blob: e5eae5bc1d4da4881276ec59084bd53568f9c451 [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
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700379 * @transport_header: Transport layer header
380 * @network_header: Network layer header
381 * @mac_header: Link layer header
382 * @tail: Tail pointer
383 * @end: End pointer
384 * @head: Head of buffer
385 * @data: Data head pointer
386 * @truesize: Buffer size
387 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700388 */
389
390struct sk_buff {
391 /* These two members must be first. */
392 struct sk_buff *next;
393 struct sk_buff *prev;
394
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700395 ktime_t tstamp;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000396
397 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400 /*
401 * This is the control buffer. It is free to use for every
402 * layer. Please put your private variables there. If you
403 * want to keep them across layers you have to do a skb_clone()
404 * first. This is owned by whoever has the skb queued ATM.
405 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000406 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407
Eric Dumazet7fee2262010-05-11 23:19:48 +0000408 unsigned long _skb_refdst;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000409#ifdef CONFIG_XFRM
410 struct sec_path *sp;
411#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700413 data_len;
414 __u16 mac_len,
415 hdr_len;
Al Viroff1dcad2006-11-20 18:07:29 -0800416 union {
417 __wsum csum;
Herbert Xu663ead32007-04-09 11:59:07 -0700418 struct {
419 __u16 csum_start;
420 __u16 csum_offset;
421 };
Al Viroff1dcad2006-11-20 18:07:29 -0800422 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423 __u32 priority;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200424 kmemcheck_bitfield_begin(flags1);
Thomas Graf1cbb3382005-07-05 14:13:41 -0700425 __u8 local_df:1,
426 cloned:1,
427 ip_summed:2,
Harald Welte6869c4d2005-08-09 19:24:19 -0700428 nohdr:1,
429 nfctinfo:3;
David S. Millerd179cd12005-08-17 14:57:30 -0700430 __u8 pkt_type:3,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800431 fclone:2,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700432 ipvs_property:1,
Herbert Xua59322b2007-12-05 01:53:40 -0800433 peeked:1,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700434 nf_trace:1;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200435 kmemcheck_bitfield_end(flags1);
Eric Dumazet4ab408d2010-03-01 03:09:26 +0000436 __be16 protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437
438 void (*destructor)(struct sk_buff *skb);
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800439#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700440 struct nf_conntrack *nfct;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100441#endif
442#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800443 struct sk_buff *nfct_reasm;
444#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445#ifdef CONFIG_BRIDGE_NETFILTER
446 struct nf_bridge_info *nf_bridge;
447#endif
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700448
Eric Dumazet8964be42009-11-20 15:35:04 -0800449 int skb_iif;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000450
451 __u32 rxhash;
452
453 __u16 vlan_tci;
454
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700456 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700458 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200461
Eric Dumazet0a148422011-04-20 09:27:32 +0000462 __u16 queue_mapping;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200463 kmemcheck_bitfield_begin(flags2);
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400464#ifdef CONFIG_IPV6_NDISC_NODETYPE
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000465 __u8 ndisc_nodetype:2;
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400466#endif
Mel Gormanc93bdd02012-07-31 16:44:19 -0700467 __u8 pfmemalloc:1;
Tom Herbert3853b582010-11-21 13:17:29 +0000468 __u8 ooo_okay:1;
Tom Herbertbdeab992011-08-14 19:45:55 +0000469 __u8 l4_rxhash:1;
Johannes Berg6e3e9392011-11-09 10:15:42 +0100470 __u8 wifi_acked_valid:1;
471 __u8 wifi_acked:1;
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000472 __u8 no_fcs:1;
Eric Dumazetd3836f22012-04-27 00:33:38 +0000473 __u8 head_frag:1;
474 /* 8/10 bit hole (depending on ndisc_nodetype presence) */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200475 kmemcheck_bitfield_end(flags2);
476
Chris Leech97fc2f02006-05-23 17:55:33 -0700477#ifdef CONFIG_NET_DMA
478 dma_cookie_t dma_cookie;
479#endif
James Morris984bc162006-06-09 00:29:17 -0700480#ifdef CONFIG_NETWORK_SECMARK
481 __u32 secmark;
482#endif
Neil Horman3b885782009-10-12 13:26:31 -0700483 union {
484 __u32 mark;
485 __u32 dropcount;
Eric Dumazeta21d4572012-04-10 20:30:48 +0000486 __u32 avail_size;
Neil Horman3b885782009-10-12 13:26:31 -0700487 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700489 sk_buff_data_t transport_header;
490 sk_buff_data_t network_header;
491 sk_buff_data_t mac_header;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700493 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700494 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700496 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700497 unsigned int truesize;
498 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499};
500
501#ifdef __KERNEL__
502/*
503 * Handling routines are only of interest to the kernel
504 */
505#include <linux/slab.h>
506
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507
Mel Gormanc93bdd02012-07-31 16:44:19 -0700508#define SKB_ALLOC_FCLONE 0x01
509#define SKB_ALLOC_RX 0x02
510
511/* Returns true if the skb was allocated from PFMEMALLOC reserves */
512static inline bool skb_pfmemalloc(const struct sk_buff *skb)
513{
514 return unlikely(skb->pfmemalloc);
515}
516
Eric Dumazet7fee2262010-05-11 23:19:48 +0000517/*
518 * skb might have a dst pointer attached, refcounted or not.
519 * _skb_refdst low order bit is set if refcount was _not_ taken
520 */
521#define SKB_DST_NOREF 1UL
522#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
523
524/**
525 * skb_dst - returns skb dst_entry
526 * @skb: buffer
527 *
528 * Returns skb dst_entry, regardless of reference taken or not.
529 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000530static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
531{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000532 /* If refdst was not refcounted, check we still are in a
533 * rcu_read_lock section
534 */
535 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
536 !rcu_read_lock_held() &&
537 !rcu_read_lock_bh_held());
538 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000539}
540
Eric Dumazet7fee2262010-05-11 23:19:48 +0000541/**
542 * skb_dst_set - sets skb dst
543 * @skb: buffer
544 * @dst: dst entry
545 *
546 * Sets skb dst, assuming a reference was taken on dst and should
547 * be released by skb_dst_drop()
548 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000549static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
550{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000551 skb->_skb_refdst = (unsigned long)dst;
552}
553
Eric Dumazet27b75c92010-10-15 05:44:11 +0000554extern void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst);
Eric Dumazet7fee2262010-05-11 23:19:48 +0000555
556/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300557 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000558 * @skb: buffer
559 */
560static inline bool skb_dst_is_noref(const struct sk_buff *skb)
561{
562 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000563}
564
Eric Dumazet511c3f92009-06-02 05:14:27 +0000565static inline struct rtable *skb_rtable(const struct sk_buff *skb)
566{
Eric Dumazetadf30902009-06-02 05:19:30 +0000567 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000568}
569
Jörn Engel231d06a2006-03-20 21:28:35 -0800570extern void kfree_skb(struct sk_buff *skb);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000571extern void consume_skb(struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572extern void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000573extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000574
575extern void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
576extern bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
577 bool *fragstolen, int *delta_truesize);
578
David S. Millerd179cd12005-08-17 14:57:30 -0700579extern struct sk_buff *__alloc_skb(unsigned int size,
Mel Gormanc93bdd02012-07-31 16:44:19 -0700580 gfp_t priority, int flags, int node);
Eric Dumazetd3836f22012-04-27 00:33:38 +0000581extern struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700582static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100583 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700584{
Eric Dumazet564824b2010-10-11 19:05:25 +0000585 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700586}
587
588static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100589 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700590{
Mel Gormanc93bdd02012-07-31 16:44:19 -0700591 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700592}
593
Herbert Xue0053ec2007-10-14 00:37:52 -0700594extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
Michael S. Tsirkin48c83012011-08-31 08:03:29 +0000595extern int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700596extern struct sk_buff *skb_clone(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100597 gfp_t priority);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700598extern struct sk_buff *skb_copy(const struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100599 gfp_t priority);
Eric Dumazet117632e2011-12-03 21:39:53 +0000600extern struct sk_buff *__pskb_copy(struct sk_buff *skb,
601 int headroom, gfp_t gfp_mask);
602
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603extern int pskb_expand_head(struct sk_buff *skb,
Victor Fusco86a76ca2005-07-08 14:57:47 -0700604 int nhead, int ntail,
Al Virodd0fc662005-10-07 07:46:04 +0100605 gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
607 unsigned int headroom);
608extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
609 int newheadroom, int newtailroom,
Al Virodd0fc662005-10-07 07:46:04 +0100610 gfp_t priority);
David Howells716ea3a2007-04-02 20:19:53 -0700611extern int skb_to_sgvec(struct sk_buff *skb,
612 struct scatterlist *sg, int offset,
613 int len);
614extern int skb_cow_data(struct sk_buff *skb, int tailbits,
615 struct sk_buff **trailer);
Herbert Xu5b057c62006-06-23 02:06:41 -0700616extern int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000617#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700619extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
620 int getfrag(void *from, char *to, int offset,
621 int len,int odd, struct sk_buff *skb),
622 void *from, int length);
623
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800624struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700625 __u32 lower_offset;
626 __u32 upper_offset;
627 __u32 frag_idx;
628 __u32 stepped_offset;
629 struct sk_buff *root_skb;
630 struct sk_buff *cur_skb;
631 __u8 *frag_data;
632};
633
634extern void skb_prepare_seq_read(struct sk_buff *skb,
635 unsigned int from, unsigned int to,
636 struct skb_seq_state *st);
637extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
638 struct skb_seq_state *st);
639extern void skb_abort_seq_read(struct skb_seq_state *st);
640
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700641extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
642 unsigned int to, struct ts_config *config,
643 struct ts_state *state);
644
Tom Herbertbdeab992011-08-14 19:45:55 +0000645extern void __skb_get_rxhash(struct sk_buff *skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000646static inline __u32 skb_get_rxhash(struct sk_buff *skb)
647{
Willem de Bruijnecd5cf52012-10-26 11:52:08 +0000648 if (!skb->l4_rxhash)
Tom Herbertbdeab992011-08-14 19:45:55 +0000649 __skb_get_rxhash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000650
651 return skb->rxhash;
652}
653
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700654#ifdef NET_SKBUFF_DATA_USES_OFFSET
655static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
656{
657 return skb->head + skb->end;
658}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000659
660static inline unsigned int skb_end_offset(const struct sk_buff *skb)
661{
662 return skb->end;
663}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700664#else
665static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
666{
667 return skb->end;
668}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000669
670static inline unsigned int skb_end_offset(const struct sk_buff *skb)
671{
672 return skb->end - skb->head;
673}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700674#endif
675
Linus Torvalds1da177e2005-04-16 15:20:36 -0700676/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700677#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700678
Patrick Ohlyac45f602009-02-12 05:03:37 +0000679static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
680{
681 return &skb_shinfo(skb)->hwtstamps;
682}
683
Linus Torvalds1da177e2005-04-16 15:20:36 -0700684/**
685 * skb_queue_empty - check if a queue is empty
686 * @list: queue head
687 *
688 * Returns true if the queue is empty, false otherwise.
689 */
690static inline int skb_queue_empty(const struct sk_buff_head *list)
691{
692 return list->next == (struct sk_buff *)list;
693}
694
695/**
David S. Millerfc7ebb22008-09-23 00:34:07 -0700696 * skb_queue_is_last - check if skb is the last entry in the queue
697 * @list: queue head
698 * @skb: buffer
699 *
700 * Returns true if @skb is the last buffer on the list.
701 */
702static inline bool skb_queue_is_last(const struct sk_buff_head *list,
703 const struct sk_buff *skb)
704{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000705 return skb->next == (struct sk_buff *)list;
David S. Millerfc7ebb22008-09-23 00:34:07 -0700706}
707
708/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800709 * skb_queue_is_first - check if skb is the first entry in the queue
710 * @list: queue head
711 * @skb: buffer
712 *
713 * Returns true if @skb is the first buffer on the list.
714 */
715static inline bool skb_queue_is_first(const struct sk_buff_head *list,
716 const struct sk_buff *skb)
717{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000718 return skb->prev == (struct sk_buff *)list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800719}
720
721/**
David S. Miller249c8b42008-09-23 00:44:42 -0700722 * skb_queue_next - return the next packet in the queue
723 * @list: queue head
724 * @skb: current buffer
725 *
726 * Return the next packet in @list after @skb. It is only valid to
727 * call this if skb_queue_is_last() evaluates to false.
728 */
729static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
730 const struct sk_buff *skb)
731{
732 /* This BUG_ON may seem severe, but if we just return then we
733 * are going to dereference garbage.
734 */
735 BUG_ON(skb_queue_is_last(list, skb));
736 return skb->next;
737}
738
739/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800740 * skb_queue_prev - return the prev packet in the queue
741 * @list: queue head
742 * @skb: current buffer
743 *
744 * Return the prev packet in @list before @skb. It is only valid to
745 * call this if skb_queue_is_first() evaluates to false.
746 */
747static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
748 const struct sk_buff *skb)
749{
750 /* This BUG_ON may seem severe, but if we just return then we
751 * are going to dereference garbage.
752 */
753 BUG_ON(skb_queue_is_first(list, skb));
754 return skb->prev;
755}
756
757/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758 * skb_get - reference buffer
759 * @skb: buffer to reference
760 *
761 * Makes another reference to a socket buffer and returns a pointer
762 * to the buffer.
763 */
764static inline struct sk_buff *skb_get(struct sk_buff *skb)
765{
766 atomic_inc(&skb->users);
767 return skb;
768}
769
770/*
771 * If users == 1, we are the only owner and are can avoid redundant
772 * atomic change.
773 */
774
775/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776 * skb_cloned - is the buffer a clone
777 * @skb: buffer to check
778 *
779 * Returns true if the buffer was generated with skb_clone() and is
780 * one of multiple shared copies of the buffer. Cloned buffers are
781 * shared data so must not be written to under normal circumstances.
782 */
783static inline int skb_cloned(const struct sk_buff *skb)
784{
785 return skb->cloned &&
786 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
787}
788
789/**
790 * skb_header_cloned - is the header a clone
791 * @skb: buffer to check
792 *
793 * Returns true if modifying the header part of the buffer requires
794 * the data to be copied.
795 */
796static inline int skb_header_cloned(const struct sk_buff *skb)
797{
798 int dataref;
799
800 if (!skb->cloned)
801 return 0;
802
803 dataref = atomic_read(&skb_shinfo(skb)->dataref);
804 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
805 return dataref != 1;
806}
807
808/**
809 * skb_header_release - release reference to header
810 * @skb: buffer to operate on
811 *
812 * Drop a reference to the header part of the buffer. This is done
813 * by acquiring a payload reference. You must not read from the header
814 * part of skb->data after this.
815 */
816static inline void skb_header_release(struct sk_buff *skb)
817{
818 BUG_ON(skb->nohdr);
819 skb->nohdr = 1;
820 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
821}
822
823/**
824 * skb_shared - is the buffer shared
825 * @skb: buffer to check
826 *
827 * Returns true if more than one person has a reference to this
828 * buffer.
829 */
830static inline int skb_shared(const struct sk_buff *skb)
831{
832 return atomic_read(&skb->users) != 1;
833}
834
835/**
836 * skb_share_check - check if buffer is shared and if so clone it
837 * @skb: buffer to check
838 * @pri: priority for memory allocation
839 *
840 * If the buffer is shared the buffer is cloned and the old copy
841 * drops a reference. A new clone with a single reference is returned.
842 * If the buffer is not shared the original buffer is returned. When
843 * being called from interrupt status or with spinlocks held pri must
844 * be GFP_ATOMIC.
845 *
846 * NULL is returned on a memory allocation failure.
847 */
Eric Dumazet47061bc2012-08-03 20:54:15 +0000848static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700849{
850 might_sleep_if(pri & __GFP_WAIT);
851 if (skb_shared(skb)) {
852 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +0000853
854 if (likely(nskb))
855 consume_skb(skb);
856 else
857 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858 skb = nskb;
859 }
860 return skb;
861}
862
863/*
864 * Copy shared buffers into a new sk_buff. We effectively do COW on
865 * packets to handle cases where we have a local reader and forward
866 * and a couple of other messy ones. The normal one is tcpdumping
867 * a packet thats being forwarded.
868 */
869
870/**
871 * skb_unshare - make a copy of a shared buffer
872 * @skb: buffer to check
873 * @pri: priority for memory allocation
874 *
875 * If the socket buffer is a clone then this function creates a new
876 * copy of the data, drops a reference count on the old copy and returns
877 * the new copy with the reference count at 1. If the buffer is not a clone
878 * the original buffer is returned. When called with a spinlock held or
879 * from interrupt state @pri must be %GFP_ATOMIC
880 *
881 * %NULL is returned on a memory allocation failure.
882 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -0700883static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100884 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885{
886 might_sleep_if(pri & __GFP_WAIT);
887 if (skb_cloned(skb)) {
888 struct sk_buff *nskb = skb_copy(skb, pri);
889 kfree_skb(skb); /* Free our shared copy */
890 skb = nskb;
891 }
892 return skb;
893}
894
895/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -0800896 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897 * @list_: list to peek at
898 *
899 * Peek an &sk_buff. Unlike most other operations you _MUST_
900 * be careful with this one. A peek leaves the buffer on the
901 * list and someone else may run off with it. You must hold
902 * the appropriate locks or have a private queue to do this.
903 *
904 * Returns %NULL for an empty list or a pointer to the head element.
905 * The reference count is not incremented and the reference is therefore
906 * volatile. Use with caution.
907 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -0400908static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700909{
Eric Dumazet18d07002012-04-30 16:31:46 +0000910 struct sk_buff *skb = list_->next;
911
912 if (skb == (struct sk_buff *)list_)
913 skb = NULL;
914 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700915}
916
917/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +0000918 * skb_peek_next - peek skb following the given one from a queue
919 * @skb: skb to start from
920 * @list_: list to peek at
921 *
922 * Returns %NULL when the end of the list is met or a pointer to the
923 * next element. The reference count is not incremented and the
924 * reference is therefore volatile. Use with caution.
925 */
926static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
927 const struct sk_buff_head *list_)
928{
929 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +0000930
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +0000931 if (next == (struct sk_buff *)list_)
932 next = NULL;
933 return next;
934}
935
936/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -0800937 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938 * @list_: list to peek at
939 *
940 * Peek an &sk_buff. Unlike most other operations you _MUST_
941 * be careful with this one. A peek leaves the buffer on the
942 * list and someone else may run off with it. You must hold
943 * the appropriate locks or have a private queue to do this.
944 *
945 * Returns %NULL for an empty list or a pointer to the tail element.
946 * The reference count is not incremented and the reference is therefore
947 * volatile. Use with caution.
948 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -0400949static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700950{
Eric Dumazet18d07002012-04-30 16:31:46 +0000951 struct sk_buff *skb = list_->prev;
952
953 if (skb == (struct sk_buff *)list_)
954 skb = NULL;
955 return skb;
956
Linus Torvalds1da177e2005-04-16 15:20:36 -0700957}
958
959/**
960 * skb_queue_len - get queue length
961 * @list_: list to measure
962 *
963 * Return the length of an &sk_buff queue.
964 */
965static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
966{
967 return list_->qlen;
968}
969
David S. Miller67fed452008-09-21 22:36:24 -0700970/**
971 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
972 * @list: queue to initialize
973 *
974 * This initializes only the list and queue length aspects of
975 * an sk_buff_head object. This allows to initialize the list
976 * aspects of an sk_buff_head without reinitializing things like
977 * the spinlock. It can also be used for on-stack sk_buff_head
978 * objects where the spinlock is known to not be used.
979 */
980static inline void __skb_queue_head_init(struct sk_buff_head *list)
981{
982 list->prev = list->next = (struct sk_buff *)list;
983 list->qlen = 0;
984}
985
Arjan van de Ven76f10ad2006-08-02 14:06:55 -0700986/*
987 * This function creates a split out lock class for each invocation;
988 * this is needed for now since a whole lot of users of the skb-queue
989 * infrastructure in drivers have different locking usage (in hardirq)
990 * than the networking core (in softirq only). In the long run either the
991 * network layer or drivers should need annotation to consolidate the
992 * main types of usage into 3 classes.
993 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700994static inline void skb_queue_head_init(struct sk_buff_head *list)
995{
996 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -0700997 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700998}
999
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001000static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1001 struct lock_class_key *class)
1002{
1003 skb_queue_head_init(list);
1004 lockdep_set_class(&list->lock, class);
1005}
1006
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001008 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009 *
1010 * The "__skb_xxxx()" functions are the non-atomic ones that
1011 * can only be called with interrupts disabled.
1012 */
Gerrit Renkerbf299272008-04-14 00:04:51 -07001013extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
1014static inline void __skb_insert(struct sk_buff *newsk,
1015 struct sk_buff *prev, struct sk_buff *next,
1016 struct sk_buff_head *list)
1017{
1018 newsk->next = next;
1019 newsk->prev = prev;
1020 next->prev = prev->next = newsk;
1021 list->qlen++;
1022}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001023
David S. Miller67fed452008-09-21 22:36:24 -07001024static inline void __skb_queue_splice(const struct sk_buff_head *list,
1025 struct sk_buff *prev,
1026 struct sk_buff *next)
1027{
1028 struct sk_buff *first = list->next;
1029 struct sk_buff *last = list->prev;
1030
1031 first->prev = prev;
1032 prev->next = first;
1033
1034 last->next = next;
1035 next->prev = last;
1036}
1037
1038/**
1039 * skb_queue_splice - join two skb lists, this is designed for stacks
1040 * @list: the new list to add
1041 * @head: the place to add it in the first list
1042 */
1043static inline void skb_queue_splice(const struct sk_buff_head *list,
1044 struct sk_buff_head *head)
1045{
1046 if (!skb_queue_empty(list)) {
1047 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001048 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001049 }
1050}
1051
1052/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001053 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001054 * @list: the new list to add
1055 * @head: the place to add it in the first list
1056 *
1057 * The list at @list is reinitialised
1058 */
1059static inline void skb_queue_splice_init(struct sk_buff_head *list,
1060 struct sk_buff_head *head)
1061{
1062 if (!skb_queue_empty(list)) {
1063 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001064 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001065 __skb_queue_head_init(list);
1066 }
1067}
1068
1069/**
1070 * skb_queue_splice_tail - join two skb lists, each list being a queue
1071 * @list: the new list to add
1072 * @head: the place to add it in the first list
1073 */
1074static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1075 struct sk_buff_head *head)
1076{
1077 if (!skb_queue_empty(list)) {
1078 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001079 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001080 }
1081}
1082
1083/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001084 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001085 * @list: the new list to add
1086 * @head: the place to add it in the first list
1087 *
1088 * Each of the lists is a queue.
1089 * The list at @list is reinitialised
1090 */
1091static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1092 struct sk_buff_head *head)
1093{
1094 if (!skb_queue_empty(list)) {
1095 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001096 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001097 __skb_queue_head_init(list);
1098 }
1099}
1100
Linus Torvalds1da177e2005-04-16 15:20:36 -07001101/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001102 * __skb_queue_after - queue a buffer at the list head
1103 * @list: list to use
1104 * @prev: place after this buffer
1105 * @newsk: buffer to queue
1106 *
1107 * Queue a buffer int the middle of a list. This function takes no locks
1108 * and you must therefore hold required locks before calling it.
1109 *
1110 * A buffer cannot be placed on two lists at the same time.
1111 */
1112static inline void __skb_queue_after(struct sk_buff_head *list,
1113 struct sk_buff *prev,
1114 struct sk_buff *newsk)
1115{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001116 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001117}
1118
Gerrit Renker7de6c032008-04-14 00:05:09 -07001119extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1120 struct sk_buff_head *list);
1121
Gerrit Renkerf5572852008-04-14 00:05:28 -07001122static inline void __skb_queue_before(struct sk_buff_head *list,
1123 struct sk_buff *next,
1124 struct sk_buff *newsk)
1125{
1126 __skb_insert(newsk, next->prev, next, list);
1127}
1128
Stephen Hemminger300ce172005-10-30 13:47:34 -08001129/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130 * __skb_queue_head - queue a buffer at the list head
1131 * @list: list to use
1132 * @newsk: buffer to queue
1133 *
1134 * Queue a buffer at the start of a list. This function takes no locks
1135 * and you must therefore hold required locks before calling it.
1136 *
1137 * A buffer cannot be placed on two lists at the same time.
1138 */
1139extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1140static inline void __skb_queue_head(struct sk_buff_head *list,
1141 struct sk_buff *newsk)
1142{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001143 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001144}
1145
1146/**
1147 * __skb_queue_tail - queue a buffer at the list tail
1148 * @list: list to use
1149 * @newsk: buffer to queue
1150 *
1151 * Queue a buffer at the end of a list. This function takes no locks
1152 * and you must therefore hold required locks before calling it.
1153 *
1154 * A buffer cannot be placed on two lists at the same time.
1155 */
1156extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1157static inline void __skb_queue_tail(struct sk_buff_head *list,
1158 struct sk_buff *newsk)
1159{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001160 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161}
1162
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164 * remove sk_buff from list. _Must_ be called atomically, and with
1165 * the list known..
1166 */
David S. Miller8728b832005-08-09 19:25:21 -07001167extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1169{
1170 struct sk_buff *next, *prev;
1171
1172 list->qlen--;
1173 next = skb->next;
1174 prev = skb->prev;
1175 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001176 next->prev = prev;
1177 prev->next = next;
1178}
1179
Gerrit Renkerf525c062008-04-14 00:04:12 -07001180/**
1181 * __skb_dequeue - remove from the head of the queue
1182 * @list: list to dequeue from
1183 *
1184 * Remove the head of the list. This function does not take any locks
1185 * so must be used with appropriate locks held only. The head item is
1186 * returned or %NULL if the list is empty.
1187 */
1188extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1189static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1190{
1191 struct sk_buff *skb = skb_peek(list);
1192 if (skb)
1193 __skb_unlink(skb, list);
1194 return skb;
1195}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196
1197/**
1198 * __skb_dequeue_tail - remove from the tail of the queue
1199 * @list: list to dequeue from
1200 *
1201 * Remove the tail of the list. This function does not take any locks
1202 * so must be used with appropriate locks held only. The tail item is
1203 * returned or %NULL if the list is empty.
1204 */
1205extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1206static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1207{
1208 struct sk_buff *skb = skb_peek_tail(list);
1209 if (skb)
1210 __skb_unlink(skb, list);
1211 return skb;
1212}
1213
1214
David S. Millerbdcc0922012-03-07 20:53:36 -05001215static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001216{
1217 return skb->data_len;
1218}
1219
1220static inline unsigned int skb_headlen(const struct sk_buff *skb)
1221{
1222 return skb->len - skb->data_len;
1223}
1224
1225static inline int skb_pagelen(const struct sk_buff *skb)
1226{
1227 int i, len = 0;
1228
1229 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001230 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001231 return len + skb_headlen(skb);
1232}
1233
Ian Campbell131ea662011-08-19 06:25:00 +00001234/**
1235 * __skb_fill_page_desc - initialise a paged fragment in an skb
1236 * @skb: buffer containing fragment to be initialised
1237 * @i: paged fragment index to initialise
1238 * @page: the page to use for this fragment
1239 * @off: the offset to the data with @page
1240 * @size: the length of the data
1241 *
1242 * Initialises the @i'th fragment of @skb to point to &size bytes at
1243 * offset @off within @page.
1244 *
1245 * Does not take any additional reference on the fragment.
1246 */
1247static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1248 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001249{
1250 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1251
Mel Gormanc48a11c2012-07-31 16:44:23 -07001252 /*
1253 * Propagate page->pfmemalloc to the skb if we can. The problem is
1254 * that not all callers have unique ownership of the page. If
1255 * pfmemalloc is set, we check the mapping as a mapping implies
1256 * page->index is set (index and pfmemalloc share space).
1257 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1258 * do not lose pfmemalloc information as the pages would not be
1259 * allocated using __GFP_MEMALLOC.
1260 */
1261 if (page->pfmemalloc && !page->mapping)
1262 skb->pfmemalloc = true;
Ian Campbella8605c62011-10-19 23:01:49 +00001263 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001264 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001265 skb_frag_size_set(frag, size);
Ian Campbell131ea662011-08-19 06:25:00 +00001266}
1267
1268/**
1269 * skb_fill_page_desc - initialise a paged fragment in an skb
1270 * @skb: buffer containing fragment to be initialised
1271 * @i: paged fragment index to initialise
1272 * @page: the page to use for this fragment
1273 * @off: the offset to the data with @page
1274 * @size: the length of the data
1275 *
1276 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1277 * @skb to point to &size bytes at offset @off within @page. In
1278 * addition updates @skb such that @i is the last fragment.
1279 *
1280 * Does not take any additional reference on the fragment.
1281 */
1282static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1283 struct page *page, int off, int size)
1284{
1285 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001286 skb_shinfo(skb)->nr_frags = i + 1;
1287}
1288
Peter Zijlstra654bed12008-10-07 14:22:33 -07001289extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
Eric Dumazet50269e12012-03-23 23:59:33 +00001290 int off, int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001291
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001293#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001294#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1295
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001296#ifdef NET_SKBUFF_DATA_USES_OFFSET
1297static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1298{
1299 return skb->head + skb->tail;
1300}
1301
1302static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1303{
1304 skb->tail = skb->data - skb->head;
1305}
1306
1307static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1308{
1309 skb_reset_tail_pointer(skb);
1310 skb->tail += offset;
1311}
1312#else /* NET_SKBUFF_DATA_USES_OFFSET */
1313static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1314{
1315 return skb->tail;
1316}
1317
1318static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1319{
1320 skb->tail = skb->data;
1321}
1322
1323static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1324{
1325 skb->tail = skb->data + offset;
1326}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001327
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001328#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1329
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330/*
1331 * Add data to an sk_buff
1332 */
Ilpo Järvinen0dde3e12008-03-27 17:43:41 -07001333extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001334static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1335{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001336 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001337 SKB_LINEAR_ASSERT(skb);
1338 skb->tail += len;
1339 skb->len += len;
1340 return tmp;
1341}
1342
Ilpo Järvinenc2aa2702008-03-27 17:52:40 -07001343extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001344static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1345{
1346 skb->data -= len;
1347 skb->len += len;
1348 return skb->data;
1349}
1350
Ilpo Järvinen6be8ac22008-03-27 17:47:24 -07001351extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1353{
1354 skb->len -= len;
1355 BUG_ON(skb->len < skb->data_len);
1356 return skb->data += len;
1357}
1358
David S. Miller47d29642010-05-02 02:21:44 -07001359static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1360{
1361 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1362}
1363
Linus Torvalds1da177e2005-04-16 15:20:36 -07001364extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1365
1366static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1367{
1368 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001369 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370 return NULL;
1371 skb->len -= len;
1372 return skb->data += len;
1373}
1374
1375static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1376{
1377 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1378}
1379
1380static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1381{
1382 if (likely(len <= skb_headlen(skb)))
1383 return 1;
1384 if (unlikely(len > skb->len))
1385 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001386 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001387}
1388
1389/**
1390 * skb_headroom - bytes at buffer head
1391 * @skb: buffer to check
1392 *
1393 * Return the number of bytes of free space at the head of an &sk_buff.
1394 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001395static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001396{
1397 return skb->data - skb->head;
1398}
1399
1400/**
1401 * skb_tailroom - bytes at buffer end
1402 * @skb: buffer to check
1403 *
1404 * Return the number of bytes of free space at the tail of an sk_buff
1405 */
1406static inline int skb_tailroom(const struct sk_buff *skb)
1407{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001408 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001409}
1410
1411/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00001412 * skb_availroom - 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 * allocated by sk_stream_alloc()
1417 */
1418static inline int skb_availroom(const struct sk_buff *skb)
1419{
1420 return skb_is_nonlinear(skb) ? 0 : skb->avail_size - skb->len;
1421}
1422
1423/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001424 * skb_reserve - adjust headroom
1425 * @skb: buffer to alter
1426 * @len: bytes to move
1427 *
1428 * Increase the headroom of an empty &sk_buff by reducing the tail
1429 * room. This is only allowed for an empty buffer.
1430 */
David S. Miller8243126c2006-01-17 02:54:21 -08001431static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001432{
1433 skb->data += len;
1434 skb->tail += len;
1435}
1436
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00001437static inline void skb_reset_mac_len(struct sk_buff *skb)
1438{
1439 skb->mac_len = skb->network_header - skb->mac_header;
1440}
1441
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001442#ifdef NET_SKBUFF_DATA_USES_OFFSET
1443static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1444{
1445 return skb->head + skb->transport_header;
1446}
1447
1448static inline void skb_reset_transport_header(struct sk_buff *skb)
1449{
1450 skb->transport_header = skb->data - skb->head;
1451}
1452
1453static inline void skb_set_transport_header(struct sk_buff *skb,
1454 const int offset)
1455{
1456 skb_reset_transport_header(skb);
1457 skb->transport_header += offset;
1458}
1459
1460static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1461{
1462 return skb->head + skb->network_header;
1463}
1464
1465static inline void skb_reset_network_header(struct sk_buff *skb)
1466{
1467 skb->network_header = skb->data - skb->head;
1468}
1469
1470static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1471{
1472 skb_reset_network_header(skb);
1473 skb->network_header += offset;
1474}
1475
1476static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1477{
1478 return skb->head + skb->mac_header;
1479}
1480
1481static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1482{
1483 return skb->mac_header != ~0U;
1484}
1485
1486static inline void skb_reset_mac_header(struct sk_buff *skb)
1487{
1488 skb->mac_header = skb->data - skb->head;
1489}
1490
1491static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1492{
1493 skb_reset_mac_header(skb);
1494 skb->mac_header += offset;
1495}
1496
1497#else /* NET_SKBUFF_DATA_USES_OFFSET */
1498
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001499static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1500{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001501 return skb->transport_header;
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001502}
1503
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001504static inline void skb_reset_transport_header(struct sk_buff *skb)
1505{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001506 skb->transport_header = skb->data;
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001507}
1508
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001509static inline void skb_set_transport_header(struct sk_buff *skb,
1510 const int offset)
1511{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001512 skb->transport_header = skb->data + offset;
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001513}
1514
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001515static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1516{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001517 return skb->network_header;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001518}
1519
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001520static inline void skb_reset_network_header(struct sk_buff *skb)
1521{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001522 skb->network_header = skb->data;
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001523}
1524
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001525static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1526{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001527 skb->network_header = skb->data + offset;
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001528}
1529
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001530static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1531{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001532 return skb->mac_header;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001533}
1534
1535static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1536{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001537 return skb->mac_header != NULL;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001538}
1539
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001540static inline void skb_reset_mac_header(struct sk_buff *skb)
1541{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001542 skb->mac_header = skb->data;
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001543}
1544
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001545static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1546{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001547 skb->mac_header = skb->data + offset;
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001548}
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001549#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1550
Eric Dumazet03606892012-02-23 10:55:02 +00001551static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1552{
1553 if (skb_mac_header_was_set(skb)) {
1554 const unsigned char *old_mac = skb_mac_header(skb);
1555
1556 skb_set_mac_header(skb, -skb->mac_len);
1557 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1558 }
1559}
1560
Michał Mirosław04fb4512010-12-14 15:24:08 +00001561static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1562{
1563 return skb->csum_start - skb_headroom(skb);
1564}
1565
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001566static inline int skb_transport_offset(const struct sk_buff *skb)
1567{
1568 return skb_transport_header(skb) - skb->data;
1569}
1570
1571static inline u32 skb_network_header_len(const struct sk_buff *skb)
1572{
1573 return skb->transport_header - skb->network_header;
1574}
1575
1576static inline int skb_network_offset(const struct sk_buff *skb)
1577{
1578 return skb_network_header(skb) - skb->data;
1579}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001580
Changli Gaof9599ce2010-08-04 04:43:44 +00001581static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1582{
1583 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1584}
1585
Linus Torvalds1da177e2005-04-16 15:20:36 -07001586/*
1587 * CPUs often take a performance hit when accessing unaligned memory
1588 * locations. The actual performance hit varies, it can be small if the
1589 * hardware handles it or large if we have to take an exception and fix it
1590 * in software.
1591 *
1592 * Since an ethernet header is 14 bytes network drivers often end up with
1593 * the IP header at an unaligned offset. The IP header can be aligned by
1594 * shifting the start of the packet by 2 bytes. Drivers should do this
1595 * with:
1596 *
Tobias Klauser8660c122009-07-13 22:48:16 +00001597 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001598 *
1599 * The downside to this alignment of the IP header is that the DMA is now
1600 * unaligned. On some architectures the cost of an unaligned DMA is high
1601 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00001602 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001603 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1604 * to be overridden.
1605 */
1606#ifndef NET_IP_ALIGN
1607#define NET_IP_ALIGN 2
1608#endif
1609
Anton Blanchard025be812006-03-31 02:27:06 -08001610/*
1611 * The networking layer reserves some headroom in skb data (via
1612 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1613 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08001614 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08001615 *
1616 * Unfortunately this headroom changes the DMA alignment of the resulting
1617 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1618 * on some architectures. An architecture can override this value,
1619 * perhaps setting it to a cacheline in size (since that will maintain
1620 * cacheline alignment of the DMA). It must be a power of 2.
1621 *
David S. Millerd6301d32009-02-08 19:24:13 -08001622 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08001623 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07001624 *
1625 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1626 * to reduce average number of cache lines per packet.
1627 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07001628 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08001629 */
1630#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07001631#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08001632#endif
1633
Herbert Xu3cc0e872006-06-09 16:13:38 -07001634extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001635
1636static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1637{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00001638 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07001639 WARN_ON(1);
1640 return;
1641 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001642 skb->len = len;
1643 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001644}
1645
Ilpo Järvinen419ae742008-03-27 17:54:01 -07001646extern void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647
1648static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1649{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001650 if (skb->data_len)
1651 return ___pskb_trim(skb, len);
1652 __skb_trim(skb, len);
1653 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001654}
1655
1656static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1657{
1658 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1659}
1660
1661/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07001662 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1663 * @skb: buffer to alter
1664 * @len: new length
1665 *
1666 * This is identical to pskb_trim except that the caller knows that
1667 * the skb is not cloned so we should never get an error due to out-
1668 * of-memory.
1669 */
1670static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1671{
1672 int err = pskb_trim(skb, len);
1673 BUG_ON(err);
1674}
1675
1676/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677 * skb_orphan - orphan a buffer
1678 * @skb: buffer to orphan
1679 *
1680 * If a buffer currently has an owner then we call the owner's
1681 * destructor function and make the @skb unowned. The buffer continues
1682 * to exist but is no longer charged to its former owner.
1683 */
1684static inline void skb_orphan(struct sk_buff *skb)
1685{
1686 if (skb->destructor)
1687 skb->destructor(skb);
1688 skb->destructor = NULL;
1689 skb->sk = NULL;
1690}
1691
1692/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00001693 * skb_orphan_frags - orphan the frags contained in a buffer
1694 * @skb: buffer to orphan frags from
1695 * @gfp_mask: allocation mask for replacement pages
1696 *
1697 * For each frag in the SKB which needs a destructor (i.e. has an
1698 * owner) create a copy of that frag and release the original
1699 * page by calling the destructor.
1700 */
1701static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1702{
1703 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1704 return 0;
1705 return skb_copy_ubufs(skb, gfp_mask);
1706}
1707
1708/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001709 * __skb_queue_purge - empty a list
1710 * @list: list to empty
1711 *
1712 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1713 * the list and one reference dropped. This function does not take the
1714 * list lock and the caller must hold the relevant locks to use it.
1715 */
1716extern void skb_queue_purge(struct sk_buff_head *list);
1717static inline void __skb_queue_purge(struct sk_buff_head *list)
1718{
1719 struct sk_buff *skb;
1720 while ((skb = __skb_dequeue(list)) != NULL)
1721 kfree_skb(skb);
1722}
1723
Eric Dumazet6f532612012-05-18 05:12:12 +00001724extern void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001725
Christoph Hellwig8af27452006-07-31 22:35:23 -07001726extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00001727 unsigned int length,
1728 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07001729
1730/**
1731 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1732 * @dev: network device to receive on
1733 * @length: length to allocate
1734 *
1735 * Allocate a new &sk_buff and assign it a usage count of one. The
1736 * buffer has unspecified headroom built in. Users should allocate
1737 * the headroom they think they need without accounting for the
1738 * built in space. The built in space is used for optimisations.
1739 *
1740 * %NULL is returned if there is no free memory. Although this function
1741 * allocates memory it can be called from an interrupt.
1742 */
1743static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00001744 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07001745{
1746 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1747}
1748
Eric Dumazet6f532612012-05-18 05:12:12 +00001749/* legacy helper around __netdev_alloc_skb() */
1750static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1751 gfp_t gfp_mask)
1752{
1753 return __netdev_alloc_skb(NULL, length, gfp_mask);
1754}
1755
1756/* legacy helper around netdev_alloc_skb() */
1757static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1758{
1759 return netdev_alloc_skb(NULL, length);
1760}
1761
1762
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001763static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
1764 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00001765{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001766 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00001767
1768 if (NET_IP_ALIGN && skb)
1769 skb_reserve(skb, NET_IP_ALIGN);
1770 return skb;
1771}
1772
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001773static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1774 unsigned int length)
1775{
1776 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
1777}
1778
Mel Gorman06140022012-07-31 16:44:24 -07001779/*
1780 * __skb_alloc_page - allocate pages for ps-rx on a skb and preserve pfmemalloc data
1781 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1782 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1783 * @order: size of the allocation
1784 *
1785 * Allocate a new page.
1786 *
1787 * %NULL is returned if there is no free memory.
1788*/
1789static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
1790 struct sk_buff *skb,
1791 unsigned int order)
1792{
1793 struct page *page;
1794
1795 gfp_mask |= __GFP_COLD;
1796
1797 if (!(gfp_mask & __GFP_NOMEMALLOC))
1798 gfp_mask |= __GFP_MEMALLOC;
1799
1800 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
1801 if (skb && page && page->pfmemalloc)
1802 skb->pfmemalloc = true;
1803
1804 return page;
1805}
1806
1807/**
1808 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
1809 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1810 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1811 *
1812 * Allocate a new page.
1813 *
1814 * %NULL is returned if there is no free memory.
1815 */
1816static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
1817 struct sk_buff *skb)
1818{
1819 return __skb_alloc_pages(gfp_mask, skb, 0);
1820}
1821
1822/**
1823 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
1824 * @page: The page that was allocated from skb_alloc_page
1825 * @skb: The skb that may need pfmemalloc set
1826 */
1827static inline void skb_propagate_pfmemalloc(struct page *page,
1828 struct sk_buff *skb)
1829{
1830 if (page && page->pfmemalloc)
1831 skb->pfmemalloc = true;
1832}
1833
Eric Dumazet564824b2010-10-11 19:05:25 +00001834/**
Ian Campbell131ea662011-08-19 06:25:00 +00001835 * skb_frag_page - retrieve the page refered to by a paged fragment
1836 * @frag: the paged fragment
1837 *
1838 * Returns the &struct page associated with @frag.
1839 */
1840static inline struct page *skb_frag_page(const skb_frag_t *frag)
1841{
Ian Campbella8605c62011-10-19 23:01:49 +00001842 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00001843}
1844
1845/**
1846 * __skb_frag_ref - take an addition reference on a paged fragment.
1847 * @frag: the paged fragment
1848 *
1849 * Takes an additional reference on the paged fragment @frag.
1850 */
1851static inline void __skb_frag_ref(skb_frag_t *frag)
1852{
1853 get_page(skb_frag_page(frag));
1854}
1855
1856/**
1857 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
1858 * @skb: the buffer
1859 * @f: the fragment offset.
1860 *
1861 * Takes an additional reference on the @f'th paged fragment of @skb.
1862 */
1863static inline void skb_frag_ref(struct sk_buff *skb, int f)
1864{
1865 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
1866}
1867
1868/**
1869 * __skb_frag_unref - release a reference on a paged fragment.
1870 * @frag: the paged fragment
1871 *
1872 * Releases a reference on the paged fragment @frag.
1873 */
1874static inline void __skb_frag_unref(skb_frag_t *frag)
1875{
1876 put_page(skb_frag_page(frag));
1877}
1878
1879/**
1880 * skb_frag_unref - release a reference on a paged fragment of an skb.
1881 * @skb: the buffer
1882 * @f: the fragment offset
1883 *
1884 * Releases a reference on the @f'th paged fragment of @skb.
1885 */
1886static inline void skb_frag_unref(struct sk_buff *skb, int f)
1887{
1888 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
1889}
1890
1891/**
1892 * skb_frag_address - gets the address of the data contained in a paged fragment
1893 * @frag: the paged fragment buffer
1894 *
1895 * Returns the address of the data within @frag. The page must already
1896 * be mapped.
1897 */
1898static inline void *skb_frag_address(const skb_frag_t *frag)
1899{
1900 return page_address(skb_frag_page(frag)) + frag->page_offset;
1901}
1902
1903/**
1904 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
1905 * @frag: the paged fragment buffer
1906 *
1907 * Returns the address of the data within @frag. Checks that the page
1908 * is mapped and returns %NULL otherwise.
1909 */
1910static inline void *skb_frag_address_safe(const skb_frag_t *frag)
1911{
1912 void *ptr = page_address(skb_frag_page(frag));
1913 if (unlikely(!ptr))
1914 return NULL;
1915
1916 return ptr + frag->page_offset;
1917}
1918
1919/**
1920 * __skb_frag_set_page - sets the page contained in a paged fragment
1921 * @frag: the paged fragment
1922 * @page: the page to set
1923 *
1924 * Sets the fragment @frag to contain @page.
1925 */
1926static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
1927{
Ian Campbella8605c62011-10-19 23:01:49 +00001928 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00001929}
1930
1931/**
1932 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
1933 * @skb: the buffer
1934 * @f: the fragment offset
1935 * @page: the page to set
1936 *
1937 * Sets the @f'th fragment of @skb to contain @page.
1938 */
1939static inline void skb_frag_set_page(struct sk_buff *skb, int f,
1940 struct page *page)
1941{
1942 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
1943}
1944
1945/**
1946 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00001947 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00001948 * @frag: the paged fragment to map
1949 * @offset: the offset within the fragment (starting at the
1950 * fragment's own offset)
1951 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00001952 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00001953 *
1954 * Maps the page associated with @frag to @device.
1955 */
1956static inline dma_addr_t skb_frag_dma_map(struct device *dev,
1957 const skb_frag_t *frag,
1958 size_t offset, size_t size,
1959 enum dma_data_direction dir)
1960{
1961 return dma_map_page(dev, skb_frag_page(frag),
1962 frag->page_offset + offset, size, dir);
1963}
1964
Eric Dumazet117632e2011-12-03 21:39:53 +00001965static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
1966 gfp_t gfp_mask)
1967{
1968 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
1969}
1970
Ian Campbell131ea662011-08-19 06:25:00 +00001971/**
Patrick McHardy334a8132007-06-25 04:35:20 -07001972 * skb_clone_writable - is the header of a clone writable
1973 * @skb: buffer to check
1974 * @len: length up to which to write
1975 *
1976 * Returns true if modifying the header part of the cloned buffer
1977 * does not requires the data to be copied.
1978 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001979static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07001980{
1981 return !skb_header_cloned(skb) &&
1982 skb_headroom(skb) + len <= skb->hdr_len;
1983}
1984
Herbert Xud9cc2042007-09-16 16:21:16 -07001985static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
1986 int cloned)
1987{
1988 int delta = 0;
1989
Herbert Xud9cc2042007-09-16 16:21:16 -07001990 if (headroom > skb_headroom(skb))
1991 delta = headroom - skb_headroom(skb);
1992
1993 if (delta || cloned)
1994 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
1995 GFP_ATOMIC);
1996 return 0;
1997}
1998
Patrick McHardy334a8132007-06-25 04:35:20 -07001999/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002000 * skb_cow - copy header of skb when it is required
2001 * @skb: buffer to cow
2002 * @headroom: needed headroom
2003 *
2004 * If the skb passed lacks sufficient headroom or its data part
2005 * is shared, data is reallocated. If reallocation fails, an error
2006 * is returned and original skb is not changed.
2007 *
2008 * The result is skb with writable area skb->head...skb->tail
2009 * and at least @headroom of space at head.
2010 */
2011static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2012{
Herbert Xud9cc2042007-09-16 16:21:16 -07002013 return __skb_cow(skb, headroom, skb_cloned(skb));
2014}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002015
Herbert Xud9cc2042007-09-16 16:21:16 -07002016/**
2017 * skb_cow_head - skb_cow but only making the head writable
2018 * @skb: buffer to cow
2019 * @headroom: needed headroom
2020 *
2021 * This function is identical to skb_cow except that we replace the
2022 * skb_cloned check by skb_header_cloned. It should be used when
2023 * you only need to push on some header and do not need to modify
2024 * the data.
2025 */
2026static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2027{
2028 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002029}
2030
2031/**
2032 * skb_padto - pad an skbuff up to a minimal size
2033 * @skb: buffer to pad
2034 * @len: minimal length
2035 *
2036 * Pads up a buffer to ensure the trailing bytes exist and are
2037 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002038 * is untouched. Otherwise it is extended. Returns zero on
2039 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002040 */
2041
Herbert Xu5b057c62006-06-23 02:06:41 -07002042static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002043{
2044 unsigned int size = skb->len;
2045 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002046 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002047 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002048}
2049
2050static inline int skb_add_data(struct sk_buff *skb,
2051 char __user *from, int copy)
2052{
2053 const int off = skb->len;
2054
2055 if (skb->ip_summed == CHECKSUM_NONE) {
2056 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08002057 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
Linus Torvalds1da177e2005-04-16 15:20:36 -07002058 copy, 0, &err);
2059 if (!err) {
2060 skb->csum = csum_block_add(skb->csum, csum, off);
2061 return 0;
2062 }
2063 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2064 return 0;
2065
2066 __skb_trim(skb, off);
2067 return -EFAULT;
2068}
2069
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002070static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2071 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002072{
2073 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002074 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002075
Ian Campbellea2ab692011-08-22 23:44:58 +00002076 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002077 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002078 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002079 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002080}
2081
Herbert Xu364c6ba2006-06-09 16:10:40 -07002082static inline int __skb_linearize(struct sk_buff *skb)
2083{
2084 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2085}
2086
Linus Torvalds1da177e2005-04-16 15:20:36 -07002087/**
2088 * skb_linearize - convert paged skb to linear one
2089 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002090 *
2091 * If there is no free memory -ENOMEM is returned, otherwise zero
2092 * is returned and the old skb data released.
2093 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002094static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002095{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002096 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2097}
2098
2099/**
2100 * skb_linearize_cow - make sure skb is linear and writable
2101 * @skb: buffer to process
2102 *
2103 * If there is no free memory -ENOMEM is returned, otherwise zero
2104 * is returned and the old skb data released.
2105 */
2106static inline int skb_linearize_cow(struct sk_buff *skb)
2107{
2108 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2109 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002110}
2111
2112/**
2113 * skb_postpull_rcsum - update checksum for received skb after pull
2114 * @skb: buffer to update
2115 * @start: start of data before pull
2116 * @len: length of data pulled
2117 *
2118 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002119 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2120 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002121 */
2122
2123static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002124 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002125{
Patrick McHardy84fa7932006-08-29 16:44:56 -07002126 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002127 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2128}
2129
Herbert Xucbb042f2006-03-20 22:43:56 -08002130unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2131
Linus Torvalds1da177e2005-04-16 15:20:36 -07002132/**
2133 * pskb_trim_rcsum - trim received skb and update checksum
2134 * @skb: buffer to trim
2135 * @len: new length
2136 *
2137 * This is exactly the same as pskb_trim except that it ensures the
2138 * checksum of received packets are still valid after the operation.
2139 */
2140
2141static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2142{
Stephen Hemminger0e4e4222005-09-08 12:32:03 -07002143 if (likely(len >= skb->len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002144 return 0;
Patrick McHardy84fa7932006-08-29 16:44:56 -07002145 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002146 skb->ip_summed = CHECKSUM_NONE;
2147 return __pskb_trim(skb, len);
2148}
2149
Linus Torvalds1da177e2005-04-16 15:20:36 -07002150#define skb_queue_walk(queue, skb) \
2151 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002152 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002153 skb = skb->next)
2154
James Chapman46f89142007-04-30 00:07:31 -07002155#define skb_queue_walk_safe(queue, skb, tmp) \
2156 for (skb = (queue)->next, tmp = skb->next; \
2157 skb != (struct sk_buff *)(queue); \
2158 skb = tmp, tmp = skb->next)
2159
David S. Miller1164f522008-09-23 00:49:44 -07002160#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002161 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002162 skb = skb->next)
2163
2164#define skb_queue_walk_from_safe(queue, skb, tmp) \
2165 for (tmp = skb->next; \
2166 skb != (struct sk_buff *)(queue); \
2167 skb = tmp, tmp = skb->next)
2168
Stephen Hemminger300ce172005-10-30 13:47:34 -08002169#define skb_queue_reverse_walk(queue, skb) \
2170 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002171 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08002172 skb = skb->prev)
2173
David S. Miller686a2952011-01-20 22:47:32 -08002174#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2175 for (skb = (queue)->prev, tmp = skb->prev; \
2176 skb != (struct sk_buff *)(queue); \
2177 skb = tmp, tmp = skb->prev)
2178
2179#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2180 for (tmp = skb->prev; \
2181 skb != (struct sk_buff *)(queue); \
2182 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002183
David S. Miller21dc3302010-08-23 00:13:46 -07002184static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07002185{
2186 return skb_shinfo(skb)->frag_list != NULL;
2187}
2188
2189static inline void skb_frag_list_init(struct sk_buff *skb)
2190{
2191 skb_shinfo(skb)->frag_list = NULL;
2192}
2193
2194static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2195{
2196 frag->next = skb_shinfo(skb)->frag_list;
2197 skb_shinfo(skb)->frag_list = frag;
2198}
2199
2200#define skb_walk_frags(skb, iter) \
2201 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2202
Herbert Xua59322b2007-12-05 01:53:40 -08002203extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
Pavel Emelyanov3f518bf2012-02-21 07:30:58 +00002204 int *peeked, int *off, int *err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002205extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
2206 int noblock, int *err);
2207extern unsigned int datagram_poll(struct file *file, struct socket *sock,
2208 struct poll_table_struct *wait);
2209extern int skb_copy_datagram_iovec(const struct sk_buff *from,
2210 int offset, struct iovec *to,
2211 int size);
Herbert Xufb286bb2005-11-10 13:01:24 -08002212extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002213 int hlen,
2214 struct iovec *iov);
Rusty Russelldb543c12008-08-15 15:13:53 -07002215extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
2216 int offset,
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +00002217 const struct iovec *from,
2218 int from_offset,
Rusty Russelldb543c12008-08-15 15:13:53 -07002219 int len);
Michael S. Tsirkin0a1ec072009-04-20 01:25:46 +00002220extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
2221 int offset,
2222 const struct iovec *to,
2223 int to_offset,
2224 int size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002225extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
Eric Dumazet9d410c72009-10-30 05:03:53 +00002226extern void skb_free_datagram_locked(struct sock *sk,
2227 struct sk_buff *skb);
Herbert Xu27ab2562007-12-05 01:51:58 -08002228extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
Herbert Xu3305b802005-12-13 23:16:37 -08002229 unsigned int flags);
Al Viro2bbbc862006-11-14 21:37:14 -08002230extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
2231 int len, __wsum csum);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002232extern int skb_copy_bits(const struct sk_buff *skb, int offset,
2233 void *to, int len);
Stephen Hemminger0c6fcc82007-04-20 16:40:01 -07002234extern int skb_store_bits(struct sk_buff *skb, int offset,
2235 const void *from, int len);
Al Viro81d77662006-11-14 21:37:33 -08002236extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002237 int offset, u8 *to, int len,
Al Viro81d77662006-11-14 21:37:33 -08002238 __wsum csum);
Jens Axboe9c55e012007-11-06 23:30:13 -08002239extern int skb_splice_bits(struct sk_buff *skb,
2240 unsigned int offset,
2241 struct pipe_inode_info *pipe,
2242 unsigned int len,
2243 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002244extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
2245extern void skb_split(struct sk_buff *skb,
2246 struct sk_buff *skb1, const u32 len);
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08002247extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
2248 int shiftlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002249
Michał Mirosławc8f44af2011-11-15 15:29:55 +00002250extern struct sk_buff *skb_segment(struct sk_buff *skb,
2251 netdev_features_t features);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002252
Linus Torvalds1da177e2005-04-16 15:20:36 -07002253static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2254 int len, void *buffer)
2255{
2256 int hlen = skb_headlen(skb);
2257
Patrick McHardy55820ee2005-07-05 14:08:10 -07002258 if (hlen - offset >= len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002259 return skb->data + offset;
2260
2261 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2262 return NULL;
2263
2264 return buffer;
2265}
2266
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03002267static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2268 void *to,
2269 const unsigned int len)
2270{
2271 memcpy(to, skb->data, len);
2272}
2273
2274static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2275 const int offset, void *to,
2276 const unsigned int len)
2277{
2278 memcpy(to, skb->data + offset, len);
2279}
2280
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03002281static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2282 const void *from,
2283 const unsigned int len)
2284{
2285 memcpy(skb->data, from, len);
2286}
2287
2288static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2289 const int offset,
2290 const void *from,
2291 const unsigned int len)
2292{
2293 memcpy(skb->data + offset, from, len);
2294}
2295
Linus Torvalds1da177e2005-04-16 15:20:36 -07002296extern void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002297
Patrick Ohlyac45f602009-02-12 05:03:37 +00002298static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2299{
2300 return skb->tstamp;
2301}
2302
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002303/**
2304 * skb_get_timestamp - get timestamp from a skb
2305 * @skb: skb to get stamp from
2306 * @stamp: pointer to struct timeval to store stamp in
2307 *
2308 * Timestamps are stored in the skb as offsets to a base timestamp.
2309 * This function converts the offset back to a struct timeval and stores
2310 * it in stamp.
2311 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00002312static inline void skb_get_timestamp(const struct sk_buff *skb,
2313 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002314{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002315 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002316}
2317
Patrick Ohlyac45f602009-02-12 05:03:37 +00002318static inline void skb_get_timestampns(const struct sk_buff *skb,
2319 struct timespec *stamp)
2320{
2321 *stamp = ktime_to_timespec(skb->tstamp);
2322}
2323
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002324static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002325{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002326 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002327}
2328
Stephen Hemminger164891a2007-04-23 22:26:16 -07002329static inline ktime_t net_timedelta(ktime_t t)
2330{
2331 return ktime_sub(ktime_get_real(), t);
2332}
2333
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07002334static inline ktime_t net_invalid_timestamp(void)
2335{
2336 return ktime_set(0, 0);
2337}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002338
Richard Cochranc1f19b52010-07-17 08:49:36 +00002339extern void skb_timestamping_init(void);
2340
2341#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2342
2343extern void skb_clone_tx_timestamp(struct sk_buff *skb);
2344extern bool skb_defer_rx_timestamp(struct sk_buff *skb);
2345
2346#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2347
2348static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2349{
2350}
2351
2352static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2353{
2354 return false;
2355}
2356
2357#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2358
2359/**
2360 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2361 *
Richard Cochranda92b192011-10-21 00:49:15 +00002362 * PHY drivers may accept clones of transmitted packets for
2363 * timestamping via their phy_driver.txtstamp method. These drivers
2364 * must call this function to return the skb back to the stack, with
2365 * or without a timestamp.
2366 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00002367 * @skb: clone of the the original outgoing packet
Richard Cochranda92b192011-10-21 00:49:15 +00002368 * @hwtstamps: hardware time stamps, may be NULL if not available
Richard Cochranc1f19b52010-07-17 08:49:36 +00002369 *
2370 */
2371void skb_complete_tx_timestamp(struct sk_buff *skb,
2372 struct skb_shared_hwtstamps *hwtstamps);
2373
Patrick Ohlyac45f602009-02-12 05:03:37 +00002374/**
2375 * skb_tstamp_tx - queue clone of skb with send time stamps
2376 * @orig_skb: the original outgoing packet
2377 * @hwtstamps: hardware time stamps, may be NULL if not available
2378 *
2379 * If the skb has a socket associated, then this function clones the
2380 * skb (thus sharing the actual data and optional structures), stores
2381 * the optional hardware time stamping information (if non NULL) or
2382 * generates a software time stamp (otherwise), then queues the clone
2383 * to the error queue of the socket. Errors are silently ignored.
2384 */
2385extern void skb_tstamp_tx(struct sk_buff *orig_skb,
2386 struct skb_shared_hwtstamps *hwtstamps);
2387
Richard Cochran4507a712010-07-17 08:48:28 +00002388static inline void sw_tx_timestamp(struct sk_buff *skb)
2389{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002390 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2391 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00002392 skb_tstamp_tx(skb, NULL);
2393}
2394
2395/**
2396 * skb_tx_timestamp() - Driver hook for transmit timestamping
2397 *
2398 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00002399 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00002400 *
2401 * @skb: A socket buffer.
2402 */
2403static inline void skb_tx_timestamp(struct sk_buff *skb)
2404{
Richard Cochranc1f19b52010-07-17 08:49:36 +00002405 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002406 sw_tx_timestamp(skb);
2407}
2408
Johannes Berg6e3e9392011-11-09 10:15:42 +01002409/**
2410 * skb_complete_wifi_ack - deliver skb with wifi status
2411 *
2412 * @skb: the original outgoing packet
2413 * @acked: ack status
2414 *
2415 */
2416void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2417
Herbert Xu759e5d02007-03-25 20:10:56 -07002418extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
Al Virob51655b2006-11-14 21:40:42 -08002419extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002420
Herbert Xu60476372007-04-09 11:59:39 -07002421static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2422{
2423 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2424}
2425
Herbert Xufb286bb2005-11-10 13:01:24 -08002426/**
2427 * skb_checksum_complete - Calculate checksum of an entire packet
2428 * @skb: packet to process
2429 *
2430 * This function calculates the checksum over the entire packet plus
2431 * the value of skb->csum. The latter can be used to supply the
2432 * checksum of a pseudo header as used by TCP/UDP. It returns the
2433 * checksum.
2434 *
2435 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2436 * this function can be used to verify that checksum on received
2437 * packets. In that case the function should return zero if the
2438 * checksum is correct. In particular, this function will return zero
2439 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2440 * hardware has already verified the correctness of the checksum.
2441 */
Al Viro4381ca32007-07-15 21:00:11 +01002442static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002443{
Herbert Xu60476372007-04-09 11:59:39 -07002444 return skb_csum_unnecessary(skb) ?
2445 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002446}
2447
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002448#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002449extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002450static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2451{
2452 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002453 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002454}
2455static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2456{
2457 if (nfct)
2458 atomic_inc(&nfct->use);
2459}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002460#endif
2461#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -08002462static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
2463{
2464 if (skb)
2465 atomic_inc(&skb->users);
2466}
2467static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
2468{
2469 if (skb)
2470 kfree_skb(skb);
2471}
2472#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002473#ifdef CONFIG_BRIDGE_NETFILTER
2474static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2475{
2476 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2477 kfree(nf_bridge);
2478}
2479static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2480{
2481 if (nf_bridge)
2482 atomic_inc(&nf_bridge->use);
2483}
2484#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08002485static inline void nf_reset(struct sk_buff *skb)
2486{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002487#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08002488 nf_conntrack_put(skb->nfct);
2489 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002490#endif
2491#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Patrick McHardya193a4a2006-03-20 19:23:05 -08002492 nf_conntrack_put_reasm(skb->nfct_reasm);
2493 skb->nfct_reasm = NULL;
2494#endif
2495#ifdef CONFIG_BRIDGE_NETFILTER
2496 nf_bridge_put(skb->nf_bridge);
2497 skb->nf_bridge = NULL;
2498#endif
2499}
2500
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002501/* Note: This doesn't put any conntrack and bridge info in dst. */
2502static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2503{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002504#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002505 dst->nfct = src->nfct;
2506 nf_conntrack_get(src->nfct);
2507 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002508#endif
2509#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002510 dst->nfct_reasm = src->nfct_reasm;
2511 nf_conntrack_get_reasm(src->nfct_reasm);
2512#endif
2513#ifdef CONFIG_BRIDGE_NETFILTER
2514 dst->nf_bridge = src->nf_bridge;
2515 nf_bridge_get(src->nf_bridge);
2516#endif
2517}
2518
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002519static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2520{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002521#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002522 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002523#endif
2524#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002525 nf_conntrack_put_reasm(dst->nfct_reasm);
2526#endif
2527#ifdef CONFIG_BRIDGE_NETFILTER
2528 nf_bridge_put(dst->nf_bridge);
2529#endif
2530 __nf_copy(dst, src);
2531}
2532
James Morris984bc162006-06-09 00:29:17 -07002533#ifdef CONFIG_NETWORK_SECMARK
2534static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2535{
2536 to->secmark = from->secmark;
2537}
2538
2539static inline void skb_init_secmark(struct sk_buff *skb)
2540{
2541 skb->secmark = 0;
2542}
2543#else
2544static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2545{ }
2546
2547static inline void skb_init_secmark(struct sk_buff *skb)
2548{ }
2549#endif
2550
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002551static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2552{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002553 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002554}
2555
Stephen Hemminger92477442009-03-21 13:39:26 -07002556static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002557{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002558 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002559}
2560
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002561static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2562{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002563 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002564}
2565
David S. Millerd5a9e242009-01-27 16:22:11 -08002566static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2567{
2568 skb->queue_mapping = rx_queue + 1;
2569}
2570
Stephen Hemminger92477442009-03-21 13:39:26 -07002571static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002572{
2573 return skb->queue_mapping - 1;
2574}
2575
Stephen Hemminger92477442009-03-21 13:39:26 -07002576static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002577{
Eric Dumazeta02cec22010-09-22 20:43:57 +00002578 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08002579}
2580
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00002581extern u16 __skb_tx_hash(const struct net_device *dev,
2582 const struct sk_buff *skb,
2583 unsigned int num_tx_queues);
Stephen Hemminger92477442009-03-21 13:39:26 -07002584
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002585#ifdef CONFIG_XFRM
2586static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2587{
2588 return skb->sp;
2589}
2590#else
2591static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2592{
2593 return NULL;
2594}
2595#endif
2596
David S. Millerbdcc0922012-03-07 20:53:36 -05002597static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07002598{
2599 return skb_shinfo(skb)->gso_size;
2600}
2601
David S. Millerbdcc0922012-03-07 20:53:36 -05002602static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02002603{
2604 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2605}
2606
Ben Hutchings4497b072008-06-19 16:22:28 -07002607extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2608
2609static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2610{
2611 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2612 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002613 const struct skb_shared_info *shinfo = skb_shinfo(skb);
2614
Alexander Duyckb78462e2010-06-02 12:24:37 +00002615 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2616 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07002617 __skb_warn_lro_forwarding(skb);
2618 return true;
2619 }
2620 return false;
2621}
2622
Herbert Xu35fc92a2007-03-26 23:22:20 -07002623static inline void skb_forward_csum(struct sk_buff *skb)
2624{
2625 /* Unfortunately we don't support this one. Any brave souls? */
2626 if (skb->ip_summed == CHECKSUM_COMPLETE)
2627 skb->ip_summed = CHECKSUM_NONE;
2628}
2629
Eric Dumazetbc8acf22010-09-02 13:07:41 -07002630/**
2631 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2632 * @skb: skb to check
2633 *
2634 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2635 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2636 * use this helper, to document places where we make this assertion.
2637 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002638static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07002639{
2640#ifdef DEBUG
2641 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2642#endif
2643}
2644
Rusty Russellf35d9d82008-02-04 23:49:54 -05002645bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00002646
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00002647/**
2648 * skb_head_is_locked - Determine if the skb->head is locked down
2649 * @skb: skb to check
2650 *
2651 * The head on skbs build around a head frag can be removed if they are
2652 * not cloned. This function returns true if the skb head is locked down
2653 * due to either being allocated via kmalloc, or by being a clone with
2654 * multiple references to the head.
2655 */
2656static inline bool skb_head_is_locked(const struct sk_buff *skb)
2657{
2658 return !skb->head_frag || skb_cloned(skb);
2659}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002660#endif /* __KERNEL__ */
2661#endif /* _LINUX_SKBUFF_H */