43 -0400883 struct sk_buff *list = ((const struct sk_buff *)list_)->prev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884 if (list == (struct sk_buff *)list_)
885 list = NULL;
886 return list;
887}
888
889/**
890 * skb_queue_len - get queue length
891 * @list_: list to measure
892 *
893 * Return the length of an &sk_buff queue.
894 */
895static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
896{
897 return list_->qlen;
898}
899
David S. Miller67fed452008-09-21 22:36:24 -0700900/**
901 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
902 * @list: queue to initialize
903 *
904 * This initializes only the list and queue length aspects of
905 * an sk_buff_head object. This allows to initialize the list
906 * aspects of an sk_buff_head without reinitializing things like
907 * the spinlock. It can also be used for on-stack sk_buff_head
908 * objects where the spinlock is known to not be used.
909 */
910static inline void __skb_queue_head_init(struct sk_buff_head *list)
911{
912 list->prev = list->next = (struct sk_buff *)list;
913 list->qlen = 0;
914}
915
Arjan van de Ven76f10ad2006-08-02 14:06:55 -0700916/*
917 * This function creates a split out lock class for each invocation;
918 * this is needed for now since a whole lot of users of the skb-queue
919 * infrastructure in drivers have different locking usage (in hardirq)
920 * than the networking core (in softirq only). In the long run either the
921 * network layer or drivers should need annotation to consolidate the
922 * main types of usage into 3 classes.
923 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924static inline void skb_queue_head_init(struct sk_buff_head *list)
925{
926 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -0700927 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700928}
929
Pavel Emelianovc2ecba72007-04-17 12:45:31 -0700930static inline void skb_queue_head_init_class(struct sk_buff_head *list,
931 struct lock_class_key *class)
932{
933 skb_queue_head_init(list);
934 lockdep_set_class(&list->lock, class);
935}
936
Linus Torvalds1da177e2005-04-16 15:20:36 -0700937/*
Gerrit Renkerbf299272008-04-14 00:04:51 -0700938 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939 *
940 * The "__skb_xxxx()" functions are the non-atomic ones that
941 * can only be called with interrupts disabled.
942 */
Gerrit Renkerbf299272008-04-14 00:04:51 -0700943extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
944static inline void __skb_insert(struct sk_buff *newsk,
945 struct sk_buff *prev, struct sk_buff *next,
946 struct sk_buff_head *list)
947{
948 newsk->next = next;
949 newsk->prev = prev;
950 next->prev = prev->next = newsk;
951 list->qlen++;
952}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700953
David S. Miller67fed452008-09-21 22:36:24 -0700954static inline void __skb_queue_splice(const struct sk_buff_head *list,
955 struct sk_buff *prev,
956 struct sk_buff *next)
957{
958 struct sk_buff *first = list->next;
959 struct sk_buff *last = list->prev;
960
961 first->prev = prev;
962 prev->next = first;
963
964 last->next = next;
965 next->prev = last;
966}
967
968/**
969 * skb_queue_splice - join two skb lists, this is designed for stacks
970 * @list: the new list to add
971 * @head: the place to add it in the first list
972 */
973static inline void skb_queue_splice(const struct sk_buff_head *list,
974 struct sk_buff_head *head)
975{
976 if (!skb_queue_empty(list)) {
977 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -0700978 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -0700979 }
980}
981
982/**
983 * skb_queue_splice - join two skb lists and reinitialise the emptied list
984 * @list: the new list to add
985 * @head: the place to add it in the first list
986 *
987 * The list at @list is reinitialised
988 */
989static inline void skb_queue_splice_init(struct sk_buff_head *list,
990 struct sk_buff_head *head)
991{
992 if (!skb_queue_empty(list)) {
993 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -0700994 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -0700995 __skb_queue_head_init(list);
996 }
997}
998
999/**
1000 * skb_queue_splice_tail - join two skb lists, each list being a queue
1001 * @list: the new list to add
1002 * @head: the place to add it in the first list
1003 */
1004static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1005 struct sk_buff_head *head)
1006{
1007 if (!skb_queue_empty(list)) {
1008 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001009 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001010 }
1011}
1012
1013/**
1014 * skb_queue_splice_tail - join two skb lists and reinitialise the emptied list
1015 * @list: the new list to add
1016 * @head: the place to add it in the first list
1017 *
1018 * Each of the lists is a queue.
1019 * The list at @list is reinitialised
1020 */
1021static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1022 struct sk_buff_head *head)
1023{
1024 if (!skb_queue_empty(list)) {
1025 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001026 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001027 __skb_queue_head_init(list);
1028 }
1029}
1030
Linus Torvalds1da177e2005-04-16 15:20:36 -07001031/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001032 * __skb_queue_after - queue a buffer at the list head
1033 * @list: list to use
1034 * @prev: place after this buffer
1035 * @newsk: buffer to queue
1036 *
1037 * Queue a buffer int the middle of a list. This function takes no locks
1038 * and you must therefore hold required locks before calling it.
1039 *
1040 * A buffer cannot be placed on two lists at the same time.
1041 */
1042static inline void __skb_queue_after(struct sk_buff_head *list,
1043 struct sk_buff *prev,
1044 struct sk_buff *newsk)
1045{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001046 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001047}
1048
Gerrit Renker7de6c032008-04-14 00:05:09 -07001049extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1050 struct sk_buff_head *list);
1051
Gerrit Renkerf5572852008-04-14 00:05:28 -07001052static inline void __skb_queue_before(struct sk_buff_head *list,
1053 struct sk_buff *next,
1054 struct sk_buff *newsk)
1055{
1056 __skb_insert(newsk, next->prev, next, list);
1057}
1058
Stephen Hemminger300ce172005-10-30 13:47:34 -08001059/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060 * __skb_queue_head - queue a buffer at the list head
1061 * @list: list to use
1062 * @newsk: buffer to queue
1063 *
1064 * Queue a buffer at the start of a list. This function takes no locks
1065 * and you must therefore hold required locks before calling it.
1066 *
1067 * A buffer cannot be placed on two lists at the same time.
1068 */
1069extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1070static inline void __skb_queue_head(struct sk_buff_head *list,
1071 struct sk_buff *newsk)
1072{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001073 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074}
1075
1076/**
1077 * __skb_queue_tail - queue a buffer at the list tail
1078 * @list: list to use
1079 * @newsk: buffer to queue
1080 *
1081 * Queue a buffer at the end of a list. This function takes no locks
1082 * and you must therefore hold required locks before calling it.
1083 *
1084 * A buffer cannot be placed on two lists at the same time.
1085 */
1086extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1087static inline void __skb_queue_tail(struct sk_buff_head *list,
1088 struct sk_buff *newsk)
1089{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001090 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001091}
1092
Linus Torvalds1da177e2005-04-16 15:20:36 -07001093/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001094 * remove sk_buff from list. _Must_ be called atomically, and with
1095 * the list known..
1096 */
David S. Miller8728b832005-08-09 19:25:21 -07001097extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1099{
1100 struct sk_buff *next, *prev;
1101
1102 list->qlen--;
1103 next = skb->next;
1104 prev = skb->prev;
1105 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106 next->prev = prev;
1107 prev->next = next;
1108}
1109
Gerrit Renkerf525c062008-04-14 00:04:12 -07001110/**
1111 * __skb_dequeue - remove from the head of the queue
1112 * @list: list to dequeue from
1113 *
1114 * Remove the head of the list. This function does not take any locks
1115 * so must be used with appropriate locks held only. The head item is
1116 * returned or %NULL if the list is empty.
1117 */
1118extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1119static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1120{
1121 struct sk_buff *skb = skb_peek(list);
1122 if (skb)
1123 __skb_unlink(skb, list);
1124 return skb;
1125}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001126
1127/**
1128 * __skb_dequeue_tail - remove from the tail of the queue
1129 * @list: list to dequeue from
1130 *
1131 * Remove the tail of the list. This function does not take any locks
1132 * so must be used with appropriate locks held only. The tail item is
1133 * returned or %NULL if the list is empty.
1134 */
1135extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1136static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1137{
1138 struct sk_buff *skb = skb_peek_tail(list);
1139 if (skb)
1140 __skb_unlink(skb, list);
1141 return skb;
1142}
1143
1144
1145static inline int skb_is_nonlinear(const struct sk_buff *skb)
1146{
1147 return skb->data_len;
1148}
1149
1150static inline unsigned int skb_headlen(const struct sk_buff *skb)
1151{
1152 return skb->len - skb->data_len;
1153}
1154
1155static inline int skb_pagelen(const struct sk_buff *skb)
1156{
1157 int i, len = 0;
1158
1159 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001160 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161 return len + skb_headlen(skb);
1162}
1163
Ian Campbell131ea662011-08-19 06:25:00 +00001164/**
1165 * __skb_fill_page_desc - initialise a paged fragment in an skb
1166 * @skb: buffer containing fragment to be initialised
1167 * @i: paged fragment index to initialise
1168 * @page: the page to use for this fragment
1169 * @off: the offset to the data with @page
1170 * @size: the length of the data
1171 *
1172 * Initialises the @i'th fragment of @skb to point to &size bytes at
1173 * offset @off within @page.
1174 *
1175 * Does not take any additional reference on the fragment.
1176 */
1177static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1178 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001179{
1180 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1181
Ian Campbella8605c62011-10-19 23:01:49 +00001182 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001184 skb_frag_size_set(frag, size);
Ian Campbell131ea662011-08-19 06:25:00 +00001185}
1186
1187/**
1188 * skb_fill_page_desc - initialise a paged fragment in an skb
1189 * @skb: buffer containing fragment to be initialised
1190 * @i: paged fragment index to initialise
1191 * @page: the page to use for this fragment
1192 * @off: the offset to the data with @page
1193 * @size: the length of the data
1194 *
1195 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1196 * @skb to point to &size bytes at offset @off within @page. In
1197 * addition updates @skb such that @i is the last fragment.
1198 *
1199 * Does not take any additional reference on the fragment.
1200 */
1201static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1202 struct page *page, int off, int size)
1203{
1204 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001205 skb_shinfo(skb)->nr_frags = i + 1;
1206}
1207
Peter Zijlstra654bed12008-10-07 14:22:33 -07001208extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
1209 int off, int size);
1210
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001212#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001213#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1214
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001215#ifdef NET_SKBUFF_DATA_USES_OFFSET
1216static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1217{
1218 return skb->head + skb->tail;
1219}
1220
1221static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1222{
1223 skb->tail = skb->data - skb->head;
1224}
1225
1226static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1227{
1228 skb_reset_tail_pointer(skb);
1229 skb->tail += offset;
1230}
1231#else /* NET_SKBUFF_DATA_USES_OFFSET */
1232static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1233{
1234 return skb->tail;
1235}
1236
1237static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1238{
1239 skb->tail = skb->data;
1240}
1241
1242static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1243{
1244 skb->tail = skb->data + offset;
1245}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001246
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001247#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1248
Linus Torvalds1da177e2005-04-16 15:20:36 -07001249/*
1250 * Add data to an sk_buff
1251 */
Ilpo Järvinen0dde3e12008-03-27 17:43:41 -07001252extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1254{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001255 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001256 SKB_LINEAR_ASSERT(skb);
1257 skb->tail += len;
1258 skb->len += len;
1259 return tmp;
1260}
1261
Ilpo Järvinenc2aa2702008-03-27 17:52:40 -07001262extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001263static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1264{
1265 skb->data -= len;
1266 skb->len += len;
1267 return skb->data;
1268}
1269
Ilpo Järvinen6be8ac22008-03-27 17:47:24 -07001270extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001271static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1272{
1273 skb->len -= len;
1274 BUG_ON(skb->len < skb->data_len);
1275 return skb->data += len;
1276}
1277
David S. Miller47d29642010-05-02 02:21:44 -07001278static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1279{
1280 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1281}
1282
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1284
1285static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1286{
1287 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001288 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289 return NULL;
1290 skb->len -= len;
1291 return skb->data += len;
1292}
1293
1294static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1295{
1296 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1297}
1298
1299static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1300{
1301 if (likely(len <= skb_headlen(skb)))
1302 return 1;
1303 if (unlikely(len > skb->len))
1304 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001305 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001306}
1307
1308/**
1309 * skb_headroom - bytes at buffer head
1310 * @skb: buffer to check
1311 *
1312 * Return the number of bytes of free space at the head of an &sk_buff.
1313 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001314static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001315{
1316 return skb->data - skb->head;
1317}
1318
1319/**
1320 * skb_tailroom - bytes at buffer end
1321 * @skb: buffer to check
1322 *
1323 * Return the number of bytes of free space at the tail of an sk_buff
1324 */
1325static inline int skb_tailroom(const struct sk_buff *skb)
1326{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001327 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328}
1329
1330/**
1331 * skb_reserve - adjust headroom
1332 * @skb: buffer to alter
1333 * @len: bytes to move
1334 *
1335 * Increase the headroom of an empty &sk_buff by reducing the tail
1336 * room. This is only allowed for an empty buffer.
1337 */
David S. Miller8243126c2006-01-17 02:54:21 -08001338static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339{
1340 skb->data += len;
1341 skb->tail += len;
1342}
1343
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00001344static inline void skb_reset_mac_len(struct sk_buff *skb)
1345{
1346 skb->mac_len = skb->network_header - skb->mac_header;
1347}
1348
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001349#ifdef NET_SKBUFF_DATA_USES_OFFSET
1350static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1351{
1352 return skb->head + skb->transport_header;
1353}
1354
1355static inline void skb_reset_transport_header(struct sk_buff *skb)
1356{
1357 skb->transport_header = skb->data - skb->head;
1358}
1359
1360static inline void skb_set_transport_header(struct sk_buff *skb,
1361 const int offset)
1362{
1363 skb_reset_transport_header(skb);
1364 skb->transport_header += offset;
1365}
1366
1367static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1368{
1369 return skb->head + skb->network_header;
1370}
1371
1372static inline void skb_reset_network_header(struct sk_buff *skb)
1373{
1374 skb->network_header = skb->data - skb->head;
1375}
1376
1377static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1378{
1379 skb_reset_network_header(skb);
1380 skb->network_header += offset;
1381}
1382
1383static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1384{
1385 return skb->head + skb->mac_header;
1386}
1387
1388static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1389{
1390 return skb->mac_header != ~0U;
1391}
1392
1393static inline void skb_reset_mac_header(struct sk_buff *skb)
1394{
1395 skb->mac_header = skb->data - skb->head;
1396}
1397
1398static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1399{
1400 skb_reset_mac_header(skb);
1401 skb->mac_header += offset;
1402}
1403
1404#else /* NET_SKBUFF_DATA_USES_OFFSET */
1405
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001406static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1407{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001408 return skb->transport_header;
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001409}
1410
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001411static inline void skb_reset_transport_header(struct sk_buff *skb)
1412{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001413 skb->transport_header = skb->data;
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001414}
1415
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001416static inline void skb_set_transport_header(struct sk_buff *skb,
1417 const int offset)
1418{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001419 skb->transport_header = skb->data + offset;
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001420}
1421
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001422static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1423{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001424 return skb->network_header;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001425}
1426
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001427static inline void skb_reset_network_header(struct sk_buff *skb)
1428{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001429 skb->network_header = skb->data;
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001430}
1431
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001432static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1433{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001434 skb->network_header = skb->data + offset;
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001435}
1436
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001437static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1438{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001439 return skb->mac_header;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001440}
1441
1442static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1443{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001444 return skb->mac_header != NULL;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001445}
1446
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001447static inline void skb_reset_mac_header(struct sk_buff *skb)
1448{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001449 skb->mac_header = skb->data;
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001450}
1451
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001452static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1453{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001454 skb->mac_header = skb->data + offset;
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001455}
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001456#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1457
Michał Mirosław04fb4512010-12-14 15:24:08 +00001458static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1459{
1460 return skb->csum_start - skb_headroom(skb);
1461}
1462
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001463static inline int skb_transport_offset(const struct sk_buff *skb)
1464{
1465 return skb_transport_header(skb) - skb->data;
1466}
1467
1468static inline u32 skb_network_header_len(const struct sk_buff *skb)
1469{
1470 return skb->transport_header - skb->network_header;
1471}
1472
1473static inline int skb_network_offset(const struct sk_buff *skb)
1474{
1475 return skb_network_header(skb) - skb->data;
1476}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001477
Changli Gaof9599ce2010-08-04 04:43:44 +00001478static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1479{
1480 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1481}
1482
Linus Torvalds1da177e2005-04-16 15:20:36 -07001483/*
1484 * CPUs often take a performance hit when accessing unaligned memory
1485 * locations. The actual performance hit varies, it can be small if the
1486 * hardware handles it or large if we have to take an exception and fix it
1487 * in software.
1488 *
1489 * Since an ethernet header is 14 bytes network drivers often end up with
1490 * the IP header at an unaligned offset. The IP header can be aligned by
1491 * shifting the start of the packet by 2 bytes. Drivers should do this
1492 * with:
1493 *
Tobias Klauser8660c122009-07-13 22:48:16 +00001494 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001495 *
1496 * The downside to this alignment of the IP header is that the DMA is now
1497 * unaligned. On some architectures the cost of an unaligned DMA is high
1498 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00001499 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001500 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1501 * to be overridden.
1502 */
1503#ifndef NET_IP_ALIGN
1504#define NET_IP_ALIGN 2
1505#endif
1506
Anton Blanchard025be812006-03-31 02:27:06 -08001507/*
1508 * The networking layer reserves some headroom in skb data (via
1509 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1510 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08001511 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08001512 *
1513 * Unfortunately this headroom changes the DMA alignment of the resulting
1514 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1515 * on some architectures. An architecture can override this value,
1516 * perhaps setting it to a cacheline in size (since that will maintain
1517 * cacheline alignment of the DMA). It must be a power of 2.
1518 *
David S. Millerd6301d32009-02-08 19:24:13 -08001519 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08001520 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07001521 *
1522 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1523 * to reduce average number of cache lines per packet.
1524 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07001525 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08001526 */
1527#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07001528#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08001529#endif
1530
Herbert Xu3cc0e872006-06-09 16:13:38 -07001531extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001532
1533static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1534{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00001535 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07001536 WARN_ON(1);
1537 return;
1538 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001539 skb->len = len;
1540 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001541}
1542
Ilpo Järvinen419ae742008-03-27 17:54:01 -07001543extern void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001544
1545static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1546{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001547 if (skb->data_len)
1548 return ___pskb_trim(skb, len);
1549 __skb_trim(skb, len);
1550 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001551}
1552
1553static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1554{
1555 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1556}
1557
1558/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07001559 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1560 * @skb: buffer to alter
1561 * @len: new length
1562 *
1563 * This is identical to pskb_trim except that the caller knows that
1564 * the skb is not cloned so we should never get an error due to out-
1565 * of-memory.
1566 */
1567static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1568{
1569 int err = pskb_trim(skb, len);
1570 BUG_ON(err);
1571}
1572
1573/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574 * skb_orphan - orphan a buffer
1575 * @skb: buffer to orphan
1576 *
1577 * If a buffer currently has an owner then we call the owner's
1578 * destructor function and make the @skb unowned. The buffer continues
1579 * to exist but is no longer charged to its former owner.
1580 */
1581static inline void skb_orphan(struct sk_buff *skb)
1582{
1583 if (skb->destructor)
1584 skb->destructor(skb);
1585 skb->destructor = NULL;
1586 skb->sk = NULL;
1587}
1588
1589/**
1590 * __skb_queue_purge - empty a list
1591 * @list: list to empty
1592 *
1593 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1594 * the list and one reference dropped. This function does not take the
1595 * list lock and the caller must hold the relevant locks to use it.
1596 */
1597extern void skb_queue_purge(struct sk_buff_head *list);
1598static inline void __skb_queue_purge(struct sk_buff_head *list)
1599{
1600 struct sk_buff *skb;
1601 while ((skb = __skb_dequeue(list)) != NULL)
1602 kfree_skb(skb);
1603}
1604
1605/**
Christoph Hellwigb4e54de2006-07-24 15:31:14 -07001606 * __dev_alloc_skb - allocate an skbuff for receiving
Linus Torvalds1da177e2005-04-16 15:20:36 -07001607 * @length: length to allocate
1608 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1609 *
1610 * Allocate a new &sk_buff and assign it a usage count of one. The
1611 * buffer has unspecified headroom built in. Users should allocate
1612 * the headroom they think they need without accounting for the
1613 * built in space. The built in space is used for optimisations.
1614 *
Christoph Hellwig766ea8c2006-08-07 15:49:53 -07001615 * %NULL is returned if there is no free memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001616 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001617static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
Al Virodd0fc662005-10-07 07:46:04 +01001618 gfp_t gfp_mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001619{
Anton Blanchard025be812006-03-31 02:27:06 -08001620 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001621 if (likely(skb))
Anton Blanchard025be812006-03-31 02:27:06 -08001622 skb_reserve(skb, NET_SKB_PAD);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001623 return skb;
1624}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001625
Ilpo Järvinenf58518e2008-03-27 17:51:31 -07001626extern struct sk_buff *dev_alloc_skb(unsigned int length);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001627
Christoph Hellwig8af27452006-07-31 22:35:23 -07001628extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1629 unsigned int length, gfp_t gfp_mask);
1630
1631/**
1632 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1633 * @dev: network device to receive on
1634 * @length: length to allocate
1635 *
1636 * Allocate a new &sk_buff and assign it a usage count of one. The
1637 * buffer has unspecified headroom built in. Users should allocate
1638 * the headroom they think they need without accounting for the
1639 * built in space. The built in space is used for optimisations.
1640 *
1641 * %NULL is returned if there is no free memory. Although this function
1642 * allocates memory it can be called from an interrupt.
1643 */
1644static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1645 unsigned int length)
1646{
1647 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1648}
1649
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001650static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
1651 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00001652{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001653 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00001654
1655 if (NET_IP_ALIGN && skb)
1656 skb_reserve(skb, NET_IP_ALIGN);
1657 return skb;
1658}
1659
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001660static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1661 unsigned int length)
1662{
1663 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
1664}
1665
Eric Dumazet564824b2010-10-11 19:05:25 +00001666/**
1667 * __netdev_alloc_page - allocate a page for ps-rx on a specific device
1668 * @dev: network device to receive on
1669 * @gfp_mask: alloc_pages_node mask
1670 *
1671 * Allocate a new page. dev currently unused.
1672 *
1673 * %NULL is returned if there is no free memory.
1674 */
1675static inline struct page *__netdev_alloc_page(struct net_device *dev, gfp_t gfp_mask)
1676{
1677 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, 0);
1678}
Peter Zijlstra654bed12008-10-07 14:22:33 -07001679
1680/**
1681 * netdev_alloc_page - allocate a page for ps-rx on a specific device
1682 * @dev: network device to receive on
1683 *
Eric Dumazet564824b2010-10-11 19:05:25 +00001684 * Allocate a new page. dev currently unused.
Peter Zijlstra654bed12008-10-07 14:22:33 -07001685 *
1686 * %NULL is returned if there is no free memory.
1687 */
1688static inline struct page *netdev_alloc_page(struct net_device *dev)
1689{
1690 return __netdev_alloc_page(dev, GFP_ATOMIC);
1691}
1692
1693static inline void netdev_free_page(struct net_device *dev, struct page *page)
1694{
1695 __free_page(page);
1696}
1697
Linus Torvalds1da177e2005-04-16 15:20:36 -07001698/**
Ian Campbell131ea662011-08-19 06:25:00 +00001699 * skb_frag_page - retrieve the page refered to by a paged fragment
1700 * @frag: the paged fragment
1701 *
1702 * Returns the &struct page associated with @frag.
1703 */
1704static inline struct page *skb_frag_page(const skb_frag_t *frag)
1705{
Ian Campbella8605c62011-10-19 23:01:49 +00001706 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00001707}
1708
1709/**
1710 * __skb_frag_ref - take an addition reference on a paged fragment.
1711 * @frag: the paged fragment
1712 *
1713 * Takes an additional reference on the paged fragment @frag.
1714 */
1715static inline void __skb_frag_ref(skb_frag_t *frag)
1716{
1717 get_page(skb_frag_page(frag));
1718}
1719
1720/**
1721 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
1722 * @skb: the buffer
1723 * @f: the fragment offset.
1724 *
1725 * Takes an additional reference on the @f'th paged fragment of @skb.
1726 */
1727static inline void skb_frag_ref(struct sk_buff *skb, int f)
1728{
1729 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
1730}
1731
1732/**
1733 * __skb_frag_unref - release a reference on a paged fragment.
1734 * @frag: the paged fragment
1735 *
1736 * Releases a reference on the paged fragment @frag.
1737 */
1738static inline void __skb_frag_unref(skb_frag_t *frag)
1739{
1740 put_page(skb_frag_page(frag));
1741}
1742
1743/**
1744 * skb_frag_unref - release a reference on a paged fragment of an skb.
1745 * @skb: the buffer
1746 * @f: the fragment offset
1747 *
1748 * Releases a reference on the @f'th paged fragment of @skb.
1749 */
1750static inline void skb_frag_unref(struct sk_buff *skb, int f)
1751{
1752 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
1753}
1754
1755/**
1756 * skb_frag_address - gets the address of the data contained in a paged fragment
1757 * @frag: the paged fragment buffer
1758 *
1759 * Returns the address of the data within @frag. The page must already
1760 * be mapped.
1761 */
1762static inline void *skb_frag_address(const skb_frag_t *frag)
1763{
1764 return page_address(skb_frag_page(frag)) + frag->page_offset;
1765}
1766
1767/**
1768 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
1769 * @frag: the paged fragment buffer
1770 *
1771 * Returns the address of the data within @frag. Checks that the page
1772 * is mapped and returns %NULL otherwise.
1773 */
1774static inline void *skb_frag_address_safe(const skb_frag_t *frag)
1775{
1776 void *ptr = page_address(skb_frag_page(frag));
1777 if (unlikely(!ptr))
1778 return NULL;
1779
1780 return ptr + frag->page_offset;
1781}
1782
1783/**
1784 * __skb_frag_set_page - sets the page contained in a paged fragment
1785 * @frag: the paged fragment
1786 * @page: the page to set
1787 *
1788 * Sets the fragment @frag to contain @page.
1789 */
1790static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
1791{
Ian Campbella8605c62011-10-19 23:01:49 +00001792 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00001793}
1794
1795/**
1796 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
1797 * @skb: the buffer
1798 * @f: the fragment offset
1799 * @page: the page to set
1800 *
1801 * Sets the @f'th fragment of @skb to contain @page.
1802 */
1803static inline void skb_frag_set_page(struct sk_buff *skb, int f,
1804 struct page *page)
1805{
1806 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
1807}
1808
1809/**
1810 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00001811 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00001812 * @frag: the paged fragment to map
1813 * @offset: the offset within the fragment (starting at the
1814 * fragment's own offset)
1815 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00001816 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00001817 *
1818 * Maps the page associated with @frag to @device.
1819 */
1820static inline dma_addr_t skb_frag_dma_map(struct device *dev,
1821 const skb_frag_t *frag,
1822 size_t offset, size_t size,
1823 enum dma_data_direction dir)
1824{
1825 return dma_map_page(dev, skb_frag_page(frag),
1826 frag->page_offset + offset, size, dir);
1827}
1828
1829/**
Patrick McHardy334a8132007-06-25 04:35:20 -07001830 * skb_clone_writable - is the header of a clone writable
1831 * @skb: buffer to check
1832 * @len: length up to which to write
1833 *
1834 * Returns true if modifying the header part of the cloned buffer
1835 * does not requires the data to be copied.
1836 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001837static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07001838{
1839 return !skb_header_cloned(skb) &&
1840 skb_headroom(skb) + len <= skb->hdr_len;
1841}
1842
Herbert Xud9cc2042007-09-16 16:21:16 -07001843static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
1844 int cloned)
1845{
1846 int delta = 0;
1847
1848 if (headroom < NET_SKB_PAD)
1849 headroom = NET_SKB_PAD;
1850 if (headroom > skb_headroom(skb))
1851 delta = headroom - skb_headroom(skb);
1852
1853 if (delta || cloned)
1854 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
1855 GFP_ATOMIC);
1856 return 0;
1857}
1858
Patrick McHardy334a8132007-06-25 04:35:20 -07001859/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001860 * skb_cow - copy header of skb when it is required
1861 * @skb: buffer to cow
1862 * @headroom: needed headroom
1863 *
1864 * If the skb passed lacks sufficient headroom or its data part
1865 * is shared, data is reallocated. If reallocation fails, an error
1866 * is returned and original skb is not changed.
1867 *
1868 * The result is skb with writable area skb->head...skb->tail
1869 * and at least @headroom of space at head.
1870 */
1871static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1872{
Herbert Xud9cc2042007-09-16 16:21:16 -07001873 return __skb_cow(skb, headroom, skb_cloned(skb));
1874}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001875
Herbert Xud9cc2042007-09-16 16:21:16 -07001876/**
1877 * skb_cow_head - skb_cow but only making the head writable
1878 * @skb: buffer to cow
1879 * @headroom: needed headroom
1880 *
1881 * This function is identical to skb_cow except that we replace the
1882 * skb_cloned check by skb_header_cloned. It should be used when
1883 * you only need to push on some header and do not need to modify
1884 * the data.
1885 */
1886static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
1887{
1888 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001889}
1890
1891/**
1892 * skb_padto - pad an skbuff up to a minimal size
1893 * @skb: buffer to pad
1894 * @len: minimal length
1895 *
1896 * Pads up a buffer to ensure the trailing bytes exist and are
1897 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07001898 * is untouched. Otherwise it is extended. Returns zero on
1899 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001900 */
1901
Herbert Xu5b057c62006-06-23 02:06:41 -07001902static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001903{
1904 unsigned int size = skb->len;
1905 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07001906 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001907 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001908}
1909
1910static inline int skb_add_data(struct sk_buff *skb,
1911 char __user *from, int copy)
1912{
1913 const int off = skb->len;
1914
1915 if (skb->ip_summed == CHECKSUM_NONE) {
1916 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08001917 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
Linus Torvalds1da177e2005-04-16 15:20:36 -07001918 copy, 0, &err);
1919 if (!err) {
1920 skb->csum = csum_block_add(skb->csum, csum, off);
1921 return 0;
1922 }
1923 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1924 return 0;
1925
1926 __skb_trim(skb, off);
1927 return -EFAULT;
1928}
1929
1930static inline int skb_can_coalesce(struct sk_buff *skb, int i,
Ian Campbellea2ab692011-08-22 23:44:58 +00001931 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001932{
1933 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00001934 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001935
Ian Campbellea2ab692011-08-22 23:44:58 +00001936 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00001937 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001938 }
1939 return 0;
1940}
1941
Herbert Xu364c6ba2006-06-09 16:10:40 -07001942static inline int __skb_linearize(struct sk_buff *skb)
1943{
1944 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1945}
1946
Linus Torvalds1da177e2005-04-16 15:20:36 -07001947/**
1948 * skb_linearize - convert paged skb to linear one
1949 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07001950 *
1951 * If there is no free memory -ENOMEM is returned, otherwise zero
1952 * is returned and the old skb data released.
1953 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07001954static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001955{
Herbert Xu364c6ba2006-06-09 16:10:40 -07001956 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1957}
1958
1959/**
1960 * skb_linearize_cow - make sure skb is linear and writable
1961 * @skb: buffer to process
1962 *
1963 * If there is no free memory -ENOMEM is returned, otherwise zero
1964 * is returned and the old skb data released.
1965 */
1966static inline int skb_linearize_cow(struct sk_buff *skb)
1967{
1968 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1969 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001970}
1971
1972/**
1973 * skb_postpull_rcsum - update checksum for received skb after pull
1974 * @skb: buffer to update
1975 * @start: start of data before pull
1976 * @len: length of data pulled
1977 *
1978 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07001979 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
1980 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001981 */
1982
1983static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08001984 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001985{
Patrick McHardy84fa7932006-08-29 16:44:56 -07001986 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001987 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
1988}
1989
Herbert Xucbb042f2006-03-20 22:43:56 -08001990unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
1991
Linus Torvalds1da177e2005-04-16 15:20:36 -07001992/**
1993 * pskb_trim_rcsum - trim received skb and update checksum
1994 * @skb: buffer to trim
1995 * @len: new length
1996 *
1997 * This is exactly the same as pskb_trim except that it ensures the
1998 * checksum of received packets are still valid after the operation.
1999 */
2000
2001static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2002{
Stephen Hemminger0e4e4222005-09-08 12:32:03 -07002003 if (likely(len >= skb->len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002004 return 0;
Patrick McHardy84fa7932006-08-29 16:44:56 -07002005 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002006 skb->ip_summed = CHECKSUM_NONE;
2007 return __pskb_trim(skb, len);
2008}
2009
Linus Torvalds1da177e2005-04-16 15:20:36 -07002010#define skb_queue_walk(queue, skb) \
2011 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002012 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002013 skb = skb->next)
2014
James Chapman46f89142007-04-30 00:07:31 -07002015#define skb_queue_walk_safe(queue, skb, tmp) \
2016 for (skb = (queue)->next, tmp = skb->next; \
2017 skb != (struct sk_buff *)(queue); \
2018 skb = tmp, tmp = skb->next)
2019
David S. Miller1164f522008-09-23 00:49:44 -07002020#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002021 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002022 skb = skb->next)
2023
2024#define skb_queue_walk_from_safe(queue, skb, tmp) \
2025 for (tmp = skb->next; \
2026 skb != (struct sk_buff *)(queue); \
2027 skb = tmp, tmp = skb->next)
2028
Stephen Hemminger300ce172005-10-30 13:47:34 -08002029#define skb_queue_reverse_walk(queue, skb) \
2030 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002031 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08002032 skb = skb->prev)
2033
David S. Miller686a2952011-01-20 22:47:32 -08002034#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2035 for (skb = (queue)->prev, tmp = skb->prev; \
2036 skb != (struct sk_buff *)(queue); \
2037 skb = tmp, tmp = skb->prev)
2038
2039#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2040 for (tmp = skb->prev; \
2041 skb != (struct sk_buff *)(queue); \
2042 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002043
David S. Miller21dc3302010-08-23 00:13:46 -07002044static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07002045{
2046 return skb_shinfo(skb)->frag_list != NULL;
2047}
2048
2049static inline void skb_frag_list_init(struct sk_buff *skb)
2050{
2051 skb_shinfo(skb)->frag_list = NULL;
2052}
2053
2054static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2055{
2056 frag->next = skb_shinfo(skb)->frag_list;
2057 skb_shinfo(skb)->frag_list = frag;
2058}
2059
2060#define skb_walk_frags(skb, iter) \
2061 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2062
Herbert Xua59322b2007-12-05 01:53:40 -08002063extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2064 int *peeked, int *err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002065extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
2066 int noblock, int *err);
2067extern unsigned int datagram_poll(struct file *file, struct socket *sock,
2068 struct poll_table_struct *wait);
2069extern int skb_copy_datagram_iovec(const struct sk_buff *from,
2070 int offset, struct iovec *to,
2071 int size);
Herbert Xufb286bb2005-11-10 13:01:24 -08002072extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002073 int hlen,
2074 struct iovec *iov);
Rusty Russelldb543c12008-08-15 15:13:53 -07002075extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
2076 int offset,
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +00002077 const struct iovec *from,
2078 int from_offset,
Rusty Russelldb543c12008-08-15 15:13:53 -07002079 int len);
Michael S. Tsirkin0a1ec072009-04-20 01:25:46 +00002080extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
2081 int offset,
2082 const struct iovec *to,
2083 int to_offset,
2084 int size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002085extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
Eric Dumazet9d410c72009-10-30 05:03:53 +00002086extern void skb_free_datagram_locked(struct sock *sk,
2087 struct sk_buff *skb);
Herbert Xu27ab2562007-12-05 01:51:58 -08002088extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
Herbert Xu3305b802005-12-13 23:16:37 -08002089 unsigned int flags);
Al Viro2bbbc862006-11-14 21:37:14 -08002090extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
2091 int len, __wsum csum);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002092extern int skb_copy_bits(const struct sk_buff *skb, int offset,
2093 void *to, int len);
Stephen Hemminger0c6fcc82007-04-20 16:40:01 -07002094extern int skb_store_bits(struct sk_buff *skb, int offset,
2095 const void *from, int len);
Al Viro81d77662006-11-14 21:37:33 -08002096extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002097 int offset, u8 *to, int len,
Al Viro81d77662006-11-14 21:37:33 -08002098 __wsum csum);
Jens Axboe9c55e012007-11-06 23:30:13 -08002099extern int skb_splice_bits(struct sk_buff *skb,
2100 unsigned int offset,
2101 struct pipe_inode_info *pipe,
2102 unsigned int len,
2103 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002104extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
2105extern void skb_split(struct sk_buff *skb,
2106 struct sk_buff *skb1, const u32 len);
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08002107extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
2108 int shiftlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002109
Michał Mirosławc8f44af2011-11-15 15:29:55 +00002110extern struct sk_buff *skb_segment(struct sk_buff *skb,
2111 netdev_features_t features);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002112
Linus Torvalds1da177e2005-04-16 15:20:36 -07002113static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2114 int len, void *buffer)
2115{
2116 int hlen = skb_headlen(skb);
2117
Patrick McHardy55820ee2005-07-05 14:08:10 -07002118 if (hlen - offset >= len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002119 return skb->data + offset;
2120
2121 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2122 return NULL;
2123
2124 return buffer;
2125}
2126
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03002127static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2128 void *to,
2129 const unsigned int len)
2130{
2131 memcpy(to, skb->data, len);
2132}
2133
2134static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2135 const int offset, void *to,
2136 const unsigned int len)
2137{
2138 memcpy(to, skb->data + offset, len);
2139}
2140
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03002141static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2142 const void *from,
2143 const unsigned int len)
2144{
2145 memcpy(skb->data, from, len);
2146}
2147
2148static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2149 const int offset,
2150 const void *from,
2151 const unsigned int len)
2152{
2153 memcpy(skb->data + offset, from, len);
2154}
2155
Linus Torvalds1da177e2005-04-16 15:20:36 -07002156extern void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002157
Patrick Ohlyac45f602009-02-12 05:03:37 +00002158static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2159{
2160 return skb->tstamp;
2161}
2162
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002163/**
2164 * skb_get_timestamp - get timestamp from a skb
2165 * @skb: skb to get stamp from
2166 * @stamp: pointer to struct timeval to store stamp in
2167 *
2168 * Timestamps are stored in the skb as offsets to a base timestamp.
2169 * This function converts the offset back to a struct timeval and stores
2170 * it in stamp.
2171 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00002172static inline void skb_get_timestamp(const struct sk_buff *skb,
2173 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002174{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002175 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002176}
2177
Patrick Ohlyac45f602009-02-12 05:03:37 +00002178static inline void skb_get_timestampns(const struct sk_buff *skb,
2179 struct timespec *stamp)
2180{
2181 *stamp = ktime_to_timespec(skb->tstamp);
2182}
2183
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002184static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002185{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002186 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002187}
2188
Stephen Hemminger164891a2007-04-23 22:26:16 -07002189static inline ktime_t net_timedelta(ktime_t t)
2190{
2191 return ktime_sub(ktime_get_real(), t);
2192}
2193
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07002194static inline ktime_t net_invalid_timestamp(void)
2195{
2196 return ktime_set(0, 0);
2197}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002198
Richard Cochranc1f19b52010-07-17 08:49:36 +00002199extern void skb_timestamping_init(void);
2200
2201#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2202
2203extern void skb_clone_tx_timestamp(struct sk_buff *skb);
2204extern bool skb_defer_rx_timestamp(struct sk_buff *skb);
2205
2206#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2207
2208static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2209{
2210}
2211
2212static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2213{
2214 return false;
2215}
2216
2217#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2218
2219/**
2220 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2221 *
Richard Cochranda92b192011-10-21 00:49:15 +00002222 * PHY drivers may accept clones of transmitted packets for
2223 * timestamping via their phy_driver.txtstamp method. These drivers
2224 * must call this function to return the skb back to the stack, with
2225 * or without a timestamp.
2226 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00002227 * @skb: clone of the the original outgoing packet
Richard Cochranda92b192011-10-21 00:49:15 +00002228 * @hwtstamps: hardware time stamps, may be NULL if not available
Richard Cochranc1f19b52010-07-17 08:49:36 +00002229 *
2230 */
2231void skb_complete_tx_timestamp(struct sk_buff *skb,
2232 struct skb_shared_hwtstamps *hwtstamps);
2233
Patrick Ohlyac45f602009-02-12 05:03:37 +00002234/**
2235 * skb_tstamp_tx - queue clone of skb with send time stamps
2236 * @orig_skb: the original outgoing packet
2237 * @hwtstamps: hardware time stamps, may be NULL if not available
2238 *
2239 * If the skb has a socket associated, then this function clones the
2240 * skb (thus sharing the actual data and optional structures), stores
2241 * the optional hardware time stamping information (if non NULL) or
2242 * generates a software time stamp (otherwise), then queues the clone
2243 * to the error queue of the socket. Errors are silently ignored.
2244 */
2245extern void skb_tstamp_tx(struct sk_buff *orig_skb,
2246 struct skb_shared_hwtstamps *hwtstamps);
2247
Richard Cochran4507a712010-07-17 08:48:28 +00002248static inline void sw_tx_timestamp(struct sk_buff *skb)
2249{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002250 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2251 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00002252 skb_tstamp_tx(skb, NULL);
2253}
2254
2255/**
2256 * skb_tx_timestamp() - Driver hook for transmit timestamping
2257 *
2258 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00002259 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00002260 *
2261 * @skb: A socket buffer.
2262 */
2263static inline void skb_tx_timestamp(struct sk_buff *skb)
2264{
Richard Cochranc1f19b52010-07-17 08:49:36 +00002265 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002266 sw_tx_timestamp(skb);
2267}
2268
Herbert Xu759e5d02007-03-25 20:10:56 -07002269extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
Al Virob51655b2006-11-14 21:40:42 -08002270extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002271
Herbert Xu60476372007-04-09 11:59:39 -07002272static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2273{
2274 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2275}
2276
Herbert Xufb286bb2005-11-10 13:01:24 -08002277/**
2278 * skb_checksum_complete - Calculate checksum of an entire packet
2279 * @skb: packet to process
2280 *
2281 * This function calculates the checksum over the entire packet plus
2282 * the value of skb->csum. The latter can be used to supply the
2283 * checksum of a pseudo header as used by TCP/UDP. It returns the
2284 * checksum.
2285 *
2286 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2287 * this function can be used to verify that checksum on received
2288 * packets. In that case the function should return zero if the
2289 * checksum is correct. In particular, this function will return zero
2290 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2291 * hardware has already verified the correctness of the checksum.
2292 */
Al Viro4381ca32007-07-15 21:00:11 +01002293static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002294{
Herbert Xu60476372007-04-09 11:59:39 -07002295 return skb_csum_unnecessary(skb) ?
2296 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002297}
2298
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002299#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002300extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002301static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2302{
2303 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002304 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002305}
2306static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2307{
2308 if (nfct)
2309 atomic_inc(&nfct->use);
2310}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002311#endif
2312#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -08002313static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
2314{
2315 if (skb)
2316 atomic_inc(&skb->users);
2317}
2318static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
2319{
2320 if (skb)
2321 kfree_skb(skb);
2322}
2323#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002324#ifdef CONFIG_BRIDGE_NETFILTER
2325static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2326{
2327 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2328 kfree(nf_bridge);
2329}
2330static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2331{
2332 if (nf_bridge)
2333 atomic_inc(&nf_bridge->use);
2334}
2335#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08002336static inline void nf_reset(struct sk_buff *skb)
2337{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002338#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08002339 nf_conntrack_put(skb->nfct);
2340 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002341#endif
2342#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Patrick McHardya193a4a2006-03-20 19:23:05 -08002343 nf_conntrack_put_reasm(skb->nfct_reasm);
2344 skb->nfct_reasm = NULL;
2345#endif
2346#ifdef CONFIG_BRIDGE_NETFILTER
2347 nf_bridge_put(skb->nf_bridge);
2348 skb->nf_bridge = NULL;
2349#endif
2350}
2351
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002352/* Note: This doesn't put any conntrack and bridge info in dst. */
2353static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2354{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002355#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002356 dst->nfct = src->nfct;
2357 nf_conntrack_get(src->nfct);
2358 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002359#endif
2360#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002361 dst->nfct_reasm = src->nfct_reasm;
2362 nf_conntrack_get_reasm(src->nfct_reasm);
2363#endif
2364#ifdef CONFIG_BRIDGE_NETFILTER
2365 dst->nf_bridge = src->nf_bridge;
2366 nf_bridge_get(src->nf_bridge);
2367#endif
2368}
2369
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002370static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2371{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002372#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002373 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002374#endif
2375#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002376 nf_conntrack_put_reasm(dst->nfct_reasm);
2377#endif
2378#ifdef CONFIG_BRIDGE_NETFILTER
2379 nf_bridge_put(dst->nf_bridge);
2380#endif
2381 __nf_copy(dst, src);
2382}
2383
James Morris984bc162006-06-09 00:29:17 -07002384#ifdef CONFIG_NETWORK_SECMARK
2385static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2386{
2387 to->secmark = from->secmark;
2388}
2389
2390static inline void skb_init_secmark(struct sk_buff *skb)
2391{
2392 skb->secmark = 0;
2393}
2394#else
2395static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2396{ }
2397
2398static inline void skb_init_secmark(struct sk_buff *skb)
2399{ }
2400#endif
2401
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002402static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2403{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002404 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002405}
2406
Stephen Hemminger92477442009-03-21 13:39:26 -07002407static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002408{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002409 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002410}
2411
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002412static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2413{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002414 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002415}
2416
David S. Millerd5a9e242009-01-27 16:22:11 -08002417static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2418{
2419 skb->queue_mapping = rx_queue + 1;
2420}
2421
Stephen Hemminger92477442009-03-21 13:39:26 -07002422static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002423{
2424 return skb->queue_mapping - 1;
2425}
2426
Stephen Hemminger92477442009-03-21 13:39:26 -07002427static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002428{
Eric Dumazeta02cec22010-09-22 20:43:57 +00002429 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08002430}
2431
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00002432extern u16 __skb_tx_hash(const struct net_device *dev,
2433 const struct sk_buff *skb,
2434 unsigned int num_tx_queues);
Stephen Hemminger92477442009-03-21 13:39:26 -07002435
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002436#ifdef CONFIG_XFRM
2437static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2438{
2439 return skb->sp;
2440}
2441#else
2442static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2443{
2444 return NULL;
2445}
2446#endif
2447
Herbert Xu89114af2006-07-08 13:34:32 -07002448static inline int skb_is_gso(const struct sk_buff *skb)
2449{
2450 return skb_shinfo(skb)->gso_size;
2451}
2452
Brice Goglineabd7e32007-10-13 12:33:32 +02002453static inline int skb_is_gso_v6(const struct sk_buff *skb)
2454{
2455 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2456}
2457
Ben Hutchings4497b072008-06-19 16:22:28 -07002458extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2459
2460static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2461{
2462 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2463 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002464 const struct skb_shared_info *shinfo = skb_shinfo(skb);
2465
Alexander Duyckb78462e2010-06-02 12:24:37 +00002466 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2467 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07002468 __skb_warn_lro_forwarding(skb);
2469 return true;
2470 }
2471 return false;
2472}
2473
Herbert Xu35fc92a2007-03-26 23:22:20 -07002474static inline void skb_forward_csum(struct sk_buff *skb)
2475{
2476 /* Unfortunately we don't support this one. Any brave souls? */
2477 if (skb->ip_summed == CHECKSUM_COMPLETE)
2478 skb->ip_summed = CHECKSUM_NONE;
2479}
2480
Eric Dumazetbc8acf22010-09-02 13:07:41 -07002481/**
2482 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2483 * @skb: skb to check
2484 *
2485 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2486 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2487 * use this helper, to document places where we make this assertion.
2488 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002489static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07002490{
2491#ifdef DEBUG
2492 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2493#endif
2494}
2495
Rusty Russellf35d9d82008-02-04 23:49:54 -05002496bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00002497
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002498static inline bool skb_is_recycleable(const struct sk_buff *skb, int skb_size)
Andy Fleming3d153a72011-10-13 04:33:54 +00002499{
2500 if (irqs_disabled())
2501 return false;
2502
2503 if (skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)
2504 return false;
2505
2506 if (skb_is_nonlinear(skb) || skb->fclone != SKB_FCLONE_UNAVAILABLE)
2507 return false;
2508
2509 skb_size = SKB_DATA_ALIGN(skb_size + NET_SKB_PAD);
2510 if (skb_end_pointer(skb) - skb->head < skb_size)
2511 return false;
2512
2513 if (skb_shared(skb) || skb_cloned(skb))
2514 return false;
2515
2516 return true;
2517}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002518#endif /* __KERNEL__ */
2519#endif /* _LINUX_SKBUFF_H */