blob: 8b2256e880e0ea135892b2deb5f5957dfc7568e9 [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 {
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000120 atomic_t use;
121 unsigned int mask;
122 struct net_device *physindev;
123 struct net_device *physoutdev;
124 unsigned long data[32 / sizeof(unsigned long)];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125};
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
Shirley Maa6686f22011-07-06 12:22:12 +0000228 /* device driver supports TX zero-copy buffers */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000229 SKBTX_DEV_ZEROCOPY = 1 << 3,
Johannes Berg6e3e9392011-11-09 10:15:42 +0100230
231 /* generate wifi status information (where possible) */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000232 SKBTX_WIFI_STATUS = 1 << 4,
Shirley Maa6686f22011-07-06 12:22:12 +0000233};
234
235/*
236 * The callback notifies userspace to release buffers when skb DMA is done in
237 * lower device, the skb last reference should be 0 when calling this.
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000238 * The zerocopy_success argument is true if zero copy transmit occurred,
239 * false on data copy or out of memory error caused by data copy attempt.
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000240 * The ctx field is used to track device context.
241 * The desc field is used to track userspace buffer index.
Shirley Maa6686f22011-07-06 12:22:12 +0000242 */
243struct ubuf_info {
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000244 void (*callback)(struct ubuf_info *, bool zerocopy_success);
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000245 void *ctx;
Shirley Maa6686f22011-07-06 12:22:12 +0000246 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
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000379 * @inner_transport_header: Inner transport layer header (encapsulation)
380 * @inner_network_header: Network layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700381 * @transport_header: Transport layer header
382 * @network_header: Network layer header
383 * @mac_header: Link layer header
384 * @tail: Tail pointer
385 * @end: End pointer
386 * @head: Head of buffer
387 * @data: Data head pointer
388 * @truesize: Buffer size
389 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700390 */
391
392struct sk_buff {
393 /* These two members must be first. */
394 struct sk_buff *next;
395 struct sk_buff *prev;
396
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700397 ktime_t tstamp;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000398
399 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401
Linus Torvalds1da177e2005-04-16 15:20:36 -0700402 /*
403 * This is the control buffer. It is free to use for every
404 * layer. Please put your private variables there. If you
405 * want to keep them across layers you have to do a skb_clone()
406 * first. This is owned by whoever has the skb queued ATM.
407 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000408 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409
Eric Dumazet7fee2262010-05-11 23:19:48 +0000410 unsigned long _skb_refdst;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000411#ifdef CONFIG_XFRM
412 struct sec_path *sp;
413#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700415 data_len;
416 __u16 mac_len,
417 hdr_len;
Al Viroff1dcad2006-11-20 18:07:29 -0800418 union {
419 __wsum csum;
Herbert Xu663ead32007-04-09 11:59:07 -0700420 struct {
421 __u16 csum_start;
422 __u16 csum_offset;
423 };
Al Viroff1dcad2006-11-20 18:07:29 -0800424 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425 __u32 priority;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200426 kmemcheck_bitfield_begin(flags1);
Thomas Graf1cbb3382005-07-05 14:13:41 -0700427 __u8 local_df:1,
428 cloned:1,
429 ip_summed:2,
Harald Welte6869c4d2005-08-09 19:24:19 -0700430 nohdr:1,
431 nfctinfo:3;
David S. Millerd179cd12005-08-17 14:57:30 -0700432 __u8 pkt_type:3,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800433 fclone:2,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700434 ipvs_property:1,
Herbert Xua59322b2007-12-05 01:53:40 -0800435 peeked:1,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700436 nf_trace:1;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200437 kmemcheck_bitfield_end(flags1);
Eric Dumazet4ab408d2010-03-01 03:09:26 +0000438 __be16 protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439
440 void (*destructor)(struct sk_buff *skb);
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800441#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700442 struct nf_conntrack *nfct;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100443#endif
444#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800445 struct sk_buff *nfct_reasm;
446#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447#ifdef CONFIG_BRIDGE_NETFILTER
448 struct nf_bridge_info *nf_bridge;
449#endif
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700450
Eric Dumazet8964be42009-11-20 15:35:04 -0800451 int skb_iif;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000452
453 __u32 rxhash;
454
455 __u16 vlan_tci;
456
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700458 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700460 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700462#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200463
Eric Dumazet0a148422011-04-20 09:27:32 +0000464 __u16 queue_mapping;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200465 kmemcheck_bitfield_begin(flags2);
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400466#ifdef CONFIG_IPV6_NDISC_NODETYPE
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000467 __u8 ndisc_nodetype:2;
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400468#endif
Mel Gormanc93bdd02012-07-31 16:44:19 -0700469 __u8 pfmemalloc:1;
Tom Herbert3853b582010-11-21 13:17:29 +0000470 __u8 ooo_okay:1;
Tom Herbertbdeab992011-08-14 19:45:55 +0000471 __u8 l4_rxhash:1;
Johannes Berg6e3e9392011-11-09 10:15:42 +0100472 __u8 wifi_acked_valid:1;
473 __u8 wifi_acked:1;
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000474 __u8 no_fcs:1;
Eric Dumazetd3836f22012-04-27 00:33:38 +0000475 __u8 head_frag:1;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000476 /* Encapsulation protocol and NIC drivers should use
477 * this flag to indicate to each other if the skb contains
478 * encapsulated packet or not and maybe use the inner packet
479 * headers if needed
480 */
481 __u8 encapsulation:1;
482 /* 7/9 bit hole (depending on ndisc_nodetype presence) */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200483 kmemcheck_bitfield_end(flags2);
484
Chris Leech97fc2f02006-05-23 17:55:33 -0700485#ifdef CONFIG_NET_DMA
486 dma_cookie_t dma_cookie;
487#endif
James Morris984bc162006-06-09 00:29:17 -0700488#ifdef CONFIG_NETWORK_SECMARK
489 __u32 secmark;
490#endif
Neil Horman3b885782009-10-12 13:26:31 -0700491 union {
492 __u32 mark;
493 __u32 dropcount;
Eric Dumazeta21d4572012-04-10 20:30:48 +0000494 __u32 avail_size;
Neil Horman3b885782009-10-12 13:26:31 -0700495 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000497 sk_buff_data_t inner_transport_header;
498 sk_buff_data_t inner_network_header;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700499 sk_buff_data_t transport_header;
500 sk_buff_data_t network_header;
501 sk_buff_data_t mac_header;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700502 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700503 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700504 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700506 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700507 unsigned int truesize;
508 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700509};
510
511#ifdef __KERNEL__
512/*
513 * Handling routines are only of interest to the kernel
514 */
515#include <linux/slab.h>
516
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517
Mel Gormanc93bdd02012-07-31 16:44:19 -0700518#define SKB_ALLOC_FCLONE 0x01
519#define SKB_ALLOC_RX 0x02
520
521/* Returns true if the skb was allocated from PFMEMALLOC reserves */
522static inline bool skb_pfmemalloc(const struct sk_buff *skb)
523{
524 return unlikely(skb->pfmemalloc);
525}
526
Eric Dumazet7fee2262010-05-11 23:19:48 +0000527/*
528 * skb might have a dst pointer attached, refcounted or not.
529 * _skb_refdst low order bit is set if refcount was _not_ taken
530 */
531#define SKB_DST_NOREF 1UL
532#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
533
534/**
535 * skb_dst - returns skb dst_entry
536 * @skb: buffer
537 *
538 * Returns skb dst_entry, regardless of reference taken or not.
539 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000540static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
541{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000542 /* If refdst was not refcounted, check we still are in a
543 * rcu_read_lock section
544 */
545 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
546 !rcu_read_lock_held() &&
547 !rcu_read_lock_bh_held());
548 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000549}
550
Eric Dumazet7fee2262010-05-11 23:19:48 +0000551/**
552 * skb_dst_set - sets skb dst
553 * @skb: buffer
554 * @dst: dst entry
555 *
556 * Sets skb dst, assuming a reference was taken on dst and should
557 * be released by skb_dst_drop()
558 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000559static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
560{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000561 skb->_skb_refdst = (unsigned long)dst;
562}
563
Eric Dumazet27b75c92010-10-15 05:44:11 +0000564extern void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst);
Eric Dumazet7fee2262010-05-11 23:19:48 +0000565
566/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300567 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000568 * @skb: buffer
569 */
570static inline bool skb_dst_is_noref(const struct sk_buff *skb)
571{
572 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000573}
574
Eric Dumazet511c3f92009-06-02 05:14:27 +0000575static inline struct rtable *skb_rtable(const struct sk_buff *skb)
576{
Eric Dumazetadf30902009-06-02 05:19:30 +0000577 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000578}
579
Jörn Engel231d06a2006-03-20 21:28:35 -0800580extern void kfree_skb(struct sk_buff *skb);
Michael S. Tsirkin25121172012-11-01 09:16:28 +0000581extern void skb_tx_error(struct sk_buff *skb);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000582extern void consume_skb(struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583extern void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000584extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000585
586extern void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
587extern bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
588 bool *fragstolen, int *delta_truesize);
589
David S. Millerd179cd12005-08-17 14:57:30 -0700590extern struct sk_buff *__alloc_skb(unsigned int size,
Mel Gormanc93bdd02012-07-31 16:44:19 -0700591 gfp_t priority, int flags, int node);
Eric Dumazetd3836f22012-04-27 00:33:38 +0000592extern struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700593static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100594 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700595{
Eric Dumazet564824b2010-10-11 19:05:25 +0000596 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700597}
598
599static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100600 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700601{
Mel Gormanc93bdd02012-07-31 16:44:19 -0700602 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700603}
604
Herbert Xue0053ec2007-10-14 00:37:52 -0700605extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
Michael S. Tsirkin48c83012011-08-31 08:03:29 +0000606extern int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700607extern struct sk_buff *skb_clone(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100608 gfp_t priority);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700609extern struct sk_buff *skb_copy(const struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100610 gfp_t priority);
Eric Dumazet117632e2011-12-03 21:39:53 +0000611extern struct sk_buff *__pskb_copy(struct sk_buff *skb,
612 int headroom, gfp_t gfp_mask);
613
Linus Torvalds1da177e2005-04-16 15:20:36 -0700614extern int pskb_expand_head(struct sk_buff *skb,
Victor Fusco86a76ca2005-07-08 14:57:47 -0700615 int nhead, int ntail,
Al Virodd0fc662005-10-07 07:46:04 +0100616 gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
618 unsigned int headroom);
619extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
620 int newheadroom, int newtailroom,
Al Virodd0fc662005-10-07 07:46:04 +0100621 gfp_t priority);
David Howells716ea3a2007-04-02 20:19:53 -0700622extern int skb_to_sgvec(struct sk_buff *skb,
623 struct scatterlist *sg, int offset,
624 int len);
625extern int skb_cow_data(struct sk_buff *skb, int tailbits,
626 struct sk_buff **trailer);
Herbert Xu5b057c62006-06-23 02:06:41 -0700627extern int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000628#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700630extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
631 int getfrag(void *from, char *to, int offset,
632 int len,int odd, struct sk_buff *skb),
633 void *from, int length);
634
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800635struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700636 __u32 lower_offset;
637 __u32 upper_offset;
638 __u32 frag_idx;
639 __u32 stepped_offset;
640 struct sk_buff *root_skb;
641 struct sk_buff *cur_skb;
642 __u8 *frag_data;
643};
644
645extern void skb_prepare_seq_read(struct sk_buff *skb,
646 unsigned int from, unsigned int to,
647 struct skb_seq_state *st);
648extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
649 struct skb_seq_state *st);
650extern void skb_abort_seq_read(struct skb_seq_state *st);
651
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700652extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
653 unsigned int to, struct ts_config *config,
654 struct ts_state *state);
655
Tom Herbertbdeab992011-08-14 19:45:55 +0000656extern void __skb_get_rxhash(struct sk_buff *skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000657static inline __u32 skb_get_rxhash(struct sk_buff *skb)
658{
Willem de Bruijnecd5cf52012-10-26 11:52:08 +0000659 if (!skb->l4_rxhash)
Tom Herbertbdeab992011-08-14 19:45:55 +0000660 __skb_get_rxhash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000661
662 return skb->rxhash;
663}
664
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700665#ifdef NET_SKBUFF_DATA_USES_OFFSET
666static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
667{
668 return skb->head + skb->end;
669}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000670
671static inline unsigned int skb_end_offset(const struct sk_buff *skb)
672{
673 return skb->end;
674}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700675#else
676static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
677{
678 return skb->end;
679}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000680
681static inline unsigned int skb_end_offset(const struct sk_buff *skb)
682{
683 return skb->end - skb->head;
684}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700685#endif
686
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700688#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700689
Patrick Ohlyac45f602009-02-12 05:03:37 +0000690static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
691{
692 return &skb_shinfo(skb)->hwtstamps;
693}
694
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695/**
696 * skb_queue_empty - check if a queue is empty
697 * @list: queue head
698 *
699 * Returns true if the queue is empty, false otherwise.
700 */
701static inline int skb_queue_empty(const struct sk_buff_head *list)
702{
703 return list->next == (struct sk_buff *)list;
704}
705
706/**
David S. Millerfc7ebb22008-09-23 00:34:07 -0700707 * skb_queue_is_last - check if skb is the last entry in the queue
708 * @list: queue head
709 * @skb: buffer
710 *
711 * Returns true if @skb is the last buffer on the list.
712 */
713static inline bool skb_queue_is_last(const struct sk_buff_head *list,
714 const struct sk_buff *skb)
715{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000716 return skb->next == (struct sk_buff *)list;
David S. Millerfc7ebb22008-09-23 00:34:07 -0700717}
718
719/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800720 * skb_queue_is_first - check if skb is the first entry in the queue
721 * @list: queue head
722 * @skb: buffer
723 *
724 * Returns true if @skb is the first buffer on the list.
725 */
726static inline bool skb_queue_is_first(const struct sk_buff_head *list,
727 const struct sk_buff *skb)
728{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000729 return skb->prev == (struct sk_buff *)list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800730}
731
732/**
David S. Miller249c8b42008-09-23 00:44:42 -0700733 * skb_queue_next - return the next packet in the queue
734 * @list: queue head
735 * @skb: current buffer
736 *
737 * Return the next packet in @list after @skb. It is only valid to
738 * call this if skb_queue_is_last() evaluates to false.
739 */
740static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
741 const struct sk_buff *skb)
742{
743 /* This BUG_ON may seem severe, but if we just return then we
744 * are going to dereference garbage.
745 */
746 BUG_ON(skb_queue_is_last(list, skb));
747 return skb->next;
748}
749
750/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800751 * skb_queue_prev - return the prev packet in the queue
752 * @list: queue head
753 * @skb: current buffer
754 *
755 * Return the prev packet in @list before @skb. It is only valid to
756 * call this if skb_queue_is_first() evaluates to false.
757 */
758static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
759 const struct sk_buff *skb)
760{
761 /* This BUG_ON may seem severe, but if we just return then we
762 * are going to dereference garbage.
763 */
764 BUG_ON(skb_queue_is_first(list, skb));
765 return skb->prev;
766}
767
768/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769 * skb_get - reference buffer
770 * @skb: buffer to reference
771 *
772 * Makes another reference to a socket buffer and returns a pointer
773 * to the buffer.
774 */
775static inline struct sk_buff *skb_get(struct sk_buff *skb)
776{
777 atomic_inc(&skb->users);
778 return skb;
779}
780
781/*
782 * If users == 1, we are the only owner and are can avoid redundant
783 * atomic change.
784 */
785
786/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700787 * skb_cloned - is the buffer a clone
788 * @skb: buffer to check
789 *
790 * Returns true if the buffer was generated with skb_clone() and is
791 * one of multiple shared copies of the buffer. Cloned buffers are
792 * shared data so must not be written to under normal circumstances.
793 */
794static inline int skb_cloned(const struct sk_buff *skb)
795{
796 return skb->cloned &&
797 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
798}
799
800/**
801 * skb_header_cloned - is the header a clone
802 * @skb: buffer to check
803 *
804 * Returns true if modifying the header part of the buffer requires
805 * the data to be copied.
806 */
807static inline int skb_header_cloned(const struct sk_buff *skb)
808{
809 int dataref;
810
811 if (!skb->cloned)
812 return 0;
813
814 dataref = atomic_read(&skb_shinfo(skb)->dataref);
815 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
816 return dataref != 1;
817}
818
819/**
820 * skb_header_release - release reference to header
821 * @skb: buffer to operate on
822 *
823 * Drop a reference to the header part of the buffer. This is done
824 * by acquiring a payload reference. You must not read from the header
825 * part of skb->data after this.
826 */
827static inline void skb_header_release(struct sk_buff *skb)
828{
829 BUG_ON(skb->nohdr);
830 skb->nohdr = 1;
831 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
832}
833
834/**
835 * skb_shared - is the buffer shared
836 * @skb: buffer to check
837 *
838 * Returns true if more than one person has a reference to this
839 * buffer.
840 */
841static inline int skb_shared(const struct sk_buff *skb)
842{
843 return atomic_read(&skb->users) != 1;
844}
845
846/**
847 * skb_share_check - check if buffer is shared and if so clone it
848 * @skb: buffer to check
849 * @pri: priority for memory allocation
850 *
851 * If the buffer is shared the buffer is cloned and the old copy
852 * drops a reference. A new clone with a single reference is returned.
853 * If the buffer is not shared the original buffer is returned. When
854 * being called from interrupt status or with spinlocks held pri must
855 * be GFP_ATOMIC.
856 *
857 * NULL is returned on a memory allocation failure.
858 */
Eric Dumazet47061bc2012-08-03 20:54:15 +0000859static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860{
861 might_sleep_if(pri & __GFP_WAIT);
862 if (skb_shared(skb)) {
863 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +0000864
865 if (likely(nskb))
866 consume_skb(skb);
867 else
868 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869 skb = nskb;
870 }
871 return skb;
872}
873
874/*
875 * Copy shared buffers into a new sk_buff. We effectively do COW on
876 * packets to handle cases where we have a local reader and forward
877 * and a couple of other messy ones. The normal one is tcpdumping
878 * a packet thats being forwarded.
879 */
880
881/**
882 * skb_unshare - make a copy of a shared buffer
883 * @skb: buffer to check
884 * @pri: priority for memory allocation
885 *
886 * If the socket buffer is a clone then this function creates a new
887 * copy of the data, drops a reference count on the old copy and returns
888 * the new copy with the reference count at 1. If the buffer is not a clone
889 * the original buffer is returned. When called with a spinlock held or
890 * from interrupt state @pri must be %GFP_ATOMIC
891 *
892 * %NULL is returned on a memory allocation failure.
893 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -0700894static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100895 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896{
897 might_sleep_if(pri & __GFP_WAIT);
898 if (skb_cloned(skb)) {
899 struct sk_buff *nskb = skb_copy(skb, pri);
900 kfree_skb(skb); /* Free our shared copy */
901 skb = nskb;
902 }
903 return skb;
904}
905
906/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -0800907 * skb_peek - peek at the head 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 head 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(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920{
Eric Dumazet18d07002012-04-30 16:31:46 +0000921 struct sk_buff *skb = list_->next;
922
923 if (skb == (struct sk_buff *)list_)
924 skb = NULL;
925 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700926}
927
928/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +0000929 * skb_peek_next - peek skb following the given one from a queue
930 * @skb: skb to start from
931 * @list_: list to peek at
932 *
933 * Returns %NULL when the end of the list is met or a pointer to the
934 * next element. The reference count is not incremented and the
935 * reference is therefore volatile. Use with caution.
936 */
937static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
938 const struct sk_buff_head *list_)
939{
940 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +0000941
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +0000942 if (next == (struct sk_buff *)list_)
943 next = NULL;
944 return next;
945}
946
947/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -0800948 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -0700949 * @list_: list to peek at
950 *
951 * Peek an &sk_buff. Unlike most other operations you _MUST_
952 * be careful with this one. A peek leaves the buffer on the
953 * list and someone else may run off with it. You must hold
954 * the appropriate locks or have a private queue to do this.
955 *
956 * Returns %NULL for an empty list or a pointer to the tail element.
957 * The reference count is not incremented and the reference is therefore
958 * volatile. Use with caution.
959 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -0400960static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700961{
Eric Dumazet18d07002012-04-30 16:31:46 +0000962 struct sk_buff *skb = list_->prev;
963
964 if (skb == (struct sk_buff *)list_)
965 skb = NULL;
966 return skb;
967
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968}
969
970/**
971 * skb_queue_len - get queue length
972 * @list_: list to measure
973 *
974 * Return the length of an &sk_buff queue.
975 */
976static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
977{
978 return list_->qlen;
979}
980
David S. Miller67fed452008-09-21 22:36:24 -0700981/**
982 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
983 * @list: queue to initialize
984 *
985 * This initializes only the list and queue length aspects of
986 * an sk_buff_head object. This allows to initialize the list
987 * aspects of an sk_buff_head without reinitializing things like
988 * the spinlock. It can also be used for on-stack sk_buff_head
989 * objects where the spinlock is known to not be used.
990 */
991static inline void __skb_queue_head_init(struct sk_buff_head *list)
992{
993 list->prev = list->next = (struct sk_buff *)list;
994 list->qlen = 0;
995}
996
Arjan van de Ven76f10ad2006-08-02 14:06:55 -0700997/*
998 * This function creates a split out lock class for each invocation;
999 * this is needed for now since a whole lot of users of the skb-queue
1000 * infrastructure in drivers have different locking usage (in hardirq)
1001 * than the networking core (in softirq only). In the long run either the
1002 * network layer or drivers should need annotation to consolidate the
1003 * main types of usage into 3 classes.
1004 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005static inline void skb_queue_head_init(struct sk_buff_head *list)
1006{
1007 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001008 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009}
1010
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001011static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1012 struct lock_class_key *class)
1013{
1014 skb_queue_head_init(list);
1015 lockdep_set_class(&list->lock, class);
1016}
1017
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001019 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020 *
1021 * The "__skb_xxxx()" functions are the non-atomic ones that
1022 * can only be called with interrupts disabled.
1023 */
Gerrit Renkerbf299272008-04-14 00:04:51 -07001024extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
1025static inline void __skb_insert(struct sk_buff *newsk,
1026 struct sk_buff *prev, struct sk_buff *next,
1027 struct sk_buff_head *list)
1028{
1029 newsk->next = next;
1030 newsk->prev = prev;
1031 next->prev = prev->next = newsk;
1032 list->qlen++;
1033}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001034
David S. Miller67fed452008-09-21 22:36:24 -07001035static inline void __skb_queue_splice(const struct sk_buff_head *list,
1036 struct sk_buff *prev,
1037 struct sk_buff *next)
1038{
1039 struct sk_buff *first = list->next;
1040 struct sk_buff *last = list->prev;
1041
1042 first->prev = prev;
1043 prev->next = first;
1044
1045 last->next = next;
1046 next->prev = last;
1047}
1048
1049/**
1050 * skb_queue_splice - join two skb lists, this is designed for stacks
1051 * @list: the new list to add
1052 * @head: the place to add it in the first list
1053 */
1054static inline void skb_queue_splice(const struct sk_buff_head *list,
1055 struct sk_buff_head *head)
1056{
1057 if (!skb_queue_empty(list)) {
1058 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001059 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001060 }
1061}
1062
1063/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001064 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001065 * @list: the new list to add
1066 * @head: the place to add it in the first list
1067 *
1068 * The list at @list is reinitialised
1069 */
1070static inline void skb_queue_splice_init(struct sk_buff_head *list,
1071 struct sk_buff_head *head)
1072{
1073 if (!skb_queue_empty(list)) {
1074 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001075 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001076 __skb_queue_head_init(list);
1077 }
1078}
1079
1080/**
1081 * skb_queue_splice_tail - join two skb lists, each list being a queue
1082 * @list: the new list to add
1083 * @head: the place to add it in the first list
1084 */
1085static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1086 struct sk_buff_head *head)
1087{
1088 if (!skb_queue_empty(list)) {
1089 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001090 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001091 }
1092}
1093
1094/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001095 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001096 * @list: the new list to add
1097 * @head: the place to add it in the first list
1098 *
1099 * Each of the lists is a queue.
1100 * The list at @list is reinitialised
1101 */
1102static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1103 struct sk_buff_head *head)
1104{
1105 if (!skb_queue_empty(list)) {
1106 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001107 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001108 __skb_queue_head_init(list);
1109 }
1110}
1111
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001113 * __skb_queue_after - queue a buffer at the list head
1114 * @list: list to use
1115 * @prev: place after this buffer
1116 * @newsk: buffer to queue
1117 *
1118 * Queue a buffer int the middle of a list. This function takes no locks
1119 * and you must therefore hold required locks before calling it.
1120 *
1121 * A buffer cannot be placed on two lists at the same time.
1122 */
1123static inline void __skb_queue_after(struct sk_buff_head *list,
1124 struct sk_buff *prev,
1125 struct sk_buff *newsk)
1126{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001127 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001128}
1129
Gerrit Renker7de6c032008-04-14 00:05:09 -07001130extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1131 struct sk_buff_head *list);
1132
Gerrit Renkerf5572852008-04-14 00:05:28 -07001133static inline void __skb_queue_before(struct sk_buff_head *list,
1134 struct sk_buff *next,
1135 struct sk_buff *newsk)
1136{
1137 __skb_insert(newsk, next->prev, next, list);
1138}
1139
Stephen Hemminger300ce172005-10-30 13:47:34 -08001140/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001141 * __skb_queue_head - queue a buffer at the list head
1142 * @list: list to use
1143 * @newsk: buffer to queue
1144 *
1145 * Queue a buffer at the start of a list. This function takes no locks
1146 * and you must therefore hold required locks before calling it.
1147 *
1148 * A buffer cannot be placed on two lists at the same time.
1149 */
1150extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1151static inline void __skb_queue_head(struct sk_buff_head *list,
1152 struct sk_buff *newsk)
1153{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001154 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001155}
1156
1157/**
1158 * __skb_queue_tail - queue a buffer at the list tail
1159 * @list: list to use
1160 * @newsk: buffer to queue
1161 *
1162 * Queue a buffer at the end of a list. This function takes no locks
1163 * and you must therefore hold required locks before calling it.
1164 *
1165 * A buffer cannot be placed on two lists at the same time.
1166 */
1167extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1168static inline void __skb_queue_tail(struct sk_buff_head *list,
1169 struct sk_buff *newsk)
1170{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001171 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172}
1173
Linus Torvalds1da177e2005-04-16 15:20:36 -07001174/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175 * remove sk_buff from list. _Must_ be called atomically, and with
1176 * the list known..
1177 */
David S. Miller8728b832005-08-09 19:25:21 -07001178extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001179static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1180{
1181 struct sk_buff *next, *prev;
1182
1183 list->qlen--;
1184 next = skb->next;
1185 prev = skb->prev;
1186 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001187 next->prev = prev;
1188 prev->next = next;
1189}
1190
Gerrit Renkerf525c062008-04-14 00:04:12 -07001191/**
1192 * __skb_dequeue - remove from the head of the queue
1193 * @list: list to dequeue from
1194 *
1195 * Remove the head of the list. This function does not take any locks
1196 * so must be used with appropriate locks held only. The head item is
1197 * returned or %NULL if the list is empty.
1198 */
1199extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1200static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1201{
1202 struct sk_buff *skb = skb_peek(list);
1203 if (skb)
1204 __skb_unlink(skb, list);
1205 return skb;
1206}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001207
1208/**
1209 * __skb_dequeue_tail - remove from the tail of the queue
1210 * @list: list to dequeue from
1211 *
1212 * Remove the tail of the list. This function does not take any locks
1213 * so must be used with appropriate locks held only. The tail item is
1214 * returned or %NULL if the list is empty.
1215 */
1216extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1217static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1218{
1219 struct sk_buff *skb = skb_peek_tail(list);
1220 if (skb)
1221 __skb_unlink(skb, list);
1222 return skb;
1223}
1224
1225
David S. Millerbdcc0922012-03-07 20:53:36 -05001226static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001227{
1228 return skb->data_len;
1229}
1230
1231static inline unsigned int skb_headlen(const struct sk_buff *skb)
1232{
1233 return skb->len - skb->data_len;
1234}
1235
1236static inline int skb_pagelen(const struct sk_buff *skb)
1237{
1238 int i, len = 0;
1239
1240 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001241 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001242 return len + skb_headlen(skb);
1243}
1244
Ian Campbell131ea662011-08-19 06:25:00 +00001245/**
1246 * __skb_fill_page_desc - initialise a paged fragment in an skb
1247 * @skb: buffer containing fragment to be initialised
1248 * @i: paged fragment index to initialise
1249 * @page: the page to use for this fragment
1250 * @off: the offset to the data with @page
1251 * @size: the length of the data
1252 *
1253 * Initialises the @i'th fragment of @skb to point to &size bytes at
1254 * offset @off within @page.
1255 *
1256 * Does not take any additional reference on the fragment.
1257 */
1258static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1259 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001260{
1261 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1262
Mel Gormanc48a11c2012-07-31 16:44:23 -07001263 /*
1264 * Propagate page->pfmemalloc to the skb if we can. The problem is
1265 * that not all callers have unique ownership of the page. If
1266 * pfmemalloc is set, we check the mapping as a mapping implies
1267 * page->index is set (index and pfmemalloc share space).
1268 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1269 * do not lose pfmemalloc information as the pages would not be
1270 * allocated using __GFP_MEMALLOC.
1271 */
1272 if (page->pfmemalloc && !page->mapping)
1273 skb->pfmemalloc = true;
Ian Campbella8605c62011-10-19 23:01:49 +00001274 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001275 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001276 skb_frag_size_set(frag, size);
Ian Campbell131ea662011-08-19 06:25:00 +00001277}
1278
1279/**
1280 * skb_fill_page_desc - initialise a paged fragment in an skb
1281 * @skb: buffer containing fragment to be initialised
1282 * @i: paged fragment index to initialise
1283 * @page: the page to use for this fragment
1284 * @off: the offset to the data with @page
1285 * @size: the length of the data
1286 *
1287 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1288 * @skb to point to &size bytes at offset @off within @page. In
1289 * addition updates @skb such that @i is the last fragment.
1290 *
1291 * Does not take any additional reference on the fragment.
1292 */
1293static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1294 struct page *page, int off, int size)
1295{
1296 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001297 skb_shinfo(skb)->nr_frags = i + 1;
1298}
1299
Peter Zijlstra654bed12008-10-07 14:22:33 -07001300extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
Eric Dumazet50269e12012-03-23 23:59:33 +00001301 int off, int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001302
Linus Torvalds1da177e2005-04-16 15:20:36 -07001303#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001304#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001305#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1306
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001307#ifdef NET_SKBUFF_DATA_USES_OFFSET
1308static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1309{
1310 return skb->head + skb->tail;
1311}
1312
1313static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1314{
1315 skb->tail = skb->data - skb->head;
1316}
1317
1318static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1319{
1320 skb_reset_tail_pointer(skb);
1321 skb->tail += offset;
1322}
1323#else /* NET_SKBUFF_DATA_USES_OFFSET */
1324static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1325{
1326 return skb->tail;
1327}
1328
1329static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1330{
1331 skb->tail = skb->data;
1332}
1333
1334static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1335{
1336 skb->tail = skb->data + offset;
1337}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001338
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001339#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1340
Linus Torvalds1da177e2005-04-16 15:20:36 -07001341/*
1342 * Add data to an sk_buff
1343 */
Ilpo Järvinen0dde3e12008-03-27 17:43:41 -07001344extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1346{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001347 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348 SKB_LINEAR_ASSERT(skb);
1349 skb->tail += len;
1350 skb->len += len;
1351 return tmp;
1352}
1353
Ilpo Järvinenc2aa2702008-03-27 17:52:40 -07001354extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001355static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1356{
1357 skb->data -= len;
1358 skb->len += len;
1359 return skb->data;
1360}
1361
Ilpo Järvinen6be8ac22008-03-27 17:47:24 -07001362extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1364{
1365 skb->len -= len;
1366 BUG_ON(skb->len < skb->data_len);
1367 return skb->data += len;
1368}
1369
David S. Miller47d29642010-05-02 02:21:44 -07001370static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1371{
1372 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1373}
1374
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1376
1377static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1378{
1379 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001380 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001381 return NULL;
1382 skb->len -= len;
1383 return skb->data += len;
1384}
1385
1386static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1387{
1388 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1389}
1390
1391static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1392{
1393 if (likely(len <= skb_headlen(skb)))
1394 return 1;
1395 if (unlikely(len > skb->len))
1396 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001397 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398}
1399
1400/**
1401 * skb_headroom - bytes at buffer head
1402 * @skb: buffer to check
1403 *
1404 * Return the number of bytes of free space at the head of an &sk_buff.
1405 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001406static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407{
1408 return skb->data - skb->head;
1409}
1410
1411/**
1412 * skb_tailroom - bytes at buffer end
1413 * @skb: buffer to check
1414 *
1415 * Return the number of bytes of free space at the tail of an sk_buff
1416 */
1417static inline int skb_tailroom(const struct sk_buff *skb)
1418{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001419 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420}
1421
1422/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00001423 * skb_availroom - bytes at buffer end
1424 * @skb: buffer to check
1425 *
1426 * Return the number of bytes of free space at the tail of an sk_buff
1427 * allocated by sk_stream_alloc()
1428 */
1429static inline int skb_availroom(const struct sk_buff *skb)
1430{
1431 return skb_is_nonlinear(skb) ? 0 : skb->avail_size - skb->len;
1432}
1433
1434/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001435 * skb_reserve - adjust headroom
1436 * @skb: buffer to alter
1437 * @len: bytes to move
1438 *
1439 * Increase the headroom of an empty &sk_buff by reducing the tail
1440 * room. This is only allowed for an empty buffer.
1441 */
David S. Miller8243126c2006-01-17 02:54:21 -08001442static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001443{
1444 skb->data += len;
1445 skb->tail += len;
1446}
1447
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001448static inline void skb_reset_inner_headers(struct sk_buff *skb)
1449{
1450 skb->inner_network_header = skb->network_header;
1451 skb->inner_transport_header = skb->transport_header;
1452}
1453
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00001454static inline void skb_reset_mac_len(struct sk_buff *skb)
1455{
1456 skb->mac_len = skb->network_header - skb->mac_header;
1457}
1458
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001459#ifdef NET_SKBUFF_DATA_USES_OFFSET
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001460static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1461 *skb)
1462{
1463 return skb->head + skb->inner_transport_header;
1464}
1465
1466static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1467{
1468 skb->inner_transport_header = skb->data - skb->head;
1469}
1470
1471static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1472 const int offset)
1473{
1474 skb_reset_inner_transport_header(skb);
1475 skb->inner_transport_header += offset;
1476}
1477
1478static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1479{
1480 return skb->head + skb->inner_network_header;
1481}
1482
1483static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1484{
1485 skb->inner_network_header = skb->data - skb->head;
1486}
1487
1488static inline void skb_set_inner_network_header(struct sk_buff *skb,
1489 const int offset)
1490{
1491 skb_reset_inner_network_header(skb);
1492 skb->inner_network_header += offset;
1493}
1494
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001495static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1496{
1497 return skb->transport_header != ~0U;
1498}
1499
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001500static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1501{
1502 return skb->head + skb->transport_header;
1503}
1504
1505static inline void skb_reset_transport_header(struct sk_buff *skb)
1506{
1507 skb->transport_header = skb->data - skb->head;
1508}
1509
1510static inline void skb_set_transport_header(struct sk_buff *skb,
1511 const int offset)
1512{
1513 skb_reset_transport_header(skb);
1514 skb->transport_header += offset;
1515}
1516
1517static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1518{
1519 return skb->head + skb->network_header;
1520}
1521
1522static inline void skb_reset_network_header(struct sk_buff *skb)
1523{
1524 skb->network_header = skb->data - skb->head;
1525}
1526
1527static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1528{
1529 skb_reset_network_header(skb);
1530 skb->network_header += offset;
1531}
1532
1533static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1534{
1535 return skb->head + skb->mac_header;
1536}
1537
1538static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1539{
1540 return skb->mac_header != ~0U;
1541}
1542
1543static inline void skb_reset_mac_header(struct sk_buff *skb)
1544{
1545 skb->mac_header = skb->data - skb->head;
1546}
1547
1548static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1549{
1550 skb_reset_mac_header(skb);
1551 skb->mac_header += offset;
1552}
1553
1554#else /* NET_SKBUFF_DATA_USES_OFFSET */
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001555static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1556 *skb)
1557{
1558 return skb->inner_transport_header;
1559}
1560
1561static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1562{
1563 skb->inner_transport_header = skb->data;
1564}
1565
1566static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1567 const int offset)
1568{
1569 skb->inner_transport_header = skb->data + offset;
1570}
1571
1572static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1573{
1574 return skb->inner_network_header;
1575}
1576
1577static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1578{
1579 skb->inner_network_header = skb->data;
1580}
1581
1582static inline void skb_set_inner_network_header(struct sk_buff *skb,
1583 const int offset)
1584{
1585 skb->inner_network_header = skb->data + offset;
1586}
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001587
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001588static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1589{
1590 return skb->transport_header != NULL;
1591}
1592
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001593static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1594{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001595 return skb->transport_header;
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001596}
1597
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001598static inline void skb_reset_transport_header(struct sk_buff *skb)
1599{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001600 skb->transport_header = skb->data;
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001601}
1602
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001603static inline void skb_set_transport_header(struct sk_buff *skb,
1604 const int offset)
1605{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001606 skb->transport_header = skb->data + offset;
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001607}
1608
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001609static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1610{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001611 return skb->network_header;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001612}
1613
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001614static inline void skb_reset_network_header(struct sk_buff *skb)
1615{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001616 skb->network_header = skb->data;
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001617}
1618
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001619static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1620{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001621 skb->network_header = skb->data + offset;
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001622}
1623
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001624static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1625{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001626 return skb->mac_header;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001627}
1628
1629static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1630{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001631 return skb->mac_header != NULL;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001632}
1633
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001634static inline void skb_reset_mac_header(struct sk_buff *skb)
1635{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001636 skb->mac_header = skb->data;
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001637}
1638
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001639static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1640{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001641 skb->mac_header = skb->data + offset;
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001642}
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001643#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1644
Eric Dumazet03606892012-02-23 10:55:02 +00001645static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1646{
1647 if (skb_mac_header_was_set(skb)) {
1648 const unsigned char *old_mac = skb_mac_header(skb);
1649
1650 skb_set_mac_header(skb, -skb->mac_len);
1651 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1652 }
1653}
1654
Michał Mirosław04fb4512010-12-14 15:24:08 +00001655static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1656{
1657 return skb->csum_start - skb_headroom(skb);
1658}
1659
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001660static inline int skb_transport_offset(const struct sk_buff *skb)
1661{
1662 return skb_transport_header(skb) - skb->data;
1663}
1664
1665static inline u32 skb_network_header_len(const struct sk_buff *skb)
1666{
1667 return skb->transport_header - skb->network_header;
1668}
1669
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001670static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1671{
1672 return skb->inner_transport_header - skb->inner_network_header;
1673}
1674
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001675static inline int skb_network_offset(const struct sk_buff *skb)
1676{
1677 return skb_network_header(skb) - skb->data;
1678}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001679
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001680static inline int skb_inner_network_offset(const struct sk_buff *skb)
1681{
1682 return skb_inner_network_header(skb) - skb->data;
1683}
1684
Changli Gaof9599ce2010-08-04 04:43:44 +00001685static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1686{
1687 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1688}
1689
Linus Torvalds1da177e2005-04-16 15:20:36 -07001690/*
1691 * CPUs often take a performance hit when accessing unaligned memory
1692 * locations. The actual performance hit varies, it can be small if the
1693 * hardware handles it or large if we have to take an exception and fix it
1694 * in software.
1695 *
1696 * Since an ethernet header is 14 bytes network drivers often end up with
1697 * the IP header at an unaligned offset. The IP header can be aligned by
1698 * shifting the start of the packet by 2 bytes. Drivers should do this
1699 * with:
1700 *
Tobias Klauser8660c122009-07-13 22:48:16 +00001701 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001702 *
1703 * The downside to this alignment of the IP header is that the DMA is now
1704 * unaligned. On some architectures the cost of an unaligned DMA is high
1705 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00001706 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001707 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1708 * to be overridden.
1709 */
1710#ifndef NET_IP_ALIGN
1711#define NET_IP_ALIGN 2
1712#endif
1713
Anton Blanchard025be812006-03-31 02:27:06 -08001714/*
1715 * The networking layer reserves some headroom in skb data (via
1716 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1717 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08001718 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08001719 *
1720 * Unfortunately this headroom changes the DMA alignment of the resulting
1721 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1722 * on some architectures. An architecture can override this value,
1723 * perhaps setting it to a cacheline in size (since that will maintain
1724 * cacheline alignment of the DMA). It must be a power of 2.
1725 *
David S. Millerd6301d32009-02-08 19:24:13 -08001726 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08001727 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07001728 *
1729 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1730 * to reduce average number of cache lines per packet.
1731 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07001732 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08001733 */
1734#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07001735#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08001736#endif
1737
Herbert Xu3cc0e872006-06-09 16:13:38 -07001738extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001739
1740static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1741{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00001742 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07001743 WARN_ON(1);
1744 return;
1745 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001746 skb->len = len;
1747 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001748}
1749
Ilpo Järvinen419ae742008-03-27 17:54:01 -07001750extern void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751
1752static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1753{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001754 if (skb->data_len)
1755 return ___pskb_trim(skb, len);
1756 __skb_trim(skb, len);
1757 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001758}
1759
1760static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1761{
1762 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1763}
1764
1765/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07001766 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1767 * @skb: buffer to alter
1768 * @len: new length
1769 *
1770 * This is identical to pskb_trim except that the caller knows that
1771 * the skb is not cloned so we should never get an error due to out-
1772 * of-memory.
1773 */
1774static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1775{
1776 int err = pskb_trim(skb, len);
1777 BUG_ON(err);
1778}
1779
1780/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001781 * skb_orphan - orphan a buffer
1782 * @skb: buffer to orphan
1783 *
1784 * If a buffer currently has an owner then we call the owner's
1785 * destructor function and make the @skb unowned. The buffer continues
1786 * to exist but is no longer charged to its former owner.
1787 */
1788static inline void skb_orphan(struct sk_buff *skb)
1789{
1790 if (skb->destructor)
1791 skb->destructor(skb);
1792 skb->destructor = NULL;
1793 skb->sk = NULL;
1794}
1795
1796/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00001797 * skb_orphan_frags - orphan the frags contained in a buffer
1798 * @skb: buffer to orphan frags from
1799 * @gfp_mask: allocation mask for replacement pages
1800 *
1801 * For each frag in the SKB which needs a destructor (i.e. has an
1802 * owner) create a copy of that frag and release the original
1803 * page by calling the destructor.
1804 */
1805static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1806{
1807 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1808 return 0;
1809 return skb_copy_ubufs(skb, gfp_mask);
1810}
1811
1812/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001813 * __skb_queue_purge - empty a list
1814 * @list: list to empty
1815 *
1816 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1817 * the list and one reference dropped. This function does not take the
1818 * list lock and the caller must hold the relevant locks to use it.
1819 */
1820extern void skb_queue_purge(struct sk_buff_head *list);
1821static inline void __skb_queue_purge(struct sk_buff_head *list)
1822{
1823 struct sk_buff *skb;
1824 while ((skb = __skb_dequeue(list)) != NULL)
1825 kfree_skb(skb);
1826}
1827
Eric Dumazet6f532612012-05-18 05:12:12 +00001828extern void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001829
Christoph Hellwig8af27452006-07-31 22:35:23 -07001830extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00001831 unsigned int length,
1832 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07001833
1834/**
1835 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1836 * @dev: network device to receive on
1837 * @length: length to allocate
1838 *
1839 * Allocate a new &sk_buff and assign it a usage count of one. The
1840 * buffer has unspecified headroom built in. Users should allocate
1841 * the headroom they think they need without accounting for the
1842 * built in space. The built in space is used for optimisations.
1843 *
1844 * %NULL is returned if there is no free memory. Although this function
1845 * allocates memory it can be called from an interrupt.
1846 */
1847static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00001848 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07001849{
1850 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1851}
1852
Eric Dumazet6f532612012-05-18 05:12:12 +00001853/* legacy helper around __netdev_alloc_skb() */
1854static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1855 gfp_t gfp_mask)
1856{
1857 return __netdev_alloc_skb(NULL, length, gfp_mask);
1858}
1859
1860/* legacy helper around netdev_alloc_skb() */
1861static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1862{
1863 return netdev_alloc_skb(NULL, length);
1864}
1865
1866
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001867static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
1868 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00001869{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001870 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00001871
1872 if (NET_IP_ALIGN && skb)
1873 skb_reserve(skb, NET_IP_ALIGN);
1874 return skb;
1875}
1876
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001877static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1878 unsigned int length)
1879{
1880 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
1881}
1882
Mel Gorman06140022012-07-31 16:44:24 -07001883/*
1884 * __skb_alloc_page - allocate pages for ps-rx on a skb and preserve pfmemalloc data
1885 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1886 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1887 * @order: size of the allocation
1888 *
1889 * Allocate a new page.
1890 *
1891 * %NULL is returned if there is no free memory.
1892*/
1893static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
1894 struct sk_buff *skb,
1895 unsigned int order)
1896{
1897 struct page *page;
1898
1899 gfp_mask |= __GFP_COLD;
1900
1901 if (!(gfp_mask & __GFP_NOMEMALLOC))
1902 gfp_mask |= __GFP_MEMALLOC;
1903
1904 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
1905 if (skb && page && page->pfmemalloc)
1906 skb->pfmemalloc = true;
1907
1908 return page;
1909}
1910
1911/**
1912 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
1913 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1914 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1915 *
1916 * Allocate a new page.
1917 *
1918 * %NULL is returned if there is no free memory.
1919 */
1920static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
1921 struct sk_buff *skb)
1922{
1923 return __skb_alloc_pages(gfp_mask, skb, 0);
1924}
1925
1926/**
1927 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
1928 * @page: The page that was allocated from skb_alloc_page
1929 * @skb: The skb that may need pfmemalloc set
1930 */
1931static inline void skb_propagate_pfmemalloc(struct page *page,
1932 struct sk_buff *skb)
1933{
1934 if (page && page->pfmemalloc)
1935 skb->pfmemalloc = true;
1936}
1937
Eric Dumazet564824b2010-10-11 19:05:25 +00001938/**
Ian Campbell131ea662011-08-19 06:25:00 +00001939 * skb_frag_page - retrieve the page refered to by a paged fragment
1940 * @frag: the paged fragment
1941 *
1942 * Returns the &struct page associated with @frag.
1943 */
1944static inline struct page *skb_frag_page(const skb_frag_t *frag)
1945{
Ian Campbella8605c62011-10-19 23:01:49 +00001946 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00001947}
1948
1949/**
1950 * __skb_frag_ref - take an addition reference on a paged fragment.
1951 * @frag: the paged fragment
1952 *
1953 * Takes an additional reference on the paged fragment @frag.
1954 */
1955static inline void __skb_frag_ref(skb_frag_t *frag)
1956{
1957 get_page(skb_frag_page(frag));
1958}
1959
1960/**
1961 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
1962 * @skb: the buffer
1963 * @f: the fragment offset.
1964 *
1965 * Takes an additional reference on the @f'th paged fragment of @skb.
1966 */
1967static inline void skb_frag_ref(struct sk_buff *skb, int f)
1968{
1969 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
1970}
1971
1972/**
1973 * __skb_frag_unref - release a reference on a paged fragment.
1974 * @frag: the paged fragment
1975 *
1976 * Releases a reference on the paged fragment @frag.
1977 */
1978static inline void __skb_frag_unref(skb_frag_t *frag)
1979{
1980 put_page(skb_frag_page(frag));
1981}
1982
1983/**
1984 * skb_frag_unref - release a reference on a paged fragment of an skb.
1985 * @skb: the buffer
1986 * @f: the fragment offset
1987 *
1988 * Releases a reference on the @f'th paged fragment of @skb.
1989 */
1990static inline void skb_frag_unref(struct sk_buff *skb, int f)
1991{
1992 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
1993}
1994
1995/**
1996 * skb_frag_address - gets the address of the data contained in a paged fragment
1997 * @frag: the paged fragment buffer
1998 *
1999 * Returns the address of the data within @frag. The page must already
2000 * be mapped.
2001 */
2002static inline void *skb_frag_address(const skb_frag_t *frag)
2003{
2004 return page_address(skb_frag_page(frag)) + frag->page_offset;
2005}
2006
2007/**
2008 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2009 * @frag: the paged fragment buffer
2010 *
2011 * Returns the address of the data within @frag. Checks that the page
2012 * is mapped and returns %NULL otherwise.
2013 */
2014static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2015{
2016 void *ptr = page_address(skb_frag_page(frag));
2017 if (unlikely(!ptr))
2018 return NULL;
2019
2020 return ptr + frag->page_offset;
2021}
2022
2023/**
2024 * __skb_frag_set_page - sets the page contained in a paged fragment
2025 * @frag: the paged fragment
2026 * @page: the page to set
2027 *
2028 * Sets the fragment @frag to contain @page.
2029 */
2030static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2031{
Ian Campbella8605c62011-10-19 23:01:49 +00002032 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002033}
2034
2035/**
2036 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2037 * @skb: the buffer
2038 * @f: the fragment offset
2039 * @page: the page to set
2040 *
2041 * Sets the @f'th fragment of @skb to contain @page.
2042 */
2043static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2044 struct page *page)
2045{
2046 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2047}
2048
2049/**
2050 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002051 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002052 * @frag: the paged fragment to map
2053 * @offset: the offset within the fragment (starting at the
2054 * fragment's own offset)
2055 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002056 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00002057 *
2058 * Maps the page associated with @frag to @device.
2059 */
2060static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2061 const skb_frag_t *frag,
2062 size_t offset, size_t size,
2063 enum dma_data_direction dir)
2064{
2065 return dma_map_page(dev, skb_frag_page(frag),
2066 frag->page_offset + offset, size, dir);
2067}
2068
Eric Dumazet117632e2011-12-03 21:39:53 +00002069static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2070 gfp_t gfp_mask)
2071{
2072 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2073}
2074
Ian Campbell131ea662011-08-19 06:25:00 +00002075/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002076 * skb_clone_writable - is the header of a clone writable
2077 * @skb: buffer to check
2078 * @len: length up to which to write
2079 *
2080 * Returns true if modifying the header part of the cloned buffer
2081 * does not requires the data to be copied.
2082 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002083static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002084{
2085 return !skb_header_cloned(skb) &&
2086 skb_headroom(skb) + len <= skb->hdr_len;
2087}
2088
Herbert Xud9cc2042007-09-16 16:21:16 -07002089static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2090 int cloned)
2091{
2092 int delta = 0;
2093
Herbert Xud9cc2042007-09-16 16:21:16 -07002094 if (headroom > skb_headroom(skb))
2095 delta = headroom - skb_headroom(skb);
2096
2097 if (delta || cloned)
2098 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2099 GFP_ATOMIC);
2100 return 0;
2101}
2102
Patrick McHardy334a8132007-06-25 04:35:20 -07002103/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002104 * skb_cow - copy header of skb when it is required
2105 * @skb: buffer to cow
2106 * @headroom: needed headroom
2107 *
2108 * If the skb passed lacks sufficient headroom or its data part
2109 * is shared, data is reallocated. If reallocation fails, an error
2110 * is returned and original skb is not changed.
2111 *
2112 * The result is skb with writable area skb->head...skb->tail
2113 * and at least @headroom of space at head.
2114 */
2115static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2116{
Herbert Xud9cc2042007-09-16 16:21:16 -07002117 return __skb_cow(skb, headroom, skb_cloned(skb));
2118}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002119
Herbert Xud9cc2042007-09-16 16:21:16 -07002120/**
2121 * skb_cow_head - skb_cow but only making the head writable
2122 * @skb: buffer to cow
2123 * @headroom: needed headroom
2124 *
2125 * This function is identical to skb_cow except that we replace the
2126 * skb_cloned check by skb_header_cloned. It should be used when
2127 * you only need to push on some header and do not need to modify
2128 * the data.
2129 */
2130static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2131{
2132 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002133}
2134
2135/**
2136 * skb_padto - pad an skbuff up to a minimal size
2137 * @skb: buffer to pad
2138 * @len: minimal length
2139 *
2140 * Pads up a buffer to ensure the trailing bytes exist and are
2141 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002142 * is untouched. Otherwise it is extended. Returns zero on
2143 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002144 */
2145
Herbert Xu5b057c62006-06-23 02:06:41 -07002146static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002147{
2148 unsigned int size = skb->len;
2149 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002150 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002151 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002152}
2153
2154static inline int skb_add_data(struct sk_buff *skb,
2155 char __user *from, int copy)
2156{
2157 const int off = skb->len;
2158
2159 if (skb->ip_summed == CHECKSUM_NONE) {
2160 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08002161 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
Linus Torvalds1da177e2005-04-16 15:20:36 -07002162 copy, 0, &err);
2163 if (!err) {
2164 skb->csum = csum_block_add(skb->csum, csum, off);
2165 return 0;
2166 }
2167 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2168 return 0;
2169
2170 __skb_trim(skb, off);
2171 return -EFAULT;
2172}
2173
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002174static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2175 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002176{
2177 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002178 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002179
Ian Campbellea2ab692011-08-22 23:44:58 +00002180 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002181 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002182 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002183 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002184}
2185
Herbert Xu364c6ba2006-06-09 16:10:40 -07002186static inline int __skb_linearize(struct sk_buff *skb)
2187{
2188 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2189}
2190
Linus Torvalds1da177e2005-04-16 15:20:36 -07002191/**
2192 * skb_linearize - convert paged skb to linear one
2193 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002194 *
2195 * If there is no free memory -ENOMEM is returned, otherwise zero
2196 * is returned and the old skb data released.
2197 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002198static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002199{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002200 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2201}
2202
2203/**
2204 * skb_linearize_cow - make sure skb is linear and writable
2205 * @skb: buffer to process
2206 *
2207 * If there is no free memory -ENOMEM is returned, otherwise zero
2208 * is returned and the old skb data released.
2209 */
2210static inline int skb_linearize_cow(struct sk_buff *skb)
2211{
2212 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2213 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002214}
2215
2216/**
2217 * skb_postpull_rcsum - update checksum for received skb after pull
2218 * @skb: buffer to update
2219 * @start: start of data before pull
2220 * @len: length of data pulled
2221 *
2222 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002223 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2224 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002225 */
2226
2227static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002228 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002229{
Patrick McHardy84fa7932006-08-29 16:44:56 -07002230 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002231 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2232}
2233
Herbert Xucbb042f2006-03-20 22:43:56 -08002234unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2235
Linus Torvalds1da177e2005-04-16 15:20:36 -07002236/**
2237 * pskb_trim_rcsum - trim received skb and update checksum
2238 * @skb: buffer to trim
2239 * @len: new length
2240 *
2241 * This is exactly the same as pskb_trim except that it ensures the
2242 * checksum of received packets are still valid after the operation.
2243 */
2244
2245static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2246{
Stephen Hemminger0e4e4222005-09-08 12:32:03 -07002247 if (likely(len >= skb->len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002248 return 0;
Patrick McHardy84fa7932006-08-29 16:44:56 -07002249 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002250 skb->ip_summed = CHECKSUM_NONE;
2251 return __pskb_trim(skb, len);
2252}
2253
Linus Torvalds1da177e2005-04-16 15:20:36 -07002254#define skb_queue_walk(queue, skb) \
2255 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002256 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002257 skb = skb->next)
2258
James Chapman46f89142007-04-30 00:07:31 -07002259#define skb_queue_walk_safe(queue, skb, tmp) \
2260 for (skb = (queue)->next, tmp = skb->next; \
2261 skb != (struct sk_buff *)(queue); \
2262 skb = tmp, tmp = skb->next)
2263
David S. Miller1164f522008-09-23 00:49:44 -07002264#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002265 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002266 skb = skb->next)
2267
2268#define skb_queue_walk_from_safe(queue, skb, tmp) \
2269 for (tmp = skb->next; \
2270 skb != (struct sk_buff *)(queue); \
2271 skb = tmp, tmp = skb->next)
2272
Stephen Hemminger300ce172005-10-30 13:47:34 -08002273#define skb_queue_reverse_walk(queue, skb) \
2274 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002275 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08002276 skb = skb->prev)
2277
David S. Miller686a2952011-01-20 22:47:32 -08002278#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2279 for (skb = (queue)->prev, tmp = skb->prev; \
2280 skb != (struct sk_buff *)(queue); \
2281 skb = tmp, tmp = skb->prev)
2282
2283#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2284 for (tmp = skb->prev; \
2285 skb != (struct sk_buff *)(queue); \
2286 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002287
David S. Miller21dc3302010-08-23 00:13:46 -07002288static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07002289{
2290 return skb_shinfo(skb)->frag_list != NULL;
2291}
2292
2293static inline void skb_frag_list_init(struct sk_buff *skb)
2294{
2295 skb_shinfo(skb)->frag_list = NULL;
2296}
2297
2298static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2299{
2300 frag->next = skb_shinfo(skb)->frag_list;
2301 skb_shinfo(skb)->frag_list = frag;
2302}
2303
2304#define skb_walk_frags(skb, iter) \
2305 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2306
Herbert Xua59322b2007-12-05 01:53:40 -08002307extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
Pavel Emelyanov3f518bf2012-02-21 07:30:58 +00002308 int *peeked, int *off, int *err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002309extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
2310 int noblock, int *err);
2311extern unsigned int datagram_poll(struct file *file, struct socket *sock,
2312 struct poll_table_struct *wait);
2313extern int skb_copy_datagram_iovec(const struct sk_buff *from,
2314 int offset, struct iovec *to,
2315 int size);
Herbert Xufb286bb2005-11-10 13:01:24 -08002316extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002317 int hlen,
2318 struct iovec *iov);
Rusty Russelldb543c12008-08-15 15:13:53 -07002319extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
2320 int offset,
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +00002321 const struct iovec *from,
2322 int from_offset,
Rusty Russelldb543c12008-08-15 15:13:53 -07002323 int len);
Michael S. Tsirkin0a1ec072009-04-20 01:25:46 +00002324extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
2325 int offset,
2326 const struct iovec *to,
2327 int to_offset,
2328 int size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002329extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
Eric Dumazet9d410c72009-10-30 05:03:53 +00002330extern void skb_free_datagram_locked(struct sock *sk,
2331 struct sk_buff *skb);
Herbert Xu27ab2562007-12-05 01:51:58 -08002332extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
Herbert Xu3305b802005-12-13 23:16:37 -08002333 unsigned int flags);
Al Viro2bbbc862006-11-14 21:37:14 -08002334extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
2335 int len, __wsum csum);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002336extern int skb_copy_bits(const struct sk_buff *skb, int offset,
2337 void *to, int len);
Stephen Hemminger0c6fcc82007-04-20 16:40:01 -07002338extern int skb_store_bits(struct sk_buff *skb, int offset,
2339 const void *from, int len);
Al Viro81d77662006-11-14 21:37:33 -08002340extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002341 int offset, u8 *to, int len,
Al Viro81d77662006-11-14 21:37:33 -08002342 __wsum csum);
Jens Axboe9c55e012007-11-06 23:30:13 -08002343extern int skb_splice_bits(struct sk_buff *skb,
2344 unsigned int offset,
2345 struct pipe_inode_info *pipe,
2346 unsigned int len,
2347 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002348extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
2349extern void skb_split(struct sk_buff *skb,
2350 struct sk_buff *skb1, const u32 len);
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08002351extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
2352 int shiftlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002353
Michał Mirosławc8f44af2011-11-15 15:29:55 +00002354extern struct sk_buff *skb_segment(struct sk_buff *skb,
2355 netdev_features_t features);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002356
Linus Torvalds1da177e2005-04-16 15:20:36 -07002357static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2358 int len, void *buffer)
2359{
2360 int hlen = skb_headlen(skb);
2361
Patrick McHardy55820ee2005-07-05 14:08:10 -07002362 if (hlen - offset >= len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002363 return skb->data + offset;
2364
2365 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2366 return NULL;
2367
2368 return buffer;
2369}
2370
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03002371static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2372 void *to,
2373 const unsigned int len)
2374{
2375 memcpy(to, skb->data, len);
2376}
2377
2378static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2379 const int offset, void *to,
2380 const unsigned int len)
2381{
2382 memcpy(to, skb->data + offset, len);
2383}
2384
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03002385static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2386 const void *from,
2387 const unsigned int len)
2388{
2389 memcpy(skb->data, from, len);
2390}
2391
2392static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2393 const int offset,
2394 const void *from,
2395 const unsigned int len)
2396{
2397 memcpy(skb->data + offset, from, len);
2398}
2399
Linus Torvalds1da177e2005-04-16 15:20:36 -07002400extern void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002401
Patrick Ohlyac45f602009-02-12 05:03:37 +00002402static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2403{
2404 return skb->tstamp;
2405}
2406
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002407/**
2408 * skb_get_timestamp - get timestamp from a skb
2409 * @skb: skb to get stamp from
2410 * @stamp: pointer to struct timeval to store stamp in
2411 *
2412 * Timestamps are stored in the skb as offsets to a base timestamp.
2413 * This function converts the offset back to a struct timeval and stores
2414 * it in stamp.
2415 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00002416static inline void skb_get_timestamp(const struct sk_buff *skb,
2417 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002418{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002419 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002420}
2421
Patrick Ohlyac45f602009-02-12 05:03:37 +00002422static inline void skb_get_timestampns(const struct sk_buff *skb,
2423 struct timespec *stamp)
2424{
2425 *stamp = ktime_to_timespec(skb->tstamp);
2426}
2427
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002428static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002429{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002430 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002431}
2432
Stephen Hemminger164891a2007-04-23 22:26:16 -07002433static inline ktime_t net_timedelta(ktime_t t)
2434{
2435 return ktime_sub(ktime_get_real(), t);
2436}
2437
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07002438static inline ktime_t net_invalid_timestamp(void)
2439{
2440 return ktime_set(0, 0);
2441}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002442
Richard Cochranc1f19b52010-07-17 08:49:36 +00002443extern void skb_timestamping_init(void);
2444
2445#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2446
2447extern void skb_clone_tx_timestamp(struct sk_buff *skb);
2448extern bool skb_defer_rx_timestamp(struct sk_buff *skb);
2449
2450#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2451
2452static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2453{
2454}
2455
2456static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2457{
2458 return false;
2459}
2460
2461#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2462
2463/**
2464 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2465 *
Richard Cochranda92b192011-10-21 00:49:15 +00002466 * PHY drivers may accept clones of transmitted packets for
2467 * timestamping via their phy_driver.txtstamp method. These drivers
2468 * must call this function to return the skb back to the stack, with
2469 * or without a timestamp.
2470 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00002471 * @skb: clone of the the original outgoing packet
Richard Cochranda92b192011-10-21 00:49:15 +00002472 * @hwtstamps: hardware time stamps, may be NULL if not available
Richard Cochranc1f19b52010-07-17 08:49:36 +00002473 *
2474 */
2475void skb_complete_tx_timestamp(struct sk_buff *skb,
2476 struct skb_shared_hwtstamps *hwtstamps);
2477
Patrick Ohlyac45f602009-02-12 05:03:37 +00002478/**
2479 * skb_tstamp_tx - queue clone of skb with send time stamps
2480 * @orig_skb: the original outgoing packet
2481 * @hwtstamps: hardware time stamps, may be NULL if not available
2482 *
2483 * If the skb has a socket associated, then this function clones the
2484 * skb (thus sharing the actual data and optional structures), stores
2485 * the optional hardware time stamping information (if non NULL) or
2486 * generates a software time stamp (otherwise), then queues the clone
2487 * to the error queue of the socket. Errors are silently ignored.
2488 */
2489extern void skb_tstamp_tx(struct sk_buff *orig_skb,
2490 struct skb_shared_hwtstamps *hwtstamps);
2491
Richard Cochran4507a712010-07-17 08:48:28 +00002492static inline void sw_tx_timestamp(struct sk_buff *skb)
2493{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002494 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2495 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00002496 skb_tstamp_tx(skb, NULL);
2497}
2498
2499/**
2500 * skb_tx_timestamp() - Driver hook for transmit timestamping
2501 *
2502 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00002503 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00002504 *
2505 * @skb: A socket buffer.
2506 */
2507static inline void skb_tx_timestamp(struct sk_buff *skb)
2508{
Richard Cochranc1f19b52010-07-17 08:49:36 +00002509 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002510 sw_tx_timestamp(skb);
2511}
2512
Johannes Berg6e3e9392011-11-09 10:15:42 +01002513/**
2514 * skb_complete_wifi_ack - deliver skb with wifi status
2515 *
2516 * @skb: the original outgoing packet
2517 * @acked: ack status
2518 *
2519 */
2520void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2521
Herbert Xu759e5d02007-03-25 20:10:56 -07002522extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
Al Virob51655b2006-11-14 21:40:42 -08002523extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002524
Herbert Xu60476372007-04-09 11:59:39 -07002525static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2526{
2527 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2528}
2529
Herbert Xufb286bb2005-11-10 13:01:24 -08002530/**
2531 * skb_checksum_complete - Calculate checksum of an entire packet
2532 * @skb: packet to process
2533 *
2534 * This function calculates the checksum over the entire packet plus
2535 * the value of skb->csum. The latter can be used to supply the
2536 * checksum of a pseudo header as used by TCP/UDP. It returns the
2537 * checksum.
2538 *
2539 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2540 * this function can be used to verify that checksum on received
2541 * packets. In that case the function should return zero if the
2542 * checksum is correct. In particular, this function will return zero
2543 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2544 * hardware has already verified the correctness of the checksum.
2545 */
Al Viro4381ca32007-07-15 21:00:11 +01002546static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002547{
Herbert Xu60476372007-04-09 11:59:39 -07002548 return skb_csum_unnecessary(skb) ?
2549 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002550}
2551
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002552#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002553extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002554static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2555{
2556 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002557 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002558}
2559static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2560{
2561 if (nfct)
2562 atomic_inc(&nfct->use);
2563}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002564#endif
2565#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -08002566static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
2567{
2568 if (skb)
2569 atomic_inc(&skb->users);
2570}
2571static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
2572{
2573 if (skb)
2574 kfree_skb(skb);
2575}
2576#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002577#ifdef CONFIG_BRIDGE_NETFILTER
2578static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2579{
2580 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2581 kfree(nf_bridge);
2582}
2583static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2584{
2585 if (nf_bridge)
2586 atomic_inc(&nf_bridge->use);
2587}
2588#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08002589static inline void nf_reset(struct sk_buff *skb)
2590{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002591#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08002592 nf_conntrack_put(skb->nfct);
2593 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002594#endif
2595#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Patrick McHardya193a4a2006-03-20 19:23:05 -08002596 nf_conntrack_put_reasm(skb->nfct_reasm);
2597 skb->nfct_reasm = NULL;
2598#endif
2599#ifdef CONFIG_BRIDGE_NETFILTER
2600 nf_bridge_put(skb->nf_bridge);
2601 skb->nf_bridge = NULL;
2602#endif
2603}
2604
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002605/* Note: This doesn't put any conntrack and bridge info in dst. */
2606static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2607{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002608#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002609 dst->nfct = src->nfct;
2610 nf_conntrack_get(src->nfct);
2611 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002612#endif
2613#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002614 dst->nfct_reasm = src->nfct_reasm;
2615 nf_conntrack_get_reasm(src->nfct_reasm);
2616#endif
2617#ifdef CONFIG_BRIDGE_NETFILTER
2618 dst->nf_bridge = src->nf_bridge;
2619 nf_bridge_get(src->nf_bridge);
2620#endif
2621}
2622
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002623static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2624{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002625#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002626 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002627#endif
2628#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002629 nf_conntrack_put_reasm(dst->nfct_reasm);
2630#endif
2631#ifdef CONFIG_BRIDGE_NETFILTER
2632 nf_bridge_put(dst->nf_bridge);
2633#endif
2634 __nf_copy(dst, src);
2635}
2636
James Morris984bc162006-06-09 00:29:17 -07002637#ifdef CONFIG_NETWORK_SECMARK
2638static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2639{
2640 to->secmark = from->secmark;
2641}
2642
2643static inline void skb_init_secmark(struct sk_buff *skb)
2644{
2645 skb->secmark = 0;
2646}
2647#else
2648static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2649{ }
2650
2651static inline void skb_init_secmark(struct sk_buff *skb)
2652{ }
2653#endif
2654
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002655static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2656{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002657 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002658}
2659
Stephen Hemminger92477442009-03-21 13:39:26 -07002660static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002661{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002662 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002663}
2664
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002665static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2666{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002667 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002668}
2669
David S. Millerd5a9e242009-01-27 16:22:11 -08002670static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2671{
2672 skb->queue_mapping = rx_queue + 1;
2673}
2674
Stephen Hemminger92477442009-03-21 13:39:26 -07002675static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002676{
2677 return skb->queue_mapping - 1;
2678}
2679
Stephen Hemminger92477442009-03-21 13:39:26 -07002680static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002681{
Eric Dumazeta02cec22010-09-22 20:43:57 +00002682 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08002683}
2684
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00002685extern u16 __skb_tx_hash(const struct net_device *dev,
2686 const struct sk_buff *skb,
2687 unsigned int num_tx_queues);
Stephen Hemminger92477442009-03-21 13:39:26 -07002688
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002689#ifdef CONFIG_XFRM
2690static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2691{
2692 return skb->sp;
2693}
2694#else
2695static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2696{
2697 return NULL;
2698}
2699#endif
2700
David S. Millerbdcc0922012-03-07 20:53:36 -05002701static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07002702{
2703 return skb_shinfo(skb)->gso_size;
2704}
2705
David S. Millerbdcc0922012-03-07 20:53:36 -05002706static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02002707{
2708 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2709}
2710
Ben Hutchings4497b072008-06-19 16:22:28 -07002711extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2712
2713static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2714{
2715 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2716 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002717 const struct skb_shared_info *shinfo = skb_shinfo(skb);
2718
Alexander Duyckb78462e2010-06-02 12:24:37 +00002719 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2720 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07002721 __skb_warn_lro_forwarding(skb);
2722 return true;
2723 }
2724 return false;
2725}
2726
Herbert Xu35fc92a2007-03-26 23:22:20 -07002727static inline void skb_forward_csum(struct sk_buff *skb)
2728{
2729 /* Unfortunately we don't support this one. Any brave souls? */
2730 if (skb->ip_summed == CHECKSUM_COMPLETE)
2731 skb->ip_summed = CHECKSUM_NONE;
2732}
2733
Eric Dumazetbc8acf22010-09-02 13:07:41 -07002734/**
2735 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2736 * @skb: skb to check
2737 *
2738 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2739 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2740 * use this helper, to document places where we make this assertion.
2741 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002742static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07002743{
2744#ifdef DEBUG
2745 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2746#endif
2747}
2748
Rusty Russellf35d9d82008-02-04 23:49:54 -05002749bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00002750
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00002751/**
2752 * skb_head_is_locked - Determine if the skb->head is locked down
2753 * @skb: skb to check
2754 *
2755 * The head on skbs build around a head frag can be removed if they are
2756 * not cloned. This function returns true if the skb head is locked down
2757 * due to either being allocated via kmalloc, or by being a clone with
2758 * multiple references to the head.
2759 */
2760static inline bool skb_head_is_locked(const struct sk_buff *skb)
2761{
2762 return !skb->head_frag || skb_cloned(skb);
2763}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002764#endif /* __KERNEL__ */
2765#endif /* _LINUX_SKBUFF_H */