blob: 33370271b8b2c43ac65925619d15305d27a7018c [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>
Paul Gortmaker187f1882011-11-23 20:12:59 -050021#include <linux/bug.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070022#include <linux/cache.h>
23
Arun Sharma600634972011-07-26 16:09:06 -070024#include <linux/atomic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <asm/types.h>
26#include <linux/spinlock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <linux/net.h>
Thomas Graf3fc7e8a2005-06-23 21:00:17 -070028#include <linux/textsearch.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070029#include <net/checksum.h>
Al Viroa80958f2006-12-04 20:41:19 +000030#include <linux/rcupdate.h>
Chris Leech97fc2f02006-05-23 17:55:33 -070031#include <linux/dmaengine.h>
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -070032#include <linux/hrtimer.h>
Ian Campbell131ea662011-08-19 06:25:00 +000033#include <linux/dma-mapping.h>
Michał Mirosławc8f44af2011-11-15 15:29:55 +000034#include <linux/netdev_features.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070035
Herbert Xu60476372007-04-09 11:59:39 -070036/* Don't change this without changing skb_csum_unnecessary! */
Linus Torvalds1da177e2005-04-16 15:20:36 -070037#define CHECKSUM_NONE 0
Herbert Xu60476372007-04-09 11:59:39 -070038#define CHECKSUM_UNNECESSARY 1
39#define CHECKSUM_COMPLETE 2
40#define CHECKSUM_PARTIAL 3
Linus Torvalds1da177e2005-04-16 15:20:36 -070041
42#define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
43 ~(SMP_CACHE_BYTES - 1))
David S. Millerfc910a22007-03-25 20:27:59 -070044#define SKB_WITH_OVERHEAD(X) \
Herbert Xudeea84b2007-10-21 16:27:46 -070045 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
David S. Millerfc910a22007-03-25 20:27:59 -070046#define SKB_MAX_ORDER(X, ORDER) \
47 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
Linus Torvalds1da177e2005-04-16 15:20:36 -070048#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
49#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
50
Eric Dumazet87fb4b72011-10-13 07:28:54 +000051/* return minimum truesize of one skb containing X bytes of data */
52#define SKB_TRUESIZE(X) ((X) + \
53 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
54 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
55
Linus Torvalds1da177e2005-04-16 15:20:36 -070056/* A. Checksumming of received packets by device.
57 *
58 * NONE: device failed to checksum this packet.
59 * skb->csum is undefined.
60 *
61 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
62 * skb->csum is undefined.
63 * It is bad option, but, unfortunately, many of vendors do this.
64 * Apparently with secret goal to sell you new device, when you
65 * will add new protocol to your host. F.e. IPv6. 8)
66 *
Patrick McHardy84fa7932006-08-29 16:44:56 -070067 * COMPLETE: the most generic way. Device supplied checksum of _all_
Linus Torvalds1da177e2005-04-16 15:20:36 -070068 * the packet as seen by netif_rx in skb->csum.
69 * NOTE: Even if device supports only some protocols, but
Patrick McHardy84fa7932006-08-29 16:44:56 -070070 * is able to produce some skb->csum, it MUST use COMPLETE,
Linus Torvalds1da177e2005-04-16 15:20:36 -070071 * not UNNECESSARY.
72 *
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070073 * PARTIAL: identical to the case for output below. This may occur
74 * on a packet received directly from another Linux OS, e.g.,
75 * a virtualised Linux kernel on the same host. The packet can
76 * be treated in the same way as UNNECESSARY except that on
77 * output (i.e., forwarding) the checksum must be filled in
78 * by the OS or the hardware.
79 *
Linus Torvalds1da177e2005-04-16 15:20:36 -070080 * B. Checksumming on output.
81 *
82 * NONE: skb is checksummed by protocol or csum is not required.
83 *
Patrick McHardy84fa7932006-08-29 16:44:56 -070084 * PARTIAL: device is required to csum packet as seen by hard_start_xmit
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070085 * from skb->csum_start to the end and to record the checksum
86 * at skb->csum_start + skb->csum_offset.
Linus Torvalds1da177e2005-04-16 15:20:36 -070087 *
88 * Device must show its capabilities in dev->features, set
89 * at device setup time.
90 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
91 * everything.
Linus Torvalds1da177e2005-04-16 15:20:36 -070092 * 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 *
Yi Zou3af79302012-03-19 11:12:41 +000098 * UNNECESSARY: device will do per protocol specific csum. Protocol drivers
99 * that do not want net to perform the checksum calculation should use
100 * this flag in their outgoing skbs.
101 * NETIF_F_FCOE_CRC this indicates the device can do FCoE FC CRC
102 * offload. Correspondingly, the FCoE protocol driver
103 * stack should use CHECKSUM_UNNECESSARY.
104 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105 * Any questions? No questions, good. --ANK
106 */
107
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108struct net_device;
David Howells716ea3a2007-04-02 20:19:53 -0700109struct scatterlist;
Jens Axboe9c55e012007-11-06 23:30:13 -0800110struct pipe_inode_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700112#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113struct nf_conntrack {
114 atomic_t use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115};
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700116#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117
118#ifdef CONFIG_BRIDGE_NETFILTER
119struct nf_bridge_info {
120 atomic_t use;
121 struct net_device *physindev;
122 struct net_device *physoutdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123 unsigned int mask;
124 unsigned long data[32 / sizeof(unsigned long)];
125};
126#endif
127
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128struct sk_buff_head {
129 /* These two members must be first. */
130 struct sk_buff *next;
131 struct sk_buff *prev;
132
133 __u32 qlen;
134 spinlock_t lock;
135};
136
137struct sk_buff;
138
Ian Campbell9d4dde52011-12-22 23:39:14 +0000139/* To allow 64K frame to be packed as single skb without frag_list we
140 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
141 * buffers which do not start on a page boundary.
142 *
143 * Since GRO uses frags we allocate at least 16 regardless of page
144 * size.
Anton Blancharda715dea2011-03-27 14:57:26 +0000145 */
Ian Campbell9d4dde52011-12-22 23:39:14 +0000146#if (65536/PAGE_SIZE + 1) < 16
David S. Millereec00952011-03-29 23:34:08 -0700147#define MAX_SKB_FRAGS 16UL
Anton Blancharda715dea2011-03-27 14:57:26 +0000148#else
Ian Campbell9d4dde52011-12-22 23:39:14 +0000149#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
Anton Blancharda715dea2011-03-27 14:57:26 +0000150#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151
152typedef struct skb_frag_struct skb_frag_t;
153
154struct skb_frag_struct {
Ian Campbella8605c62011-10-19 23:01:49 +0000155 struct {
156 struct page *p;
157 } page;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000158#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
David S. Millera309bb02007-07-30 18:47:03 -0700159 __u32 page_offset;
160 __u32 size;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000161#else
162 __u16 page_offset;
163 __u16 size;
164#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165};
166
Eric Dumazet9e903e02011-10-18 21:00:24 +0000167static inline unsigned int skb_frag_size(const skb_frag_t *frag)
168{
169 return frag->size;
170}
171
172static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
173{
174 frag->size = size;
175}
176
177static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
178{
179 frag->size += delta;
180}
181
182static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
183{
184 frag->size -= delta;
185}
186
Patrick Ohlyac45f602009-02-12 05:03:37 +0000187#define HAVE_HW_TIME_STAMP
188
189/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800190 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000191 * @hwtstamp: hardware time stamp transformed into duration
192 * since arbitrary point in time
193 * @syststamp: hwtstamp transformed to system time base
194 *
195 * Software time stamps generated by ktime_get_real() are stored in
196 * skb->tstamp. The relation between the different kinds of time
197 * stamps is as follows:
198 *
199 * syststamp and tstamp can be compared against each other in
200 * arbitrary combinations. The accuracy of a
201 * syststamp/tstamp/"syststamp from other device" comparison is
202 * limited by the accuracy of the transformation into system time
203 * base. This depends on the device driver and its underlying
204 * hardware.
205 *
206 * hwtstamps can only be compared against other hwtstamps from
207 * the same device.
208 *
209 * This structure is attached to packets as part of the
210 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
211 */
212struct skb_shared_hwtstamps {
213 ktime_t hwtstamp;
214 ktime_t syststamp;
215};
216
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000217/* Definitions for tx_flags in struct skb_shared_info */
218enum {
219 /* generate hardware time stamp */
220 SKBTX_HW_TSTAMP = 1 << 0,
221
222 /* generate software time stamp */
223 SKBTX_SW_TSTAMP = 1 << 1,
224
225 /* device driver is going to provide hardware time stamp */
226 SKBTX_IN_PROGRESS = 1 << 2,
227
228 /* ensure the originating sk reference is available on driver level */
229 SKBTX_DRV_NEEDS_SK_REF = 1 << 3,
Shirley Maa6686f22011-07-06 12:22:12 +0000230
231 /* device driver supports TX zero-copy buffers */
232 SKBTX_DEV_ZEROCOPY = 1 << 4,
Johannes Berg6e3e9392011-11-09 10:15:42 +0100233
234 /* generate wifi status information (where possible) */
235 SKBTX_WIFI_STATUS = 1 << 5,
Shirley Maa6686f22011-07-06 12:22:12 +0000236};
237
238/*
239 * The callback notifies userspace to release buffers when skb DMA is done in
240 * lower device, the skb last reference should be 0 when calling this.
241 * The desc is used to track userspace buffer index.
242 */
243struct ubuf_info {
244 void (*callback)(void *);
245 void *arg;
246 unsigned long desc;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000247};
248
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249/* This data is invariant across clones and lives at
250 * the end of the header data, ie. at skb->end.
251 */
252struct skb_shared_info {
Ian Campbell9f42f122012-01-05 07:13:39 +0000253 unsigned char nr_frags;
254 __u8 tx_flags;
Herbert Xu79671682006-06-22 02:40:14 -0700255 unsigned short gso_size;
256 /* Warning: this field is not always filled in (UFO)! */
257 unsigned short gso_segs;
258 unsigned short gso_type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000260 struct skb_shared_hwtstamps hwtstamps;
Ian Campbell9f42f122012-01-05 07:13:39 +0000261 __be32 ip6_frag_id;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700262
263 /*
264 * Warning : all fields before dataref are cleared in __alloc_skb()
265 */
266 atomic_t dataref;
267
Johann Baudy69e3c752009-05-18 22:11:22 -0700268 /* Intermediate layers must ensure that destructor_arg
269 * remains valid until skb destructor */
270 void * destructor_arg;
Shirley Maa6686f22011-07-06 12:22:12 +0000271
Eric Dumazetfed66382010-07-22 19:09:08 +0000272 /* must be last field, see pskb_expand_head() */
273 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700274};
275
276/* We divide dataref into two halves. The higher 16 bits hold references
277 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700278 * the entire skb->data. A clone of a headerless skb holds the length of
279 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280 *
281 * All users must obey the rule that the skb->data reference count must be
282 * greater than or equal to the payload reference count.
283 *
284 * Holding a reference to the payload part means that the user does not
285 * care about modifications to the header part of skb->data.
286 */
287#define SKB_DATAREF_SHIFT 16
288#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
289
David S. Millerd179cd12005-08-17 14:57:30 -0700290
291enum {
292 SKB_FCLONE_UNAVAILABLE,
293 SKB_FCLONE_ORIG,
294 SKB_FCLONE_CLONE,
295};
296
Herbert Xu79671682006-06-22 02:40:14 -0700297enum {
298 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700299 SKB_GSO_UDP = 1 << 1,
Herbert Xu576a30e2006-06-27 13:22:38 -0700300
301 /* This indicates the skb is from an untrusted source. */
302 SKB_GSO_DODGY = 1 << 2,
Michael Chanb0da85372006-06-29 12:30:00 -0700303
304 /* This indicates the tcp segment has CWR set. */
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700305 SKB_GSO_TCP_ECN = 1 << 3,
306
307 SKB_GSO_TCPV6 = 1 << 4,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800308
309 SKB_GSO_FCOE = 1 << 5,
Herbert Xu79671682006-06-22 02:40:14 -0700310};
311
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700312#if BITS_PER_LONG > 32
313#define NET_SKBUFF_DATA_USES_OFFSET 1
314#endif
315
316#ifdef NET_SKBUFF_DATA_USES_OFFSET
317typedef unsigned int sk_buff_data_t;
318#else
319typedef unsigned char *sk_buff_data_t;
320#endif
321
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100322#if defined(CONFIG_NF_DEFRAG_IPV4) || defined(CONFIG_NF_DEFRAG_IPV4_MODULE) || \
323 defined(CONFIG_NF_DEFRAG_IPV6) || defined(CONFIG_NF_DEFRAG_IPV6_MODULE)
324#define NET_SKBUFF_NF_DEFRAG_NEEDED 1
325#endif
326
Linus Torvalds1da177e2005-04-16 15:20:36 -0700327/**
328 * struct sk_buff - socket buffer
329 * @next: Next buffer in list
330 * @prev: Previous buffer in list
Herbert Xu325ed822005-10-03 13:57:23 -0700331 * @tstamp: Time we arrived
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700332 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700334 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000335 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700336 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700337 * @len: Length of actual data
338 * @data_len: Data length
339 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700340 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700341 * @csum: Checksum (must include start/offset pair)
342 * @csum_start: Offset from skb->head where checksumming should start
343 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700344 * @priority: Packet queueing priority
Martin Waitz67be2dd2005-05-01 08:59:26 -0700345 * @local_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700347 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348 * @nohdr: Payload reference only, must not modify header
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700349 * @nfctinfo: Relationship of this skb to the connection
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700351 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700352 * @ipvs_property: skbuff is owned by ipvs
Randy Dunlap31729362008-02-18 20:52:13 -0800353 * @peeked: this packet has been seen already, so stats have been
354 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700355 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700356 * @protocol: Packet protocol from driver
357 * @destructor: Destruct function
358 * @nfct: Associated connection, if any
Randy Dunlap461ddf32005-11-20 21:25:15 -0800359 * @nfct_reasm: netfilter conntrack re-assembly pointer
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800361 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362 * @tc_index: Traffic control index
363 * @tc_verd: traffic control verdict
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700364 * @rxhash: the packet hash computed on receive
365 * @queue_mapping: Queue mapping for multiqueue devices
Randy Dunlap553a5672008-04-20 10:51:01 -0700366 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700367 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Changli Gao4ca24622011-08-19 07:26:44 -0700368 * @l4_rxhash: indicate rxhash is a canonical 4-tuple hash over transport
369 * ports.
Johannes Berg6e3e9392011-11-09 10:15:42 +0100370 * @wifi_acked_valid: wifi_acked was set
371 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000372 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Randy Dunlapf4b8ea72006-06-22 16:00:11 -0700373 * @dma_cookie: a cookie to one of several possible DMA operations
374 * done by skb DMA functions
James Morris984bc162006-06-09 00:29:17 -0700375 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700376 * @mark: Generic packet mark
377 * @dropcount: total number of sk_receive_queue overflows
Patrick McHardy6aa895b02008-07-14 22:49:06 -0700378 * @vlan_tci: vlan tag control information
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700379 * @transport_header: Transport layer header
380 * @network_header: Network layer header
381 * @mac_header: Link layer header
382 * @tail: Tail pointer
383 * @end: End pointer
384 * @head: Head of buffer
385 * @data: Data head pointer
386 * @truesize: Buffer size
387 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388 */
389
390struct sk_buff {
391 /* These two members must be first. */
392 struct sk_buff *next;
393 struct sk_buff *prev;
394
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700395 ktime_t tstamp;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000396
397 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400 /*
401 * This is the control buffer. It is free to use for every
402 * layer. Please put your private variables there. If you
403 * want to keep them across layers you have to do a skb_clone()
404 * first. This is owned by whoever has the skb queued ATM.
405 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000406 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407
Eric Dumazet7fee2262010-05-11 23:19:48 +0000408 unsigned long _skb_refdst;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000409#ifdef CONFIG_XFRM
410 struct sec_path *sp;
411#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700413 data_len;
414 __u16 mac_len,
415 hdr_len;
Al Viroff1dcad2006-11-20 18:07:29 -0800416 union {
417 __wsum csum;
Herbert Xu663ead32007-04-09 11:59:07 -0700418 struct {
419 __u16 csum_start;
420 __u16 csum_offset;
421 };
Al Viroff1dcad2006-11-20 18:07:29 -0800422 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423 __u32 priority;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200424 kmemcheck_bitfield_begin(flags1);
Thomas Graf1cbb3382005-07-05 14:13:41 -0700425 __u8 local_df:1,
426 cloned:1,
427 ip_summed:2,
Harald Welte6869c4d2005-08-09 19:24:19 -0700428 nohdr:1,
429 nfctinfo:3;
David S. Millerd179cd12005-08-17 14:57:30 -0700430 __u8 pkt_type:3,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800431 fclone:2,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700432 ipvs_property:1,
Herbert Xua59322b2007-12-05 01:53:40 -0800433 peeked:1,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700434 nf_trace:1;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200435 kmemcheck_bitfield_end(flags1);
Eric Dumazet4ab408d2010-03-01 03:09:26 +0000436 __be16 protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437
438 void (*destructor)(struct sk_buff *skb);
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800439#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700440 struct nf_conntrack *nfct;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100441#endif
442#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800443 struct sk_buff *nfct_reasm;
444#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445#ifdef CONFIG_BRIDGE_NETFILTER
446 struct nf_bridge_info *nf_bridge;
447#endif
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700448
Eric Dumazet8964be42009-11-20 15:35:04 -0800449 int skb_iif;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000450
451 __u32 rxhash;
452
453 __u16 vlan_tci;
454
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700456 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700458 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200461
Eric Dumazet0a148422011-04-20 09:27:32 +0000462 __u16 queue_mapping;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200463 kmemcheck_bitfield_begin(flags2);
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400464#ifdef CONFIG_IPV6_NDISC_NODETYPE
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000465 __u8 ndisc_nodetype:2;
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400466#endif
Tom Herbert3853b582010-11-21 13:17:29 +0000467 __u8 ooo_okay:1;
Tom Herbertbdeab992011-08-14 19:45:55 +0000468 __u8 l4_rxhash:1;
Johannes Berg6e3e9392011-11-09 10:15:42 +0100469 __u8 wifi_acked_valid:1;
470 __u8 wifi_acked:1;
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000471 __u8 no_fcs:1;
472 /* 9/11 bit hole (depending on ndisc_nodetype presence) */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200473 kmemcheck_bitfield_end(flags2);
474
Chris Leech97fc2f02006-05-23 17:55:33 -0700475#ifdef CONFIG_NET_DMA
476 dma_cookie_t dma_cookie;
477#endif
James Morris984bc162006-06-09 00:29:17 -0700478#ifdef CONFIG_NETWORK_SECMARK
479 __u32 secmark;
480#endif
Neil Horman3b885782009-10-12 13:26:31 -0700481 union {
482 __u32 mark;
483 __u32 dropcount;
484 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700486 sk_buff_data_t transport_header;
487 sk_buff_data_t network_header;
488 sk_buff_data_t mac_header;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700490 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700491 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700493 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700494 unsigned int truesize;
495 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496};
497
498#ifdef __KERNEL__
499/*
500 * Handling routines are only of interest to the kernel
501 */
502#include <linux/slab.h>
503
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504
Eric Dumazet7fee2262010-05-11 23:19:48 +0000505/*
506 * skb might have a dst pointer attached, refcounted or not.
507 * _skb_refdst low order bit is set if refcount was _not_ taken
508 */
509#define SKB_DST_NOREF 1UL
510#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
511
512/**
513 * skb_dst - returns skb dst_entry
514 * @skb: buffer
515 *
516 * Returns skb dst_entry, regardless of reference taken or not.
517 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000518static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
519{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000520 /* If refdst was not refcounted, check we still are in a
521 * rcu_read_lock section
522 */
523 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
524 !rcu_read_lock_held() &&
525 !rcu_read_lock_bh_held());
526 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000527}
528
Eric Dumazet7fee2262010-05-11 23:19:48 +0000529/**
530 * skb_dst_set - sets skb dst
531 * @skb: buffer
532 * @dst: dst entry
533 *
534 * Sets skb dst, assuming a reference was taken on dst and should
535 * be released by skb_dst_drop()
536 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000537static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
538{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000539 skb->_skb_refdst = (unsigned long)dst;
540}
541
Eric Dumazet27b75c92010-10-15 05:44:11 +0000542extern void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst);
Eric Dumazet7fee2262010-05-11 23:19:48 +0000543
544/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300545 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000546 * @skb: buffer
547 */
548static inline bool skb_dst_is_noref(const struct sk_buff *skb)
549{
550 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000551}
552
Eric Dumazet511c3f92009-06-02 05:14:27 +0000553static inline struct rtable *skb_rtable(const struct sk_buff *skb)
554{
Eric Dumazetadf30902009-06-02 05:19:30 +0000555 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000556}
557
Jörn Engel231d06a2006-03-20 21:28:35 -0800558extern void kfree_skb(struct sk_buff *skb);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000559extern void consume_skb(struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700560extern void __kfree_skb(struct sk_buff *skb);
David S. Millerd179cd12005-08-17 14:57:30 -0700561extern struct sk_buff *__alloc_skb(unsigned int size,
Christoph Hellwigb30973f2006-12-06 20:32:36 -0800562 gfp_t priority, int fclone, int node);
Eric Dumazetb2b5ce92011-11-14 06:03:34 +0000563extern struct sk_buff *build_skb(void *data);
David S. Millerd179cd12005-08-17 14:57:30 -0700564static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100565 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700566{
Eric Dumazet564824b2010-10-11 19:05:25 +0000567 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700568}
569
570static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100571 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700572{
Eric Dumazet564824b2010-10-11 19:05:25 +0000573 return __alloc_skb(size, priority, 1, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700574}
575
Andy Fleming3d153a72011-10-13 04:33:54 +0000576extern void skb_recycle(struct sk_buff *skb);
Changli Gao5b0daa32010-05-29 00:12:13 -0700577extern bool skb_recycle_check(struct sk_buff *skb, int skb_size);
Lennert Buytenhek04a4bb52008-10-01 02:33:12 -0700578
Herbert Xue0053ec2007-10-14 00:37:52 -0700579extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
Michael S. Tsirkin48c83012011-08-31 08:03:29 +0000580extern int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700581extern struct sk_buff *skb_clone(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100582 gfp_t priority);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700583extern struct sk_buff *skb_copy(const struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100584 gfp_t priority);
Eric Dumazet117632e2011-12-03 21:39:53 +0000585extern struct sk_buff *__pskb_copy(struct sk_buff *skb,
586 int headroom, gfp_t gfp_mask);
587
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588extern int pskb_expand_head(struct sk_buff *skb,
Victor Fusco86a76ca2005-07-08 14:57:47 -0700589 int nhead, int ntail,
Al Virodd0fc662005-10-07 07:46:04 +0100590 gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
592 unsigned int headroom);
593extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
594 int newheadroom, int newtailroom,
Al Virodd0fc662005-10-07 07:46:04 +0100595 gfp_t priority);
David Howells716ea3a2007-04-02 20:19:53 -0700596extern int skb_to_sgvec(struct sk_buff *skb,
597 struct scatterlist *sg, int offset,
598 int len);
599extern int skb_cow_data(struct sk_buff *skb, int tailbits,
600 struct sk_buff **trailer);
Herbert Xu5b057c62006-06-23 02:06:41 -0700601extern int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000602#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700604extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
605 int getfrag(void *from, char *to, int offset,
606 int len,int odd, struct sk_buff *skb),
607 void *from, int length);
608
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800609struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700610 __u32 lower_offset;
611 __u32 upper_offset;
612 __u32 frag_idx;
613 __u32 stepped_offset;
614 struct sk_buff *root_skb;
615 struct sk_buff *cur_skb;
616 __u8 *frag_data;
617};
618
619extern void skb_prepare_seq_read(struct sk_buff *skb,
620 unsigned int from, unsigned int to,
621 struct skb_seq_state *st);
622extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
623 struct skb_seq_state *st);
624extern void skb_abort_seq_read(struct skb_seq_state *st);
625
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700626extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
627 unsigned int to, struct ts_config *config,
628 struct ts_state *state);
629
Tom Herbertbdeab992011-08-14 19:45:55 +0000630extern void __skb_get_rxhash(struct sk_buff *skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000631static inline __u32 skb_get_rxhash(struct sk_buff *skb)
632{
633 if (!skb->rxhash)
Tom Herbertbdeab992011-08-14 19:45:55 +0000634 __skb_get_rxhash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000635
636 return skb->rxhash;
637}
638
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700639#ifdef NET_SKBUFF_DATA_USES_OFFSET
640static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
641{
642 return skb->head + skb->end;
643}
644#else
645static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
646{
647 return skb->end;
648}
649#endif
650
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700652#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653
Patrick Ohlyac45f602009-02-12 05:03:37 +0000654static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
655{
656 return &skb_shinfo(skb)->hwtstamps;
657}
658
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659/**
660 * skb_queue_empty - check if a queue is empty
661 * @list: queue head
662 *
663 * Returns true if the queue is empty, false otherwise.
664 */
665static inline int skb_queue_empty(const struct sk_buff_head *list)
666{
667 return list->next == (struct sk_buff *)list;
668}
669
670/**
David S. Millerfc7ebb22008-09-23 00:34:07 -0700671 * skb_queue_is_last - check if skb is the last entry in the queue
672 * @list: queue head
673 * @skb: buffer
674 *
675 * Returns true if @skb is the last buffer on the list.
676 */
677static inline bool skb_queue_is_last(const struct sk_buff_head *list,
678 const struct sk_buff *skb)
679{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000680 return skb->next == (struct sk_buff *)list;
David S. Millerfc7ebb22008-09-23 00:34:07 -0700681}
682
683/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800684 * skb_queue_is_first - check if skb is the first entry in the queue
685 * @list: queue head
686 * @skb: buffer
687 *
688 * Returns true if @skb is the first buffer on the list.
689 */
690static inline bool skb_queue_is_first(const struct sk_buff_head *list,
691 const struct sk_buff *skb)
692{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000693 return skb->prev == (struct sk_buff *)list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800694}
695
696/**
David S. Miller249c8b42008-09-23 00:44:42 -0700697 * skb_queue_next - return the next packet in the queue
698 * @list: queue head
699 * @skb: current buffer
700 *
701 * Return the next packet in @list after @skb. It is only valid to
702 * call this if skb_queue_is_last() evaluates to false.
703 */
704static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
705 const struct sk_buff *skb)
706{
707 /* This BUG_ON may seem severe, but if we just return then we
708 * are going to dereference garbage.
709 */
710 BUG_ON(skb_queue_is_last(list, skb));
711 return skb->next;
712}
713
714/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800715 * skb_queue_prev - return the prev packet in the queue
716 * @list: queue head
717 * @skb: current buffer
718 *
719 * Return the prev packet in @list before @skb. It is only valid to
720 * call this if skb_queue_is_first() evaluates to false.
721 */
722static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
723 const struct sk_buff *skb)
724{
725 /* This BUG_ON may seem severe, but if we just return then we
726 * are going to dereference garbage.
727 */
728 BUG_ON(skb_queue_is_first(list, skb));
729 return skb->prev;
730}
731
732/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700733 * skb_get - reference buffer
734 * @skb: buffer to reference
735 *
736 * Makes another reference to a socket buffer and returns a pointer
737 * to the buffer.
738 */
739static inline struct sk_buff *skb_get(struct sk_buff *skb)
740{
741 atomic_inc(&skb->users);
742 return skb;
743}
744
745/*
746 * If users == 1, we are the only owner and are can avoid redundant
747 * atomic change.
748 */
749
750/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751 * skb_cloned - is the buffer a clone
752 * @skb: buffer to check
753 *
754 * Returns true if the buffer was generated with skb_clone() and is
755 * one of multiple shared copies of the buffer. Cloned buffers are
756 * shared data so must not be written to under normal circumstances.
757 */
758static inline int skb_cloned(const struct sk_buff *skb)
759{
760 return skb->cloned &&
761 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
762}
763
764/**
765 * skb_header_cloned - is the header a clone
766 * @skb: buffer to check
767 *
768 * Returns true if modifying the header part of the buffer requires
769 * the data to be copied.
770 */
771static inline int skb_header_cloned(const struct sk_buff *skb)
772{
773 int dataref;
774
775 if (!skb->cloned)
776 return 0;
777
778 dataref = atomic_read(&skb_shinfo(skb)->dataref);
779 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
780 return dataref != 1;
781}
782
783/**
784 * skb_header_release - release reference to header
785 * @skb: buffer to operate on
786 *
787 * Drop a reference to the header part of the buffer. This is done
788 * by acquiring a payload reference. You must not read from the header
789 * part of skb->data after this.
790 */
791static inline void skb_header_release(struct sk_buff *skb)
792{
793 BUG_ON(skb->nohdr);
794 skb->nohdr = 1;
795 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
796}
797
798/**
799 * skb_shared - is the buffer shared
800 * @skb: buffer to check
801 *
802 * Returns true if more than one person has a reference to this
803 * buffer.
804 */
805static inline int skb_shared(const struct sk_buff *skb)
806{
807 return atomic_read(&skb->users) != 1;
808}
809
810/**
811 * skb_share_check - check if buffer is shared and if so clone it
812 * @skb: buffer to check
813 * @pri: priority for memory allocation
814 *
815 * If the buffer is shared the buffer is cloned and the old copy
816 * drops a reference. A new clone with a single reference is returned.
817 * If the buffer is not shared the original buffer is returned. When
818 * being called from interrupt status or with spinlocks held pri must
819 * be GFP_ATOMIC.
820 *
821 * NULL is returned on a memory allocation failure.
822 */
Victor Fusco86a76ca2005-07-08 14:57:47 -0700823static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100824 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700825{
826 might_sleep_if(pri & __GFP_WAIT);
827 if (skb_shared(skb)) {
828 struct sk_buff *nskb = skb_clone(skb, pri);
829 kfree_skb(skb);
830 skb = nskb;
831 }
832 return skb;
833}
834
835/*
836 * Copy shared buffers into a new sk_buff. We effectively do COW on
837 * packets to handle cases where we have a local reader and forward
838 * and a couple of other messy ones. The normal one is tcpdumping
839 * a packet thats being forwarded.
840 */
841
842/**
843 * skb_unshare - make a copy of a shared buffer
844 * @skb: buffer to check
845 * @pri: priority for memory allocation
846 *
847 * If the socket buffer is a clone then this function creates a new
848 * copy of the data, drops a reference count on the old copy and returns
849 * the new copy with the reference count at 1. If the buffer is not a clone
850 * the original buffer is returned. When called with a spinlock held or
851 * from interrupt state @pri must be %GFP_ATOMIC
852 *
853 * %NULL is returned on a memory allocation failure.
854 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -0700855static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100856 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857{
858 might_sleep_if(pri & __GFP_WAIT);
859 if (skb_cloned(skb)) {
860 struct sk_buff *nskb = skb_copy(skb, pri);
861 kfree_skb(skb); /* Free our shared copy */
862 skb = nskb;
863 }
864 return skb;
865}
866
867/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -0800868 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869 * @list_: list to peek at
870 *
871 * Peek an &sk_buff. Unlike most other operations you _MUST_
872 * be careful with this one. A peek leaves the buffer on the
873 * list and someone else may run off with it. You must hold
874 * the appropriate locks or have a private queue to do this.
875 *
876 * Returns %NULL for an empty list or a pointer to the head element.
877 * The reference count is not incremented and the reference is therefore
878 * volatile. Use with caution.
879 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -0400880static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881{
Eric Dumazet05bdd2f2011-10-20 17:45:43 -0400882 struct sk_buff *list = ((const struct sk_buff *)list_)->next;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883 if (list == (struct sk_buff *)list_)
884 list = NULL;
885 return list;
886}
887
888/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +0000889 * skb_peek_next - peek skb following the given one from a queue
890 * @skb: skb to start from
891 * @list_: list to peek at
892 *
893 * Returns %NULL when the end of the list is met or a pointer to the
894 * next element. The reference count is not incremented and the
895 * reference is therefore volatile. Use with caution.
896 */
897static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
898 const struct sk_buff_head *list_)
899{
900 struct sk_buff *next = skb->next;
901 if (next == (struct sk_buff *)list_)
902 next = NULL;
903 return next;
904}
905
906/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -0800907 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908 * @list_: list to peek at
909 *
910 * Peek an &sk_buff. Unlike most other operations you _MUST_
911 * be careful with this one. A peek leaves the buffer on the
912 * list and someone else may run off with it. You must hold
913 * the appropriate locks or have a private queue to do this.
914 *
915 * Returns %NULL for an empty list or a pointer to the tail element.
916 * The reference count is not incremented and the reference is therefore
917 * volatile. Use with caution.
918 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -0400919static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920{
Eric Dumazet05bdd2f2011-10-20 17:45:43 -0400921 struct sk_buff *list = ((const struct sk_buff *)list_)->prev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922 if (list == (struct sk_buff *)list_)
923 list = NULL;
924 return list;
925}
926
927/**
928 * skb_queue_len - get queue length
929 * @list_: list to measure
930 *
931 * Return the length of an &sk_buff queue.
932 */
933static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
934{
935 return list_->qlen;
936}
937
David S. Miller67fed452008-09-21 22:36:24 -0700938/**
939 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
940 * @list: queue to initialize
941 *
942 * This initializes only the list and queue length aspects of
943 * an sk_buff_head object. This allows to initialize the list
944 * aspects of an sk_buff_head without reinitializing things like
945 * the spinlock. It can also be used for on-stack sk_buff_head
946 * objects where the spinlock is known to not be used.
947 */
948static inline void __skb_queue_head_init(struct sk_buff_head *list)
949{
950 list->prev = list->next = (struct sk_buff *)list;
951 list->qlen = 0;
952}
953
Arjan van de Ven76f10ad2006-08-02 14:06:55 -0700954/*
955 * This function creates a split out lock class for each invocation;
956 * this is needed for now since a whole lot of users of the skb-queue
957 * infrastructure in drivers have different locking usage (in hardirq)
958 * than the networking core (in softirq only). In the long run either the
959 * network layer or drivers should need annotation to consolidate the
960 * main types of usage into 3 classes.
961 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700962static inline void skb_queue_head_init(struct sk_buff_head *list)
963{
964 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -0700965 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966}
967
Pavel Emelianovc2ecba72007-04-17 12:45:31 -0700968static inline void skb_queue_head_init_class(struct sk_buff_head *list,
969 struct lock_class_key *class)
970{
971 skb_queue_head_init(list);
972 lockdep_set_class(&list->lock, class);
973}
974
Linus Torvalds1da177e2005-04-16 15:20:36 -0700975/*
Gerrit Renkerbf299272008-04-14 00:04:51 -0700976 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977 *
978 * The "__skb_xxxx()" functions are the non-atomic ones that
979 * can only be called with interrupts disabled.
980 */
Gerrit Renkerbf299272008-04-14 00:04:51 -0700981extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
982static inline void __skb_insert(struct sk_buff *newsk,
983 struct sk_buff *prev, struct sk_buff *next,
984 struct sk_buff_head *list)
985{
986 newsk->next = next;
987 newsk->prev = prev;
988 next->prev = prev->next = newsk;
989 list->qlen++;
990}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700991
David S. Miller67fed452008-09-21 22:36:24 -0700992static inline void __skb_queue_splice(const struct sk_buff_head *list,
993 struct sk_buff *prev,
994 struct sk_buff *next)
995{
996 struct sk_buff *first = list->next;
997 struct sk_buff *last = list->prev;
998
999 first->prev = prev;
1000 prev->next = first;
1001
1002 last->next = next;
1003 next->prev = last;
1004}
1005
1006/**
1007 * skb_queue_splice - join two skb lists, this is designed for stacks
1008 * @list: the new list to add
1009 * @head: the place to add it in the first list
1010 */
1011static inline void skb_queue_splice(const struct sk_buff_head *list,
1012 struct sk_buff_head *head)
1013{
1014 if (!skb_queue_empty(list)) {
1015 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001016 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001017 }
1018}
1019
1020/**
1021 * skb_queue_splice - join two skb lists and reinitialise the emptied list
1022 * @list: the new list to add
1023 * @head: the place to add it in the first list
1024 *
1025 * The list at @list is reinitialised
1026 */
1027static inline void skb_queue_splice_init(struct sk_buff_head *list,
1028 struct sk_buff_head *head)
1029{
1030 if (!skb_queue_empty(list)) {
1031 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001032 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001033 __skb_queue_head_init(list);
1034 }
1035}
1036
1037/**
1038 * skb_queue_splice_tail - join two skb lists, each list being a queue
1039 * @list: the new list to add
1040 * @head: the place to add it in the first list
1041 */
1042static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1043 struct sk_buff_head *head)
1044{
1045 if (!skb_queue_empty(list)) {
1046 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001047 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001048 }
1049}
1050
1051/**
1052 * skb_queue_splice_tail - join two skb lists and reinitialise the emptied list
1053 * @list: the new list to add
1054 * @head: the place to add it in the first list
1055 *
1056 * Each of the lists is a queue.
1057 * The list at @list is reinitialised
1058 */
1059static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1060 struct sk_buff_head *head)
1061{
1062 if (!skb_queue_empty(list)) {
1063 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001064 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001065 __skb_queue_head_init(list);
1066 }
1067}
1068
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001070 * __skb_queue_after - queue a buffer at the list head
1071 * @list: list to use
1072 * @prev: place after this buffer
1073 * @newsk: buffer to queue
1074 *
1075 * Queue a buffer int the middle of a list. This function takes no locks
1076 * and you must therefore hold required locks before calling it.
1077 *
1078 * A buffer cannot be placed on two lists at the same time.
1079 */
1080static inline void __skb_queue_after(struct sk_buff_head *list,
1081 struct sk_buff *prev,
1082 struct sk_buff *newsk)
1083{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001084 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001085}
1086
Gerrit Renker7de6c032008-04-14 00:05:09 -07001087extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1088 struct sk_buff_head *list);
1089
Gerrit Renkerf5572852008-04-14 00:05:28 -07001090static inline void __skb_queue_before(struct sk_buff_head *list,
1091 struct sk_buff *next,
1092 struct sk_buff *newsk)
1093{
1094 __skb_insert(newsk, next->prev, next, list);
1095}
1096
Stephen Hemminger300ce172005-10-30 13:47:34 -08001097/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098 * __skb_queue_head - queue a buffer at the list head
1099 * @list: list to use
1100 * @newsk: buffer to queue
1101 *
1102 * Queue a buffer at the start of a list. This function takes no locks
1103 * and you must therefore hold required locks before calling it.
1104 *
1105 * A buffer cannot be placed on two lists at the same time.
1106 */
1107extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1108static inline void __skb_queue_head(struct sk_buff_head *list,
1109 struct sk_buff *newsk)
1110{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001111 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112}
1113
1114/**
1115 * __skb_queue_tail - queue a buffer at the list tail
1116 * @list: list to use
1117 * @newsk: buffer to queue
1118 *
1119 * Queue a buffer at the end of a list. This function takes no locks
1120 * and you must therefore hold required locks before calling it.
1121 *
1122 * A buffer cannot be placed on two lists at the same time.
1123 */
1124extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1125static inline void __skb_queue_tail(struct sk_buff_head *list,
1126 struct sk_buff *newsk)
1127{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001128 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001129}
1130
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132 * remove sk_buff from list. _Must_ be called atomically, and with
1133 * the list known..
1134 */
David S. Miller8728b832005-08-09 19:25:21 -07001135extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1137{
1138 struct sk_buff *next, *prev;
1139
1140 list->qlen--;
1141 next = skb->next;
1142 prev = skb->prev;
1143 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144 next->prev = prev;
1145 prev->next = next;
1146}
1147
Gerrit Renkerf525c062008-04-14 00:04:12 -07001148/**
1149 * __skb_dequeue - remove from the head of the queue
1150 * @list: list to dequeue from
1151 *
1152 * Remove the head of the list. This function does not take any locks
1153 * so must be used with appropriate locks held only. The head item is
1154 * returned or %NULL if the list is empty.
1155 */
1156extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1157static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1158{
1159 struct sk_buff *skb = skb_peek(list);
1160 if (skb)
1161 __skb_unlink(skb, list);
1162 return skb;
1163}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164
1165/**
1166 * __skb_dequeue_tail - remove from the tail of the queue
1167 * @list: list to dequeue from
1168 *
1169 * Remove the tail of the list. This function does not take any locks
1170 * so must be used with appropriate locks held only. The tail item is
1171 * returned or %NULL if the list is empty.
1172 */
1173extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1174static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1175{
1176 struct sk_buff *skb = skb_peek_tail(list);
1177 if (skb)
1178 __skb_unlink(skb, list);
1179 return skb;
1180}
1181
1182
David S. Millerbdcc0922012-03-07 20:53:36 -05001183static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001184{
1185 return skb->data_len;
1186}
1187
1188static inline unsigned int skb_headlen(const struct sk_buff *skb)
1189{
1190 return skb->len - skb->data_len;
1191}
1192
1193static inline int skb_pagelen(const struct sk_buff *skb)
1194{
1195 int i, len = 0;
1196
1197 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001198 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001199 return len + skb_headlen(skb);
1200}
1201
Ian Campbell131ea662011-08-19 06:25:00 +00001202/**
1203 * __skb_fill_page_desc - initialise a paged fragment in an skb
1204 * @skb: buffer containing fragment to be initialised
1205 * @i: paged fragment index to initialise
1206 * @page: the page to use for this fragment
1207 * @off: the offset to the data with @page
1208 * @size: the length of the data
1209 *
1210 * Initialises the @i'th fragment of @skb to point to &size bytes at
1211 * offset @off within @page.
1212 *
1213 * Does not take any additional reference on the fragment.
1214 */
1215static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1216 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001217{
1218 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1219
Ian Campbella8605c62011-10-19 23:01:49 +00001220 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001221 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001222 skb_frag_size_set(frag, size);
Ian Campbell131ea662011-08-19 06:25:00 +00001223}
1224
1225/**
1226 * skb_fill_page_desc - initialise a paged fragment in an skb
1227 * @skb: buffer containing fragment to be initialised
1228 * @i: paged fragment index to initialise
1229 * @page: the page to use for this fragment
1230 * @off: the offset to the data with @page
1231 * @size: the length of the data
1232 *
1233 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1234 * @skb to point to &size bytes at offset @off within @page. In
1235 * addition updates @skb such that @i is the last fragment.
1236 *
1237 * Does not take any additional reference on the fragment.
1238 */
1239static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1240 struct page *page, int off, int size)
1241{
1242 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001243 skb_shinfo(skb)->nr_frags = i + 1;
1244}
1245
Peter Zijlstra654bed12008-10-07 14:22:33 -07001246extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
Eric Dumazet50269e12012-03-23 23:59:33 +00001247 int off, int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001248
Linus Torvalds1da177e2005-04-16 15:20:36 -07001249#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001250#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001251#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1252
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001253#ifdef NET_SKBUFF_DATA_USES_OFFSET
1254static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1255{
1256 return skb->head + skb->tail;
1257}
1258
1259static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1260{
1261 skb->tail = skb->data - skb->head;
1262}
1263
1264static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1265{
1266 skb_reset_tail_pointer(skb);
1267 skb->tail += offset;
1268}
1269#else /* NET_SKBUFF_DATA_USES_OFFSET */
1270static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1271{
1272 return skb->tail;
1273}
1274
1275static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1276{
1277 skb->tail = skb->data;
1278}
1279
1280static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1281{
1282 skb->tail = skb->data + offset;
1283}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001284
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001285#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1286
Linus Torvalds1da177e2005-04-16 15:20:36 -07001287/*
1288 * Add data to an sk_buff
1289 */
Ilpo Järvinen0dde3e12008-03-27 17:43:41 -07001290extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001291static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1292{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001293 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001294 SKB_LINEAR_ASSERT(skb);
1295 skb->tail += len;
1296 skb->len += len;
1297 return tmp;
1298}
1299
Ilpo Järvinenc2aa2702008-03-27 17:52:40 -07001300extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001301static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1302{
1303 skb->data -= len;
1304 skb->len += len;
1305 return skb->data;
1306}
1307
Ilpo Järvinen6be8ac22008-03-27 17:47:24 -07001308extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001309static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1310{
1311 skb->len -= len;
1312 BUG_ON(skb->len < skb->data_len);
1313 return skb->data += len;
1314}
1315
David S. Miller47d29642010-05-02 02:21:44 -07001316static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1317{
1318 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1319}
1320
Linus Torvalds1da177e2005-04-16 15:20:36 -07001321extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1322
1323static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1324{
1325 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001326 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001327 return NULL;
1328 skb->len -= len;
1329 return skb->data += len;
1330}
1331
1332static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1333{
1334 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1335}
1336
1337static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1338{
1339 if (likely(len <= skb_headlen(skb)))
1340 return 1;
1341 if (unlikely(len > skb->len))
1342 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001343 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344}
1345
1346/**
1347 * skb_headroom - bytes at buffer head
1348 * @skb: buffer to check
1349 *
1350 * Return the number of bytes of free space at the head of an &sk_buff.
1351 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001352static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001353{
1354 return skb->data - skb->head;
1355}
1356
1357/**
1358 * skb_tailroom - bytes at buffer end
1359 * @skb: buffer to check
1360 *
1361 * Return the number of bytes of free space at the tail of an sk_buff
1362 */
1363static inline int skb_tailroom(const struct sk_buff *skb)
1364{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001365 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001366}
1367
1368/**
1369 * skb_reserve - adjust headroom
1370 * @skb: buffer to alter
1371 * @len: bytes to move
1372 *
1373 * Increase the headroom of an empty &sk_buff by reducing the tail
1374 * room. This is only allowed for an empty buffer.
1375 */
David S. Miller8243126c2006-01-17 02:54:21 -08001376static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001377{
1378 skb->data += len;
1379 skb->tail += len;
1380}
1381
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00001382static inline void skb_reset_mac_len(struct sk_buff *skb)
1383{
1384 skb->mac_len = skb->network_header - skb->mac_header;
1385}
1386
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001387#ifdef NET_SKBUFF_DATA_USES_OFFSET
1388static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1389{
1390 return skb->head + skb->transport_header;
1391}
1392
1393static inline void skb_reset_transport_header(struct sk_buff *skb)
1394{
1395 skb->transport_header = skb->data - skb->head;
1396}
1397
1398static inline void skb_set_transport_header(struct sk_buff *skb,
1399 const int offset)
1400{
1401 skb_reset_transport_header(skb);
1402 skb->transport_header += offset;
1403}
1404
1405static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1406{
1407 return skb->head + skb->network_header;
1408}
1409
1410static inline void skb_reset_network_header(struct sk_buff *skb)
1411{
1412 skb->network_header = skb->data - skb->head;
1413}
1414
1415static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1416{
1417 skb_reset_network_header(skb);
1418 skb->network_header += offset;
1419}
1420
1421static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1422{
1423 return skb->head + skb->mac_header;
1424}
1425
1426static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1427{
1428 return skb->mac_header != ~0U;
1429}
1430
1431static inline void skb_reset_mac_header(struct sk_buff *skb)
1432{
1433 skb->mac_header = skb->data - skb->head;
1434}
1435
1436static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1437{
1438 skb_reset_mac_header(skb);
1439 skb->mac_header += offset;
1440}
1441
1442#else /* NET_SKBUFF_DATA_USES_OFFSET */
1443
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001444static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1445{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001446 return skb->transport_header;
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001447}
1448
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001449static inline void skb_reset_transport_header(struct sk_buff *skb)
1450{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001451 skb->transport_header = skb->data;
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001452}
1453
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001454static inline void skb_set_transport_header(struct sk_buff *skb,
1455 const int offset)
1456{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001457 skb->transport_header = skb->data + offset;
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001458}
1459
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001460static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1461{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001462 return skb->network_header;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001463}
1464
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001465static inline void skb_reset_network_header(struct sk_buff *skb)
1466{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001467 skb->network_header = skb->data;
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001468}
1469
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001470static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1471{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001472 skb->network_header = skb->data + offset;
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001473}
1474
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001475static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1476{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001477 return skb->mac_header;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001478}
1479
1480static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1481{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001482 return skb->mac_header != NULL;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001483}
1484
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001485static inline void skb_reset_mac_header(struct sk_buff *skb)
1486{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001487 skb->mac_header = skb->data;
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001488}
1489
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001490static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1491{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001492 skb->mac_header = skb->data + offset;
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001493}
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001494#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1495
Eric Dumazet03606892012-02-23 10:55:02 +00001496static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1497{
1498 if (skb_mac_header_was_set(skb)) {
1499 const unsigned char *old_mac = skb_mac_header(skb);
1500
1501 skb_set_mac_header(skb, -skb->mac_len);
1502 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1503 }
1504}
1505
Michał Mirosław04fb4512010-12-14 15:24:08 +00001506static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1507{
1508 return skb->csum_start - skb_headroom(skb);
1509}
1510
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001511static inline int skb_transport_offset(const struct sk_buff *skb)
1512{
1513 return skb_transport_header(skb) - skb->data;
1514}
1515
1516static inline u32 skb_network_header_len(const struct sk_buff *skb)
1517{
1518 return skb->transport_header - skb->network_header;
1519}
1520
1521static inline int skb_network_offset(const struct sk_buff *skb)
1522{
1523 return skb_network_header(skb) - skb->data;
1524}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001525
Changli Gaof9599ce2010-08-04 04:43:44 +00001526static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1527{
1528 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1529}
1530
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531/*
1532 * CPUs often take a performance hit when accessing unaligned memory
1533 * locations. The actual performance hit varies, it can be small if the
1534 * hardware handles it or large if we have to take an exception and fix it
1535 * in software.
1536 *
1537 * Since an ethernet header is 14 bytes network drivers often end up with
1538 * the IP header at an unaligned offset. The IP header can be aligned by
1539 * shifting the start of the packet by 2 bytes. Drivers should do this
1540 * with:
1541 *
Tobias Klauser8660c122009-07-13 22:48:16 +00001542 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001543 *
1544 * The downside to this alignment of the IP header is that the DMA is now
1545 * unaligned. On some architectures the cost of an unaligned DMA is high
1546 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00001547 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001548 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1549 * to be overridden.
1550 */
1551#ifndef NET_IP_ALIGN
1552#define NET_IP_ALIGN 2
1553#endif
1554
Anton Blanchard025be812006-03-31 02:27:06 -08001555/*
1556 * The networking layer reserves some headroom in skb data (via
1557 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1558 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08001559 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08001560 *
1561 * Unfortunately this headroom changes the DMA alignment of the resulting
1562 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1563 * on some architectures. An architecture can override this value,
1564 * perhaps setting it to a cacheline in size (since that will maintain
1565 * cacheline alignment of the DMA). It must be a power of 2.
1566 *
David S. Millerd6301d32009-02-08 19:24:13 -08001567 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08001568 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07001569 *
1570 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1571 * to reduce average number of cache lines per packet.
1572 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07001573 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08001574 */
1575#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07001576#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08001577#endif
1578
Herbert Xu3cc0e872006-06-09 16:13:38 -07001579extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001580
1581static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1582{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00001583 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07001584 WARN_ON(1);
1585 return;
1586 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001587 skb->len = len;
1588 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001589}
1590
Ilpo Järvinen419ae742008-03-27 17:54:01 -07001591extern void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001592
1593static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1594{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001595 if (skb->data_len)
1596 return ___pskb_trim(skb, len);
1597 __skb_trim(skb, len);
1598 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001599}
1600
1601static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1602{
1603 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1604}
1605
1606/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07001607 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1608 * @skb: buffer to alter
1609 * @len: new length
1610 *
1611 * This is identical to pskb_trim except that the caller knows that
1612 * the skb is not cloned so we should never get an error due to out-
1613 * of-memory.
1614 */
1615static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1616{
1617 int err = pskb_trim(skb, len);
1618 BUG_ON(err);
1619}
1620
1621/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001622 * skb_orphan - orphan a buffer
1623 * @skb: buffer to orphan
1624 *
1625 * If a buffer currently has an owner then we call the owner's
1626 * destructor function and make the @skb unowned. The buffer continues
1627 * to exist but is no longer charged to its former owner.
1628 */
1629static inline void skb_orphan(struct sk_buff *skb)
1630{
1631 if (skb->destructor)
1632 skb->destructor(skb);
1633 skb->destructor = NULL;
1634 skb->sk = NULL;
1635}
1636
1637/**
1638 * __skb_queue_purge - empty a list
1639 * @list: list to empty
1640 *
1641 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1642 * the list and one reference dropped. This function does not take the
1643 * list lock and the caller must hold the relevant locks to use it.
1644 */
1645extern void skb_queue_purge(struct sk_buff_head *list);
1646static inline void __skb_queue_purge(struct sk_buff_head *list)
1647{
1648 struct sk_buff *skb;
1649 while ((skb = __skb_dequeue(list)) != NULL)
1650 kfree_skb(skb);
1651}
1652
1653/**
Christoph Hellwigb4e54de2006-07-24 15:31:14 -07001654 * __dev_alloc_skb - allocate an skbuff for receiving
Linus Torvalds1da177e2005-04-16 15:20:36 -07001655 * @length: length to allocate
1656 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1657 *
1658 * Allocate a new &sk_buff and assign it a usage count of one. The
1659 * buffer has unspecified headroom built in. Users should allocate
1660 * the headroom they think they need without accounting for the
1661 * built in space. The built in space is used for optimisations.
1662 *
Christoph Hellwig766ea8c2006-08-07 15:49:53 -07001663 * %NULL is returned if there is no free memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001664 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001665static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
Al Virodd0fc662005-10-07 07:46:04 +01001666 gfp_t gfp_mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001667{
Anton Blanchard025be812006-03-31 02:27:06 -08001668 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669 if (likely(skb))
Anton Blanchard025be812006-03-31 02:27:06 -08001670 skb_reserve(skb, NET_SKB_PAD);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001671 return skb;
1672}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001673
Ilpo Järvinenf58518e2008-03-27 17:51:31 -07001674extern struct sk_buff *dev_alloc_skb(unsigned int length);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001675
Christoph Hellwig8af27452006-07-31 22:35:23 -07001676extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1677 unsigned int length, gfp_t gfp_mask);
1678
1679/**
1680 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1681 * @dev: network device to receive on
1682 * @length: length to allocate
1683 *
1684 * Allocate a new &sk_buff and assign it a usage count of one. The
1685 * buffer has unspecified headroom built in. Users should allocate
1686 * the headroom they think they need without accounting for the
1687 * built in space. The built in space is used for optimisations.
1688 *
1689 * %NULL is returned if there is no free memory. Although this function
1690 * allocates memory it can be called from an interrupt.
1691 */
1692static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1693 unsigned int length)
1694{
1695 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1696}
1697
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001698static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
1699 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00001700{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001701 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00001702
1703 if (NET_IP_ALIGN && skb)
1704 skb_reserve(skb, NET_IP_ALIGN);
1705 return skb;
1706}
1707
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001708static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1709 unsigned int length)
1710{
1711 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
1712}
1713
Eric Dumazet564824b2010-10-11 19:05:25 +00001714/**
Ian Campbell131ea662011-08-19 06:25:00 +00001715 * skb_frag_page - retrieve the page refered to by a paged fragment
1716 * @frag: the paged fragment
1717 *
1718 * Returns the &struct page associated with @frag.
1719 */
1720static inline struct page *skb_frag_page(const skb_frag_t *frag)
1721{
Ian Campbella8605c62011-10-19 23:01:49 +00001722 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00001723}
1724
1725/**
1726 * __skb_frag_ref - take an addition reference on a paged fragment.
1727 * @frag: the paged fragment
1728 *
1729 * Takes an additional reference on the paged fragment @frag.
1730 */
1731static inline void __skb_frag_ref(skb_frag_t *frag)
1732{
1733 get_page(skb_frag_page(frag));
1734}
1735
1736/**
1737 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
1738 * @skb: the buffer
1739 * @f: the fragment offset.
1740 *
1741 * Takes an additional reference on the @f'th paged fragment of @skb.
1742 */
1743static inline void skb_frag_ref(struct sk_buff *skb, int f)
1744{
1745 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
1746}
1747
1748/**
1749 * __skb_frag_unref - release a reference on a paged fragment.
1750 * @frag: the paged fragment
1751 *
1752 * Releases a reference on the paged fragment @frag.
1753 */
1754static inline void __skb_frag_unref(skb_frag_t *frag)
1755{
1756 put_page(skb_frag_page(frag));
1757}
1758
1759/**
1760 * skb_frag_unref - release a reference on a paged fragment of an skb.
1761 * @skb: the buffer
1762 * @f: the fragment offset
1763 *
1764 * Releases a reference on the @f'th paged fragment of @skb.
1765 */
1766static inline void skb_frag_unref(struct sk_buff *skb, int f)
1767{
1768 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
1769}
1770
1771/**
1772 * skb_frag_address - gets the address of the data contained in a paged fragment
1773 * @frag: the paged fragment buffer
1774 *
1775 * Returns the address of the data within @frag. The page must already
1776 * be mapped.
1777 */
1778static inline void *skb_frag_address(const skb_frag_t *frag)
1779{
1780 return page_address(skb_frag_page(frag)) + frag->page_offset;
1781}
1782
1783/**
1784 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
1785 * @frag: the paged fragment buffer
1786 *
1787 * Returns the address of the data within @frag. Checks that the page
1788 * is mapped and returns %NULL otherwise.
1789 */
1790static inline void *skb_frag_address_safe(const skb_frag_t *frag)
1791{
1792 void *ptr = page_address(skb_frag_page(frag));
1793 if (unlikely(!ptr))
1794 return NULL;
1795
1796 return ptr + frag->page_offset;
1797}
1798
1799/**
1800 * __skb_frag_set_page - sets the page contained in a paged fragment
1801 * @frag: the paged fragment
1802 * @page: the page to set
1803 *
1804 * Sets the fragment @frag to contain @page.
1805 */
1806static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
1807{
Ian Campbella8605c62011-10-19 23:01:49 +00001808 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00001809}
1810
1811/**
1812 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
1813 * @skb: the buffer
1814 * @f: the fragment offset
1815 * @page: the page to set
1816 *
1817 * Sets the @f'th fragment of @skb to contain @page.
1818 */
1819static inline void skb_frag_set_page(struct sk_buff *skb, int f,
1820 struct page *page)
1821{
1822 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
1823}
1824
1825/**
1826 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00001827 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00001828 * @frag: the paged fragment to map
1829 * @offset: the offset within the fragment (starting at the
1830 * fragment's own offset)
1831 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00001832 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00001833 *
1834 * Maps the page associated with @frag to @device.
1835 */
1836static inline dma_addr_t skb_frag_dma_map(struct device *dev,
1837 const skb_frag_t *frag,
1838 size_t offset, size_t size,
1839 enum dma_data_direction dir)
1840{
1841 return dma_map_page(dev, skb_frag_page(frag),
1842 frag->page_offset + offset, size, dir);
1843}
1844
Eric Dumazet117632e2011-12-03 21:39:53 +00001845static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
1846 gfp_t gfp_mask)
1847{
1848 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
1849}
1850
Ian Campbell131ea662011-08-19 06:25:00 +00001851/**
Patrick McHardy334a8132007-06-25 04:35:20 -07001852 * skb_clone_writable - is the header of a clone writable
1853 * @skb: buffer to check
1854 * @len: length up to which to write
1855 *
1856 * Returns true if modifying the header part of the cloned buffer
1857 * does not requires the data to be copied.
1858 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001859static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07001860{
1861 return !skb_header_cloned(skb) &&
1862 skb_headroom(skb) + len <= skb->hdr_len;
1863}
1864
Herbert Xud9cc2042007-09-16 16:21:16 -07001865static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
1866 int cloned)
1867{
1868 int delta = 0;
1869
1870 if (headroom < NET_SKB_PAD)
1871 headroom = NET_SKB_PAD;
1872 if (headroom > skb_headroom(skb))
1873 delta = headroom - skb_headroom(skb);
1874
1875 if (delta || cloned)
1876 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
1877 GFP_ATOMIC);
1878 return 0;
1879}
1880
Patrick McHardy334a8132007-06-25 04:35:20 -07001881/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001882 * skb_cow - copy header of skb when it is required
1883 * @skb: buffer to cow
1884 * @headroom: needed headroom
1885 *
1886 * If the skb passed lacks sufficient headroom or its data part
1887 * is shared, data is reallocated. If reallocation fails, an error
1888 * is returned and original skb is not changed.
1889 *
1890 * The result is skb with writable area skb->head...skb->tail
1891 * and at least @headroom of space at head.
1892 */
1893static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1894{
Herbert Xud9cc2042007-09-16 16:21:16 -07001895 return __skb_cow(skb, headroom, skb_cloned(skb));
1896}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001897
Herbert Xud9cc2042007-09-16 16:21:16 -07001898/**
1899 * skb_cow_head - skb_cow but only making the head writable
1900 * @skb: buffer to cow
1901 * @headroom: needed headroom
1902 *
1903 * This function is identical to skb_cow except that we replace the
1904 * skb_cloned check by skb_header_cloned. It should be used when
1905 * you only need to push on some header and do not need to modify
1906 * the data.
1907 */
1908static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
1909{
1910 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001911}
1912
1913/**
1914 * skb_padto - pad an skbuff up to a minimal size
1915 * @skb: buffer to pad
1916 * @len: minimal length
1917 *
1918 * Pads up a buffer to ensure the trailing bytes exist and are
1919 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07001920 * is untouched. Otherwise it is extended. Returns zero on
1921 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001922 */
1923
Herbert Xu5b057c62006-06-23 02:06:41 -07001924static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001925{
1926 unsigned int size = skb->len;
1927 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07001928 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001929 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001930}
1931
1932static inline int skb_add_data(struct sk_buff *skb,
1933 char __user *from, int copy)
1934{
1935 const int off = skb->len;
1936
1937 if (skb->ip_summed == CHECKSUM_NONE) {
1938 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08001939 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
Linus Torvalds1da177e2005-04-16 15:20:36 -07001940 copy, 0, &err);
1941 if (!err) {
1942 skb->csum = csum_block_add(skb->csum, csum, off);
1943 return 0;
1944 }
1945 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1946 return 0;
1947
1948 __skb_trim(skb, off);
1949 return -EFAULT;
1950}
1951
1952static inline int skb_can_coalesce(struct sk_buff *skb, int i,
Ian Campbellea2ab692011-08-22 23:44:58 +00001953 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001954{
1955 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00001956 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957
Ian Campbellea2ab692011-08-22 23:44:58 +00001958 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00001959 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001960 }
1961 return 0;
1962}
1963
Herbert Xu364c6ba2006-06-09 16:10:40 -07001964static inline int __skb_linearize(struct sk_buff *skb)
1965{
1966 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1967}
1968
Linus Torvalds1da177e2005-04-16 15:20:36 -07001969/**
1970 * skb_linearize - convert paged skb to linear one
1971 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07001972 *
1973 * If there is no free memory -ENOMEM is returned, otherwise zero
1974 * is returned and the old skb data released.
1975 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07001976static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001977{
Herbert Xu364c6ba2006-06-09 16:10:40 -07001978 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1979}
1980
1981/**
1982 * skb_linearize_cow - make sure skb is linear and writable
1983 * @skb: buffer to process
1984 *
1985 * If there is no free memory -ENOMEM is returned, otherwise zero
1986 * is returned and the old skb data released.
1987 */
1988static inline int skb_linearize_cow(struct sk_buff *skb)
1989{
1990 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
1991 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001992}
1993
1994/**
1995 * skb_postpull_rcsum - update checksum for received skb after pull
1996 * @skb: buffer to update
1997 * @start: start of data before pull
1998 * @len: length of data pulled
1999 *
2000 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002001 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2002 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002003 */
2004
2005static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002006 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002007{
Patrick McHardy84fa7932006-08-29 16:44:56 -07002008 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2010}
2011
Herbert Xucbb042f2006-03-20 22:43:56 -08002012unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2013
Linus Torvalds1da177e2005-04-16 15:20:36 -07002014/**
2015 * pskb_trim_rcsum - trim received skb and update checksum
2016 * @skb: buffer to trim
2017 * @len: new length
2018 *
2019 * This is exactly the same as pskb_trim except that it ensures the
2020 * checksum of received packets are still valid after the operation.
2021 */
2022
2023static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2024{
Stephen Hemminger0e4e4222005-09-08 12:32:03 -07002025 if (likely(len >= skb->len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002026 return 0;
Patrick McHardy84fa7932006-08-29 16:44:56 -07002027 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002028 skb->ip_summed = CHECKSUM_NONE;
2029 return __pskb_trim(skb, len);
2030}
2031
Linus Torvalds1da177e2005-04-16 15:20:36 -07002032#define skb_queue_walk(queue, skb) \
2033 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002034 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002035 skb = skb->next)
2036
James Chapman46f89142007-04-30 00:07:31 -07002037#define skb_queue_walk_safe(queue, skb, tmp) \
2038 for (skb = (queue)->next, tmp = skb->next; \
2039 skb != (struct sk_buff *)(queue); \
2040 skb = tmp, tmp = skb->next)
2041
David S. Miller1164f522008-09-23 00:49:44 -07002042#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002043 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002044 skb = skb->next)
2045
2046#define skb_queue_walk_from_safe(queue, skb, tmp) \
2047 for (tmp = skb->next; \
2048 skb != (struct sk_buff *)(queue); \
2049 skb = tmp, tmp = skb->next)
2050
Stephen Hemminger300ce172005-10-30 13:47:34 -08002051#define skb_queue_reverse_walk(queue, skb) \
2052 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002053 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08002054 skb = skb->prev)
2055
David S. Miller686a2952011-01-20 22:47:32 -08002056#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2057 for (skb = (queue)->prev, tmp = skb->prev; \
2058 skb != (struct sk_buff *)(queue); \
2059 skb = tmp, tmp = skb->prev)
2060
2061#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2062 for (tmp = skb->prev; \
2063 skb != (struct sk_buff *)(queue); \
2064 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002065
David S. Miller21dc3302010-08-23 00:13:46 -07002066static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07002067{
2068 return skb_shinfo(skb)->frag_list != NULL;
2069}
2070
2071static inline void skb_frag_list_init(struct sk_buff *skb)
2072{
2073 skb_shinfo(skb)->frag_list = NULL;
2074}
2075
2076static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2077{
2078 frag->next = skb_shinfo(skb)->frag_list;
2079 skb_shinfo(skb)->frag_list = frag;
2080}
2081
2082#define skb_walk_frags(skb, iter) \
2083 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2084
Herbert Xua59322b2007-12-05 01:53:40 -08002085extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
Pavel Emelyanov3f518bf2012-02-21 07:30:58 +00002086 int *peeked, int *off, int *err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002087extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
2088 int noblock, int *err);
2089extern unsigned int datagram_poll(struct file *file, struct socket *sock,
2090 struct poll_table_struct *wait);
2091extern int skb_copy_datagram_iovec(const struct sk_buff *from,
2092 int offset, struct iovec *to,
2093 int size);
Herbert Xufb286bb2005-11-10 13:01:24 -08002094extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002095 int hlen,
2096 struct iovec *iov);
Rusty Russelldb543c12008-08-15 15:13:53 -07002097extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
2098 int offset,
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +00002099 const struct iovec *from,
2100 int from_offset,
Rusty Russelldb543c12008-08-15 15:13:53 -07002101 int len);
Michael S. Tsirkin0a1ec072009-04-20 01:25:46 +00002102extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
2103 int offset,
2104 const struct iovec *to,
2105 int to_offset,
2106 int size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002107extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
Eric Dumazet9d410c72009-10-30 05:03:53 +00002108extern void skb_free_datagram_locked(struct sock *sk,
2109 struct sk_buff *skb);
Herbert Xu27ab2562007-12-05 01:51:58 -08002110extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
Herbert Xu3305b802005-12-13 23:16:37 -08002111 unsigned int flags);
Al Viro2bbbc862006-11-14 21:37:14 -08002112extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
2113 int len, __wsum csum);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002114extern int skb_copy_bits(const struct sk_buff *skb, int offset,
2115 void *to, int len);
Stephen Hemminger0c6fcc82007-04-20 16:40:01 -07002116extern int skb_store_bits(struct sk_buff *skb, int offset,
2117 const void *from, int len);
Al Viro81d77662006-11-14 21:37:33 -08002118extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002119 int offset, u8 *to, int len,
Al Viro81d77662006-11-14 21:37:33 -08002120 __wsum csum);
Jens Axboe9c55e012007-11-06 23:30:13 -08002121extern int skb_splice_bits(struct sk_buff *skb,
2122 unsigned int offset,
2123 struct pipe_inode_info *pipe,
2124 unsigned int len,
2125 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002126extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
2127extern void skb_split(struct sk_buff *skb,
2128 struct sk_buff *skb1, const u32 len);
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08002129extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
2130 int shiftlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002131
Michał Mirosławc8f44af2011-11-15 15:29:55 +00002132extern struct sk_buff *skb_segment(struct sk_buff *skb,
2133 netdev_features_t features);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002134
Linus Torvalds1da177e2005-04-16 15:20:36 -07002135static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2136 int len, void *buffer)
2137{
2138 int hlen = skb_headlen(skb);
2139
Patrick McHardy55820ee2005-07-05 14:08:10 -07002140 if (hlen - offset >= len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002141 return skb->data + offset;
2142
2143 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2144 return NULL;
2145
2146 return buffer;
2147}
2148
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03002149static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2150 void *to,
2151 const unsigned int len)
2152{
2153 memcpy(to, skb->data, len);
2154}
2155
2156static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2157 const int offset, void *to,
2158 const unsigned int len)
2159{
2160 memcpy(to, skb->data + offset, len);
2161}
2162
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03002163static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2164 const void *from,
2165 const unsigned int len)
2166{
2167 memcpy(skb->data, from, len);
2168}
2169
2170static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2171 const int offset,
2172 const void *from,
2173 const unsigned int len)
2174{
2175 memcpy(skb->data + offset, from, len);
2176}
2177
Linus Torvalds1da177e2005-04-16 15:20:36 -07002178extern void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002179
Patrick Ohlyac45f602009-02-12 05:03:37 +00002180static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2181{
2182 return skb->tstamp;
2183}
2184
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002185/**
2186 * skb_get_timestamp - get timestamp from a skb
2187 * @skb: skb to get stamp from
2188 * @stamp: pointer to struct timeval to store stamp in
2189 *
2190 * Timestamps are stored in the skb as offsets to a base timestamp.
2191 * This function converts the offset back to a struct timeval and stores
2192 * it in stamp.
2193 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00002194static inline void skb_get_timestamp(const struct sk_buff *skb,
2195 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002196{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002197 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002198}
2199
Patrick Ohlyac45f602009-02-12 05:03:37 +00002200static inline void skb_get_timestampns(const struct sk_buff *skb,
2201 struct timespec *stamp)
2202{
2203 *stamp = ktime_to_timespec(skb->tstamp);
2204}
2205
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002206static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002207{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002208 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002209}
2210
Stephen Hemminger164891a2007-04-23 22:26:16 -07002211static inline ktime_t net_timedelta(ktime_t t)
2212{
2213 return ktime_sub(ktime_get_real(), t);
2214}
2215
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07002216static inline ktime_t net_invalid_timestamp(void)
2217{
2218 return ktime_set(0, 0);
2219}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002220
Richard Cochranc1f19b52010-07-17 08:49:36 +00002221extern void skb_timestamping_init(void);
2222
2223#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2224
2225extern void skb_clone_tx_timestamp(struct sk_buff *skb);
2226extern bool skb_defer_rx_timestamp(struct sk_buff *skb);
2227
2228#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2229
2230static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2231{
2232}
2233
2234static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2235{
2236 return false;
2237}
2238
2239#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2240
2241/**
2242 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2243 *
Richard Cochranda92b192011-10-21 00:49:15 +00002244 * PHY drivers may accept clones of transmitted packets for
2245 * timestamping via their phy_driver.txtstamp method. These drivers
2246 * must call this function to return the skb back to the stack, with
2247 * or without a timestamp.
2248 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00002249 * @skb: clone of the the original outgoing packet
Richard Cochranda92b192011-10-21 00:49:15 +00002250 * @hwtstamps: hardware time stamps, may be NULL if not available
Richard Cochranc1f19b52010-07-17 08:49:36 +00002251 *
2252 */
2253void skb_complete_tx_timestamp(struct sk_buff *skb,
2254 struct skb_shared_hwtstamps *hwtstamps);
2255
Patrick Ohlyac45f602009-02-12 05:03:37 +00002256/**
2257 * skb_tstamp_tx - queue clone of skb with send time stamps
2258 * @orig_skb: the original outgoing packet
2259 * @hwtstamps: hardware time stamps, may be NULL if not available
2260 *
2261 * If the skb has a socket associated, then this function clones the
2262 * skb (thus sharing the actual data and optional structures), stores
2263 * the optional hardware time stamping information (if non NULL) or
2264 * generates a software time stamp (otherwise), then queues the clone
2265 * to the error queue of the socket. Errors are silently ignored.
2266 */
2267extern void skb_tstamp_tx(struct sk_buff *orig_skb,
2268 struct skb_shared_hwtstamps *hwtstamps);
2269
Richard Cochran4507a712010-07-17 08:48:28 +00002270static inline void sw_tx_timestamp(struct sk_buff *skb)
2271{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002272 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2273 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00002274 skb_tstamp_tx(skb, NULL);
2275}
2276
2277/**
2278 * skb_tx_timestamp() - Driver hook for transmit timestamping
2279 *
2280 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00002281 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00002282 *
2283 * @skb: A socket buffer.
2284 */
2285static inline void skb_tx_timestamp(struct sk_buff *skb)
2286{
Richard Cochranc1f19b52010-07-17 08:49:36 +00002287 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002288 sw_tx_timestamp(skb);
2289}
2290
Johannes Berg6e3e9392011-11-09 10:15:42 +01002291/**
2292 * skb_complete_wifi_ack - deliver skb with wifi status
2293 *
2294 * @skb: the original outgoing packet
2295 * @acked: ack status
2296 *
2297 */
2298void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2299
Herbert Xu759e5d02007-03-25 20:10:56 -07002300extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
Al Virob51655b2006-11-14 21:40:42 -08002301extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002302
Herbert Xu60476372007-04-09 11:59:39 -07002303static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2304{
2305 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2306}
2307
Herbert Xufb286bb2005-11-10 13:01:24 -08002308/**
2309 * skb_checksum_complete - Calculate checksum of an entire packet
2310 * @skb: packet to process
2311 *
2312 * This function calculates the checksum over the entire packet plus
2313 * the value of skb->csum. The latter can be used to supply the
2314 * checksum of a pseudo header as used by TCP/UDP. It returns the
2315 * checksum.
2316 *
2317 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2318 * this function can be used to verify that checksum on received
2319 * packets. In that case the function should return zero if the
2320 * checksum is correct. In particular, this function will return zero
2321 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2322 * hardware has already verified the correctness of the checksum.
2323 */
Al Viro4381ca32007-07-15 21:00:11 +01002324static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002325{
Herbert Xu60476372007-04-09 11:59:39 -07002326 return skb_csum_unnecessary(skb) ?
2327 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002328}
2329
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002330#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002331extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002332static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2333{
2334 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002335 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002336}
2337static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2338{
2339 if (nfct)
2340 atomic_inc(&nfct->use);
2341}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002342#endif
2343#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -08002344static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
2345{
2346 if (skb)
2347 atomic_inc(&skb->users);
2348}
2349static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
2350{
2351 if (skb)
2352 kfree_skb(skb);
2353}
2354#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002355#ifdef CONFIG_BRIDGE_NETFILTER
2356static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2357{
2358 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2359 kfree(nf_bridge);
2360}
2361static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2362{
2363 if (nf_bridge)
2364 atomic_inc(&nf_bridge->use);
2365}
2366#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08002367static inline void nf_reset(struct sk_buff *skb)
2368{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002369#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08002370 nf_conntrack_put(skb->nfct);
2371 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002372#endif
2373#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Patrick McHardya193a4a2006-03-20 19:23:05 -08002374 nf_conntrack_put_reasm(skb->nfct_reasm);
2375 skb->nfct_reasm = NULL;
2376#endif
2377#ifdef CONFIG_BRIDGE_NETFILTER
2378 nf_bridge_put(skb->nf_bridge);
2379 skb->nf_bridge = NULL;
2380#endif
2381}
2382
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002383/* Note: This doesn't put any conntrack and bridge info in dst. */
2384static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2385{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002386#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002387 dst->nfct = src->nfct;
2388 nf_conntrack_get(src->nfct);
2389 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002390#endif
2391#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002392 dst->nfct_reasm = src->nfct_reasm;
2393 nf_conntrack_get_reasm(src->nfct_reasm);
2394#endif
2395#ifdef CONFIG_BRIDGE_NETFILTER
2396 dst->nf_bridge = src->nf_bridge;
2397 nf_bridge_get(src->nf_bridge);
2398#endif
2399}
2400
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002401static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2402{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002403#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002404 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002405#endif
2406#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002407 nf_conntrack_put_reasm(dst->nfct_reasm);
2408#endif
2409#ifdef CONFIG_BRIDGE_NETFILTER
2410 nf_bridge_put(dst->nf_bridge);
2411#endif
2412 __nf_copy(dst, src);
2413}
2414
James Morris984bc162006-06-09 00:29:17 -07002415#ifdef CONFIG_NETWORK_SECMARK
2416static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2417{
2418 to->secmark = from->secmark;
2419}
2420
2421static inline void skb_init_secmark(struct sk_buff *skb)
2422{
2423 skb->secmark = 0;
2424}
2425#else
2426static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2427{ }
2428
2429static inline void skb_init_secmark(struct sk_buff *skb)
2430{ }
2431#endif
2432
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002433static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2434{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002435 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002436}
2437
Stephen Hemminger92477442009-03-21 13:39:26 -07002438static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002439{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002440 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002441}
2442
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002443static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2444{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002445 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002446}
2447
David S. Millerd5a9e242009-01-27 16:22:11 -08002448static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2449{
2450 skb->queue_mapping = rx_queue + 1;
2451}
2452
Stephen Hemminger92477442009-03-21 13:39:26 -07002453static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002454{
2455 return skb->queue_mapping - 1;
2456}
2457
Stephen Hemminger92477442009-03-21 13:39:26 -07002458static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002459{
Eric Dumazeta02cec22010-09-22 20:43:57 +00002460 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08002461}
2462
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00002463extern u16 __skb_tx_hash(const struct net_device *dev,
2464 const struct sk_buff *skb,
2465 unsigned int num_tx_queues);
Stephen Hemminger92477442009-03-21 13:39:26 -07002466
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002467#ifdef CONFIG_XFRM
2468static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2469{
2470 return skb->sp;
2471}
2472#else
2473static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2474{
2475 return NULL;
2476}
2477#endif
2478
David S. Millerbdcc0922012-03-07 20:53:36 -05002479static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07002480{
2481 return skb_shinfo(skb)->gso_size;
2482}
2483
David S. Millerbdcc0922012-03-07 20:53:36 -05002484static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02002485{
2486 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2487}
2488
Ben Hutchings4497b072008-06-19 16:22:28 -07002489extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2490
2491static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2492{
2493 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2494 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002495 const struct skb_shared_info *shinfo = skb_shinfo(skb);
2496
Alexander Duyckb78462e2010-06-02 12:24:37 +00002497 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2498 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07002499 __skb_warn_lro_forwarding(skb);
2500 return true;
2501 }
2502 return false;
2503}
2504
Herbert Xu35fc92a2007-03-26 23:22:20 -07002505static inline void skb_forward_csum(struct sk_buff *skb)
2506{
2507 /* Unfortunately we don't support this one. Any brave souls? */
2508 if (skb->ip_summed == CHECKSUM_COMPLETE)
2509 skb->ip_summed = CHECKSUM_NONE;
2510}
2511
Eric Dumazetbc8acf22010-09-02 13:07:41 -07002512/**
2513 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2514 * @skb: skb to check
2515 *
2516 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2517 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2518 * use this helper, to document places where we make this assertion.
2519 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002520static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07002521{
2522#ifdef DEBUG
2523 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2524#endif
2525}
2526
Rusty Russellf35d9d82008-02-04 23:49:54 -05002527bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00002528
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002529static inline bool skb_is_recycleable(const struct sk_buff *skb, int skb_size)
Andy Fleming3d153a72011-10-13 04:33:54 +00002530{
2531 if (irqs_disabled())
2532 return false;
2533
2534 if (skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)
2535 return false;
2536
2537 if (skb_is_nonlinear(skb) || skb->fclone != SKB_FCLONE_UNAVAILABLE)
2538 return false;
2539
2540 skb_size = SKB_DATA_ALIGN(skb_size + NET_SKB_PAD);
2541 if (skb_end_pointer(skb) - skb->head < skb_size)
2542 return false;
2543
2544 if (skb_shared(skb) || skb_cloned(skb))
2545 return false;
2546
2547 return true;
2548}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002549#endif /* __KERNEL__ */
2550#endif /* _LINUX_SKBUFF_H */