blob: fe37c21d3a60fb4b8e21c1eed48766e0cc92dedd [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * Definitions for the 'struct sk_buff' memory handlers.
3 *
4 * Authors:
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14#ifndef _LINUX_SKBUFF_H
15#define _LINUX_SKBUFF_H
16
Linus Torvalds1da177e2005-04-16 15:20:36 -070017#include <linux/kernel.h>
Vegard Nossumfe55f6d2008-08-30 12:16:35 +020018#include <linux/kmemcheck.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070019#include <linux/compiler.h>
20#include <linux/time.h>
Paul Gortmaker187f1882011-11-23 20:12:59 -050021#include <linux/bug.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070022#include <linux/cache.h>
23
Arun Sharma600634972011-07-26 16:09:06 -070024#include <linux/atomic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <asm/types.h>
26#include <linux/spinlock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <linux/net.h>
Thomas Graf3fc7e8a2005-06-23 21:00:17 -070028#include <linux/textsearch.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070029#include <net/checksum.h>
Al Viroa80958f2006-12-04 20:41:19 +000030#include <linux/rcupdate.h>
Chris Leech97fc2f02006-05-23 17:55:33 -070031#include <linux/dmaengine.h>
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -070032#include <linux/hrtimer.h>
Ian Campbell131ea662011-08-19 06:25:00 +000033#include <linux/dma-mapping.h>
Michał Mirosławc8f44af2011-11-15 15:29:55 +000034#include <linux/netdev_features.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070035
Herbert Xu60476372007-04-09 11:59:39 -070036/* Don't change this without changing skb_csum_unnecessary! */
Linus Torvalds1da177e2005-04-16 15:20:36 -070037#define CHECKSUM_NONE 0
Herbert Xu60476372007-04-09 11:59:39 -070038#define CHECKSUM_UNNECESSARY 1
39#define CHECKSUM_COMPLETE 2
40#define CHECKSUM_PARTIAL 3
Linus Torvalds1da177e2005-04-16 15:20:36 -070041
42#define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
43 ~(SMP_CACHE_BYTES - 1))
David S. Millerfc910a22007-03-25 20:27:59 -070044#define SKB_WITH_OVERHEAD(X) \
Herbert Xudeea84b2007-10-21 16:27:46 -070045 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
David S. Millerfc910a22007-03-25 20:27:59 -070046#define SKB_MAX_ORDER(X, ORDER) \
47 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
Linus Torvalds1da177e2005-04-16 15:20:36 -070048#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
49#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
50
Eric Dumazet87fb4b72011-10-13 07:28:54 +000051/* return minimum truesize of one skb containing X bytes of data */
52#define SKB_TRUESIZE(X) ((X) + \
53 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
54 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
55
Linus Torvalds1da177e2005-04-16 15:20:36 -070056/* A. Checksumming of received packets by device.
57 *
58 * NONE: device failed to checksum this packet.
59 * skb->csum is undefined.
60 *
61 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
62 * skb->csum is undefined.
63 * It is bad option, but, unfortunately, many of vendors do this.
64 * Apparently with secret goal to sell you new device, when you
65 * will add new protocol to your host. F.e. IPv6. 8)
66 *
Patrick McHardy84fa7932006-08-29 16:44:56 -070067 * COMPLETE: the most generic way. Device supplied checksum of _all_
Linus Torvalds1da177e2005-04-16 15:20:36 -070068 * the packet as seen by netif_rx in skb->csum.
69 * NOTE: Even if device supports only some protocols, but
Patrick McHardy84fa7932006-08-29 16:44:56 -070070 * is able to produce some skb->csum, it MUST use COMPLETE,
Linus Torvalds1da177e2005-04-16 15:20:36 -070071 * not UNNECESSARY.
72 *
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070073 * PARTIAL: identical to the case for output below. This may occur
74 * on a packet received directly from another Linux OS, e.g.,
75 * a virtualised Linux kernel on the same host. The packet can
76 * be treated in the same way as UNNECESSARY except that on
77 * output (i.e., forwarding) the checksum must be filled in
78 * by the OS or the hardware.
79 *
Linus Torvalds1da177e2005-04-16 15:20:36 -070080 * B. Checksumming on output.
81 *
82 * NONE: skb is checksummed by protocol or csum is not required.
83 *
Patrick McHardy84fa7932006-08-29 16:44:56 -070084 * PARTIAL: device is required to csum packet as seen by hard_start_xmit
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070085 * from skb->csum_start to the end and to record the checksum
86 * at skb->csum_start + skb->csum_offset.
Linus Torvalds1da177e2005-04-16 15:20:36 -070087 *
88 * Device must show its capabilities in dev->features, set
89 * at device setup time.
90 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
91 * everything.
Linus Torvalds1da177e2005-04-16 15:20:36 -070092 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
93 * TCP/UDP over IPv4. Sigh. Vendors like this
94 * way by an unknown reason. Though, see comment above
95 * about CHECKSUM_UNNECESSARY. 8)
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070096 * NETIF_F_IPV6_CSUM about as dumb as the last one but does IPv6 instead.
Linus Torvalds1da177e2005-04-16 15:20:36 -070097 *
Yi Zou3af79302012-03-19 11:12:41 +000098 * UNNECESSARY: device will do per protocol specific csum. Protocol drivers
99 * that do not want net to perform the checksum calculation should use
100 * this flag in their outgoing skbs.
101 * NETIF_F_FCOE_CRC this indicates the device can do FCoE FC CRC
102 * offload. Correspondingly, the FCoE protocol driver
103 * stack should use CHECKSUM_UNNECESSARY.
104 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105 * Any questions? No questions, good. --ANK
106 */
107
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108struct net_device;
David Howells716ea3a2007-04-02 20:19:53 -0700109struct scatterlist;
Jens Axboe9c55e012007-11-06 23:30:13 -0800110struct pipe_inode_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700112#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113struct nf_conntrack {
114 atomic_t use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115};
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700116#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117
118#ifdef CONFIG_BRIDGE_NETFILTER
119struct nf_bridge_info {
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000120 atomic_t use;
121 unsigned int mask;
122 struct net_device *physindev;
123 struct net_device *physoutdev;
124 unsigned long data[32 / sizeof(unsigned long)];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125};
126#endif
127
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128struct sk_buff_head {
129 /* These two members must be first. */
130 struct sk_buff *next;
131 struct sk_buff *prev;
132
133 __u32 qlen;
134 spinlock_t lock;
135};
136
137struct sk_buff;
138
Ian Campbell9d4dde52011-12-22 23:39:14 +0000139/* To allow 64K frame to be packed as single skb without frag_list we
140 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
141 * buffers which do not start on a page boundary.
142 *
143 * Since GRO uses frags we allocate at least 16 regardless of page
144 * size.
Anton Blancharda715dea2011-03-27 14:57:26 +0000145 */
Ian Campbell9d4dde52011-12-22 23:39:14 +0000146#if (65536/PAGE_SIZE + 1) < 16
David S. Millereec00952011-03-29 23:34:08 -0700147#define MAX_SKB_FRAGS 16UL
Anton Blancharda715dea2011-03-27 14:57:26 +0000148#else
Ian Campbell9d4dde52011-12-22 23:39:14 +0000149#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
Anton Blancharda715dea2011-03-27 14:57:26 +0000150#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151
152typedef struct skb_frag_struct skb_frag_t;
153
154struct skb_frag_struct {
Ian Campbella8605c62011-10-19 23:01:49 +0000155 struct {
156 struct page *p;
157 } page;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000158#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
David S. Millera309bb02007-07-30 18:47:03 -0700159 __u32 page_offset;
160 __u32 size;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000161#else
162 __u16 page_offset;
163 __u16 size;
164#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165};
166
Eric Dumazet9e903e02011-10-18 21:00:24 +0000167static inline unsigned int skb_frag_size(const skb_frag_t *frag)
168{
169 return frag->size;
170}
171
172static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
173{
174 frag->size = size;
175}
176
177static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
178{
179 frag->size += delta;
180}
181
182static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
183{
184 frag->size -= delta;
185}
186
Patrick Ohlyac45f602009-02-12 05:03:37 +0000187#define HAVE_HW_TIME_STAMP
188
189/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800190 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000191 * @hwtstamp: hardware time stamp transformed into duration
192 * since arbitrary point in time
193 * @syststamp: hwtstamp transformed to system time base
194 *
195 * Software time stamps generated by ktime_get_real() are stored in
196 * skb->tstamp. The relation between the different kinds of time
197 * stamps is as follows:
198 *
199 * syststamp and tstamp can be compared against each other in
200 * arbitrary combinations. The accuracy of a
201 * syststamp/tstamp/"syststamp from other device" comparison is
202 * limited by the accuracy of the transformation into system time
203 * base. This depends on the device driver and its underlying
204 * hardware.
205 *
206 * hwtstamps can only be compared against other hwtstamps from
207 * the same device.
208 *
209 * This structure is attached to packets as part of the
210 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
211 */
212struct skb_shared_hwtstamps {
213 ktime_t hwtstamp;
214 ktime_t syststamp;
215};
216
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000217/* Definitions for tx_flags in struct skb_shared_info */
218enum {
219 /* generate hardware time stamp */
220 SKBTX_HW_TSTAMP = 1 << 0,
221
222 /* generate software time stamp */
223 SKBTX_SW_TSTAMP = 1 << 1,
224
225 /* device driver is going to provide hardware time stamp */
226 SKBTX_IN_PROGRESS = 1 << 2,
227
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;
Eric Dumazetd3836f22012-04-27 00:33:38 +0000473 __u8 head_frag:1;
474 /* 8/10 bit hole (depending on ndisc_nodetype presence) */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200475 kmemcheck_bitfield_end(flags2);
476
Chris Leech97fc2f02006-05-23 17:55:33 -0700477#ifdef CONFIG_NET_DMA
478 dma_cookie_t dma_cookie;
479#endif
James Morris984bc162006-06-09 00:29:17 -0700480#ifdef CONFIG_NETWORK_SECMARK
481 __u32 secmark;
482#endif
Neil Horman3b885782009-10-12 13:26:31 -0700483 union {
484 __u32 mark;
485 __u32 dropcount;
Eric Dumazeta21d4572012-04-10 20:30:48 +0000486 __u32 avail_size;
Neil Horman3b885782009-10-12 13:26:31 -0700487 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700489 sk_buff_data_t transport_header;
490 sk_buff_data_t network_header;
491 sk_buff_data_t mac_header;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700493 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700494 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700496 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700497 unsigned int truesize;
498 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499};
500
501#ifdef __KERNEL__
502/*
503 * Handling routines are only of interest to the kernel
504 */
505#include <linux/slab.h>
506
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507
Eric Dumazet7fee2262010-05-11 23:19:48 +0000508/*
509 * skb might have a dst pointer attached, refcounted or not.
510 * _skb_refdst low order bit is set if refcount was _not_ taken
511 */
512#define SKB_DST_NOREF 1UL
513#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
514
515/**
516 * skb_dst - returns skb dst_entry
517 * @skb: buffer
518 *
519 * Returns skb dst_entry, regardless of reference taken or not.
520 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000521static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
522{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000523 /* If refdst was not refcounted, check we still are in a
524 * rcu_read_lock section
525 */
526 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
527 !rcu_read_lock_held() &&
528 !rcu_read_lock_bh_held());
529 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000530}
531
Eric Dumazet7fee2262010-05-11 23:19:48 +0000532/**
533 * skb_dst_set - sets skb dst
534 * @skb: buffer
535 * @dst: dst entry
536 *
537 * Sets skb dst, assuming a reference was taken on dst and should
538 * be released by skb_dst_drop()
539 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000540static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
541{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000542 skb->_skb_refdst = (unsigned long)dst;
543}
544
Eric Dumazet27b75c92010-10-15 05:44:11 +0000545extern void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst);
Eric Dumazet7fee2262010-05-11 23:19:48 +0000546
547/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300548 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000549 * @skb: buffer
550 */
551static inline bool skb_dst_is_noref(const struct sk_buff *skb)
552{
553 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000554}
555
Eric Dumazet511c3f92009-06-02 05:14:27 +0000556static inline struct rtable *skb_rtable(const struct sk_buff *skb)
557{
Eric Dumazetadf30902009-06-02 05:19:30 +0000558 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000559}
560
Jörn Engel231d06a2006-03-20 21:28:35 -0800561extern void kfree_skb(struct sk_buff *skb);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000562extern void consume_skb(struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700563extern void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000564extern struct kmem_cache *skbuff_head_cache;
David S. Millerd179cd12005-08-17 14:57:30 -0700565extern struct sk_buff *__alloc_skb(unsigned int size,
Christoph Hellwigb30973f2006-12-06 20:32:36 -0800566 gfp_t priority, int fclone, int node);
Eric Dumazetd3836f22012-04-27 00:33:38 +0000567extern struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700568static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100569 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700570{
Eric Dumazet564824b2010-10-11 19:05:25 +0000571 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700572}
573
574static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100575 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700576{
Eric Dumazet564824b2010-10-11 19:05:25 +0000577 return __alloc_skb(size, priority, 1, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700578}
579
Andy Fleming3d153a72011-10-13 04:33:54 +0000580extern void skb_recycle(struct sk_buff *skb);
Changli Gao5b0daa32010-05-29 00:12:13 -0700581extern bool skb_recycle_check(struct sk_buff *skb, int skb_size);
Lennert Buytenhek04a4bb52008-10-01 02:33:12 -0700582
Herbert Xue0053ec2007-10-14 00:37:52 -0700583extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
Michael S. Tsirkin48c83012011-08-31 08:03:29 +0000584extern int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700585extern struct sk_buff *skb_clone(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100586 gfp_t priority);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700587extern struct sk_buff *skb_copy(const struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100588 gfp_t priority);
Eric Dumazet117632e2011-12-03 21:39:53 +0000589extern struct sk_buff *__pskb_copy(struct sk_buff *skb,
590 int headroom, gfp_t gfp_mask);
591
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592extern int pskb_expand_head(struct sk_buff *skb,
Victor Fusco86a76ca2005-07-08 14:57:47 -0700593 int nhead, int ntail,
Al Virodd0fc662005-10-07 07:46:04 +0100594 gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
596 unsigned int headroom);
597extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
598 int newheadroom, int newtailroom,
Al Virodd0fc662005-10-07 07:46:04 +0100599 gfp_t priority);
David Howells716ea3a2007-04-02 20:19:53 -0700600extern int skb_to_sgvec(struct sk_buff *skb,
601 struct scatterlist *sg, int offset,
602 int len);
603extern int skb_cow_data(struct sk_buff *skb, int tailbits,
604 struct sk_buff **trailer);
Herbert Xu5b057c62006-06-23 02:06:41 -0700605extern int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000606#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700608extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
609 int getfrag(void *from, char *to, int offset,
610 int len,int odd, struct sk_buff *skb),
611 void *from, int length);
612
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800613struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700614 __u32 lower_offset;
615 __u32 upper_offset;
616 __u32 frag_idx;
617 __u32 stepped_offset;
618 struct sk_buff *root_skb;
619 struct sk_buff *cur_skb;
620 __u8 *frag_data;
621};
622
623extern void skb_prepare_seq_read(struct sk_buff *skb,
624 unsigned int from, unsigned int to,
625 struct skb_seq_state *st);
626extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
627 struct skb_seq_state *st);
628extern void skb_abort_seq_read(struct skb_seq_state *st);
629
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700630extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
631 unsigned int to, struct ts_config *config,
632 struct ts_state *state);
633
Tom Herbertbdeab992011-08-14 19:45:55 +0000634extern void __skb_get_rxhash(struct sk_buff *skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000635static inline __u32 skb_get_rxhash(struct sk_buff *skb)
636{
637 if (!skb->rxhash)
Tom Herbertbdeab992011-08-14 19:45:55 +0000638 __skb_get_rxhash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000639
640 return skb->rxhash;
641}
642
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700643#ifdef NET_SKBUFF_DATA_USES_OFFSET
644static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
645{
646 return skb->head + skb->end;
647}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000648
649static inline unsigned int skb_end_offset(const struct sk_buff *skb)
650{
651 return skb->end;
652}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700653#else
654static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
655{
656 return skb->end;
657}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000658
659static inline unsigned int skb_end_offset(const struct sk_buff *skb)
660{
661 return skb->end - skb->head;
662}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700663#endif
664
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700666#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700667
Patrick Ohlyac45f602009-02-12 05:03:37 +0000668static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
669{
670 return &skb_shinfo(skb)->hwtstamps;
671}
672
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673/**
674 * skb_queue_empty - check if a queue is empty
675 * @list: queue head
676 *
677 * Returns true if the queue is empty, false otherwise.
678 */
679static inline int skb_queue_empty(const struct sk_buff_head *list)
680{
681 return list->next == (struct sk_buff *)list;
682}
683
684/**
David S. Millerfc7ebb22008-09-23 00:34:07 -0700685 * skb_queue_is_last - check if skb is the last entry in the queue
686 * @list: queue head
687 * @skb: buffer
688 *
689 * Returns true if @skb is the last buffer on the list.
690 */
691static inline bool skb_queue_is_last(const struct sk_buff_head *list,
692 const struct sk_buff *skb)
693{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000694 return skb->next == (struct sk_buff *)list;
David S. Millerfc7ebb22008-09-23 00:34:07 -0700695}
696
697/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800698 * skb_queue_is_first - check if skb is the first entry in the queue
699 * @list: queue head
700 * @skb: buffer
701 *
702 * Returns true if @skb is the first buffer on the list.
703 */
704static inline bool skb_queue_is_first(const struct sk_buff_head *list,
705 const struct sk_buff *skb)
706{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000707 return skb->prev == (struct sk_buff *)list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800708}
709
710/**
David S. Miller249c8b42008-09-23 00:44:42 -0700711 * skb_queue_next - return the next packet in the queue
712 * @list: queue head
713 * @skb: current buffer
714 *
715 * Return the next packet in @list after @skb. It is only valid to
716 * call this if skb_queue_is_last() evaluates to false.
717 */
718static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
719 const struct sk_buff *skb)
720{
721 /* This BUG_ON may seem severe, but if we just return then we
722 * are going to dereference garbage.
723 */
724 BUG_ON(skb_queue_is_last(list, skb));
725 return skb->next;
726}
727
728/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800729 * skb_queue_prev - return the prev packet in the queue
730 * @list: queue head
731 * @skb: current buffer
732 *
733 * Return the prev packet in @list before @skb. It is only valid to
734 * call this if skb_queue_is_first() evaluates to false.
735 */
736static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
737 const struct sk_buff *skb)
738{
739 /* This BUG_ON may seem severe, but if we just return then we
740 * are going to dereference garbage.
741 */
742 BUG_ON(skb_queue_is_first(list, skb));
743 return skb->prev;
744}
745
746/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747 * skb_get - reference buffer
748 * @skb: buffer to reference
749 *
750 * Makes another reference to a socket buffer and returns a pointer
751 * to the buffer.
752 */
753static inline struct sk_buff *skb_get(struct sk_buff *skb)
754{
755 atomic_inc(&skb->users);
756 return skb;
757}
758
759/*
760 * If users == 1, we are the only owner and are can avoid redundant
761 * atomic change.
762 */
763
764/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765 * skb_cloned - is the buffer a clone
766 * @skb: buffer to check
767 *
768 * Returns true if the buffer was generated with skb_clone() and is
769 * one of multiple shared copies of the buffer. Cloned buffers are
770 * shared data so must not be written to under normal circumstances.
771 */
772static inline int skb_cloned(const struct sk_buff *skb)
773{
774 return skb->cloned &&
775 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
776}
777
778/**
779 * skb_header_cloned - is the header a clone
780 * @skb: buffer to check
781 *
782 * Returns true if modifying the header part of the buffer requires
783 * the data to be copied.
784 */
785static inline int skb_header_cloned(const struct sk_buff *skb)
786{
787 int dataref;
788
789 if (!skb->cloned)
790 return 0;
791
792 dataref = atomic_read(&skb_shinfo(skb)->dataref);
793 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
794 return dataref != 1;
795}
796
797/**
798 * skb_header_release - release reference to header
799 * @skb: buffer to operate on
800 *
801 * Drop a reference to the header part of the buffer. This is done
802 * by acquiring a payload reference. You must not read from the header
803 * part of skb->data after this.
804 */
805static inline void skb_header_release(struct sk_buff *skb)
806{
807 BUG_ON(skb->nohdr);
808 skb->nohdr = 1;
809 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
810}
811
812/**
813 * skb_shared - is the buffer shared
814 * @skb: buffer to check
815 *
816 * Returns true if more than one person has a reference to this
817 * buffer.
818 */
819static inline int skb_shared(const struct sk_buff *skb)
820{
821 return atomic_read(&skb->users) != 1;
822}
823
824/**
825 * skb_share_check - check if buffer is shared and if so clone it
826 * @skb: buffer to check
827 * @pri: priority for memory allocation
828 *
829 * If the buffer is shared the buffer is cloned and the old copy
830 * drops a reference. A new clone with a single reference is returned.
831 * If the buffer is not shared the original buffer is returned. When
832 * being called from interrupt status or with spinlocks held pri must
833 * be GFP_ATOMIC.
834 *
835 * NULL is returned on a memory allocation failure.
836 */
Victor Fusco86a76ca2005-07-08 14:57:47 -0700837static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100838 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839{
840 might_sleep_if(pri & __GFP_WAIT);
841 if (skb_shared(skb)) {
842 struct sk_buff *nskb = skb_clone(skb, pri);
843 kfree_skb(skb);
844 skb = nskb;
845 }
846 return skb;
847}
848
849/*
850 * Copy shared buffers into a new sk_buff. We effectively do COW on
851 * packets to handle cases where we have a local reader and forward
852 * and a couple of other messy ones. The normal one is tcpdumping
853 * a packet thats being forwarded.
854 */
855
856/**
857 * skb_unshare - make a copy of a shared buffer
858 * @skb: buffer to check
859 * @pri: priority for memory allocation
860 *
861 * If the socket buffer is a clone then this function creates a new
862 * copy of the data, drops a reference count on the old copy and returns
863 * the new copy with the reference count at 1. If the buffer is not a clone
864 * the original buffer is returned. When called with a spinlock held or
865 * from interrupt state @pri must be %GFP_ATOMIC
866 *
867 * %NULL is returned on a memory allocation failure.
868 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -0700869static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100870 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871{
872 might_sleep_if(pri & __GFP_WAIT);
873 if (skb_cloned(skb)) {
874 struct sk_buff *nskb = skb_copy(skb, pri);
875 kfree_skb(skb); /* Free our shared copy */
876 skb = nskb;
877 }
878 return skb;
879}
880
881/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -0800882 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883 * @list_: list to peek at
884 *
885 * Peek an &sk_buff. Unlike most other operations you _MUST_
886 * be careful with this one. A peek leaves the buffer on the
887 * list and someone else may run off with it. You must hold
888 * the appropriate locks or have a private queue to do this.
889 *
890 * Returns %NULL for an empty list or a pointer to the head element.
891 * The reference count is not incremented and the reference is therefore
892 * volatile. Use with caution.
893 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -0400894static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895{
Eric Dumazet18d07002012-04-30 16:31:46 +0000896 struct sk_buff *skb = list_->next;
897
898 if (skb == (struct sk_buff *)list_)
899 skb = NULL;
900 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700901}
902
903/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +0000904 * skb_peek_next - peek skb following the given one from a queue
905 * @skb: skb to start from
906 * @list_: list to peek at
907 *
908 * Returns %NULL when the end of the list is met or a pointer to the
909 * next element. The reference count is not incremented and the
910 * reference is therefore volatile. Use with caution.
911 */
912static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
913 const struct sk_buff_head *list_)
914{
915 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +0000916
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +0000917 if (next == (struct sk_buff *)list_)
918 next = NULL;
919 return next;
920}
921
922/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -0800923 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -0700924 * @list_: list to peek at
925 *
926 * Peek an &sk_buff. Unlike most other operations you _MUST_
927 * be careful with this one. A peek leaves the buffer on the
928 * list and someone else may run off with it. You must hold
929 * the appropriate locks or have a private queue to do this.
930 *
931 * Returns %NULL for an empty list or a pointer to the tail element.
932 * The reference count is not incremented and the reference is therefore
933 * volatile. Use with caution.
934 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -0400935static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700936{
Eric Dumazet18d07002012-04-30 16:31:46 +0000937 struct sk_buff *skb = list_->prev;
938
939 if (skb == (struct sk_buff *)list_)
940 skb = NULL;
941 return skb;
942
Linus Torvalds1da177e2005-04-16 15:20:36 -0700943}
944
945/**
946 * skb_queue_len - get queue length
947 * @list_: list to measure
948 *
949 * Return the length of an &sk_buff queue.
950 */
951static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
952{
953 return list_->qlen;
954}
955
David S. Miller67fed452008-09-21 22:36:24 -0700956/**
957 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
958 * @list: queue to initialize
959 *
960 * This initializes only the list and queue length aspects of
961 * an sk_buff_head object. This allows to initialize the list
962 * aspects of an sk_buff_head without reinitializing things like
963 * the spinlock. It can also be used for on-stack sk_buff_head
964 * objects where the spinlock is known to not be used.
965 */
966static inline void __skb_queue_head_init(struct sk_buff_head *list)
967{
968 list->prev = list->next = (struct sk_buff *)list;
969 list->qlen = 0;
970}
971
Arjan van de Ven76f10ad2006-08-02 14:06:55 -0700972/*
973 * This function creates a split out lock class for each invocation;
974 * this is needed for now since a whole lot of users of the skb-queue
975 * infrastructure in drivers have different locking usage (in hardirq)
976 * than the networking core (in softirq only). In the long run either the
977 * network layer or drivers should need annotation to consolidate the
978 * main types of usage into 3 classes.
979 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700980static inline void skb_queue_head_init(struct sk_buff_head *list)
981{
982 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -0700983 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984}
985
Pavel Emelianovc2ecba72007-04-17 12:45:31 -0700986static inline void skb_queue_head_init_class(struct sk_buff_head *list,
987 struct lock_class_key *class)
988{
989 skb_queue_head_init(list);
990 lockdep_set_class(&list->lock, class);
991}
992
Linus Torvalds1da177e2005-04-16 15:20:36 -0700993/*
Gerrit Renkerbf299272008-04-14 00:04:51 -0700994 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995 *
996 * The "__skb_xxxx()" functions are the non-atomic ones that
997 * can only be called with interrupts disabled.
998 */
Gerrit Renkerbf299272008-04-14 00:04:51 -0700999extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
1000static inline void __skb_insert(struct sk_buff *newsk,
1001 struct sk_buff *prev, struct sk_buff *next,
1002 struct sk_buff_head *list)
1003{
1004 newsk->next = next;
1005 newsk->prev = prev;
1006 next->prev = prev->next = newsk;
1007 list->qlen++;
1008}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001009
David S. Miller67fed452008-09-21 22:36:24 -07001010static inline void __skb_queue_splice(const struct sk_buff_head *list,
1011 struct sk_buff *prev,
1012 struct sk_buff *next)
1013{
1014 struct sk_buff *first = list->next;
1015 struct sk_buff *last = list->prev;
1016
1017 first->prev = prev;
1018 prev->next = first;
1019
1020 last->next = next;
1021 next->prev = last;
1022}
1023
1024/**
1025 * skb_queue_splice - join two skb lists, this is designed for stacks
1026 * @list: the new list to add
1027 * @head: the place to add it in the first list
1028 */
1029static inline void skb_queue_splice(const 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 }
1036}
1037
1038/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001039 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001040 * @list: the new list to add
1041 * @head: the place to add it in the first list
1042 *
1043 * The list at @list is reinitialised
1044 */
1045static inline void skb_queue_splice_init(struct sk_buff_head *list,
1046 struct sk_buff_head *head)
1047{
1048 if (!skb_queue_empty(list)) {
1049 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001050 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001051 __skb_queue_head_init(list);
1052 }
1053}
1054
1055/**
1056 * skb_queue_splice_tail - join two skb lists, each list being a queue
1057 * @list: the new list to add
1058 * @head: the place to add it in the first list
1059 */
1060static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1061 struct sk_buff_head *head)
1062{
1063 if (!skb_queue_empty(list)) {
1064 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001065 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001066 }
1067}
1068
1069/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001070 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001071 * @list: the new list to add
1072 * @head: the place to add it in the first list
1073 *
1074 * Each of the lists is a queue.
1075 * The list at @list is reinitialised
1076 */
1077static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1078 struct sk_buff_head *head)
1079{
1080 if (!skb_queue_empty(list)) {
1081 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001082 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001083 __skb_queue_head_init(list);
1084 }
1085}
1086
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001088 * __skb_queue_after - queue a buffer at the list head
1089 * @list: list to use
1090 * @prev: place after this buffer
1091 * @newsk: buffer to queue
1092 *
1093 * Queue a buffer int the middle of a list. This function takes no locks
1094 * and you must therefore hold required locks before calling it.
1095 *
1096 * A buffer cannot be placed on two lists at the same time.
1097 */
1098static inline void __skb_queue_after(struct sk_buff_head *list,
1099 struct sk_buff *prev,
1100 struct sk_buff *newsk)
1101{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001102 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001103}
1104
Gerrit Renker7de6c032008-04-14 00:05:09 -07001105extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1106 struct sk_buff_head *list);
1107
Gerrit Renkerf5572852008-04-14 00:05:28 -07001108static inline void __skb_queue_before(struct sk_buff_head *list,
1109 struct sk_buff *next,
1110 struct sk_buff *newsk)
1111{
1112 __skb_insert(newsk, next->prev, next, list);
1113}
1114
Stephen Hemminger300ce172005-10-30 13:47:34 -08001115/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001116 * __skb_queue_head - queue a buffer at the list head
1117 * @list: list to use
1118 * @newsk: buffer to queue
1119 *
1120 * Queue a buffer at the start of a list. This function takes no locks
1121 * and you must therefore hold required locks before calling it.
1122 *
1123 * A buffer cannot be placed on two lists at the same time.
1124 */
1125extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1126static inline void __skb_queue_head(struct sk_buff_head *list,
1127 struct sk_buff *newsk)
1128{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001129 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130}
1131
1132/**
1133 * __skb_queue_tail - queue a buffer at the list tail
1134 * @list: list to use
1135 * @newsk: buffer to queue
1136 *
1137 * Queue a buffer at the end of a list. This function takes no locks
1138 * and you must therefore hold required locks before calling it.
1139 *
1140 * A buffer cannot be placed on two lists at the same time.
1141 */
1142extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1143static inline void __skb_queue_tail(struct sk_buff_head *list,
1144 struct sk_buff *newsk)
1145{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001146 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147}
1148
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001150 * remove sk_buff from list. _Must_ be called atomically, and with
1151 * the list known..
1152 */
David S. Miller8728b832005-08-09 19:25:21 -07001153extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001154static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1155{
1156 struct sk_buff *next, *prev;
1157
1158 list->qlen--;
1159 next = skb->next;
1160 prev = skb->prev;
1161 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162 next->prev = prev;
1163 prev->next = next;
1164}
1165
Gerrit Renkerf525c062008-04-14 00:04:12 -07001166/**
1167 * __skb_dequeue - remove from the head of the queue
1168 * @list: list to dequeue from
1169 *
1170 * Remove the head of the list. This function does not take any locks
1171 * so must be used with appropriate locks held only. The head item is
1172 * returned or %NULL if the list is empty.
1173 */
1174extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1175static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1176{
1177 struct sk_buff *skb = skb_peek(list);
1178 if (skb)
1179 __skb_unlink(skb, list);
1180 return skb;
1181}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182
1183/**
1184 * __skb_dequeue_tail - remove from the tail of the queue
1185 * @list: list to dequeue from
1186 *
1187 * Remove the tail of the list. This function does not take any locks
1188 * so must be used with appropriate locks held only. The tail item is
1189 * returned or %NULL if the list is empty.
1190 */
1191extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1192static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1193{
1194 struct sk_buff *skb = skb_peek_tail(list);
1195 if (skb)
1196 __skb_unlink(skb, list);
1197 return skb;
1198}
1199
1200
David S. Millerbdcc0922012-03-07 20:53:36 -05001201static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001202{
1203 return skb->data_len;
1204}
1205
1206static inline unsigned int skb_headlen(const struct sk_buff *skb)
1207{
1208 return skb->len - skb->data_len;
1209}
1210
1211static inline int skb_pagelen(const struct sk_buff *skb)
1212{
1213 int i, len = 0;
1214
1215 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001216 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001217 return len + skb_headlen(skb);
1218}
1219
Ian Campbell131ea662011-08-19 06:25:00 +00001220/**
1221 * __skb_fill_page_desc - initialise a paged fragment in an skb
1222 * @skb: buffer containing fragment to be initialised
1223 * @i: paged fragment index to initialise
1224 * @page: the page to use for this fragment
1225 * @off: the offset to the data with @page
1226 * @size: the length of the data
1227 *
1228 * Initialises the @i'th fragment of @skb to point to &size bytes at
1229 * offset @off within @page.
1230 *
1231 * Does not take any additional reference on the fragment.
1232 */
1233static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1234 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001235{
1236 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1237
Ian Campbella8605c62011-10-19 23:01:49 +00001238 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001239 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001240 skb_frag_size_set(frag, size);
Ian Campbell131ea662011-08-19 06:25:00 +00001241}
1242
1243/**
1244 * skb_fill_page_desc - initialise a paged fragment in an skb
1245 * @skb: buffer containing fragment to be initialised
1246 * @i: paged fragment index to initialise
1247 * @page: the page to use for this fragment
1248 * @off: the offset to the data with @page
1249 * @size: the length of the data
1250 *
1251 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1252 * @skb to point to &size bytes at offset @off within @page. In
1253 * addition updates @skb such that @i is the last fragment.
1254 *
1255 * Does not take any additional reference on the fragment.
1256 */
1257static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1258 struct page *page, int off, int size)
1259{
1260 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001261 skb_shinfo(skb)->nr_frags = i + 1;
1262}
1263
Peter Zijlstra654bed12008-10-07 14:22:33 -07001264extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
Eric Dumazet50269e12012-03-23 23:59:33 +00001265 int off, int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001266
Linus Torvalds1da177e2005-04-16 15:20:36 -07001267#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001268#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001269#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1270
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001271#ifdef NET_SKBUFF_DATA_USES_OFFSET
1272static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1273{
1274 return skb->head + skb->tail;
1275}
1276
1277static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1278{
1279 skb->tail = skb->data - skb->head;
1280}
1281
1282static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1283{
1284 skb_reset_tail_pointer(skb);
1285 skb->tail += offset;
1286}
1287#else /* NET_SKBUFF_DATA_USES_OFFSET */
1288static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1289{
1290 return skb->tail;
1291}
1292
1293static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1294{
1295 skb->tail = skb->data;
1296}
1297
1298static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1299{
1300 skb->tail = skb->data + offset;
1301}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001302
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001303#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1304
Linus Torvalds1da177e2005-04-16 15:20:36 -07001305/*
1306 * Add data to an sk_buff
1307 */
Ilpo Järvinen0dde3e12008-03-27 17:43:41 -07001308extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001309static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1310{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001311 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001312 SKB_LINEAR_ASSERT(skb);
1313 skb->tail += len;
1314 skb->len += len;
1315 return tmp;
1316}
1317
Ilpo Järvinenc2aa2702008-03-27 17:52:40 -07001318extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001319static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1320{
1321 skb->data -= len;
1322 skb->len += len;
1323 return skb->data;
1324}
1325
Ilpo Järvinen6be8ac22008-03-27 17:47:24 -07001326extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001327static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1328{
1329 skb->len -= len;
1330 BUG_ON(skb->len < skb->data_len);
1331 return skb->data += len;
1332}
1333
David S. Miller47d29642010-05-02 02:21:44 -07001334static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1335{
1336 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1337}
1338
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1340
1341static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1342{
1343 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001344 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345 return NULL;
1346 skb->len -= len;
1347 return skb->data += len;
1348}
1349
1350static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1351{
1352 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1353}
1354
1355static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1356{
1357 if (likely(len <= skb_headlen(skb)))
1358 return 1;
1359 if (unlikely(len > skb->len))
1360 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001361 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362}
1363
1364/**
1365 * skb_headroom - bytes at buffer head
1366 * @skb: buffer to check
1367 *
1368 * Return the number of bytes of free space at the head of an &sk_buff.
1369 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001370static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001371{
1372 return skb->data - skb->head;
1373}
1374
1375/**
1376 * skb_tailroom - bytes at buffer end
1377 * @skb: buffer to check
1378 *
1379 * Return the number of bytes of free space at the tail of an sk_buff
1380 */
1381static inline int skb_tailroom(const struct sk_buff *skb)
1382{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001383 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001384}
1385
1386/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00001387 * skb_availroom - bytes at buffer end
1388 * @skb: buffer to check
1389 *
1390 * Return the number of bytes of free space at the tail of an sk_buff
1391 * allocated by sk_stream_alloc()
1392 */
1393static inline int skb_availroom(const struct sk_buff *skb)
1394{
1395 return skb_is_nonlinear(skb) ? 0 : skb->avail_size - skb->len;
1396}
1397
1398/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399 * skb_reserve - adjust headroom
1400 * @skb: buffer to alter
1401 * @len: bytes to move
1402 *
1403 * Increase the headroom of an empty &sk_buff by reducing the tail
1404 * room. This is only allowed for an empty buffer.
1405 */
David S. Miller8243126c2006-01-17 02:54:21 -08001406static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407{
1408 skb->data += len;
1409 skb->tail += len;
1410}
1411
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00001412static inline void skb_reset_mac_len(struct sk_buff *skb)
1413{
1414 skb->mac_len = skb->network_header - skb->mac_header;
1415}
1416
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001417#ifdef NET_SKBUFF_DATA_USES_OFFSET
1418static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1419{
1420 return skb->head + skb->transport_header;
1421}
1422
1423static inline void skb_reset_transport_header(struct sk_buff *skb)
1424{
1425 skb->transport_header = skb->data - skb->head;
1426}
1427
1428static inline void skb_set_transport_header(struct sk_buff *skb,
1429 const int offset)
1430{
1431 skb_reset_transport_header(skb);
1432 skb->transport_header += offset;
1433}
1434
1435static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1436{
1437 return skb->head + skb->network_header;
1438}
1439
1440static inline void skb_reset_network_header(struct sk_buff *skb)
1441{
1442 skb->network_header = skb->data - skb->head;
1443}
1444
1445static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1446{
1447 skb_reset_network_header(skb);
1448 skb->network_header += offset;
1449}
1450
1451static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1452{
1453 return skb->head + skb->mac_header;
1454}
1455
1456static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1457{
1458 return skb->mac_header != ~0U;
1459}
1460
1461static inline void skb_reset_mac_header(struct sk_buff *skb)
1462{
1463 skb->mac_header = skb->data - skb->head;
1464}
1465
1466static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1467{
1468 skb_reset_mac_header(skb);
1469 skb->mac_header += offset;
1470}
1471
1472#else /* NET_SKBUFF_DATA_USES_OFFSET */
1473
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001474static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1475{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001476 return skb->transport_header;
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001477}
1478
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001479static inline void skb_reset_transport_header(struct sk_buff *skb)
1480{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001481 skb->transport_header = skb->data;
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001482}
1483
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001484static inline void skb_set_transport_header(struct sk_buff *skb,
1485 const int offset)
1486{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001487 skb->transport_header = skb->data + offset;
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001488}
1489
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001490static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1491{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001492 return skb->network_header;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001493}
1494
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001495static inline void skb_reset_network_header(struct sk_buff *skb)
1496{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001497 skb->network_header = skb->data;
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001498}
1499
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001500static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1501{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001502 skb->network_header = skb->data + offset;
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001503}
1504
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001505static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1506{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001507 return skb->mac_header;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001508}
1509
1510static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1511{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001512 return skb->mac_header != NULL;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001513}
1514
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001515static inline void skb_reset_mac_header(struct sk_buff *skb)
1516{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001517 skb->mac_header = skb->data;
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001518}
1519
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001520static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1521{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001522 skb->mac_header = skb->data + offset;
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001523}
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001524#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1525
Eric Dumazet03606892012-02-23 10:55:02 +00001526static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1527{
1528 if (skb_mac_header_was_set(skb)) {
1529 const unsigned char *old_mac = skb_mac_header(skb);
1530
1531 skb_set_mac_header(skb, -skb->mac_len);
1532 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1533 }
1534}
1535
Michał Mirosław04fb4512010-12-14 15:24:08 +00001536static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1537{
1538 return skb->csum_start - skb_headroom(skb);
1539}
1540
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001541static inline int skb_transport_offset(const struct sk_buff *skb)
1542{
1543 return skb_transport_header(skb) - skb->data;
1544}
1545
1546static inline u32 skb_network_header_len(const struct sk_buff *skb)
1547{
1548 return skb->transport_header - skb->network_header;
1549}
1550
1551static inline int skb_network_offset(const struct sk_buff *skb)
1552{
1553 return skb_network_header(skb) - skb->data;
1554}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001555
Changli Gaof9599ce2010-08-04 04:43:44 +00001556static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1557{
1558 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1559}
1560
Linus Torvalds1da177e2005-04-16 15:20:36 -07001561/*
1562 * CPUs often take a performance hit when accessing unaligned memory
1563 * locations. The actual performance hit varies, it can be small if the
1564 * hardware handles it or large if we have to take an exception and fix it
1565 * in software.
1566 *
1567 * Since an ethernet header is 14 bytes network drivers often end up with
1568 * the IP header at an unaligned offset. The IP header can be aligned by
1569 * shifting the start of the packet by 2 bytes. Drivers should do this
1570 * with:
1571 *
Tobias Klauser8660c122009-07-13 22:48:16 +00001572 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001573 *
1574 * The downside to this alignment of the IP header is that the DMA is now
1575 * unaligned. On some architectures the cost of an unaligned DMA is high
1576 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00001577 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001578 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1579 * to be overridden.
1580 */
1581#ifndef NET_IP_ALIGN
1582#define NET_IP_ALIGN 2
1583#endif
1584
Anton Blanchard025be812006-03-31 02:27:06 -08001585/*
1586 * The networking layer reserves some headroom in skb data (via
1587 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1588 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08001589 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08001590 *
1591 * Unfortunately this headroom changes the DMA alignment of the resulting
1592 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1593 * on some architectures. An architecture can override this value,
1594 * perhaps setting it to a cacheline in size (since that will maintain
1595 * cacheline alignment of the DMA). It must be a power of 2.
1596 *
David S. Millerd6301d32009-02-08 19:24:13 -08001597 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08001598 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07001599 *
1600 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1601 * to reduce average number of cache lines per packet.
1602 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07001603 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08001604 */
1605#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07001606#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08001607#endif
1608
Herbert Xu3cc0e872006-06-09 16:13:38 -07001609extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001610
1611static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1612{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00001613 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07001614 WARN_ON(1);
1615 return;
1616 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001617 skb->len = len;
1618 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001619}
1620
Ilpo Järvinen419ae742008-03-27 17:54:01 -07001621extern void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001622
1623static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1624{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001625 if (skb->data_len)
1626 return ___pskb_trim(skb, len);
1627 __skb_trim(skb, len);
1628 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001629}
1630
1631static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1632{
1633 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1634}
1635
1636/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07001637 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1638 * @skb: buffer to alter
1639 * @len: new length
1640 *
1641 * This is identical to pskb_trim except that the caller knows that
1642 * the skb is not cloned so we should never get an error due to out-
1643 * of-memory.
1644 */
1645static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1646{
1647 int err = pskb_trim(skb, len);
1648 BUG_ON(err);
1649}
1650
1651/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001652 * skb_orphan - orphan a buffer
1653 * @skb: buffer to orphan
1654 *
1655 * If a buffer currently has an owner then we call the owner's
1656 * destructor function and make the @skb unowned. The buffer continues
1657 * to exist but is no longer charged to its former owner.
1658 */
1659static inline void skb_orphan(struct sk_buff *skb)
1660{
1661 if (skb->destructor)
1662 skb->destructor(skb);
1663 skb->destructor = NULL;
1664 skb->sk = NULL;
1665}
1666
1667/**
1668 * __skb_queue_purge - empty a list
1669 * @list: list to empty
1670 *
1671 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1672 * the list and one reference dropped. This function does not take the
1673 * list lock and the caller must hold the relevant locks to use it.
1674 */
1675extern void skb_queue_purge(struct sk_buff_head *list);
1676static inline void __skb_queue_purge(struct sk_buff_head *list)
1677{
1678 struct sk_buff *skb;
1679 while ((skb = __skb_dequeue(list)) != NULL)
1680 kfree_skb(skb);
1681}
1682
Eric Dumazet6f532612012-05-18 05:12:12 +00001683extern void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001684
Christoph Hellwig8af27452006-07-31 22:35:23 -07001685extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00001686 unsigned int length,
1687 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07001688
1689/**
1690 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1691 * @dev: network device to receive on
1692 * @length: length to allocate
1693 *
1694 * Allocate a new &sk_buff and assign it a usage count of one. The
1695 * buffer has unspecified headroom built in. Users should allocate
1696 * the headroom they think they need without accounting for the
1697 * built in space. The built in space is used for optimisations.
1698 *
1699 * %NULL is returned if there is no free memory. Although this function
1700 * allocates memory it can be called from an interrupt.
1701 */
1702static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00001703 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07001704{
1705 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1706}
1707
Eric Dumazet6f532612012-05-18 05:12:12 +00001708/* legacy helper around __netdev_alloc_skb() */
1709static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1710 gfp_t gfp_mask)
1711{
1712 return __netdev_alloc_skb(NULL, length, gfp_mask);
1713}
1714
1715/* legacy helper around netdev_alloc_skb() */
1716static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1717{
1718 return netdev_alloc_skb(NULL, length);
1719}
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, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00001724{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001725 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00001726
1727 if (NET_IP_ALIGN && skb)
1728 skb_reserve(skb, NET_IP_ALIGN);
1729 return skb;
1730}
1731
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001732static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1733 unsigned int length)
1734{
1735 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
1736}
1737
Eric Dumazet564824b2010-10-11 19:05:25 +00001738/**
Ian Campbell131ea662011-08-19 06:25:00 +00001739 * skb_frag_page - retrieve the page refered to by a paged fragment
1740 * @frag: the paged fragment
1741 *
1742 * Returns the &struct page associated with @frag.
1743 */
1744static inline struct page *skb_frag_page(const skb_frag_t *frag)
1745{
Ian Campbella8605c62011-10-19 23:01:49 +00001746 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00001747}
1748
1749/**
1750 * __skb_frag_ref - take an addition reference on a paged fragment.
1751 * @frag: the paged fragment
1752 *
1753 * Takes an additional reference on the paged fragment @frag.
1754 */
1755static inline void __skb_frag_ref(skb_frag_t *frag)
1756{
1757 get_page(skb_frag_page(frag));
1758}
1759
1760/**
1761 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
1762 * @skb: the buffer
1763 * @f: the fragment offset.
1764 *
1765 * Takes an additional reference on the @f'th paged fragment of @skb.
1766 */
1767static inline void skb_frag_ref(struct sk_buff *skb, int f)
1768{
1769 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
1770}
1771
1772/**
1773 * __skb_frag_unref - release a reference on a paged fragment.
1774 * @frag: the paged fragment
1775 *
1776 * Releases a reference on the paged fragment @frag.
1777 */
1778static inline void __skb_frag_unref(skb_frag_t *frag)
1779{
1780 put_page(skb_frag_page(frag));
1781}
1782
1783/**
1784 * skb_frag_unref - release a reference on a paged fragment of an skb.
1785 * @skb: the buffer
1786 * @f: the fragment offset
1787 *
1788 * Releases a reference on the @f'th paged fragment of @skb.
1789 */
1790static inline void skb_frag_unref(struct sk_buff *skb, int f)
1791{
1792 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
1793}
1794
1795/**
1796 * skb_frag_address - gets the address of the data contained in a paged fragment
1797 * @frag: the paged fragment buffer
1798 *
1799 * Returns the address of the data within @frag. The page must already
1800 * be mapped.
1801 */
1802static inline void *skb_frag_address(const skb_frag_t *frag)
1803{
1804 return page_address(skb_frag_page(frag)) + frag->page_offset;
1805}
1806
1807/**
1808 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
1809 * @frag: the paged fragment buffer
1810 *
1811 * Returns the address of the data within @frag. Checks that the page
1812 * is mapped and returns %NULL otherwise.
1813 */
1814static inline void *skb_frag_address_safe(const skb_frag_t *frag)
1815{
1816 void *ptr = page_address(skb_frag_page(frag));
1817 if (unlikely(!ptr))
1818 return NULL;
1819
1820 return ptr + frag->page_offset;
1821}
1822
1823/**
1824 * __skb_frag_set_page - sets the page contained in a paged fragment
1825 * @frag: the paged fragment
1826 * @page: the page to set
1827 *
1828 * Sets the fragment @frag to contain @page.
1829 */
1830static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
1831{
Ian Campbella8605c62011-10-19 23:01:49 +00001832 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00001833}
1834
1835/**
1836 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
1837 * @skb: the buffer
1838 * @f: the fragment offset
1839 * @page: the page to set
1840 *
1841 * Sets the @f'th fragment of @skb to contain @page.
1842 */
1843static inline void skb_frag_set_page(struct sk_buff *skb, int f,
1844 struct page *page)
1845{
1846 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
1847}
1848
1849/**
1850 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00001851 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00001852 * @frag: the paged fragment to map
1853 * @offset: the offset within the fragment (starting at the
1854 * fragment's own offset)
1855 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00001856 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00001857 *
1858 * Maps the page associated with @frag to @device.
1859 */
1860static inline dma_addr_t skb_frag_dma_map(struct device *dev,
1861 const skb_frag_t *frag,
1862 size_t offset, size_t size,
1863 enum dma_data_direction dir)
1864{
1865 return dma_map_page(dev, skb_frag_page(frag),
1866 frag->page_offset + offset, size, dir);
1867}
1868
Eric Dumazet117632e2011-12-03 21:39:53 +00001869static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
1870 gfp_t gfp_mask)
1871{
1872 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
1873}
1874
Ian Campbell131ea662011-08-19 06:25:00 +00001875/**
Patrick McHardy334a8132007-06-25 04:35:20 -07001876 * skb_clone_writable - is the header of a clone writable
1877 * @skb: buffer to check
1878 * @len: length up to which to write
1879 *
1880 * Returns true if modifying the header part of the cloned buffer
1881 * does not requires the data to be copied.
1882 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001883static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07001884{
1885 return !skb_header_cloned(skb) &&
1886 skb_headroom(skb) + len <= skb->hdr_len;
1887}
1888
Herbert Xud9cc2042007-09-16 16:21:16 -07001889static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
1890 int cloned)
1891{
1892 int delta = 0;
1893
1894 if (headroom < NET_SKB_PAD)
1895 headroom = NET_SKB_PAD;
1896 if (headroom > skb_headroom(skb))
1897 delta = headroom - skb_headroom(skb);
1898
1899 if (delta || cloned)
1900 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
1901 GFP_ATOMIC);
1902 return 0;
1903}
1904
Patrick McHardy334a8132007-06-25 04:35:20 -07001905/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001906 * skb_cow - copy header of skb when it is required
1907 * @skb: buffer to cow
1908 * @headroom: needed headroom
1909 *
1910 * If the skb passed lacks sufficient headroom or its data part
1911 * is shared, data is reallocated. If reallocation fails, an error
1912 * is returned and original skb is not changed.
1913 *
1914 * The result is skb with writable area skb->head...skb->tail
1915 * and at least @headroom of space at head.
1916 */
1917static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1918{
Herbert Xud9cc2042007-09-16 16:21:16 -07001919 return __skb_cow(skb, headroom, skb_cloned(skb));
1920}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001921
Herbert Xud9cc2042007-09-16 16:21:16 -07001922/**
1923 * skb_cow_head - skb_cow but only making the head writable
1924 * @skb: buffer to cow
1925 * @headroom: needed headroom
1926 *
1927 * This function is identical to skb_cow except that we replace the
1928 * skb_cloned check by skb_header_cloned. It should be used when
1929 * you only need to push on some header and do not need to modify
1930 * the data.
1931 */
1932static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
1933{
1934 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001935}
1936
1937/**
1938 * skb_padto - pad an skbuff up to a minimal size
1939 * @skb: buffer to pad
1940 * @len: minimal length
1941 *
1942 * Pads up a buffer to ensure the trailing bytes exist and are
1943 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07001944 * is untouched. Otherwise it is extended. Returns zero on
1945 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001946 */
1947
Herbert Xu5b057c62006-06-23 02:06:41 -07001948static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001949{
1950 unsigned int size = skb->len;
1951 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07001952 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001953 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001954}
1955
1956static inline int skb_add_data(struct sk_buff *skb,
1957 char __user *from, int copy)
1958{
1959 const int off = skb->len;
1960
1961 if (skb->ip_summed == CHECKSUM_NONE) {
1962 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08001963 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
Linus Torvalds1da177e2005-04-16 15:20:36 -07001964 copy, 0, &err);
1965 if (!err) {
1966 skb->csum = csum_block_add(skb->csum, csum, off);
1967 return 0;
1968 }
1969 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1970 return 0;
1971
1972 __skb_trim(skb, off);
1973 return -EFAULT;
1974}
1975
Eric Dumazet38ba0a62012-04-23 17:48:27 +00001976static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
1977 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001978{
1979 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00001980 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001981
Ian Campbellea2ab692011-08-22 23:44:58 +00001982 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00001983 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001984 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00001985 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001986}
1987
Herbert Xu364c6ba2006-06-09 16:10:40 -07001988static inline int __skb_linearize(struct sk_buff *skb)
1989{
1990 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1991}
1992
Linus Torvalds1da177e2005-04-16 15:20:36 -07001993/**
1994 * skb_linearize - convert paged skb to linear one
1995 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07001996 *
1997 * If there is no free memory -ENOMEM is returned, otherwise zero
1998 * is returned and the old skb data released.
1999 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002000static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002001{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002002 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2003}
2004
2005/**
2006 * skb_linearize_cow - make sure skb is linear and writable
2007 * @skb: buffer to process
2008 *
2009 * If there is no free memory -ENOMEM is returned, otherwise zero
2010 * is returned and the old skb data released.
2011 */
2012static inline int skb_linearize_cow(struct sk_buff *skb)
2013{
2014 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2015 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002016}
2017
2018/**
2019 * skb_postpull_rcsum - update checksum for received skb after pull
2020 * @skb: buffer to update
2021 * @start: start of data before pull
2022 * @len: length of data pulled
2023 *
2024 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002025 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2026 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002027 */
2028
2029static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002030 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002031{
Patrick McHardy84fa7932006-08-29 16:44:56 -07002032 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002033 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2034}
2035
Herbert Xucbb042f2006-03-20 22:43:56 -08002036unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2037
Linus Torvalds1da177e2005-04-16 15:20:36 -07002038/**
2039 * pskb_trim_rcsum - trim received skb and update checksum
2040 * @skb: buffer to trim
2041 * @len: new length
2042 *
2043 * This is exactly the same as pskb_trim except that it ensures the
2044 * checksum of received packets are still valid after the operation.
2045 */
2046
2047static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2048{
Stephen Hemminger0e4e4222005-09-08 12:32:03 -07002049 if (likely(len >= skb->len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002050 return 0;
Patrick McHardy84fa7932006-08-29 16:44:56 -07002051 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002052 skb->ip_summed = CHECKSUM_NONE;
2053 return __pskb_trim(skb, len);
2054}
2055
Linus Torvalds1da177e2005-04-16 15:20:36 -07002056#define skb_queue_walk(queue, skb) \
2057 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002058 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002059 skb = skb->next)
2060
James Chapman46f89142007-04-30 00:07:31 -07002061#define skb_queue_walk_safe(queue, skb, tmp) \
2062 for (skb = (queue)->next, tmp = skb->next; \
2063 skb != (struct sk_buff *)(queue); \
2064 skb = tmp, tmp = skb->next)
2065
David S. Miller1164f522008-09-23 00:49:44 -07002066#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002067 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002068 skb = skb->next)
2069
2070#define skb_queue_walk_from_safe(queue, skb, tmp) \
2071 for (tmp = skb->next; \
2072 skb != (struct sk_buff *)(queue); \
2073 skb = tmp, tmp = skb->next)
2074
Stephen Hemminger300ce172005-10-30 13:47:34 -08002075#define skb_queue_reverse_walk(queue, skb) \
2076 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002077 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08002078 skb = skb->prev)
2079
David S. Miller686a2952011-01-20 22:47:32 -08002080#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2081 for (skb = (queue)->prev, tmp = skb->prev; \
2082 skb != (struct sk_buff *)(queue); \
2083 skb = tmp, tmp = skb->prev)
2084
2085#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2086 for (tmp = skb->prev; \
2087 skb != (struct sk_buff *)(queue); \
2088 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002089
David S. Miller21dc3302010-08-23 00:13:46 -07002090static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07002091{
2092 return skb_shinfo(skb)->frag_list != NULL;
2093}
2094
2095static inline void skb_frag_list_init(struct sk_buff *skb)
2096{
2097 skb_shinfo(skb)->frag_list = NULL;
2098}
2099
2100static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2101{
2102 frag->next = skb_shinfo(skb)->frag_list;
2103 skb_shinfo(skb)->frag_list = frag;
2104}
2105
2106#define skb_walk_frags(skb, iter) \
2107 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2108
Herbert Xua59322b2007-12-05 01:53:40 -08002109extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
Pavel Emelyanov3f518bf2012-02-21 07:30:58 +00002110 int *peeked, int *off, int *err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002111extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
2112 int noblock, int *err);
2113extern unsigned int datagram_poll(struct file *file, struct socket *sock,
2114 struct poll_table_struct *wait);
2115extern int skb_copy_datagram_iovec(const struct sk_buff *from,
2116 int offset, struct iovec *to,
2117 int size);
Herbert Xufb286bb2005-11-10 13:01:24 -08002118extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002119 int hlen,
2120 struct iovec *iov);
Rusty Russelldb543c12008-08-15 15:13:53 -07002121extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
2122 int offset,
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +00002123 const struct iovec *from,
2124 int from_offset,
Rusty Russelldb543c12008-08-15 15:13:53 -07002125 int len);
Michael S. Tsirkin0a1ec072009-04-20 01:25:46 +00002126extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
2127 int offset,
2128 const struct iovec *to,
2129 int to_offset,
2130 int size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002131extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
Eric Dumazet9d410c72009-10-30 05:03:53 +00002132extern void skb_free_datagram_locked(struct sock *sk,
2133 struct sk_buff *skb);
Herbert Xu27ab2562007-12-05 01:51:58 -08002134extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
Herbert Xu3305b802005-12-13 23:16:37 -08002135 unsigned int flags);
Al Viro2bbbc862006-11-14 21:37:14 -08002136extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
2137 int len, __wsum csum);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002138extern int skb_copy_bits(const struct sk_buff *skb, int offset,
2139 void *to, int len);
Stephen Hemminger0c6fcc82007-04-20 16:40:01 -07002140extern int skb_store_bits(struct sk_buff *skb, int offset,
2141 const void *from, int len);
Al Viro81d77662006-11-14 21:37:33 -08002142extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002143 int offset, u8 *to, int len,
Al Viro81d77662006-11-14 21:37:33 -08002144 __wsum csum);
Jens Axboe9c55e012007-11-06 23:30:13 -08002145extern int skb_splice_bits(struct sk_buff *skb,
2146 unsigned int offset,
2147 struct pipe_inode_info *pipe,
2148 unsigned int len,
2149 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002150extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
2151extern void skb_split(struct sk_buff *skb,
2152 struct sk_buff *skb1, const u32 len);
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08002153extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
2154 int shiftlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002155
Michał Mirosławc8f44af2011-11-15 15:29:55 +00002156extern struct sk_buff *skb_segment(struct sk_buff *skb,
2157 netdev_features_t features);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002158
Linus Torvalds1da177e2005-04-16 15:20:36 -07002159static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2160 int len, void *buffer)
2161{
2162 int hlen = skb_headlen(skb);
2163
Patrick McHardy55820ee2005-07-05 14:08:10 -07002164 if (hlen - offset >= len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002165 return skb->data + offset;
2166
2167 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2168 return NULL;
2169
2170 return buffer;
2171}
2172
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03002173static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2174 void *to,
2175 const unsigned int len)
2176{
2177 memcpy(to, skb->data, len);
2178}
2179
2180static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2181 const int offset, void *to,
2182 const unsigned int len)
2183{
2184 memcpy(to, skb->data + offset, len);
2185}
2186
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03002187static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2188 const void *from,
2189 const unsigned int len)
2190{
2191 memcpy(skb->data, from, len);
2192}
2193
2194static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2195 const int offset,
2196 const void *from,
2197 const unsigned int len)
2198{
2199 memcpy(skb->data + offset, from, len);
2200}
2201
Linus Torvalds1da177e2005-04-16 15:20:36 -07002202extern void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002203
Patrick Ohlyac45f602009-02-12 05:03:37 +00002204static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2205{
2206 return skb->tstamp;
2207}
2208
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002209/**
2210 * skb_get_timestamp - get timestamp from a skb
2211 * @skb: skb to get stamp from
2212 * @stamp: pointer to struct timeval to store stamp in
2213 *
2214 * Timestamps are stored in the skb as offsets to a base timestamp.
2215 * This function converts the offset back to a struct timeval and stores
2216 * it in stamp.
2217 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00002218static inline void skb_get_timestamp(const struct sk_buff *skb,
2219 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002220{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002221 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002222}
2223
Patrick Ohlyac45f602009-02-12 05:03:37 +00002224static inline void skb_get_timestampns(const struct sk_buff *skb,
2225 struct timespec *stamp)
2226{
2227 *stamp = ktime_to_timespec(skb->tstamp);
2228}
2229
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002230static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002231{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002232 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002233}
2234
Stephen Hemminger164891a2007-04-23 22:26:16 -07002235static inline ktime_t net_timedelta(ktime_t t)
2236{
2237 return ktime_sub(ktime_get_real(), t);
2238}
2239
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07002240static inline ktime_t net_invalid_timestamp(void)
2241{
2242 return ktime_set(0, 0);
2243}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002244
Richard Cochranc1f19b52010-07-17 08:49:36 +00002245extern void skb_timestamping_init(void);
2246
2247#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2248
2249extern void skb_clone_tx_timestamp(struct sk_buff *skb);
2250extern bool skb_defer_rx_timestamp(struct sk_buff *skb);
2251
2252#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2253
2254static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2255{
2256}
2257
2258static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2259{
2260 return false;
2261}
2262
2263#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2264
2265/**
2266 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2267 *
Richard Cochranda92b192011-10-21 00:49:15 +00002268 * PHY drivers may accept clones of transmitted packets for
2269 * timestamping via their phy_driver.txtstamp method. These drivers
2270 * must call this function to return the skb back to the stack, with
2271 * or without a timestamp.
2272 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00002273 * @skb: clone of the the original outgoing packet
Richard Cochranda92b192011-10-21 00:49:15 +00002274 * @hwtstamps: hardware time stamps, may be NULL if not available
Richard Cochranc1f19b52010-07-17 08:49:36 +00002275 *
2276 */
2277void skb_complete_tx_timestamp(struct sk_buff *skb,
2278 struct skb_shared_hwtstamps *hwtstamps);
2279
Patrick Ohlyac45f602009-02-12 05:03:37 +00002280/**
2281 * skb_tstamp_tx - queue clone of skb with send time stamps
2282 * @orig_skb: the original outgoing packet
2283 * @hwtstamps: hardware time stamps, may be NULL if not available
2284 *
2285 * If the skb has a socket associated, then this function clones the
2286 * skb (thus sharing the actual data and optional structures), stores
2287 * the optional hardware time stamping information (if non NULL) or
2288 * generates a software time stamp (otherwise), then queues the clone
2289 * to the error queue of the socket. Errors are silently ignored.
2290 */
2291extern void skb_tstamp_tx(struct sk_buff *orig_skb,
2292 struct skb_shared_hwtstamps *hwtstamps);
2293
Richard Cochran4507a712010-07-17 08:48:28 +00002294static inline void sw_tx_timestamp(struct sk_buff *skb)
2295{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002296 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2297 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00002298 skb_tstamp_tx(skb, NULL);
2299}
2300
2301/**
2302 * skb_tx_timestamp() - Driver hook for transmit timestamping
2303 *
2304 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00002305 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00002306 *
2307 * @skb: A socket buffer.
2308 */
2309static inline void skb_tx_timestamp(struct sk_buff *skb)
2310{
Richard Cochranc1f19b52010-07-17 08:49:36 +00002311 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002312 sw_tx_timestamp(skb);
2313}
2314
Johannes Berg6e3e9392011-11-09 10:15:42 +01002315/**
2316 * skb_complete_wifi_ack - deliver skb with wifi status
2317 *
2318 * @skb: the original outgoing packet
2319 * @acked: ack status
2320 *
2321 */
2322void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2323
Herbert Xu759e5d02007-03-25 20:10:56 -07002324extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
Al Virob51655b2006-11-14 21:40:42 -08002325extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002326
Herbert Xu60476372007-04-09 11:59:39 -07002327static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2328{
2329 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2330}
2331
Herbert Xufb286bb2005-11-10 13:01:24 -08002332/**
2333 * skb_checksum_complete - Calculate checksum of an entire packet
2334 * @skb: packet to process
2335 *
2336 * This function calculates the checksum over the entire packet plus
2337 * the value of skb->csum. The latter can be used to supply the
2338 * checksum of a pseudo header as used by TCP/UDP. It returns the
2339 * checksum.
2340 *
2341 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2342 * this function can be used to verify that checksum on received
2343 * packets. In that case the function should return zero if the
2344 * checksum is correct. In particular, this function will return zero
2345 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2346 * hardware has already verified the correctness of the checksum.
2347 */
Al Viro4381ca32007-07-15 21:00:11 +01002348static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002349{
Herbert Xu60476372007-04-09 11:59:39 -07002350 return skb_csum_unnecessary(skb) ?
2351 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002352}
2353
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002354#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002355extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002356static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2357{
2358 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002359 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002360}
2361static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2362{
2363 if (nfct)
2364 atomic_inc(&nfct->use);
2365}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002366#endif
2367#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -08002368static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
2369{
2370 if (skb)
2371 atomic_inc(&skb->users);
2372}
2373static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
2374{
2375 if (skb)
2376 kfree_skb(skb);
2377}
2378#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002379#ifdef CONFIG_BRIDGE_NETFILTER
2380static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2381{
2382 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2383 kfree(nf_bridge);
2384}
2385static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2386{
2387 if (nf_bridge)
2388 atomic_inc(&nf_bridge->use);
2389}
2390#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08002391static inline void nf_reset(struct sk_buff *skb)
2392{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002393#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08002394 nf_conntrack_put(skb->nfct);
2395 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002396#endif
2397#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Patrick McHardya193a4a2006-03-20 19:23:05 -08002398 nf_conntrack_put_reasm(skb->nfct_reasm);
2399 skb->nfct_reasm = NULL;
2400#endif
2401#ifdef CONFIG_BRIDGE_NETFILTER
2402 nf_bridge_put(skb->nf_bridge);
2403 skb->nf_bridge = NULL;
2404#endif
2405}
2406
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002407/* Note: This doesn't put any conntrack and bridge info in dst. */
2408static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2409{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002410#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002411 dst->nfct = src->nfct;
2412 nf_conntrack_get(src->nfct);
2413 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002414#endif
2415#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002416 dst->nfct_reasm = src->nfct_reasm;
2417 nf_conntrack_get_reasm(src->nfct_reasm);
2418#endif
2419#ifdef CONFIG_BRIDGE_NETFILTER
2420 dst->nf_bridge = src->nf_bridge;
2421 nf_bridge_get(src->nf_bridge);
2422#endif
2423}
2424
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002425static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2426{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002427#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002428 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002429#endif
2430#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002431 nf_conntrack_put_reasm(dst->nfct_reasm);
2432#endif
2433#ifdef CONFIG_BRIDGE_NETFILTER
2434 nf_bridge_put(dst->nf_bridge);
2435#endif
2436 __nf_copy(dst, src);
2437}
2438
James Morris984bc162006-06-09 00:29:17 -07002439#ifdef CONFIG_NETWORK_SECMARK
2440static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2441{
2442 to->secmark = from->secmark;
2443}
2444
2445static inline void skb_init_secmark(struct sk_buff *skb)
2446{
2447 skb->secmark = 0;
2448}
2449#else
2450static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2451{ }
2452
2453static inline void skb_init_secmark(struct sk_buff *skb)
2454{ }
2455#endif
2456
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002457static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2458{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002459 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002460}
2461
Stephen Hemminger92477442009-03-21 13:39:26 -07002462static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002463{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002464 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002465}
2466
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002467static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2468{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002469 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002470}
2471
David S. Millerd5a9e242009-01-27 16:22:11 -08002472static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2473{
2474 skb->queue_mapping = rx_queue + 1;
2475}
2476
Stephen Hemminger92477442009-03-21 13:39:26 -07002477static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002478{
2479 return skb->queue_mapping - 1;
2480}
2481
Stephen Hemminger92477442009-03-21 13:39:26 -07002482static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002483{
Eric Dumazeta02cec22010-09-22 20:43:57 +00002484 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08002485}
2486
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00002487extern u16 __skb_tx_hash(const struct net_device *dev,
2488 const struct sk_buff *skb,
2489 unsigned int num_tx_queues);
Stephen Hemminger92477442009-03-21 13:39:26 -07002490
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002491#ifdef CONFIG_XFRM
2492static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2493{
2494 return skb->sp;
2495}
2496#else
2497static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2498{
2499 return NULL;
2500}
2501#endif
2502
David S. Millerbdcc0922012-03-07 20:53:36 -05002503static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07002504{
2505 return skb_shinfo(skb)->gso_size;
2506}
2507
David S. Millerbdcc0922012-03-07 20:53:36 -05002508static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02002509{
2510 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2511}
2512
Ben Hutchings4497b072008-06-19 16:22:28 -07002513extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2514
2515static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2516{
2517 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2518 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002519 const struct skb_shared_info *shinfo = skb_shinfo(skb);
2520
Alexander Duyckb78462e2010-06-02 12:24:37 +00002521 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2522 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07002523 __skb_warn_lro_forwarding(skb);
2524 return true;
2525 }
2526 return false;
2527}
2528
Herbert Xu35fc92a2007-03-26 23:22:20 -07002529static inline void skb_forward_csum(struct sk_buff *skb)
2530{
2531 /* Unfortunately we don't support this one. Any brave souls? */
2532 if (skb->ip_summed == CHECKSUM_COMPLETE)
2533 skb->ip_summed = CHECKSUM_NONE;
2534}
2535
Eric Dumazetbc8acf22010-09-02 13:07:41 -07002536/**
2537 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2538 * @skb: skb to check
2539 *
2540 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2541 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2542 * use this helper, to document places where we make this assertion.
2543 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002544static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07002545{
2546#ifdef DEBUG
2547 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2548#endif
2549}
2550
Rusty Russellf35d9d82008-02-04 23:49:54 -05002551bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00002552
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002553static inline bool skb_is_recycleable(const struct sk_buff *skb, int skb_size)
Andy Fleming3d153a72011-10-13 04:33:54 +00002554{
2555 if (irqs_disabled())
2556 return false;
2557
2558 if (skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)
2559 return false;
2560
2561 if (skb_is_nonlinear(skb) || skb->fclone != SKB_FCLONE_UNAVAILABLE)
2562 return false;
2563
2564 skb_size = SKB_DATA_ALIGN(skb_size + NET_SKB_PAD);
Alexander Duyckec47ea82012-05-04 14:26:56 +00002565 if (skb_end_offset(skb) < skb_size)
Andy Fleming3d153a72011-10-13 04:33:54 +00002566 return false;
2567
2568 if (skb_shared(skb) || skb_cloned(skb))
2569 return false;
2570
2571 return true;
2572}
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00002573
2574/**
2575 * skb_head_is_locked - Determine if the skb->head is locked down
2576 * @skb: skb to check
2577 *
2578 * The head on skbs build around a head frag can be removed if they are
2579 * not cloned. This function returns true if the skb head is locked down
2580 * due to either being allocated via kmalloc, or by being a clone with
2581 * multiple references to the head.
2582 */
2583static inline bool skb_head_is_locked(const struct sk_buff *skb)
2584{
2585 return !skb->head_frag || skb_cloned(skb);
2586}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002587#endif /* __KERNEL__ */
2588#endif /* _LINUX_SKBUFF_H */