blob: 775292a66fa4d84615e2b3d3d15c84913cc7813b [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * Definitions for the 'struct sk_buff' memory handlers.
3 *
4 * Authors:
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14#ifndef _LINUX_SKBUFF_H
15#define _LINUX_SKBUFF_H
16
Linus Torvalds1da177e2005-04-16 15:20:36 -070017#include <linux/kernel.h>
Vegard Nossumfe55f6d2008-08-30 12:16:35 +020018#include <linux/kmemcheck.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070019#include <linux/compiler.h>
20#include <linux/time.h>
Paul Gortmaker187f1882011-11-23 20:12:59 -050021#include <linux/bug.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070022#include <linux/cache.h>
23
Arun Sharma600634972011-07-26 16:09:06 -070024#include <linux/atomic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <asm/types.h>
26#include <linux/spinlock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <linux/net.h>
Thomas Graf3fc7e8a2005-06-23 21:00:17 -070028#include <linux/textsearch.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070029#include <net/checksum.h>
Al Viroa80958f2006-12-04 20:41:19 +000030#include <linux/rcupdate.h>
Chris Leech97fc2f02006-05-23 17:55:33 -070031#include <linux/dmaengine.h>
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -070032#include <linux/hrtimer.h>
Ian Campbell131ea662011-08-19 06:25:00 +000033#include <linux/dma-mapping.h>
Michał Mirosławc8f44af2011-11-15 15:29:55 +000034#include <linux/netdev_features.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070035
Herbert Xu60476372007-04-09 11:59:39 -070036/* Don't change this without changing skb_csum_unnecessary! */
Linus Torvalds1da177e2005-04-16 15:20:36 -070037#define CHECKSUM_NONE 0
Herbert Xu60476372007-04-09 11:59:39 -070038#define CHECKSUM_UNNECESSARY 1
39#define CHECKSUM_COMPLETE 2
40#define CHECKSUM_PARTIAL 3
Linus Torvalds1da177e2005-04-16 15:20:36 -070041
42#define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
43 ~(SMP_CACHE_BYTES - 1))
David S. Millerfc910a22007-03-25 20:27:59 -070044#define SKB_WITH_OVERHEAD(X) \
Herbert Xudeea84b2007-10-21 16:27:46 -070045 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
David S. Millerfc910a22007-03-25 20:27:59 -070046#define SKB_MAX_ORDER(X, ORDER) \
47 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
Linus Torvalds1da177e2005-04-16 15:20:36 -070048#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
49#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
50
Eric Dumazet87fb4b72011-10-13 07:28:54 +000051/* return minimum truesize of one skb containing X bytes of data */
52#define SKB_TRUESIZE(X) ((X) + \
53 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
54 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
55
Linus Torvalds1da177e2005-04-16 15:20:36 -070056/* A. Checksumming of received packets by device.
57 *
58 * NONE: device failed to checksum this packet.
59 * skb->csum is undefined.
60 *
61 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
62 * skb->csum is undefined.
63 * It is bad option, but, unfortunately, many of vendors do this.
64 * Apparently with secret goal to sell you new device, when you
65 * will add new protocol to your host. F.e. IPv6. 8)
66 *
Patrick McHardy84fa7932006-08-29 16:44:56 -070067 * COMPLETE: the most generic way. Device supplied checksum of _all_
Linus Torvalds1da177e2005-04-16 15:20:36 -070068 * the packet as seen by netif_rx in skb->csum.
69 * NOTE: Even if device supports only some protocols, but
Patrick McHardy84fa7932006-08-29 16:44:56 -070070 * is able to produce some skb->csum, it MUST use COMPLETE,
Linus Torvalds1da177e2005-04-16 15:20:36 -070071 * not UNNECESSARY.
72 *
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070073 * PARTIAL: identical to the case for output below. This may occur
74 * on a packet received directly from another Linux OS, e.g.,
75 * a virtualised Linux kernel on the same host. The packet can
76 * be treated in the same way as UNNECESSARY except that on
77 * output (i.e., forwarding) the checksum must be filled in
78 * by the OS or the hardware.
79 *
Linus Torvalds1da177e2005-04-16 15:20:36 -070080 * B. Checksumming on output.
81 *
82 * NONE: skb is checksummed by protocol or csum is not required.
83 *
Patrick McHardy84fa7932006-08-29 16:44:56 -070084 * PARTIAL: device is required to csum packet as seen by hard_start_xmit
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070085 * from skb->csum_start to the end and to record the checksum
86 * at skb->csum_start + skb->csum_offset.
Linus Torvalds1da177e2005-04-16 15:20:36 -070087 *
88 * Device must show its capabilities in dev->features, set
89 * at device setup time.
90 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
91 * everything.
Linus Torvalds1da177e2005-04-16 15:20:36 -070092 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
93 * TCP/UDP over IPv4. Sigh. Vendors like this
94 * way by an unknown reason. Though, see comment above
95 * about CHECKSUM_UNNECESSARY. 8)
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070096 * NETIF_F_IPV6_CSUM about as dumb as the last one but does IPv6 instead.
Linus Torvalds1da177e2005-04-16 15:20:36 -070097 *
Yi Zou3af79302012-03-19 11:12:41 +000098 * UNNECESSARY: device will do per protocol specific csum. Protocol drivers
99 * that do not want net to perform the checksum calculation should use
100 * this flag in their outgoing skbs.
101 * NETIF_F_FCOE_CRC this indicates the device can do FCoE FC CRC
102 * offload. Correspondingly, the FCoE protocol driver
103 * stack should use CHECKSUM_UNNECESSARY.
104 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105 * Any questions? No questions, good. --ANK
106 */
107
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108struct net_device;
David Howells716ea3a2007-04-02 20:19:53 -0700109struct scatterlist;
Jens Axboe9c55e012007-11-06 23:30:13 -0800110struct pipe_inode_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700112#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113struct nf_conntrack {
114 atomic_t use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115};
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700116#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117
118#ifdef CONFIG_BRIDGE_NETFILTER
119struct nf_bridge_info {
120 atomic_t use;
121 struct net_device *physindev;
122 struct net_device *physoutdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123 unsigned int mask;
124 unsigned long data[32 / sizeof(unsigned long)];
125};
126#endif
127
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128struct sk_buff_head {
129 /* These two members must be first. */
130 struct sk_buff *next;
131 struct sk_buff *prev;
132
133 __u32 qlen;
134 spinlock_t lock;
135};
136
137struct sk_buff;
138
Ian Campbell9d4dde52011-12-22 23:39:14 +0000139/* To allow 64K frame to be packed as single skb without frag_list we
140 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
141 * buffers which do not start on a page boundary.
142 *
143 * Since GRO uses frags we allocate at least 16 regardless of page
144 * size.
Anton Blancharda715dea2011-03-27 14:57:26 +0000145 */
Ian Campbell9d4dde52011-12-22 23:39:14 +0000146#if (65536/PAGE_SIZE + 1) < 16
David S. Millereec00952011-03-29 23:34:08 -0700147#define MAX_SKB_FRAGS 16UL
Anton Blancharda715dea2011-03-27 14:57:26 +0000148#else
Ian Campbell9d4dde52011-12-22 23:39:14 +0000149#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
Anton Blancharda715dea2011-03-27 14:57:26 +0000150#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151
152typedef struct skb_frag_struct skb_frag_t;
153
154struct skb_frag_struct {
Ian Campbella8605c62011-10-19 23:01:49 +0000155 struct {
156 struct page *p;
157 } page;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000158#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
David S. Millera309bb02007-07-30 18:47:03 -0700159 __u32 page_offset;
160 __u32 size;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000161#else
162 __u16 page_offset;
163 __u16 size;
164#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165};
166
Eric Dumazet9e903e02011-10-18 21:00:24 +0000167static inline unsigned int skb_frag_size(const skb_frag_t *frag)
168{
169 return frag->size;
170}
171
172static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
173{
174 frag->size = size;
175}
176
177static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
178{
179 frag->size += delta;
180}
181
182static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
183{
184 frag->size -= delta;
185}
186
Patrick Ohlyac45f602009-02-12 05:03:37 +0000187#define HAVE_HW_TIME_STAMP
188
189/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800190 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000191 * @hwtstamp: hardware time stamp transformed into duration
192 * since arbitrary point in time
193 * @syststamp: hwtstamp transformed to system time base
194 *
195 * Software time stamps generated by ktime_get_real() are stored in
196 * skb->tstamp. The relation between the different kinds of time
197 * stamps is as follows:
198 *
199 * syststamp and tstamp can be compared against each other in
200 * arbitrary combinations. The accuracy of a
201 * syststamp/tstamp/"syststamp from other device" comparison is
202 * limited by the accuracy of the transformation into system time
203 * base. This depends on the device driver and its underlying
204 * hardware.
205 *
206 * hwtstamps can only be compared against other hwtstamps from
207 * the same device.
208 *
209 * This structure is attached to packets as part of the
210 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
211 */
212struct skb_shared_hwtstamps {
213 ktime_t hwtstamp;
214 ktime_t syststamp;
215};
216
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000217/* Definitions for tx_flags in struct skb_shared_info */
218enum {
219 /* generate hardware time stamp */
220 SKBTX_HW_TSTAMP = 1 << 0,
221
222 /* generate software time stamp */
223 SKBTX_SW_TSTAMP = 1 << 1,
224
225 /* device driver is going to provide hardware time stamp */
226 SKBTX_IN_PROGRESS = 1 << 2,
227
228 /* ensure the originating sk reference is available on driver level */
229 SKBTX_DRV_NEEDS_SK_REF = 1 << 3,
Shirley Maa6686f22011-07-06 12:22:12 +0000230
231 /* device driver supports TX zero-copy buffers */
232 SKBTX_DEV_ZEROCOPY = 1 << 4,
Johannes Berg6e3e9392011-11-09 10:15:42 +0100233
234 /* generate wifi status information (where possible) */
235 SKBTX_WIFI_STATUS = 1 << 5,
Shirley Maa6686f22011-07-06 12:22:12 +0000236};
237
238/*
239 * The callback notifies userspace to release buffers when skb DMA is done in
240 * lower device, the skb last reference should be 0 when calling this.
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000241 * The ctx field is used to track device context.
242 * The desc field is used to track userspace buffer index.
Shirley Maa6686f22011-07-06 12:22:12 +0000243 */
244struct ubuf_info {
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000245 void (*callback)(struct ubuf_info *);
246 void *ctx;
Shirley Maa6686f22011-07-06 12:22:12 +0000247 unsigned long desc;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000248};
249
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250/* This data is invariant across clones and lives at
251 * the end of the header data, ie. at skb->end.
252 */
253struct skb_shared_info {
Ian Campbell9f42f122012-01-05 07:13:39 +0000254 unsigned char nr_frags;
255 __u8 tx_flags;
Herbert Xu79671682006-06-22 02:40:14 -0700256 unsigned short gso_size;
257 /* Warning: this field is not always filled in (UFO)! */
258 unsigned short gso_segs;
259 unsigned short gso_type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700260 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000261 struct skb_shared_hwtstamps hwtstamps;
Ian Campbell9f42f122012-01-05 07:13:39 +0000262 __be32 ip6_frag_id;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700263
264 /*
265 * Warning : all fields before dataref are cleared in __alloc_skb()
266 */
267 atomic_t dataref;
268
Johann Baudy69e3c752009-05-18 22:11:22 -0700269 /* Intermediate layers must ensure that destructor_arg
270 * remains valid until skb destructor */
271 void * destructor_arg;
Shirley Maa6686f22011-07-06 12:22:12 +0000272
Eric Dumazetfed66382010-07-22 19:09:08 +0000273 /* must be last field, see pskb_expand_head() */
274 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700275};
276
277/* We divide dataref into two halves. The higher 16 bits hold references
278 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700279 * the entire skb->data. A clone of a headerless skb holds the length of
280 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281 *
282 * All users must obey the rule that the skb->data reference count must be
283 * greater than or equal to the payload reference count.
284 *
285 * Holding a reference to the payload part means that the user does not
286 * care about modifications to the header part of skb->data.
287 */
288#define SKB_DATAREF_SHIFT 16
289#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
290
David S. Millerd179cd12005-08-17 14:57:30 -0700291
292enum {
293 SKB_FCLONE_UNAVAILABLE,
294 SKB_FCLONE_ORIG,
295 SKB_FCLONE_CLONE,
296};
297
Herbert Xu79671682006-06-22 02:40:14 -0700298enum {
299 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700300 SKB_GSO_UDP = 1 << 1,
Herbert Xu576a30e2006-06-27 13:22:38 -0700301
302 /* This indicates the skb is from an untrusted source. */
303 SKB_GSO_DODGY = 1 << 2,
Michael Chanb0da85372006-06-29 12:30:00 -0700304
305 /* This indicates the tcp segment has CWR set. */
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700306 SKB_GSO_TCP_ECN = 1 << 3,
307
308 SKB_GSO_TCPV6 = 1 << 4,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800309
310 SKB_GSO_FCOE = 1 << 5,
Herbert Xu79671682006-06-22 02:40:14 -0700311};
312
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700313#if BITS_PER_LONG > 32
314#define NET_SKBUFF_DATA_USES_OFFSET 1
315#endif
316
317#ifdef NET_SKBUFF_DATA_USES_OFFSET
318typedef unsigned int sk_buff_data_t;
319#else
320typedef unsigned char *sk_buff_data_t;
321#endif
322
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100323#if defined(CONFIG_NF_DEFRAG_IPV4) || defined(CONFIG_NF_DEFRAG_IPV4_MODULE) || \
324 defined(CONFIG_NF_DEFRAG_IPV6) || defined(CONFIG_NF_DEFRAG_IPV6_MODULE)
325#define NET_SKBUFF_NF_DEFRAG_NEEDED 1
326#endif
327
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328/**
329 * struct sk_buff - socket buffer
330 * @next: Next buffer in list
331 * @prev: Previous buffer in list
Herbert Xu325ed822005-10-03 13:57:23 -0700332 * @tstamp: Time we arrived
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700333 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700335 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000336 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700337 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700338 * @len: Length of actual data
339 * @data_len: Data length
340 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700341 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700342 * @csum: Checksum (must include start/offset pair)
343 * @csum_start: Offset from skb->head where checksumming should start
344 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700345 * @priority: Packet queueing priority
Martin Waitz67be2dd2005-05-01 08:59:26 -0700346 * @local_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700348 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700349 * @nohdr: Payload reference only, must not modify header
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700350 * @nfctinfo: Relationship of this skb to the connection
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700352 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700353 * @ipvs_property: skbuff is owned by ipvs
Randy Dunlap31729362008-02-18 20:52:13 -0800354 * @peeked: this packet has been seen already, so stats have been
355 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700356 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700357 * @protocol: Packet protocol from driver
358 * @destructor: Destruct function
359 * @nfct: Associated connection, if any
Randy Dunlap461ddf32005-11-20 21:25:15 -0800360 * @nfct_reasm: netfilter conntrack re-assembly pointer
Linus Torvalds1da177e2005-04-16 15:20:36 -0700361 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800362 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700363 * @tc_index: Traffic control index
364 * @tc_verd: traffic control verdict
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700365 * @rxhash: the packet hash computed on receive
366 * @queue_mapping: Queue mapping for multiqueue devices
Randy Dunlap553a5672008-04-20 10:51:01 -0700367 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700368 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Changli Gao4ca24622011-08-19 07:26:44 -0700369 * @l4_rxhash: indicate rxhash is a canonical 4-tuple hash over transport
370 * ports.
Johannes Berg6e3e9392011-11-09 10:15:42 +0100371 * @wifi_acked_valid: wifi_acked was set
372 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000373 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Randy Dunlapf4b8ea72006-06-22 16:00:11 -0700374 * @dma_cookie: a cookie to one of several possible DMA operations
375 * done by skb DMA functions
James Morris984bc162006-06-09 00:29:17 -0700376 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700377 * @mark: Generic packet mark
378 * @dropcount: total number of sk_receive_queue overflows
Patrick McHardy6aa895b02008-07-14 22:49:06 -0700379 * @vlan_tci: vlan tag control information
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700380 * @transport_header: Transport layer header
381 * @network_header: Network layer header
382 * @mac_header: Link layer header
383 * @tail: Tail pointer
384 * @end: End pointer
385 * @head: Head of buffer
386 * @data: Data head pointer
387 * @truesize: Buffer size
388 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389 */
390
391struct sk_buff {
392 /* These two members must be first. */
393 struct sk_buff *next;
394 struct sk_buff *prev;
395
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700396 ktime_t tstamp;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000397
398 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700399 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700400
Linus Torvalds1da177e2005-04-16 15:20:36 -0700401 /*
402 * This is the control buffer. It is free to use for every
403 * layer. Please put your private variables there. If you
404 * want to keep them across layers you have to do a skb_clone()
405 * first. This is owned by whoever has the skb queued ATM.
406 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000407 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408
Eric Dumazet7fee2262010-05-11 23:19:48 +0000409 unsigned long _skb_refdst;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000410#ifdef CONFIG_XFRM
411 struct sec_path *sp;
412#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700413 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700414 data_len;
415 __u16 mac_len,
416 hdr_len;
Al Viroff1dcad2006-11-20 18:07:29 -0800417 union {
418 __wsum csum;
Herbert Xu663ead32007-04-09 11:59:07 -0700419 struct {
420 __u16 csum_start;
421 __u16 csum_offset;
422 };
Al Viroff1dcad2006-11-20 18:07:29 -0800423 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424 __u32 priority;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200425 kmemcheck_bitfield_begin(flags1);
Thomas Graf1cbb3382005-07-05 14:13:41 -0700426 __u8 local_df:1,
427 cloned:1,
428 ip_summed:2,
Harald Welte6869c4d2005-08-09 19:24:19 -0700429 nohdr:1,
430 nfctinfo:3;
David S. Millerd179cd12005-08-17 14:57:30 -0700431 __u8 pkt_type:3,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800432 fclone:2,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700433 ipvs_property:1,
Herbert Xua59322b2007-12-05 01:53:40 -0800434 peeked:1,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700435 nf_trace:1;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200436 kmemcheck_bitfield_end(flags1);
Eric Dumazet4ab408d2010-03-01 03:09:26 +0000437 __be16 protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438
439 void (*destructor)(struct sk_buff *skb);
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800440#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700441 struct nf_conntrack *nfct;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100442#endif
443#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800444 struct sk_buff *nfct_reasm;
445#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446#ifdef CONFIG_BRIDGE_NETFILTER
447 struct nf_bridge_info *nf_bridge;
448#endif
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700449
Eric Dumazet8964be42009-11-20 15:35:04 -0800450 int skb_iif;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000451
452 __u32 rxhash;
453
454 __u16 vlan_tci;
455
Linus Torvalds1da177e2005-04-16 15:20:36 -0700456#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700457 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700459 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200462
Eric Dumazet0a148422011-04-20 09:27:32 +0000463 __u16 queue_mapping;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200464 kmemcheck_bitfield_begin(flags2);
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400465#ifdef CONFIG_IPV6_NDISC_NODETYPE
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000466 __u8 ndisc_nodetype:2;
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400467#endif
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;
473 /* 9/11 bit hole (depending on ndisc_nodetype presence) */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200474 kmemcheck_bitfield_end(flags2);
475
Chris Leech97fc2f02006-05-23 17:55:33 -0700476#ifdef CONFIG_NET_DMA
477 dma_cookie_t dma_cookie;
478#endif
James Morris984bc162006-06-09 00:29:17 -0700479#ifdef CONFIG_NETWORK_SECMARK
480 __u32 secmark;
481#endif
Neil Horman3b885782009-10-12 13:26:31 -0700482 union {
483 __u32 mark;
484 __u32 dropcount;
Eric Dumazeta21d4572012-04-10 20:30:48 +0000485 __u32 avail_size;
Neil Horman3b885782009-10-12 13:26:31 -0700486 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700488 sk_buff_data_t transport_header;
489 sk_buff_data_t network_header;
490 sk_buff_data_t mac_header;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700492 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700493 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700495 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700496 unsigned int truesize;
497 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498};
499
500#ifdef __KERNEL__
501/*
502 * Handling routines are only of interest to the kernel
503 */
504#include <linux/slab.h>
505
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506
Eric Dumazet7fee2262010-05-11 23:19:48 +0000507/*
508 * skb might have a dst pointer attached, refcounted or not.
509 * _skb_refdst low order bit is set if refcount was _not_ taken
510 */
511#define SKB_DST_NOREF 1UL
512#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
513
514/**
515 * skb_dst - returns skb dst_entry
516 * @skb: buffer
517 *
518 * Returns skb dst_entry, regardless of reference taken or not.
519 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000520static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
521{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000522 /* If refdst was not refcounted, check we still are in a
523 * rcu_read_lock section
524 */
525 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
526 !rcu_read_lock_held() &&
527 !rcu_read_lock_bh_held());
528 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000529}
530
Eric Dumazet7fee2262010-05-11 23:19:48 +0000531/**
532 * skb_dst_set - sets skb dst
533 * @skb: buffer
534 * @dst: dst entry
535 *
536 * Sets skb dst, assuming a reference was taken on dst and should
537 * be released by skb_dst_drop()
538 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000539static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
540{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000541 skb->_skb_refdst = (unsigned long)dst;
542}
543
Eric Dumazet27b75c92010-10-15 05:44:11 +0000544extern void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst);
Eric Dumazet7fee2262010-05-11 23:19:48 +0000545
546/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300547 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000548 * @skb: buffer
549 */
550static inline bool skb_dst_is_noref(const struct sk_buff *skb)
551{
552 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000553}
554
Eric Dumazet511c3f92009-06-02 05:14:27 +0000555static inline struct rtable *skb_rtable(const struct sk_buff *skb)
556{
Eric Dumazetadf30902009-06-02 05:19:30 +0000557 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000558}
559
Jörn Engel231d06a2006-03-20 21:28:35 -0800560extern void kfree_skb(struct sk_buff *skb);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000561extern void consume_skb(struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700562extern void __kfree_skb(struct sk_buff *skb);
David S. Millerd179cd12005-08-17 14:57:30 -0700563extern struct sk_buff *__alloc_skb(unsigned int size,
Christoph Hellwigb30973f2006-12-06 20:32:36 -0800564 gfp_t priority, int fclone, int node);
Eric Dumazetb2b5ce92011-11-14 06:03:34 +0000565extern struct sk_buff *build_skb(void *data);
David S. Millerd179cd12005-08-17 14:57:30 -0700566static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100567 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700568{
Eric Dumazet564824b2010-10-11 19:05:25 +0000569 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700570}
571
572static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100573 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700574{
Eric Dumazet564824b2010-10-11 19:05:25 +0000575 return __alloc_skb(size, priority, 1, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700576}
577
Andy Fleming3d153a72011-10-13 04:33:54 +0000578extern void skb_recycle(struct sk_buff *skb);
Changli Gao5b0daa32010-05-29 00:12:13 -0700579extern bool skb_recycle_check(struct sk_buff *skb, int skb_size);
Lennert Buytenhek04a4bb52008-10-01 02:33:12 -0700580
Herbert Xue0053ec2007-10-14 00:37:52 -0700581extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
Michael S. Tsirkin48c83012011-08-31 08:03:29 +0000582extern int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700583extern struct sk_buff *skb_clone(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100584 gfp_t priority);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700585extern struct sk_buff *skb_copy(const struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100586 gfp_t priority);
Eric Dumazet117632e2011-12-03 21:39:53 +0000587extern struct sk_buff *__pskb_copy(struct sk_buff *skb,
588 int headroom, gfp_t gfp_mask);
589
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590extern int pskb_expand_head(struct sk_buff *skb,
Victor Fusco86a76ca2005-07-08 14:57:47 -0700591 int nhead, int ntail,
Al Virodd0fc662005-10-07 07:46:04 +0100592 gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700593extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
594 unsigned int headroom);
595extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
596 int newheadroom, int newtailroom,
Al Virodd0fc662005-10-07 07:46:04 +0100597 gfp_t priority);
David Howells716ea3a2007-04-02 20:19:53 -0700598extern int skb_to_sgvec(struct sk_buff *skb,
599 struct scatterlist *sg, int offset,
600 int len);
601extern int skb_cow_data(struct sk_buff *skb, int tailbits,
602 struct sk_buff **trailer);
Herbert Xu5b057c62006-06-23 02:06:41 -0700603extern int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000604#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700605
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700606extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
607 int getfrag(void *from, char *to, int offset,
608 int len,int odd, struct sk_buff *skb),
609 void *from, int length);
610
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800611struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700612 __u32 lower_offset;
613 __u32 upper_offset;
614 __u32 frag_idx;
615 __u32 stepped_offset;
616 struct sk_buff *root_skb;
617 struct sk_buff *cur_skb;
618 __u8 *frag_data;
619};
620
621extern void skb_prepare_seq_read(struct sk_buff *skb,
622 unsigned int from, unsigned int to,
623 struct skb_seq_state *st);
624extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
625 struct skb_seq_state *st);
626extern void skb_abort_seq_read(struct skb_seq_state *st);
627
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700628extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
629 unsigned int to, struct ts_config *config,
630 struct ts_state *state);
631
Tom Herbertbdeab992011-08-14 19:45:55 +0000632extern void __skb_get_rxhash(struct sk_buff *skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000633static inline __u32 skb_get_rxhash(struct sk_buff *skb)
634{
635 if (!skb->rxhash)
Tom Herbertbdeab992011-08-14 19:45:55 +0000636 __skb_get_rxhash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000637
638 return skb->rxhash;
639}
640
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700641#ifdef NET_SKBUFF_DATA_USES_OFFSET
642static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
643{
644 return skb->head + skb->end;
645}
646#else
647static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
648{
649 return skb->end;
650}
651#endif
652
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700654#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655
Patrick Ohlyac45f602009-02-12 05:03:37 +0000656static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
657{
658 return &skb_shinfo(skb)->hwtstamps;
659}
660
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661/**
662 * skb_queue_empty - check if a queue is empty
663 * @list: queue head
664 *
665 * Returns true if the queue is empty, false otherwise.
666 */
667static inline int skb_queue_empty(const struct sk_buff_head *list)
668{
669 return list->next == (struct sk_buff *)list;
670}
671
672/**
David S. Millerfc7ebb22008-09-23 00:34:07 -0700673 * skb_queue_is_last - check if skb is the last entry in the queue
674 * @list: queue head
675 * @skb: buffer
676 *
677 * Returns true if @skb is the last buffer on the list.
678 */
679static inline bool skb_queue_is_last(const struct sk_buff_head *list,
680 const struct sk_buff *skb)
681{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000682 return skb->next == (struct sk_buff *)list;
David S. Millerfc7ebb22008-09-23 00:34:07 -0700683}
684
685/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800686 * skb_queue_is_first - check if skb is the first entry in the queue
687 * @list: queue head
688 * @skb: buffer
689 *
690 * Returns true if @skb is the first buffer on the list.
691 */
692static inline bool skb_queue_is_first(const struct sk_buff_head *list,
693 const struct sk_buff *skb)
694{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000695 return skb->prev == (struct sk_buff *)list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800696}
697
698/**
David S. Miller249c8b42008-09-23 00:44:42 -0700699 * skb_queue_next - return the next packet in the queue
700 * @list: queue head
701 * @skb: current buffer
702 *
703 * Return the next packet in @list after @skb. It is only valid to
704 * call this if skb_queue_is_last() evaluates to false.
705 */
706static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
707 const struct sk_buff *skb)
708{
709 /* This BUG_ON may seem severe, but if we just return then we
710 * are going to dereference garbage.
711 */
712 BUG_ON(skb_queue_is_last(list, skb));
713 return skb->next;
714}
715
716/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800717 * skb_queue_prev - return the prev packet in the queue
718 * @list: queue head
719 * @skb: current buffer
720 *
721 * Return the prev packet in @list before @skb. It is only valid to
722 * call this if skb_queue_is_first() evaluates to false.
723 */
724static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
725 const struct sk_buff *skb)
726{
727 /* This BUG_ON may seem severe, but if we just return then we
728 * are going to dereference garbage.
729 */
730 BUG_ON(skb_queue_is_first(list, skb));
731 return skb->prev;
732}
733
734/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735 * skb_get - reference buffer
736 * @skb: buffer to reference
737 *
738 * Makes another reference to a socket buffer and returns a pointer
739 * to the buffer.
740 */
741static inline struct sk_buff *skb_get(struct sk_buff *skb)
742{
743 atomic_inc(&skb->users);
744 return skb;
745}
746
747/*
748 * If users == 1, we are the only owner and are can avoid redundant
749 * atomic change.
750 */
751
752/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753 * skb_cloned - is the buffer a clone
754 * @skb: buffer to check
755 *
756 * Returns true if the buffer was generated with skb_clone() and is
757 * one of multiple shared copies of the buffer. Cloned buffers are
758 * shared data so must not be written to under normal circumstances.
759 */
760static inline int skb_cloned(const struct sk_buff *skb)
761{
762 return skb->cloned &&
763 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
764}
765
766/**
767 * skb_header_cloned - is the header a clone
768 * @skb: buffer to check
769 *
770 * Returns true if modifying the header part of the buffer requires
771 * the data to be copied.
772 */
773static inline int skb_header_cloned(const struct sk_buff *skb)
774{
775 int dataref;
776
777 if (!skb->cloned)
778 return 0;
779
780 dataref = atomic_read(&skb_shinfo(skb)->dataref);
781 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
782 return dataref != 1;
783}
784
785/**
786 * skb_header_release - release reference to header
787 * @skb: buffer to operate on
788 *
789 * Drop a reference to the header part of the buffer. This is done
790 * by acquiring a payload reference. You must not read from the header
791 * part of skb->data after this.
792 */
793static inline void skb_header_release(struct sk_buff *skb)
794{
795 BUG_ON(skb->nohdr);
796 skb->nohdr = 1;
797 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
798}
799
800/**
801 * skb_shared - is the buffer shared
802 * @skb: buffer to check
803 *
804 * Returns true if more than one person has a reference to this
805 * buffer.
806 */
807static inline int skb_shared(const struct sk_buff *skb)
808{
809 return atomic_read(&skb->users) != 1;
810}
811
812/**
813 * skb_share_check - check if buffer is shared and if so clone it
814 * @skb: buffer to check
815 * @pri: priority for memory allocation
816 *
817 * If the buffer is shared the buffer is cloned and the old copy
818 * drops a reference. A new clone with a single reference is returned.
819 * If the buffer is not shared the original buffer is returned. When
820 * being called from interrupt status or with spinlocks held pri must
821 * be GFP_ATOMIC.
822 *
823 * NULL is returned on a memory allocation failure.
824 */
Victor Fusco86a76ca2005-07-08 14:57:47 -0700825static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100826 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700827{
828 might_sleep_if(pri & __GFP_WAIT);
829 if (skb_shared(skb)) {
830 struct sk_buff *nskb = skb_clone(skb, pri);
831 kfree_skb(skb);
832 skb = nskb;
833 }
834 return skb;
835}
836
837/*
838 * Copy shared buffers into a new sk_buff. We effectively do COW on
839 * packets to handle cases where we have a local reader and forward
840 * and a couple of other messy ones. The normal one is tcpdumping
841 * a packet thats being forwarded.
842 */
843
844/**
845 * skb_unshare - make a copy of a shared buffer
846 * @skb: buffer to check
847 * @pri: priority for memory allocation
848 *
849 * If the socket buffer is a clone then this function creates a new
850 * copy of the data, drops a reference count on the old copy and returns
851 * the new copy with the reference count at 1. If the buffer is not a clone
852 * the original buffer is returned. When called with a spinlock held or
853 * from interrupt state @pri must be %GFP_ATOMIC
854 *
855 * %NULL is returned on a memory allocation failure.
856 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -0700857static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100858 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859{
860 might_sleep_if(pri & __GFP_WAIT);
861 if (skb_cloned(skb)) {
862 struct sk_buff *nskb = skb_copy(skb, pri);
863 kfree_skb(skb); /* Free our shared copy */
864 skb = nskb;
865 }
866 return skb;
867}
868
869/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -0800870 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871 * @list_: list to peek at
872 *
873 * Peek an &sk_buff. Unlike most other operations you _MUST_
874 * be careful with this one. A peek leaves the buffer on the
875 * list and someone else may run off with it. You must hold
876 * the appropriate locks or have a private queue to do this.
877 *
878 * Returns %NULL for an empty list or a pointer to the head element.
879 * The reference count is not incremented and the reference is therefore
880 * volatile. Use with caution.
881 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -0400882static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883{
Eric Dumazet05bdd2f2011-10-20 17:45:43 -0400884 struct sk_buff *list = ((const struct sk_buff *)list_)->next;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885 if (list == (struct sk_buff *)list_)
886 list = NULL;
887 return list;
888}
889
890/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +0000891 * skb_peek_next - peek skb following the given one from a queue
892 * @skb: skb to start from
893 * @list_: list to peek at
894 *
895 * Returns %NULL when the end of the list is met or a pointer to the
896 * next element. The reference count is not incremented and the
897 * reference is therefore volatile. Use with caution.
898 */
899static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
900 const struct sk_buff_head *list_)
901{
902 struct sk_buff *next = skb->next;
903 if (next == (struct sk_buff *)list_)
904 next = NULL;
905 return next;
906}
907
908/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -0800909 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910 * @list_: list to peek at
911 *
912 * Peek an &sk_buff. Unlike most other operations you _MUST_
913 * be careful with this one. A peek leaves the buffer on the
914 * list and someone else may run off with it. You must hold
915 * the appropriate locks or have a private queue to do this.
916 *
917 * Returns %NULL for an empty list or a pointer to the tail element.
918 * The reference count is not incremented and the reference is therefore
919 * volatile. Use with caution.
920 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -0400921static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922{
Eric Dumazet05bdd2f2011-10-20 17:45:43 -0400923 struct sk_buff *list = ((const struct sk_buff *)list_)->prev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924 if (list == (struct sk_buff *)list_)
925 list = NULL;
926 return list;
927}
928
929/**
930 * skb_queue_len - get queue length
931 * @list_: list to measure
932 *
933 * Return the length of an &sk_buff queue.
934 */
935static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
936{
937 return list_->qlen;
938}
939
David S. Miller67fed452008-09-21 22:36:24 -0700940/**
941 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
942 * @list: queue to initialize
943 *
944 * This initializes only the list and queue length aspects of
945 * an sk_buff_head object. This allows to initialize the list
946 * aspects of an sk_buff_head without reinitializing things like
947 * the spinlock. It can also be used for on-stack sk_buff_head
948 * objects where the spinlock is known to not be used.
949 */
950static inline void __skb_queue_head_init(struct sk_buff_head *list)
951{
952 list->prev = list->next = (struct sk_buff *)list;
953 list->qlen = 0;
954}
955
Arjan van de Ven76f10ad2006-08-02 14:06:55 -0700956/*
957 * This function creates a split out lock class for each invocation;
958 * this is needed for now since a whole lot of users of the skb-queue
959 * infrastructure in drivers have different locking usage (in hardirq)
960 * than the networking core (in softirq only). In the long run either the
961 * network layer or drivers should need annotation to consolidate the
962 * main types of usage into 3 classes.
963 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700964static inline void skb_queue_head_init(struct sk_buff_head *list)
965{
966 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -0700967 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700968}
969
Pavel Emelianovc2ecba72007-04-17 12:45:31 -0700970static inline void skb_queue_head_init_class(struct sk_buff_head *list,
971 struct lock_class_key *class)
972{
973 skb_queue_head_init(list);
974 lockdep_set_class(&list->lock, class);
975}
976
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977/*
Gerrit Renkerbf299272008-04-14 00:04:51 -0700978 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700979 *
980 * The "__skb_xxxx()" functions are the non-atomic ones that
981 * can only be called with interrupts disabled.
982 */
Gerrit Renkerbf299272008-04-14 00:04:51 -0700983extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
984static inline void __skb_insert(struct sk_buff *newsk,
985 struct sk_buff *prev, struct sk_buff *next,
986 struct sk_buff_head *list)
987{
988 newsk->next = next;
989 newsk->prev = prev;
990 next->prev = prev->next = newsk;
991 list->qlen++;
992}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700993
David S. Miller67fed452008-09-21 22:36:24 -0700994static inline void __skb_queue_splice(const struct sk_buff_head *list,
995 struct sk_buff *prev,
996 struct sk_buff *next)
997{
998 struct sk_buff *first = list->next;
999 struct sk_buff *last = list->prev;
1000
1001 first->prev = prev;
1002 prev->next = first;
1003
1004 last->next = next;
1005 next->prev = last;
1006}
1007
1008/**
1009 * skb_queue_splice - join two skb lists, this is designed for stacks
1010 * @list: the new list to add
1011 * @head: the place to add it in the first list
1012 */
1013static inline void skb_queue_splice(const struct sk_buff_head *list,
1014 struct sk_buff_head *head)
1015{
1016 if (!skb_queue_empty(list)) {
1017 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001018 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001019 }
1020}
1021
1022/**
1023 * skb_queue_splice - join two skb lists and reinitialise the emptied list
1024 * @list: the new list to add
1025 * @head: the place to add it in the first list
1026 *
1027 * The list at @list is reinitialised
1028 */
1029static inline void skb_queue_splice_init(struct sk_buff_head *list,
1030 struct sk_buff_head *head)
1031{
1032 if (!skb_queue_empty(list)) {
1033 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001034 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001035 __skb_queue_head_init(list);
1036 }
1037}
1038
1039/**
1040 * skb_queue_splice_tail - join two skb lists, each list being a queue
1041 * @list: the new list to add
1042 * @head: the place to add it in the first list
1043 */
1044static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1045 struct sk_buff_head *head)
1046{
1047 if (!skb_queue_empty(list)) {
1048 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001049 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001050 }
1051}
1052
1053/**
1054 * skb_queue_splice_tail - join two skb lists and reinitialise the emptied list
1055 * @list: the new list to add
1056 * @head: the place to add it in the first list
1057 *
1058 * Each of the lists is a queue.
1059 * The list at @list is reinitialised
1060 */
1061static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1062 struct sk_buff_head *head)
1063{
1064 if (!skb_queue_empty(list)) {
1065 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001066 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001067 __skb_queue_head_init(list);
1068 }
1069}
1070
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001072 * __skb_queue_after - queue a buffer at the list head
1073 * @list: list to use
1074 * @prev: place after this buffer
1075 * @newsk: buffer to queue
1076 *
1077 * Queue a buffer int the middle of a list. This function takes no locks
1078 * and you must therefore hold required locks before calling it.
1079 *
1080 * A buffer cannot be placed on two lists at the same time.
1081 */
1082static inline void __skb_queue_after(struct sk_buff_head *list,
1083 struct sk_buff *prev,
1084 struct sk_buff *newsk)
1085{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001086 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001087}
1088
Gerrit Renker7de6c032008-04-14 00:05:09 -07001089extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1090 struct sk_buff_head *list);
1091
Gerrit Renkerf5572852008-04-14 00:05:28 -07001092static inline void __skb_queue_before(struct sk_buff_head *list,
1093 struct sk_buff *next,
1094 struct sk_buff *newsk)
1095{
1096 __skb_insert(newsk, next->prev, next, list);
1097}
1098
Stephen Hemminger300ce172005-10-30 13:47:34 -08001099/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100 * __skb_queue_head - queue a buffer at the list head
1101 * @list: list to use
1102 * @newsk: buffer to queue
1103 *
1104 * Queue a buffer at the start of a list. This function takes no locks
1105 * and you must therefore hold required locks before calling it.
1106 *
1107 * A buffer cannot be placed on two lists at the same time.
1108 */
1109extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1110static inline void __skb_queue_head(struct sk_buff_head *list,
1111 struct sk_buff *newsk)
1112{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001113 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114}
1115
1116/**
1117 * __skb_queue_tail - queue a buffer at the list tail
1118 * @list: list to use
1119 * @newsk: buffer to queue
1120 *
1121 * Queue a buffer at the end of a list. This function takes no locks
1122 * and you must therefore hold required locks before calling it.
1123 *
1124 * A buffer cannot be placed on two lists at the same time.
1125 */
1126extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1127static inline void __skb_queue_tail(struct sk_buff_head *list,
1128 struct sk_buff *newsk)
1129{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001130 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131}
1132
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134 * remove sk_buff from list. _Must_ be called atomically, and with
1135 * the list known..
1136 */
David S. Miller8728b832005-08-09 19:25:21 -07001137extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1139{
1140 struct sk_buff *next, *prev;
1141
1142 list->qlen--;
1143 next = skb->next;
1144 prev = skb->prev;
1145 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146 next->prev = prev;
1147 prev->next = next;
1148}
1149
Gerrit Renkerf525c062008-04-14 00:04:12 -07001150/**
1151 * __skb_dequeue - remove from the head of the queue
1152 * @list: list to dequeue from
1153 *
1154 * Remove the head of the list. This function does not take any locks
1155 * so must be used with appropriate locks held only. The head item is
1156 * returned or %NULL if the list is empty.
1157 */
1158extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1159static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1160{
1161 struct sk_buff *skb = skb_peek(list);
1162 if (skb)
1163 __skb_unlink(skb, list);
1164 return skb;
1165}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001166
1167/**
1168 * __skb_dequeue_tail - remove from the tail of the queue
1169 * @list: list to dequeue from
1170 *
1171 * Remove the tail of the list. This function does not take any locks
1172 * so must be used with appropriate locks held only. The tail item is
1173 * returned or %NULL if the list is empty.
1174 */
1175extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1176static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1177{
1178 struct sk_buff *skb = skb_peek_tail(list);
1179 if (skb)
1180 __skb_unlink(skb, list);
1181 return skb;
1182}
1183
1184
David S. Millerbdcc0922012-03-07 20:53:36 -05001185static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001186{
1187 return skb->data_len;
1188}
1189
1190static inline unsigned int skb_headlen(const struct sk_buff *skb)
1191{
1192 return skb->len - skb->data_len;
1193}
1194
1195static inline int skb_pagelen(const struct sk_buff *skb)
1196{
1197 int i, len = 0;
1198
1199 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001200 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001201 return len + skb_headlen(skb);
1202}
1203
Ian Campbell131ea662011-08-19 06:25:00 +00001204/**
1205 * __skb_fill_page_desc - initialise a paged fragment in an skb
1206 * @skb: buffer containing fragment to be initialised
1207 * @i: paged fragment index to initialise
1208 * @page: the page to use for this fragment
1209 * @off: the offset to the data with @page
1210 * @size: the length of the data
1211 *
1212 * Initialises the @i'th fragment of @skb to point to &size bytes at
1213 * offset @off within @page.
1214 *
1215 * Does not take any additional reference on the fragment.
1216 */
1217static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1218 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001219{
1220 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1221
Ian Campbella8605c62011-10-19 23:01:49 +00001222 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001223 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001224 skb_frag_size_set(frag, size);
Ian Campbell131ea662011-08-19 06:25:00 +00001225}
1226
1227/**
1228 * skb_fill_page_desc - initialise a paged fragment in an skb
1229 * @skb: buffer containing fragment to be initialised
1230 * @i: paged fragment index to initialise
1231 * @page: the page to use for this fragment
1232 * @off: the offset to the data with @page
1233 * @size: the length of the data
1234 *
1235 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1236 * @skb to point to &size bytes at offset @off within @page. In
1237 * addition updates @skb such that @i is the last fragment.
1238 *
1239 * Does not take any additional reference on the fragment.
1240 */
1241static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1242 struct page *page, int off, int size)
1243{
1244 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001245 skb_shinfo(skb)->nr_frags = i + 1;
1246}
1247
Peter Zijlstra654bed12008-10-07 14:22:33 -07001248extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
Eric Dumazet50269e12012-03-23 23:59:33 +00001249 int off, int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001250
Linus Torvalds1da177e2005-04-16 15:20:36 -07001251#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001252#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001253#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1254
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001255#ifdef NET_SKBUFF_DATA_USES_OFFSET
1256static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1257{
1258 return skb->head + skb->tail;
1259}
1260
1261static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1262{
1263 skb->tail = skb->data - skb->head;
1264}
1265
1266static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1267{
1268 skb_reset_tail_pointer(skb);
1269 skb->tail += offset;
1270}
1271#else /* NET_SKBUFF_DATA_USES_OFFSET */
1272static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1273{
1274 return skb->tail;
1275}
1276
1277static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1278{
1279 skb->tail = skb->data;
1280}
1281
1282static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1283{
1284 skb->tail = skb->data + offset;
1285}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001286
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001287#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1288
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289/*
1290 * Add data to an sk_buff
1291 */
Ilpo Järvinen0dde3e12008-03-27 17:43:41 -07001292extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001293static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1294{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001295 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001296 SKB_LINEAR_ASSERT(skb);
1297 skb->tail += len;
1298 skb->len += len;
1299 return tmp;
1300}
1301
Ilpo Järvinenc2aa2702008-03-27 17:52:40 -07001302extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001303static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1304{
1305 skb->data -= len;
1306 skb->len += len;
1307 return skb->data;
1308}
1309
Ilpo Järvinen6be8ac22008-03-27 17:47:24 -07001310extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001311static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1312{
1313 skb->len -= len;
1314 BUG_ON(skb->len < skb->data_len);
1315 return skb->data += len;
1316}
1317
David S. Miller47d29642010-05-02 02:21:44 -07001318static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1319{
1320 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1321}
1322
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1324
1325static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1326{
1327 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001328 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001329 return NULL;
1330 skb->len -= len;
1331 return skb->data += len;
1332}
1333
1334static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1335{
1336 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1337}
1338
1339static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1340{
1341 if (likely(len <= skb_headlen(skb)))
1342 return 1;
1343 if (unlikely(len > skb->len))
1344 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001345 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001346}
1347
1348/**
1349 * skb_headroom - bytes at buffer head
1350 * @skb: buffer to check
1351 *
1352 * Return the number of bytes of free space at the head of an &sk_buff.
1353 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001354static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001355{
1356 return skb->data - skb->head;
1357}
1358
1359/**
1360 * skb_tailroom - bytes at buffer end
1361 * @skb: buffer to check
1362 *
1363 * Return the number of bytes of free space at the tail of an sk_buff
1364 */
1365static inline int skb_tailroom(const struct sk_buff *skb)
1366{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001367 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001368}
1369
1370/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00001371 * skb_availroom - bytes at buffer end
1372 * @skb: buffer to check
1373 *
1374 * Return the number of bytes of free space at the tail of an sk_buff
1375 * allocated by sk_stream_alloc()
1376 */
1377static inline int skb_availroom(const struct sk_buff *skb)
1378{
1379 return skb_is_nonlinear(skb) ? 0 : skb->avail_size - skb->len;
1380}
1381
1382/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383 * skb_reserve - adjust headroom
1384 * @skb: buffer to alter
1385 * @len: bytes to move
1386 *
1387 * Increase the headroom of an empty &sk_buff by reducing the tail
1388 * room. This is only allowed for an empty buffer.
1389 */
David S. Miller8243126c2006-01-17 02:54:21 -08001390static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001391{
1392 skb->data += len;
1393 skb->tail += len;
1394}
1395
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00001396static inline void skb_reset_mac_len(struct sk_buff *skb)
1397{
1398 skb->mac_len = skb->network_header - skb->mac_header;
1399}
1400
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001401#ifdef NET_SKBUFF_DATA_USES_OFFSET
1402static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1403{
1404 return skb->head + skb->transport_header;
1405}
1406
1407static inline void skb_reset_transport_header(struct sk_buff *skb)
1408{
1409 skb->transport_header = skb->data - skb->head;
1410}
1411
1412static inline void skb_set_transport_header(struct sk_buff *skb,
1413 const int offset)
1414{
1415 skb_reset_transport_header(skb);
1416 skb->transport_header += offset;
1417}
1418
1419static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1420{
1421 return skb->head + skb->network_header;
1422}
1423
1424static inline void skb_reset_network_header(struct sk_buff *skb)
1425{
1426 skb->network_header = skb->data - skb->head;
1427}
1428
1429static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1430{
1431 skb_reset_network_header(skb);
1432 skb->network_header += offset;
1433}
1434
1435static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1436{
1437 return skb->head + skb->mac_header;
1438}
1439
1440static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1441{
1442 return skb->mac_header != ~0U;
1443}
1444
1445static inline void skb_reset_mac_header(struct sk_buff *skb)
1446{
1447 skb->mac_header = skb->data - skb->head;
1448}
1449
1450static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1451{
1452 skb_reset_mac_header(skb);
1453 skb->mac_header += offset;
1454}
1455
1456#else /* NET_SKBUFF_DATA_USES_OFFSET */
1457
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001458static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1459{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001460 return skb->transport_header;
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001461}
1462
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001463static inline void skb_reset_transport_header(struct sk_buff *skb)
1464{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001465 skb->transport_header = skb->data;
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001466}
1467
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001468static inline void skb_set_transport_header(struct sk_buff *skb,
1469 const int offset)
1470{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001471 skb->transport_header = skb->data + offset;
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001472}
1473
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001474static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1475{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001476 return skb->network_header;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001477}
1478
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001479static inline void skb_reset_network_header(struct sk_buff *skb)
1480{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001481 skb->network_header = skb->data;
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001482}
1483
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001484static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1485{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001486 skb->network_header = skb->data + offset;
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001487}
1488
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001489static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1490{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001491 return skb->mac_header;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001492}
1493
1494static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1495{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001496 return skb->mac_header != NULL;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001497}
1498
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001499static inline void skb_reset_mac_header(struct sk_buff *skb)
1500{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001501 skb->mac_header = skb->data;
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001502}
1503
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001504static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1505{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001506 skb->mac_header = skb->data + offset;
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001507}
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001508#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1509
Eric Dumazet03606892012-02-23 10:55:02 +00001510static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1511{
1512 if (skb_mac_header_was_set(skb)) {
1513 const unsigned char *old_mac = skb_mac_header(skb);
1514
1515 skb_set_mac_header(skb, -skb->mac_len);
1516 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1517 }
1518}
1519
Michał Mirosław04fb4512010-12-14 15:24:08 +00001520static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1521{
1522 return skb->csum_start - skb_headroom(skb);
1523}
1524
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001525static inline int skb_transport_offset(const struct sk_buff *skb)
1526{
1527 return skb_transport_header(skb) - skb->data;
1528}
1529
1530static inline u32 skb_network_header_len(const struct sk_buff *skb)
1531{
1532 return skb->transport_header - skb->network_header;
1533}
1534
1535static inline int skb_network_offset(const struct sk_buff *skb)
1536{
1537 return skb_network_header(skb) - skb->data;
1538}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001539
Changli Gaof9599ce2010-08-04 04:43:44 +00001540static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1541{
1542 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1543}
1544
Linus Torvalds1da177e2005-04-16 15:20:36 -07001545/*
1546 * CPUs often take a performance hit when accessing unaligned memory
1547 * locations. The actual performance hit varies, it can be small if the
1548 * hardware handles it or large if we have to take an exception and fix it
1549 * in software.
1550 *
1551 * Since an ethernet header is 14 bytes network drivers often end up with
1552 * the IP header at an unaligned offset. The IP header can be aligned by
1553 * shifting the start of the packet by 2 bytes. Drivers should do this
1554 * with:
1555 *
Tobias Klauser8660c122009-07-13 22:48:16 +00001556 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001557 *
1558 * The downside to this alignment of the IP header is that the DMA is now
1559 * unaligned. On some architectures the cost of an unaligned DMA is high
1560 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00001561 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001562 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1563 * to be overridden.
1564 */
1565#ifndef NET_IP_ALIGN
1566#define NET_IP_ALIGN 2
1567#endif
1568
Anton Blanchard025be812006-03-31 02:27:06 -08001569/*
1570 * The networking layer reserves some headroom in skb data (via
1571 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1572 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08001573 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08001574 *
1575 * Unfortunately this headroom changes the DMA alignment of the resulting
1576 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1577 * on some architectures. An architecture can override this value,
1578 * perhaps setting it to a cacheline in size (since that will maintain
1579 * cacheline alignment of the DMA). It must be a power of 2.
1580 *
David S. Millerd6301d32009-02-08 19:24:13 -08001581 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08001582 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07001583 *
1584 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1585 * to reduce average number of cache lines per packet.
1586 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07001587 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08001588 */
1589#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07001590#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08001591#endif
1592
Herbert Xu3cc0e872006-06-09 16:13:38 -07001593extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001594
1595static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1596{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00001597 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07001598 WARN_ON(1);
1599 return;
1600 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001601 skb->len = len;
1602 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001603}
1604
Ilpo Järvinen419ae742008-03-27 17:54:01 -07001605extern void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001606
1607static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1608{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001609 if (skb->data_len)
1610 return ___pskb_trim(skb, len);
1611 __skb_trim(skb, len);
1612 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001613}
1614
1615static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1616{
1617 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1618}
1619
1620/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07001621 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1622 * @skb: buffer to alter
1623 * @len: new length
1624 *
1625 * This is identical to pskb_trim except that the caller knows that
1626 * the skb is not cloned so we should never get an error due to out-
1627 * of-memory.
1628 */
1629static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1630{
1631 int err = pskb_trim(skb, len);
1632 BUG_ON(err);
1633}
1634
1635/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001636 * skb_orphan - orphan a buffer
1637 * @skb: buffer to orphan
1638 *
1639 * If a buffer currently has an owner then we call the owner's
1640 * destructor function and make the @skb unowned. The buffer continues
1641 * to exist but is no longer charged to its former owner.
1642 */
1643static inline void skb_orphan(struct sk_buff *skb)
1644{
1645 if (skb->destructor)
1646 skb->destructor(skb);
1647 skb->destructor = NULL;
1648 skb->sk = NULL;
1649}
1650
1651/**
1652 * __skb_queue_purge - empty a list
1653 * @list: list to empty
1654 *
1655 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1656 * the list and one reference dropped. This function does not take the
1657 * list lock and the caller must hold the relevant locks to use it.
1658 */
1659extern void skb_queue_purge(struct sk_buff_head *list);
1660static inline void __skb_queue_purge(struct sk_buff_head *list)
1661{
1662 struct sk_buff *skb;
1663 while ((skb = __skb_dequeue(list)) != NULL)
1664 kfree_skb(skb);
1665}
1666
1667/**
Christoph Hellwigb4e54de2006-07-24 15:31:14 -07001668 * __dev_alloc_skb - allocate an skbuff for receiving
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669 * @length: length to allocate
1670 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1671 *
1672 * Allocate a new &sk_buff and assign it a usage count of one. The
1673 * buffer has unspecified headroom built in. Users should allocate
1674 * the headroom they think they need without accounting for the
1675 * built in space. The built in space is used for optimisations.
1676 *
Christoph Hellwig766ea8c2006-08-07 15:49:53 -07001677 * %NULL is returned if there is no free memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001678 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001679static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
Al Virodd0fc662005-10-07 07:46:04 +01001680 gfp_t gfp_mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001681{
Anton Blanchard025be812006-03-31 02:27:06 -08001682 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001683 if (likely(skb))
Anton Blanchard025be812006-03-31 02:27:06 -08001684 skb_reserve(skb, NET_SKB_PAD);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001685 return skb;
1686}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687
Ilpo Järvinenf58518e2008-03-27 17:51:31 -07001688extern struct sk_buff *dev_alloc_skb(unsigned int length);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001689
Christoph Hellwig8af27452006-07-31 22:35:23 -07001690extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1691 unsigned int length, gfp_t gfp_mask);
1692
1693/**
1694 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1695 * @dev: network device to receive on
1696 * @length: length to allocate
1697 *
1698 * Allocate a new &sk_buff and assign it a usage count of one. The
1699 * buffer has unspecified headroom built in. Users should allocate
1700 * the headroom they think they need without accounting for the
1701 * built in space. The built in space is used for optimisations.
1702 *
1703 * %NULL is returned if there is no free memory. Although this function
1704 * allocates memory it can be called from an interrupt.
1705 */
1706static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1707 unsigned int length)
1708{
1709 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1710}
1711
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001712static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
1713 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00001714{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001715 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00001716
1717 if (NET_IP_ALIGN && skb)
1718 skb_reserve(skb, NET_IP_ALIGN);
1719 return skb;
1720}
1721
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001722static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1723 unsigned int length)
1724{
1725 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
1726}
1727
Eric Dumazet564824b2010-10-11 19:05:25 +00001728/**
Ian Campbell131ea662011-08-19 06:25:00 +00001729 * skb_frag_page - retrieve the page refered to by a paged fragment
1730 * @frag: the paged fragment
1731 *
1732 * Returns the &struct page associated with @frag.
1733 */
1734static inline struct page *skb_frag_page(const skb_frag_t *frag)
1735{
Ian Campbella8605c62011-10-19 23:01:49 +00001736 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00001737}
1738
1739/**
1740 * __skb_frag_ref - take an addition reference on a paged fragment.
1741 * @frag: the paged fragment
1742 *
1743 * Takes an additional reference on the paged fragment @frag.
1744 */
1745static inline void __skb_frag_ref(skb_frag_t *frag)
1746{
1747 get_page(skb_frag_page(frag));
1748}
1749
1750/**
1751 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
1752 * @skb: the buffer
1753 * @f: the fragment offset.
1754 *
1755 * Takes an additional reference on the @f'th paged fragment of @skb.
1756 */
1757static inline void skb_frag_ref(struct sk_buff *skb, int f)
1758{
1759 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
1760}
1761
1762/**
1763 * __skb_frag_unref - release a reference on a paged fragment.
1764 * @frag: the paged fragment
1765 *
1766 * Releases a reference on the paged fragment @frag.
1767 */
1768static inline void __skb_frag_unref(skb_frag_t *frag)
1769{
1770 put_page(skb_frag_page(frag));
1771}
1772
1773/**
1774 * skb_frag_unref - release a reference on a paged fragment of an skb.
1775 * @skb: the buffer
1776 * @f: the fragment offset
1777 *
1778 * Releases a reference on the @f'th paged fragment of @skb.
1779 */
1780static inline void skb_frag_unref(struct sk_buff *skb, int f)
1781{
1782 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
1783}
1784
1785/**
1786 * skb_frag_address - gets the address of the data contained in a paged fragment
1787 * @frag: the paged fragment buffer
1788 *
1789 * Returns the address of the data within @frag. The page must already
1790 * be mapped.
1791 */
1792static inline void *skb_frag_address(const skb_frag_t *frag)
1793{
1794 return page_address(skb_frag_page(frag)) + frag->page_offset;
1795}
1796
1797/**
1798 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
1799 * @frag: the paged fragment buffer
1800 *
1801 * Returns the address of the data within @frag. Checks that the page
1802 * is mapped and returns %NULL otherwise.
1803 */
1804static inline void *skb_frag_address_safe(const skb_frag_t *frag)
1805{
1806 void *ptr = page_address(skb_frag_page(frag));
1807 if (unlikely(!ptr))
1808 return NULL;
1809
1810 return ptr + frag->page_offset;
1811}
1812
1813/**
1814 * __skb_frag_set_page - sets the page contained in a paged fragment
1815 * @frag: the paged fragment
1816 * @page: the page to set
1817 *
1818 * Sets the fragment @frag to contain @page.
1819 */
1820static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
1821{
Ian Campbella8605c62011-10-19 23:01:49 +00001822 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00001823}
1824
1825/**
1826 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
1827 * @skb: the buffer
1828 * @f: the fragment offset
1829 * @page: the page to set
1830 *
1831 * Sets the @f'th fragment of @skb to contain @page.
1832 */
1833static inline void skb_frag_set_page(struct sk_buff *skb, int f,
1834 struct page *page)
1835{
1836 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
1837}
1838
1839/**
1840 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00001841 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00001842 * @frag: the paged fragment to map
1843 * @offset: the offset within the fragment (starting at the
1844 * fragment's own offset)
1845 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00001846 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00001847 *
1848 * Maps the page associated with @frag to @device.
1849 */
1850static inline dma_addr_t skb_frag_dma_map(struct device *dev,
1851 const skb_frag_t *frag,
1852 size_t offset, size_t size,
1853 enum dma_data_direction dir)
1854{
1855 return dma_map_page(dev, skb_frag_page(frag),
1856 frag->page_offset + offset, size, dir);
1857}
1858
Eric Dumazet117632e2011-12-03 21:39:53 +00001859static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
1860 gfp_t gfp_mask)
1861{
1862 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
1863}
1864
Ian Campbell131ea662011-08-19 06:25:00 +00001865/**
Patrick McHardy334a8132007-06-25 04:35:20 -07001866 * skb_clone_writable - is the header of a clone writable
1867 * @skb: buffer to check
1868 * @len: length up to which to write
1869 *
1870 * Returns true if modifying the header part of the cloned buffer
1871 * does not requires the data to be copied.
1872 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001873static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07001874{
1875 return !skb_header_cloned(skb) &&
1876 skb_headroom(skb) + len <= skb->hdr_len;
1877}
1878
Herbert Xud9cc2042007-09-16 16:21:16 -07001879static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
1880 int cloned)
1881{
1882 int delta = 0;
1883
1884 if (headroom < NET_SKB_PAD)
1885 headroom = NET_SKB_PAD;
1886 if (headroom > skb_headroom(skb))
1887 delta = headroom - skb_headroom(skb);
1888
1889 if (delta || cloned)
1890 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
1891 GFP_ATOMIC);
1892 return 0;
1893}
1894
Patrick McHardy334a8132007-06-25 04:35:20 -07001895/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001896 * skb_cow - copy header of skb when it is required
1897 * @skb: buffer to cow
1898 * @headroom: needed headroom
1899 *
1900 * If the skb passed lacks sufficient headroom or its data part
1901 * is shared, data is reallocated. If reallocation fails, an error
1902 * is returned and original skb is not changed.
1903 *
1904 * The result is skb with writable area skb->head...skb->tail
1905 * and at least @headroom of space at head.
1906 */
1907static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1908{
Herbert Xud9cc2042007-09-16 16:21:16 -07001909 return __skb_cow(skb, headroom, skb_cloned(skb));
1910}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001911
Herbert Xud9cc2042007-09-16 16:21:16 -07001912/**
1913 * skb_cow_head - skb_cow but only making the head writable
1914 * @skb: buffer to cow
1915 * @headroom: needed headroom
1916 *
1917 * This function is identical to skb_cow except that we replace the
1918 * skb_cloned check by skb_header_cloned. It should be used when
1919 * you only need to push on some header and do not need to modify
1920 * the data.
1921 */
1922static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
1923{
1924 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001925}
1926
1927/**
1928 * skb_padto - pad an skbuff up to a minimal size
1929 * @skb: buffer to pad
1930 * @len: minimal length
1931 *
1932 * Pads up a buffer to ensure the trailing bytes exist and are
1933 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07001934 * is untouched. Otherwise it is extended. Returns zero on
1935 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001936 */
1937
Herbert Xu5b057c62006-06-23 02:06:41 -07001938static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001939{
1940 unsigned int size = skb->len;
1941 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07001942 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001943 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001944}
1945
1946static inline int skb_add_data(struct sk_buff *skb,
1947 char __user *from, int copy)
1948{
1949 const int off = skb->len;
1950
1951 if (skb->ip_summed == CHECKSUM_NONE) {
1952 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08001953 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
Linus Torvalds1da177e2005-04-16 15:20:36 -07001954 copy, 0, &err);
1955 if (!err) {
1956 skb->csum = csum_block_add(skb->csum, csum, off);
1957 return 0;
1958 }
1959 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1960 return 0;
1961
1962 __skb_trim(skb, off);
1963 return -EFAULT;
1964}
1965
1966static inline int skb_can_coalesce(struct sk_buff *skb, int i,
Ian Campbellea2ab692011-08-22 23:44:58 +00001967 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001968{
1969 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00001970 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001971
Ian Campbellea2ab692011-08-22 23:44:58 +00001972 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00001973 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001974 }
1975 return 0;
1976}
1977
Herbert Xu364c6ba2006-06-09 16:10:40 -07001978static inline int __skb_linearize(struct sk_buff *skb)
1979{
1980 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1981}
1982
Linus Torvalds1da177e2005-04-16 15:20:36 -07001983/**
1984 * skb_linearize - convert paged skb to linear one
1985 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07001986 *
1987 * If there is no free memory -ENOMEM is returned, otherwise zero
1988 * is returned and the old skb data released.
1989 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07001990static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001991{
Herbert Xu364c6ba2006-06-09 16:10:40 -07001992 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1993}
1994
1995/**
1996 * skb_linearize_cow - make sure skb is linear and writable
1997 * @skb: buffer to process
1998 *
1999 * If there is no free memory -ENOMEM is returned, otherwise zero
2000 * is returned and the old skb data released.
2001 */
2002static inline int skb_linearize_cow(struct sk_buff *skb)
2003{
2004 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2005 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002006}
2007
2008/**
2009 * skb_postpull_rcsum - update checksum for received skb after pull
2010 * @skb: buffer to update
2011 * @start: start of data before pull
2012 * @len: length of data pulled
2013 *
2014 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002015 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2016 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002017 */
2018
2019static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002020 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021{
Patrick McHardy84fa7932006-08-29 16:44:56 -07002022 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002023 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2024}
2025
Herbert Xucbb042f2006-03-20 22:43:56 -08002026unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2027
Linus Torvalds1da177e2005-04-16 15:20:36 -07002028/**
2029 * pskb_trim_rcsum - trim received skb and update checksum
2030 * @skb: buffer to trim
2031 * @len: new length
2032 *
2033 * This is exactly the same as pskb_trim except that it ensures the
2034 * checksum of received packets are still valid after the operation.
2035 */
2036
2037static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2038{
Stephen Hemminger0e4e4222005-09-08 12:32:03 -07002039 if (likely(len >= skb->len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002040 return 0;
Patrick McHardy84fa7932006-08-29 16:44:56 -07002041 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002042 skb->ip_summed = CHECKSUM_NONE;
2043 return __pskb_trim(skb, len);
2044}
2045
Linus Torvalds1da177e2005-04-16 15:20:36 -07002046#define skb_queue_walk(queue, skb) \
2047 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002048 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002049 skb = skb->next)
2050
James Chapman46f89142007-04-30 00:07:31 -07002051#define skb_queue_walk_safe(queue, skb, tmp) \
2052 for (skb = (queue)->next, tmp = skb->next; \
2053 skb != (struct sk_buff *)(queue); \
2054 skb = tmp, tmp = skb->next)
2055
David S. Miller1164f522008-09-23 00:49:44 -07002056#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002057 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002058 skb = skb->next)
2059
2060#define skb_queue_walk_from_safe(queue, skb, tmp) \
2061 for (tmp = skb->next; \
2062 skb != (struct sk_buff *)(queue); \
2063 skb = tmp, tmp = skb->next)
2064
Stephen Hemminger300ce172005-10-30 13:47:34 -08002065#define skb_queue_reverse_walk(queue, skb) \
2066 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002067 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08002068 skb = skb->prev)
2069
David S. Miller686a2952011-01-20 22:47:32 -08002070#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2071 for (skb = (queue)->prev, tmp = skb->prev; \
2072 skb != (struct sk_buff *)(queue); \
2073 skb = tmp, tmp = skb->prev)
2074
2075#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2076 for (tmp = skb->prev; \
2077 skb != (struct sk_buff *)(queue); \
2078 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002079
David S. Miller21dc3302010-08-23 00:13:46 -07002080static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07002081{
2082 return skb_shinfo(skb)->frag_list != NULL;
2083}
2084
2085static inline void skb_frag_list_init(struct sk_buff *skb)
2086{
2087 skb_shinfo(skb)->frag_list = NULL;
2088}
2089
2090static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2091{
2092 frag->next = skb_shinfo(skb)->frag_list;
2093 skb_shinfo(skb)->frag_list = frag;
2094}
2095
2096#define skb_walk_frags(skb, iter) \
2097 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2098
Herbert Xua59322b2007-12-05 01:53:40 -08002099extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
Pavel Emelyanov3f518bf2012-02-21 07:30:58 +00002100 int *peeked, int *off, int *err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002101extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
2102 int noblock, int *err);
2103extern unsigned int datagram_poll(struct file *file, struct socket *sock,
2104 struct poll_table_struct *wait);
2105extern int skb_copy_datagram_iovec(const struct sk_buff *from,
2106 int offset, struct iovec *to,
2107 int size);
Herbert Xufb286bb2005-11-10 13:01:24 -08002108extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002109 int hlen,
2110 struct iovec *iov);
Rusty Russelldb543c12008-08-15 15:13:53 -07002111extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
2112 int offset,
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +00002113 const struct iovec *from,
2114 int from_offset,
Rusty Russelldb543c12008-08-15 15:13:53 -07002115 int len);
Michael S. Tsirkin0a1ec072009-04-20 01:25:46 +00002116extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
2117 int offset,
2118 const struct iovec *to,
2119 int to_offset,
2120 int size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002121extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
Eric Dumazet9d410c72009-10-30 05:03:53 +00002122extern void skb_free_datagram_locked(struct sock *sk,
2123 struct sk_buff *skb);
Herbert Xu27ab2562007-12-05 01:51:58 -08002124extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
Herbert Xu3305b802005-12-13 23:16:37 -08002125 unsigned int flags);
Al Viro2bbbc862006-11-14 21:37:14 -08002126extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
2127 int len, __wsum csum);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002128extern int skb_copy_bits(const struct sk_buff *skb, int offset,
2129 void *to, int len);
Stephen Hemminger0c6fcc82007-04-20 16:40:01 -07002130extern int skb_store_bits(struct sk_buff *skb, int offset,
2131 const void *from, int len);
Al Viro81d77662006-11-14 21:37:33 -08002132extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002133 int offset, u8 *to, int len,
Al Viro81d77662006-11-14 21:37:33 -08002134 __wsum csum);
Jens Axboe9c55e012007-11-06 23:30:13 -08002135extern int skb_splice_bits(struct sk_buff *skb,
2136 unsigned int offset,
2137 struct pipe_inode_info *pipe,
2138 unsigned int len,
2139 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002140extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
2141extern void skb_split(struct sk_buff *skb,
2142 struct sk_buff *skb1, const u32 len);
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08002143extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
2144 int shiftlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002145
Michał Mirosławc8f44af2011-11-15 15:29:55 +00002146extern struct sk_buff *skb_segment(struct sk_buff *skb,
2147 netdev_features_t features);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002148
Linus Torvalds1da177e2005-04-16 15:20:36 -07002149static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2150 int len, void *buffer)
2151{
2152 int hlen = skb_headlen(skb);
2153
Patrick McHardy55820ee2005-07-05 14:08:10 -07002154 if (hlen - offset >= len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002155 return skb->data + offset;
2156
2157 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2158 return NULL;
2159
2160 return buffer;
2161}
2162
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03002163static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2164 void *to,
2165 const unsigned int len)
2166{
2167 memcpy(to, skb->data, len);
2168}
2169
2170static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2171 const int offset, void *to,
2172 const unsigned int len)
2173{
2174 memcpy(to, skb->data + offset, len);
2175}
2176
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03002177static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2178 const void *from,
2179 const unsigned int len)
2180{
2181 memcpy(skb->data, from, len);
2182}
2183
2184static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2185 const int offset,
2186 const void *from,
2187 const unsigned int len)
2188{
2189 memcpy(skb->data + offset, from, len);
2190}
2191
Linus Torvalds1da177e2005-04-16 15:20:36 -07002192extern void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002193
Patrick Ohlyac45f602009-02-12 05:03:37 +00002194static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2195{
2196 return skb->tstamp;
2197}
2198
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002199/**
2200 * skb_get_timestamp - get timestamp from a skb
2201 * @skb: skb to get stamp from
2202 * @stamp: pointer to struct timeval to store stamp in
2203 *
2204 * Timestamps are stored in the skb as offsets to a base timestamp.
2205 * This function converts the offset back to a struct timeval and stores
2206 * it in stamp.
2207 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00002208static inline void skb_get_timestamp(const struct sk_buff *skb,
2209 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002210{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002211 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002212}
2213
Patrick Ohlyac45f602009-02-12 05:03:37 +00002214static inline void skb_get_timestampns(const struct sk_buff *skb,
2215 struct timespec *stamp)
2216{
2217 *stamp = ktime_to_timespec(skb->tstamp);
2218}
2219
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002220static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002221{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002222 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002223}
2224
Stephen Hemminger164891a2007-04-23 22:26:16 -07002225static inline ktime_t net_timedelta(ktime_t t)
2226{
2227 return ktime_sub(ktime_get_real(), t);
2228}
2229
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07002230static inline ktime_t net_invalid_timestamp(void)
2231{
2232 return ktime_set(0, 0);
2233}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002234
Richard Cochranc1f19b52010-07-17 08:49:36 +00002235extern void skb_timestamping_init(void);
2236
2237#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2238
2239extern void skb_clone_tx_timestamp(struct sk_buff *skb);
2240extern bool skb_defer_rx_timestamp(struct sk_buff *skb);
2241
2242#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2243
2244static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2245{
2246}
2247
2248static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2249{
2250 return false;
2251}
2252
2253#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2254
2255/**
2256 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2257 *
Richard Cochranda92b192011-10-21 00:49:15 +00002258 * PHY drivers may accept clones of transmitted packets for
2259 * timestamping via their phy_driver.txtstamp method. These drivers
2260 * must call this function to return the skb back to the stack, with
2261 * or without a timestamp.
2262 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00002263 * @skb: clone of the the original outgoing packet
Richard Cochranda92b192011-10-21 00:49:15 +00002264 * @hwtstamps: hardware time stamps, may be NULL if not available
Richard Cochranc1f19b52010-07-17 08:49:36 +00002265 *
2266 */
2267void skb_complete_tx_timestamp(struct sk_buff *skb,
2268 struct skb_shared_hwtstamps *hwtstamps);
2269
Patrick Ohlyac45f602009-02-12 05:03:37 +00002270/**
2271 * skb_tstamp_tx - queue clone of skb with send time stamps
2272 * @orig_skb: the original outgoing packet
2273 * @hwtstamps: hardware time stamps, may be NULL if not available
2274 *
2275 * If the skb has a socket associated, then this function clones the
2276 * skb (thus sharing the actual data and optional structures), stores
2277 * the optional hardware time stamping information (if non NULL) or
2278 * generates a software time stamp (otherwise), then queues the clone
2279 * to the error queue of the socket. Errors are silently ignored.
2280 */
2281extern void skb_tstamp_tx(struct sk_buff *orig_skb,
2282 struct skb_shared_hwtstamps *hwtstamps);
2283
Richard Cochran4507a712010-07-17 08:48:28 +00002284static inline void sw_tx_timestamp(struct sk_buff *skb)
2285{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002286 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2287 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00002288 skb_tstamp_tx(skb, NULL);
2289}
2290
2291/**
2292 * skb_tx_timestamp() - Driver hook for transmit timestamping
2293 *
2294 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00002295 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00002296 *
2297 * @skb: A socket buffer.
2298 */
2299static inline void skb_tx_timestamp(struct sk_buff *skb)
2300{
Richard Cochranc1f19b52010-07-17 08:49:36 +00002301 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002302 sw_tx_timestamp(skb);
2303}
2304
Johannes Berg6e3e9392011-11-09 10:15:42 +01002305/**
2306 * skb_complete_wifi_ack - deliver skb with wifi status
2307 *
2308 * @skb: the original outgoing packet
2309 * @acked: ack status
2310 *
2311 */
2312void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2313
Herbert Xu759e5d02007-03-25 20:10:56 -07002314extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
Al Virob51655b2006-11-14 21:40:42 -08002315extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002316
Herbert Xu60476372007-04-09 11:59:39 -07002317static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2318{
2319 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2320}
2321
Herbert Xufb286bb2005-11-10 13:01:24 -08002322/**
2323 * skb_checksum_complete - Calculate checksum of an entire packet
2324 * @skb: packet to process
2325 *
2326 * This function calculates the checksum over the entire packet plus
2327 * the value of skb->csum. The latter can be used to supply the
2328 * checksum of a pseudo header as used by TCP/UDP. It returns the
2329 * checksum.
2330 *
2331 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2332 * this function can be used to verify that checksum on received
2333 * packets. In that case the function should return zero if the
2334 * checksum is correct. In particular, this function will return zero
2335 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2336 * hardware has already verified the correctness of the checksum.
2337 */
Al Viro4381ca32007-07-15 21:00:11 +01002338static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002339{
Herbert Xu60476372007-04-09 11:59:39 -07002340 return skb_csum_unnecessary(skb) ?
2341 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002342}
2343
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002344#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002345extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002346static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2347{
2348 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002349 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002350}
2351static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2352{
2353 if (nfct)
2354 atomic_inc(&nfct->use);
2355}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002356#endif
2357#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -08002358static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
2359{
2360 if (skb)
2361 atomic_inc(&skb->users);
2362}
2363static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
2364{
2365 if (skb)
2366 kfree_skb(skb);
2367}
2368#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002369#ifdef CONFIG_BRIDGE_NETFILTER
2370static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2371{
2372 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2373 kfree(nf_bridge);
2374}
2375static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2376{
2377 if (nf_bridge)
2378 atomic_inc(&nf_bridge->use);
2379}
2380#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08002381static inline void nf_reset(struct sk_buff *skb)
2382{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002383#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08002384 nf_conntrack_put(skb->nfct);
2385 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002386#endif
2387#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Patrick McHardya193a4a2006-03-20 19:23:05 -08002388 nf_conntrack_put_reasm(skb->nfct_reasm);
2389 skb->nfct_reasm = NULL;
2390#endif
2391#ifdef CONFIG_BRIDGE_NETFILTER
2392 nf_bridge_put(skb->nf_bridge);
2393 skb->nf_bridge = NULL;
2394#endif
2395}
2396
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002397/* Note: This doesn't put any conntrack and bridge info in dst. */
2398static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2399{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002400#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002401 dst->nfct = src->nfct;
2402 nf_conntrack_get(src->nfct);
2403 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002404#endif
2405#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002406 dst->nfct_reasm = src->nfct_reasm;
2407 nf_conntrack_get_reasm(src->nfct_reasm);
2408#endif
2409#ifdef CONFIG_BRIDGE_NETFILTER
2410 dst->nf_bridge = src->nf_bridge;
2411 nf_bridge_get(src->nf_bridge);
2412#endif
2413}
2414
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002415static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2416{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002417#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002418 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002419#endif
2420#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002421 nf_conntrack_put_reasm(dst->nfct_reasm);
2422#endif
2423#ifdef CONFIG_BRIDGE_NETFILTER
2424 nf_bridge_put(dst->nf_bridge);
2425#endif
2426 __nf_copy(dst, src);
2427}
2428
James Morris984bc162006-06-09 00:29:17 -07002429#ifdef CONFIG_NETWORK_SECMARK
2430static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2431{
2432 to->secmark = from->secmark;
2433}
2434
2435static inline void skb_init_secmark(struct sk_buff *skb)
2436{
2437 skb->secmark = 0;
2438}
2439#else
2440static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2441{ }
2442
2443static inline void skb_init_secmark(struct sk_buff *skb)
2444{ }
2445#endif
2446
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002447static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2448{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002449 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002450}
2451
Stephen Hemminger92477442009-03-21 13:39:26 -07002452static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002453{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002454 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002455}
2456
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002457static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2458{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002459 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002460}
2461
David S. Millerd5a9e242009-01-27 16:22:11 -08002462static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2463{
2464 skb->queue_mapping = rx_queue + 1;
2465}
2466
Stephen Hemminger92477442009-03-21 13:39:26 -07002467static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002468{
2469 return skb->queue_mapping - 1;
2470}
2471
Stephen Hemminger92477442009-03-21 13:39:26 -07002472static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002473{
Eric Dumazeta02cec22010-09-22 20:43:57 +00002474 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08002475}
2476
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00002477extern u16 __skb_tx_hash(const struct net_device *dev,
2478 const struct sk_buff *skb,
2479 unsigned int num_tx_queues);
Stephen Hemminger92477442009-03-21 13:39:26 -07002480
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002481#ifdef CONFIG_XFRM
2482static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2483{
2484 return skb->sp;
2485}
2486#else
2487static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2488{
2489 return NULL;
2490}
2491#endif
2492
David S. Millerbdcc0922012-03-07 20:53:36 -05002493static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07002494{
2495 return skb_shinfo(skb)->gso_size;
2496}
2497
David S. Millerbdcc0922012-03-07 20:53:36 -05002498static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02002499{
2500 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2501}
2502
Ben Hutchings4497b072008-06-19 16:22:28 -07002503extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2504
2505static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2506{
2507 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2508 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002509 const struct skb_shared_info *shinfo = skb_shinfo(skb);
2510
Alexander Duyckb78462e2010-06-02 12:24:37 +00002511 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2512 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07002513 __skb_warn_lro_forwarding(skb);
2514 return true;
2515 }
2516 return false;
2517}
2518
Herbert Xu35fc92a2007-03-26 23:22:20 -07002519static inline void skb_forward_csum(struct sk_buff *skb)
2520{
2521 /* Unfortunately we don't support this one. Any brave souls? */
2522 if (skb->ip_summed == CHECKSUM_COMPLETE)
2523 skb->ip_summed = CHECKSUM_NONE;
2524}
2525
Eric Dumazetbc8acf22010-09-02 13:07:41 -07002526/**
2527 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2528 * @skb: skb to check
2529 *
2530 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2531 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2532 * use this helper, to document places where we make this assertion.
2533 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002534static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07002535{
2536#ifdef DEBUG
2537 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2538#endif
2539}
2540
Rusty Russellf35d9d82008-02-04 23:49:54 -05002541bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00002542
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002543static inline bool skb_is_recycleable(const struct sk_buff *skb, int skb_size)
Andy Fleming3d153a72011-10-13 04:33:54 +00002544{
2545 if (irqs_disabled())
2546 return false;
2547
2548 if (skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)
2549 return false;
2550
2551 if (skb_is_nonlinear(skb) || skb->fclone != SKB_FCLONE_UNAVAILABLE)
2552 return false;
2553
2554 skb_size = SKB_DATA_ALIGN(skb_size + NET_SKB_PAD);
2555 if (skb_end_pointer(skb) - skb->head < skb_size)
2556 return false;
2557
2558 if (skb_shared(skb) || skb_cloned(skb))
2559 return false;
2560
2561 return true;
2562}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002563#endif /* __KERNEL__ */
2564#endif /* _LINUX_SKBUFF_H */