blob: dc4d49a0c07dd8772b64dffe3132aece431fa8b5 [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 Sharma60063492011-07-26 16:09:06 -070024#include <linux/atomic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <asm/types.h>
26#include <linux/spinlock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <linux/net.h>
Thomas Graf3fc7e8a2005-06-23 21:00:17 -070028#include <linux/textsearch.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070029#include <net/checksum.h>
Al Viroa80958f2006-12-04 20:41:19 +000030#include <linux/rcupdate.h>
Chris Leech97fc2f02006-05-23 17:55:33 -070031#include <linux/dmaengine.h>
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -070032#include <linux/hrtimer.h>
Ian Campbell131ea662011-08-19 06:25:00 +000033#include <linux/dma-mapping.h>
Michał Mirosławc8f44af2011-11-15 15:29:55 +000034#include <linux/netdev_features.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070035
Herbert Xu60476372007-04-09 11:59:39 -070036/* Don't change this without changing skb_csum_unnecessary! */
Linus Torvalds1da177e2005-04-16 15:20:36 -070037#define CHECKSUM_NONE 0
Herbert Xu60476372007-04-09 11:59:39 -070038#define CHECKSUM_UNNECESSARY 1
39#define CHECKSUM_COMPLETE 2
40#define CHECKSUM_PARTIAL 3
Linus Torvalds1da177e2005-04-16 15:20:36 -070041
42#define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
43 ~(SMP_CACHE_BYTES - 1))
David S. Millerfc910a22007-03-25 20:27:59 -070044#define SKB_WITH_OVERHEAD(X) \
Herbert Xudeea84b2007-10-21 16:27:46 -070045 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
David S. Millerfc910a22007-03-25 20:27:59 -070046#define SKB_MAX_ORDER(X, ORDER) \
47 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
Linus Torvalds1da177e2005-04-16 15:20:36 -070048#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
49#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
50
Eric Dumazet87fb4b72011-10-13 07:28:54 +000051/* return minimum truesize of one skb containing X bytes of data */
52#define SKB_TRUESIZE(X) ((X) + \
53 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
54 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
55
Linus Torvalds1da177e2005-04-16 15:20:36 -070056/* A. Checksumming of received packets by device.
57 *
58 * NONE: device failed to checksum this packet.
59 * skb->csum is undefined.
60 *
61 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
62 * skb->csum is undefined.
63 * It is bad option, but, unfortunately, many of vendors do this.
64 * Apparently with secret goal to sell you new device, when you
65 * will add new protocol to your host. F.e. IPv6. 8)
66 *
Patrick McHardy84fa7932006-08-29 16:44:56 -070067 * COMPLETE: the most generic way. Device supplied checksum of _all_
Linus Torvalds1da177e2005-04-16 15:20:36 -070068 * the packet as seen by netif_rx in skb->csum.
69 * NOTE: Even if device supports only some protocols, but
Patrick McHardy84fa7932006-08-29 16:44:56 -070070 * is able to produce some skb->csum, it MUST use COMPLETE,
Linus Torvalds1da177e2005-04-16 15:20:36 -070071 * not UNNECESSARY.
72 *
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070073 * PARTIAL: identical to the case for output below. This may occur
74 * on a packet received directly from another Linux OS, e.g.,
75 * a virtualised Linux kernel on the same host. The packet can
76 * be treated in the same way as UNNECESSARY except that on
77 * output (i.e., forwarding) the checksum must be filled in
78 * by the OS or the hardware.
79 *
Linus Torvalds1da177e2005-04-16 15:20:36 -070080 * B. Checksumming on output.
81 *
82 * NONE: skb is checksummed by protocol or csum is not required.
83 *
Patrick McHardy84fa7932006-08-29 16:44:56 -070084 * PARTIAL: device is required to csum packet as seen by hard_start_xmit
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070085 * from skb->csum_start to the end and to record the checksum
86 * at skb->csum_start + skb->csum_offset.
Linus Torvalds1da177e2005-04-16 15:20:36 -070087 *
88 * Device must show its capabilities in dev->features, set
89 * at device setup time.
90 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
91 * everything.
Linus Torvalds1da177e2005-04-16 15:20:36 -070092 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
93 * TCP/UDP over IPv4. Sigh. Vendors like this
94 * way by an unknown reason. Though, see comment above
95 * about CHECKSUM_UNNECESSARY. 8)
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070096 * NETIF_F_IPV6_CSUM about as dumb as the last one but does IPv6 instead.
Linus Torvalds1da177e2005-04-16 15:20:36 -070097 *
Yi Zou3af79302012-03-19 11:12:41 +000098 * UNNECESSARY: device will do per protocol specific csum. Protocol drivers
99 * that do not want net to perform the checksum calculation should use
100 * this flag in their outgoing skbs.
101 * NETIF_F_FCOE_CRC this indicates the device can do FCoE FC CRC
102 * offload. Correspondingly, the FCoE protocol driver
103 * stack should use CHECKSUM_UNNECESSARY.
104 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105 * Any questions? No questions, good. --ANK
106 */
107
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108struct net_device;
David Howells716ea3a2007-04-02 20:19:53 -0700109struct scatterlist;
Jens Axboe9c55e012007-11-06 23:30:13 -0800110struct pipe_inode_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700112#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113struct nf_conntrack {
114 atomic_t use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115};
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700116#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117
118#ifdef CONFIG_BRIDGE_NETFILTER
119struct nf_bridge_info {
120 atomic_t use;
121 struct net_device *physindev;
122 struct net_device *physoutdev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700123 unsigned int mask;
124 unsigned long data[32 / sizeof(unsigned long)];
125};
126#endif
127
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128struct sk_buff_head {
129 /* These two members must be first. */
130 struct sk_buff *next;
131 struct sk_buff *prev;
132
133 __u32 qlen;
134 spinlock_t lock;
135};
136
137struct sk_buff;
138
Ian Campbell9d4dde52011-12-22 23:39:14 +0000139/* To allow 64K frame to be packed as single skb without frag_list we
140 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
141 * buffers which do not start on a page boundary.
142 *
143 * Since GRO uses frags we allocate at least 16 regardless of page
144 * size.
Anton Blancharda715dea2011-03-27 14:57:26 +0000145 */
Ian Campbell9d4dde52011-12-22 23:39:14 +0000146#if (65536/PAGE_SIZE + 1) < 16
David S. Millereec00952011-03-29 23:34:08 -0700147#define MAX_SKB_FRAGS 16UL
Anton Blancharda715dea2011-03-27 14:57:26 +0000148#else
Ian Campbell9d4dde52011-12-22 23:39:14 +0000149#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
Anton Blancharda715dea2011-03-27 14:57:26 +0000150#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151
152typedef struct skb_frag_struct skb_frag_t;
153
154struct skb_frag_struct {
Ian Campbella8605c62011-10-19 23:01:49 +0000155 struct {
156 struct page *p;
157 } page;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000158#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
David S. Millera309bb02007-07-30 18:47:03 -0700159 __u32 page_offset;
160 __u32 size;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000161#else
162 __u16 page_offset;
163 __u16 size;
164#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165};
166
Eric Dumazet9e903e02011-10-18 21:00:24 +0000167static inline unsigned int skb_frag_size(const skb_frag_t *frag)
168{
169 return frag->size;
170}
171
172static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
173{
174 frag->size = size;
175}
176
177static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
178{
179 frag->size += delta;
180}
181
182static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
183{
184 frag->size -= delta;
185}
186
Patrick Ohlyac45f602009-02-12 05:03:37 +0000187#define HAVE_HW_TIME_STAMP
188
189/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800190 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000191 * @hwtstamp: hardware time stamp transformed into duration
192 * since arbitrary point in time
193 * @syststamp: hwtstamp transformed to system time base
194 *
195 * Software time stamps generated by ktime_get_real() are stored in
196 * skb->tstamp. The relation between the different kinds of time
197 * stamps is as follows:
198 *
199 * syststamp and tstamp can be compared against each other in
200 * arbitrary combinations. The accuracy of a
201 * syststamp/tstamp/"syststamp from other device" comparison is
202 * limited by the accuracy of the transformation into system time
203 * base. This depends on the device driver and its underlying
204 * hardware.
205 *
206 * hwtstamps can only be compared against other hwtstamps from
207 * the same device.
208 *
209 * This structure is attached to packets as part of the
210 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
211 */
212struct skb_shared_hwtstamps {
213 ktime_t hwtstamp;
214 ktime_t syststamp;
215};
216
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000217/* Definitions for tx_flags in struct skb_shared_info */
218enum {
219 /* generate hardware time stamp */
220 SKBTX_HW_TSTAMP = 1 << 0,
221
222 /* generate software time stamp */
223 SKBTX_SW_TSTAMP = 1 << 1,
224
225 /* device driver is going to provide hardware time stamp */
226 SKBTX_IN_PROGRESS = 1 << 2,
227
Shirley Maa6686f22011-07-06 12:22:12 +0000228 /* device driver supports TX zero-copy buffers */
Eric Dumazet993772c2012-06-14 06:42:44 +0000229 SKBTX_DEV_ZEROCOPY = 1 << 3,
Johannes Berg6e3e9392011-11-09 10:15:42 +0100230
231 /* generate wifi status information (where possible) */
Eric Dumazet993772c2012-06-14 06:42:44 +0000232 SKBTX_WIFI_STATUS = 1 << 4,
Shirley Maa6686f22011-07-06 12:22:12 +0000233};
234
235/*
236 * The callback notifies userspace to release buffers when skb DMA is done in
237 * lower device, the skb last reference should be 0 when calling this.
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000238 * The ctx field is used to track device context.
239 * The desc field is used to track userspace buffer index.
Shirley Maa6686f22011-07-06 12:22:12 +0000240 */
241struct ubuf_info {
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000242 void (*callback)(struct ubuf_info *);
243 void *ctx;
Shirley Maa6686f22011-07-06 12:22:12 +0000244 unsigned long desc;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000245};
246
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247/* This data is invariant across clones and lives at
248 * the end of the header data, ie. at skb->end.
249 */
250struct skb_shared_info {
Ian Campbell9f42f122012-01-05 07:13:39 +0000251 unsigned char nr_frags;
252 __u8 tx_flags;
Herbert Xu79671682006-06-22 02:40:14 -0700253 unsigned short gso_size;
254 /* Warning: this field is not always filled in (UFO)! */
255 unsigned short gso_segs;
256 unsigned short gso_type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000258 struct skb_shared_hwtstamps hwtstamps;
Ian Campbell9f42f122012-01-05 07:13:39 +0000259 __be32 ip6_frag_id;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700260
261 /*
262 * Warning : all fields before dataref are cleared in __alloc_skb()
263 */
264 atomic_t dataref;
265
Johann Baudy69e3c752009-05-18 22:11:22 -0700266 /* Intermediate layers must ensure that destructor_arg
267 * remains valid until skb destructor */
268 void * destructor_arg;
Shirley Maa6686f22011-07-06 12:22:12 +0000269
Eric Dumazetfed66382010-07-22 19:09:08 +0000270 /* must be last field, see pskb_expand_head() */
271 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272};
273
274/* We divide dataref into two halves. The higher 16 bits hold references
275 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700276 * the entire skb->data. A clone of a headerless skb holds the length of
277 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278 *
279 * All users must obey the rule that the skb->data reference count must be
280 * greater than or equal to the payload reference count.
281 *
282 * Holding a reference to the payload part means that the user does not
283 * care about modifications to the header part of skb->data.
284 */
285#define SKB_DATAREF_SHIFT 16
286#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
287
David S. Millerd179cd12005-08-17 14:57:30 -0700288
289enum {
290 SKB_FCLONE_UNAVAILABLE,
291 SKB_FCLONE_ORIG,
292 SKB_FCLONE_CLONE,
293};
294
Herbert Xu79671682006-06-22 02:40:14 -0700295enum {
296 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700297 SKB_GSO_UDP = 1 << 1,
Herbert Xu576a30e2006-06-27 13:22:38 -0700298
299 /* This indicates the skb is from an untrusted source. */
300 SKB_GSO_DODGY = 1 << 2,
Michael Chanb0da85372006-06-29 12:30:00 -0700301
302 /* This indicates the tcp segment has CWR set. */
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700303 SKB_GSO_TCP_ECN = 1 << 3,
304
305 SKB_GSO_TCPV6 = 1 << 4,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800306
307 SKB_GSO_FCOE = 1 << 5,
Herbert Xu79671682006-06-22 02:40:14 -0700308};
309
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700310#if BITS_PER_LONG > 32
311#define NET_SKBUFF_DATA_USES_OFFSET 1
312#endif
313
314#ifdef NET_SKBUFF_DATA_USES_OFFSET
315typedef unsigned int sk_buff_data_t;
316#else
317typedef unsigned char *sk_buff_data_t;
318#endif
319
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100320#if defined(CONFIG_NF_DEFRAG_IPV4) || defined(CONFIG_NF_DEFRAG_IPV4_MODULE) || \
321 defined(CONFIG_NF_DEFRAG_IPV6) || defined(CONFIG_NF_DEFRAG_IPV6_MODULE)
322#define NET_SKBUFF_NF_DEFRAG_NEEDED 1
323#endif
324
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325/**
326 * struct sk_buff - socket buffer
327 * @next: Next buffer in list
328 * @prev: Previous buffer in list
Herbert Xu325ed822005-10-03 13:57:23 -0700329 * @tstamp: Time we arrived
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700330 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700332 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000333 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700334 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700335 * @len: Length of actual data
336 * @data_len: Data length
337 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700338 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700339 * @csum: Checksum (must include start/offset pair)
340 * @csum_start: Offset from skb->head where checksumming should start
341 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700342 * @priority: Packet queueing priority
Martin Waitz67be2dd2005-05-01 08:59:26 -0700343 * @local_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700345 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346 * @nohdr: Payload reference only, must not modify header
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700347 * @nfctinfo: Relationship of this skb to the connection
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700349 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700350 * @ipvs_property: skbuff is owned by ipvs
Randy Dunlap31729362008-02-18 20:52:13 -0800351 * @peeked: this packet has been seen already, so stats have been
352 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700353 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700354 * @protocol: Packet protocol from driver
355 * @destructor: Destruct function
356 * @nfct: Associated connection, if any
Randy Dunlap461ddf32005-11-20 21:25:15 -0800357 * @nfct_reasm: netfilter conntrack re-assembly pointer
Linus Torvalds1da177e2005-04-16 15:20:36 -0700358 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800359 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360 * @tc_index: Traffic control index
361 * @tc_verd: traffic control verdict
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700362 * @rxhash: the packet hash computed on receive
363 * @queue_mapping: Queue mapping for multiqueue devices
Randy Dunlap553a5672008-04-20 10:51:01 -0700364 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700365 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Changli Gao4ca24622011-08-19 07:26:44 -0700366 * @l4_rxhash: indicate rxhash is a canonical 4-tuple hash over transport
367 * ports.
Johannes Berg6e3e9392011-11-09 10:15:42 +0100368 * @wifi_acked_valid: wifi_acked was set
369 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000370 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Randy Dunlapf4b8ea72006-06-22 16:00:11 -0700371 * @dma_cookie: a cookie to one of several possible DMA operations
372 * done by skb DMA functions
James Morris984bc162006-06-09 00:29:17 -0700373 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700374 * @mark: Generic packet mark
375 * @dropcount: total number of sk_receive_queue overflows
Patrick McHardy6aa895b2008-07-14 22:49:06 -0700376 * @vlan_tci: vlan tag control information
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700377 * @transport_header: Transport layer header
378 * @network_header: Network layer header
379 * @mac_header: Link layer header
380 * @tail: Tail pointer
381 * @end: End pointer
382 * @head: Head of buffer
383 * @data: Data head pointer
384 * @truesize: Buffer size
385 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386 */
387
388struct sk_buff {
389 /* These two members must be first. */
390 struct sk_buff *next;
391 struct sk_buff *prev;
392
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700393 ktime_t tstamp;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000394
395 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398 /*
399 * This is the control buffer. It is free to use for every
400 * layer. Please put your private variables there. If you
401 * want to keep them across layers you have to do a skb_clone()
402 * first. This is owned by whoever has the skb queued ATM.
403 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000404 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405
Eric Dumazet7fee2262010-05-11 23:19:48 +0000406 unsigned long _skb_refdst;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000407#ifdef CONFIG_XFRM
408 struct sec_path *sp;
409#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700411 data_len;
412 __u16 mac_len,
413 hdr_len;
Al Viroff1dcad2006-11-20 18:07:29 -0800414 union {
415 __wsum csum;
Herbert Xu663ead32007-04-09 11:59:07 -0700416 struct {
417 __u16 csum_start;
418 __u16 csum_offset;
419 };
Al Viroff1dcad2006-11-20 18:07:29 -0800420 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 __u32 priority;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200422 kmemcheck_bitfield_begin(flags1);
Thomas Graf1cbb3382005-07-05 14:13:41 -0700423 __u8 local_df:1,
424 cloned:1,
425 ip_summed:2,
Harald Welte6869c4d2005-08-09 19:24:19 -0700426 nohdr:1,
427 nfctinfo:3;
David S. Millerd179cd12005-08-17 14:57:30 -0700428 __u8 pkt_type:3,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800429 fclone:2,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700430 ipvs_property:1,
Herbert Xua59322b2007-12-05 01:53:40 -0800431 peeked:1,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700432 nf_trace:1;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200433 kmemcheck_bitfield_end(flags1);
Eric Dumazet4ab408d2010-03-01 03:09:26 +0000434 __be16 protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435
436 void (*destructor)(struct sk_buff *skb);
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800437#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700438 struct nf_conntrack *nfct;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100439#endif
440#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800441 struct sk_buff *nfct_reasm;
442#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443#ifdef CONFIG_BRIDGE_NETFILTER
444 struct nf_bridge_info *nf_bridge;
445#endif
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700446
Eric Dumazet8964be42009-11-20 15:35:04 -0800447 int skb_iif;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000448
449 __u32 rxhash;
450
451 __u16 vlan_tci;
452
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700454 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700456 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200459
Eric Dumazet0a148422011-04-20 09:27:32 +0000460 __u16 queue_mapping;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200461 kmemcheck_bitfield_begin(flags2);
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400462#ifdef CONFIG_IPV6_NDISC_NODETYPE
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000463 __u8 ndisc_nodetype:2;
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400464#endif
Tom Herbert3853b582010-11-21 13:17:29 +0000465 __u8 ooo_okay:1;
Tom Herbertbdeab992011-08-14 19:45:55 +0000466 __u8 l4_rxhash:1;
Johannes Berg6e3e9392011-11-09 10:15:42 +0100467 __u8 wifi_acked_valid:1;
468 __u8 wifi_acked:1;
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000469 __u8 no_fcs:1;
470 /* 9/11 bit hole (depending on ndisc_nodetype presence) */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200471 kmemcheck_bitfield_end(flags2);
472
Chris Leech97fc2f02006-05-23 17:55:33 -0700473#ifdef CONFIG_NET_DMA
474 dma_cookie_t dma_cookie;
475#endif
James Morris984bc162006-06-09 00:29:17 -0700476#ifdef CONFIG_NETWORK_SECMARK
477 __u32 secmark;
478#endif
Neil Horman3b885782009-10-12 13:26:31 -0700479 union {
480 __u32 mark;
481 __u32 dropcount;
Eric Dumazetb6da5782013-03-14 05:40:32 +0000482 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700483 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700485 sk_buff_data_t transport_header;
486 sk_buff_data_t network_header;
487 sk_buff_data_t mac_header;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700489 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700490 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700492 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700493 unsigned int truesize;
494 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495};
496
497#ifdef __KERNEL__
498/*
499 * Handling routines are only of interest to the kernel
500 */
501#include <linux/slab.h>
502
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503
Eric Dumazet7fee2262010-05-11 23:19:48 +0000504/*
505 * skb might have a dst pointer attached, refcounted or not.
506 * _skb_refdst low order bit is set if refcount was _not_ taken
507 */
508#define SKB_DST_NOREF 1UL
509#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
510
511/**
512 * skb_dst - returns skb dst_entry
513 * @skb: buffer
514 *
515 * Returns skb dst_entry, regardless of reference taken or not.
516 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000517static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
518{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000519 /* If refdst was not refcounted, check we still are in a
520 * rcu_read_lock section
521 */
522 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
523 !rcu_read_lock_held() &&
524 !rcu_read_lock_bh_held());
525 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000526}
527
Eric Dumazet7fee2262010-05-11 23:19:48 +0000528/**
529 * skb_dst_set - sets skb dst
530 * @skb: buffer
531 * @dst: dst entry
532 *
533 * Sets skb dst, assuming a reference was taken on dst and should
534 * be released by skb_dst_drop()
535 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000536static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
537{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000538 skb->_skb_refdst = (unsigned long)dst;
539}
540
Eric Dumazet27b75c92010-10-15 05:44:11 +0000541extern void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst);
Eric Dumazet7fee2262010-05-11 23:19:48 +0000542
543/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300544 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000545 * @skb: buffer
546 */
547static inline bool skb_dst_is_noref(const struct sk_buff *skb)
548{
549 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000550}
551
Eric Dumazet511c3f92009-06-02 05:14:27 +0000552static inline struct rtable *skb_rtable(const struct sk_buff *skb)
553{
Eric Dumazetadf30902009-06-02 05:19:30 +0000554 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000555}
556
Jörn Engel231d06a2006-03-20 21:28:35 -0800557extern void kfree_skb(struct sk_buff *skb);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000558extern void consume_skb(struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700559extern void __kfree_skb(struct sk_buff *skb);
David S. Millerd179cd12005-08-17 14:57:30 -0700560extern struct sk_buff *__alloc_skb(unsigned int size,
Christoph Hellwigb30973f2006-12-06 20:32:36 -0800561 gfp_t priority, int fclone, int node);
Eric Dumazetb2b5ce92011-11-14 06:03:34 +0000562extern struct sk_buff *build_skb(void *data);
David S. Millerd179cd12005-08-17 14:57:30 -0700563static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100564 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700565{
Eric Dumazet564824b2010-10-11 19:05:25 +0000566 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700567}
568
569static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100570 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700571{
Eric Dumazet564824b2010-10-11 19:05:25 +0000572 return __alloc_skb(size, priority, 1, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700573}
574
Andy Fleming3d153a72011-10-13 04:33:54 +0000575extern void skb_recycle(struct sk_buff *skb);
Changli Gao5b0daa32010-05-29 00:12:13 -0700576extern bool skb_recycle_check(struct sk_buff *skb, int skb_size);
Lennert Buytenhek04a4bb52008-10-01 02:33:12 -0700577
Herbert Xue0053ec2007-10-14 00:37:52 -0700578extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
Michael S. Tsirkin48c83012011-08-31 08:03:29 +0000579extern int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700580extern struct sk_buff *skb_clone(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100581 gfp_t priority);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700582extern struct sk_buff *skb_copy(const struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100583 gfp_t priority);
Eric Dumazet117632e2011-12-03 21:39:53 +0000584extern struct sk_buff *__pskb_copy(struct sk_buff *skb,
585 int headroom, gfp_t gfp_mask);
586
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587extern int pskb_expand_head(struct sk_buff *skb,
Victor Fusco86a76ca2005-07-08 14:57:47 -0700588 int nhead, int ntail,
Al Virodd0fc662005-10-07 07:46:04 +0100589 gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
591 unsigned int headroom);
592extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
593 int newheadroom, int newtailroom,
Al Virodd0fc662005-10-07 07:46:04 +0100594 gfp_t priority);
David Howells716ea3a2007-04-02 20:19:53 -0700595extern int skb_to_sgvec(struct sk_buff *skb,
596 struct scatterlist *sg, int offset,
597 int len);
598extern int skb_cow_data(struct sk_buff *skb, int tailbits,
599 struct sk_buff **trailer);
Herbert Xu5b057c62006-06-23 02:06:41 -0700600extern int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000601#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700603extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
604 int getfrag(void *from, char *to, int offset,
605 int len,int odd, struct sk_buff *skb),
606 void *from, int length);
607
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800608struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700609 __u32 lower_offset;
610 __u32 upper_offset;
611 __u32 frag_idx;
612 __u32 stepped_offset;
613 struct sk_buff *root_skb;
614 struct sk_buff *cur_skb;
615 __u8 *frag_data;
616};
617
618extern void skb_prepare_seq_read(struct sk_buff *skb,
619 unsigned int from, unsigned int to,
620 struct skb_seq_state *st);
621extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
622 struct skb_seq_state *st);
623extern void skb_abort_seq_read(struct skb_seq_state *st);
624
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700625extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
626 unsigned int to, struct ts_config *config,
627 struct ts_state *state);
628
Tom Herbertbdeab992011-08-14 19:45:55 +0000629extern void __skb_get_rxhash(struct sk_buff *skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000630static inline __u32 skb_get_rxhash(struct sk_buff *skb)
631{
632 if (!skb->rxhash)
Tom Herbertbdeab992011-08-14 19:45:55 +0000633 __skb_get_rxhash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000634
635 return skb->rxhash;
636}
637
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700638#ifdef NET_SKBUFF_DATA_USES_OFFSET
639static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
640{
641 return skb->head + skb->end;
642}
643#else
644static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
645{
646 return skb->end;
647}
648#endif
649
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700651#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700652
Patrick Ohlyac45f602009-02-12 05:03:37 +0000653static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
654{
655 return &skb_shinfo(skb)->hwtstamps;
656}
657
Linus Torvalds1da177e2005-04-16 15:20:36 -0700658/**
659 * skb_queue_empty - check if a queue is empty
660 * @list: queue head
661 *
662 * Returns true if the queue is empty, false otherwise.
663 */
664static inline int skb_queue_empty(const struct sk_buff_head *list)
665{
666 return list->next == (struct sk_buff *)list;
667}
668
669/**
David S. Millerfc7ebb22008-09-23 00:34:07 -0700670 * skb_queue_is_last - check if skb is the last entry in the queue
671 * @list: queue head
672 * @skb: buffer
673 *
674 * Returns true if @skb is the last buffer on the list.
675 */
676static inline bool skb_queue_is_last(const struct sk_buff_head *list,
677 const struct sk_buff *skb)
678{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000679 return skb->next == (struct sk_buff *)list;
David S. Millerfc7ebb22008-09-23 00:34:07 -0700680}
681
682/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800683 * skb_queue_is_first - check if skb is the first entry in the queue
684 * @list: queue head
685 * @skb: buffer
686 *
687 * Returns true if @skb is the first buffer on the list.
688 */
689static inline bool skb_queue_is_first(const struct sk_buff_head *list,
690 const struct sk_buff *skb)
691{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000692 return skb->prev == (struct sk_buff *)list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800693}
694
695/**
David S. Miller249c8b42008-09-23 00:44:42 -0700696 * skb_queue_next - return the next packet in the queue
697 * @list: queue head
698 * @skb: current buffer
699 *
700 * Return the next packet in @list after @skb. It is only valid to
701 * call this if skb_queue_is_last() evaluates to false.
702 */
703static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
704 const struct sk_buff *skb)
705{
706 /* This BUG_ON may seem severe, but if we just return then we
707 * are going to dereference garbage.
708 */
709 BUG_ON(skb_queue_is_last(list, skb));
710 return skb->next;
711}
712
713/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800714 * skb_queue_prev - return the prev packet in the queue
715 * @list: queue head
716 * @skb: current buffer
717 *
718 * Return the prev packet in @list before @skb. It is only valid to
719 * call this if skb_queue_is_first() evaluates to false.
720 */
721static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
722 const struct sk_buff *skb)
723{
724 /* This BUG_ON may seem severe, but if we just return then we
725 * are going to dereference garbage.
726 */
727 BUG_ON(skb_queue_is_first(list, skb));
728 return skb->prev;
729}
730
731/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700732 * skb_get - reference buffer
733 * @skb: buffer to reference
734 *
735 * Makes another reference to a socket buffer and returns a pointer
736 * to the buffer.
737 */
738static inline struct sk_buff *skb_get(struct sk_buff *skb)
739{
740 atomic_inc(&skb->users);
741 return skb;
742}
743
744/*
745 * If users == 1, we are the only owner and are can avoid redundant
746 * atomic change.
747 */
748
749/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700750 * skb_cloned - is the buffer a clone
751 * @skb: buffer to check
752 *
753 * Returns true if the buffer was generated with skb_clone() and is
754 * one of multiple shared copies of the buffer. Cloned buffers are
755 * shared data so must not be written to under normal circumstances.
756 */
757static inline int skb_cloned(const struct sk_buff *skb)
758{
759 return skb->cloned &&
760 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
761}
762
763/**
764 * skb_header_cloned - is the header a clone
765 * @skb: buffer to check
766 *
767 * Returns true if modifying the header part of the buffer requires
768 * the data to be copied.
769 */
770static inline int skb_header_cloned(const struct sk_buff *skb)
771{
772 int dataref;
773
774 if (!skb->cloned)
775 return 0;
776
777 dataref = atomic_read(&skb_shinfo(skb)->dataref);
778 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
779 return dataref != 1;
780}
781
782/**
783 * skb_header_release - release reference to header
784 * @skb: buffer to operate on
785 *
786 * Drop a reference to the header part of the buffer. This is done
787 * by acquiring a payload reference. You must not read from the header
788 * part of skb->data after this.
789 */
790static inline void skb_header_release(struct sk_buff *skb)
791{
792 BUG_ON(skb->nohdr);
793 skb->nohdr = 1;
794 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
795}
796
797/**
798 * skb_shared - is the buffer shared
799 * @skb: buffer to check
800 *
801 * Returns true if more than one person has a reference to this
802 * buffer.
803 */
804static inline int skb_shared(const struct sk_buff *skb)
805{
806 return atomic_read(&skb->users) != 1;
807}
808
809/**
810 * skb_share_check - check if buffer is shared and if so clone it
811 * @skb: buffer to check
812 * @pri: priority for memory allocation
813 *
814 * If the buffer is shared the buffer is cloned and the old copy
815 * drops a reference. A new clone with a single reference is returned.
816 * If the buffer is not shared the original buffer is returned. When
817 * being called from interrupt status or with spinlocks held pri must
818 * be GFP_ATOMIC.
819 *
820 * NULL is returned on a memory allocation failure.
821 */
Victor Fusco86a76ca2005-07-08 14:57:47 -0700822static inline struct sk_buff *skb_share_check(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100823 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700824{
825 might_sleep_if(pri & __GFP_WAIT);
826 if (skb_shared(skb)) {
827 struct sk_buff *nskb = skb_clone(skb, pri);
828 kfree_skb(skb);
829 skb = nskb;
830 }
831 return skb;
832}
833
834/*
835 * Copy shared buffers into a new sk_buff. We effectively do COW on
836 * packets to handle cases where we have a local reader and forward
837 * and a couple of other messy ones. The normal one is tcpdumping
838 * a packet thats being forwarded.
839 */
840
841/**
842 * skb_unshare - make a copy of a shared buffer
843 * @skb: buffer to check
844 * @pri: priority for memory allocation
845 *
846 * If the socket buffer is a clone then this function creates a new
847 * copy of the data, drops a reference count on the old copy and returns
848 * the new copy with the reference count at 1. If the buffer is not a clone
849 * the original buffer is returned. When called with a spinlock held or
850 * from interrupt state @pri must be %GFP_ATOMIC
851 *
852 * %NULL is returned on a memory allocation failure.
853 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -0700854static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100855 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856{
857 might_sleep_if(pri & __GFP_WAIT);
858 if (skb_cloned(skb)) {
859 struct sk_buff *nskb = skb_copy(skb, pri);
860 kfree_skb(skb); /* Free our shared copy */
861 skb = nskb;
862 }
863 return skb;
864}
865
866/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -0800867 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868 * @list_: list to peek at
869 *
870 * Peek an &sk_buff. Unlike most other operations you _MUST_
871 * be careful with this one. A peek leaves the buffer on the
872 * list and someone else may run off with it. You must hold
873 * the appropriate locks or have a private queue to do this.
874 *
875 * Returns %NULL for an empty list or a pointer to the head element.
876 * The reference count is not incremented and the reference is therefore
877 * volatile. Use with caution.
878 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -0400879static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880{
Eric Dumazet05bdd2f2011-10-20 17:45:43 -0400881 struct sk_buff *list = ((const struct sk_buff *)list_)->next;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700882 if (list == (struct sk_buff *)list_)
883 list = NULL;
884 return list;
885}
886
887/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +0000888 * skb_peek_next - peek skb following the given one from a queue
889 * @skb: skb to start from
890 * @list_: list to peek at
891 *
892 * Returns %NULL when the end of the list is met or a pointer to the
893 * next element. The reference count is not incremented and the
894 * reference is therefore volatile. Use with caution.
895 */
896static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
897 const struct sk_buff_head *list_)
898{
899 struct sk_buff *next = skb->next;
900 if (next == (struct sk_buff *)list_)
901 next = NULL;
902 return next;
903}
904
905/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -0800906 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907 * @list_: list to peek at
908 *
909 * Peek an &sk_buff. Unlike most other operations you _MUST_
910 * be careful with this one. A peek leaves the buffer on the
911 * list and someone else may run off with it. You must hold
912 * the appropriate locks or have a private queue to do this.
913 *
914 * Returns %NULL for an empty list or a pointer to the tail element.
915 * The reference count is not incremented and the reference is therefore
916 * volatile. Use with caution.
917 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -0400918static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700919{
Eric Dumazet05bdd2f2011-10-20 17:45:43 -0400920 struct sk_buff *list = ((const struct sk_buff *)list_)->prev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700921 if (list == (struct sk_buff *)list_)
922 list = NULL;
923 return list;
924}
925
926/**
927 * skb_queue_len - get queue length
928 * @list_: list to measure
929 *
930 * Return the length of an &sk_buff queue.
931 */
932static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
933{
934 return list_->qlen;
935}
936
David S. Miller67fed452008-09-21 22:36:24 -0700937/**
938 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
939 * @list: queue to initialize
940 *
941 * This initializes only the list and queue length aspects of
942 * an sk_buff_head object. This allows to initialize the list
943 * aspects of an sk_buff_head without reinitializing things like
944 * the spinlock. It can also be used for on-stack sk_buff_head
945 * objects where the spinlock is known to not be used.
946 */
947static inline void __skb_queue_head_init(struct sk_buff_head *list)
948{
949 list->prev = list->next = (struct sk_buff *)list;
950 list->qlen = 0;
951}
952
Arjan van de Ven76f10ad2006-08-02 14:06:55 -0700953/*
954 * This function creates a split out lock class for each invocation;
955 * this is needed for now since a whole lot of users of the skb-queue
956 * infrastructure in drivers have different locking usage (in hardirq)
957 * than the networking core (in softirq only). In the long run either the
958 * network layer or drivers should need annotation to consolidate the
959 * main types of usage into 3 classes.
960 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700961static inline void skb_queue_head_init(struct sk_buff_head *list)
962{
963 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -0700964 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965}
966
Pavel Emelianovc2ecba72007-04-17 12:45:31 -0700967static inline void skb_queue_head_init_class(struct sk_buff_head *list,
968 struct lock_class_key *class)
969{
970 skb_queue_head_init(list);
971 lockdep_set_class(&list->lock, class);
972}
973
Linus Torvalds1da177e2005-04-16 15:20:36 -0700974/*
Gerrit Renkerbf299272008-04-14 00:04:51 -0700975 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700976 *
977 * The "__skb_xxxx()" functions are the non-atomic ones that
978 * can only be called with interrupts disabled.
979 */
Gerrit Renkerbf299272008-04-14 00:04:51 -0700980extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
981static inline void __skb_insert(struct sk_buff *newsk,
982 struct sk_buff *prev, struct sk_buff *next,
983 struct sk_buff_head *list)
984{
985 newsk->next = next;
986 newsk->prev = prev;
987 next->prev = prev->next = newsk;
988 list->qlen++;
989}
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990
David S. Miller67fed452008-09-21 22:36:24 -0700991static inline void __skb_queue_splice(const struct sk_buff_head *list,
992 struct sk_buff *prev,
993 struct sk_buff *next)
994{
995 struct sk_buff *first = list->next;
996 struct sk_buff *last = list->prev;
997
998 first->prev = prev;
999 prev->next = first;
1000
1001 last->next = next;
1002 next->prev = last;
1003}
1004
1005/**
1006 * skb_queue_splice - join two skb lists, this is designed for stacks
1007 * @list: the new list to add
1008 * @head: the place to add it in the first list
1009 */
1010static inline void skb_queue_splice(const struct sk_buff_head *list,
1011 struct sk_buff_head *head)
1012{
1013 if (!skb_queue_empty(list)) {
1014 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001015 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001016 }
1017}
1018
1019/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001020 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001021 * @list: the new list to add
1022 * @head: the place to add it in the first list
1023 *
1024 * The list at @list is reinitialised
1025 */
1026static inline void skb_queue_splice_init(struct sk_buff_head *list,
1027 struct sk_buff_head *head)
1028{
1029 if (!skb_queue_empty(list)) {
1030 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001031 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001032 __skb_queue_head_init(list);
1033 }
1034}
1035
1036/**
1037 * skb_queue_splice_tail - join two skb lists, each list being a queue
1038 * @list: the new list to add
1039 * @head: the place to add it in the first list
1040 */
1041static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1042 struct sk_buff_head *head)
1043{
1044 if (!skb_queue_empty(list)) {
1045 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001046 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001047 }
1048}
1049
1050/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001051 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001052 * @list: the new list to add
1053 * @head: the place to add it in the first list
1054 *
1055 * Each of the lists is a queue.
1056 * The list at @list is reinitialised
1057 */
1058static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1059 struct sk_buff_head *head)
1060{
1061 if (!skb_queue_empty(list)) {
1062 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001063 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001064 __skb_queue_head_init(list);
1065 }
1066}
1067
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001069 * __skb_queue_after - queue a buffer at the list head
1070 * @list: list to use
1071 * @prev: place after this buffer
1072 * @newsk: buffer to queue
1073 *
1074 * Queue a buffer int the middle of a list. This function takes no locks
1075 * and you must therefore hold required locks before calling it.
1076 *
1077 * A buffer cannot be placed on two lists at the same time.
1078 */
1079static inline void __skb_queue_after(struct sk_buff_head *list,
1080 struct sk_buff *prev,
1081 struct sk_buff *newsk)
1082{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001083 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001084}
1085
Gerrit Renker7de6c032008-04-14 00:05:09 -07001086extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1087 struct sk_buff_head *list);
1088
Gerrit Renkerf5572852008-04-14 00:05:28 -07001089static inline void __skb_queue_before(struct sk_buff_head *list,
1090 struct sk_buff *next,
1091 struct sk_buff *newsk)
1092{
1093 __skb_insert(newsk, next->prev, next, list);
1094}
1095
Stephen Hemminger300ce172005-10-30 13:47:34 -08001096/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001097 * __skb_queue_head - queue a buffer at the list head
1098 * @list: list to use
1099 * @newsk: buffer to queue
1100 *
1101 * Queue a buffer at the start of a list. This function takes no locks
1102 * and you must therefore hold required locks before calling it.
1103 *
1104 * A buffer cannot be placed on two lists at the same time.
1105 */
1106extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1107static inline void __skb_queue_head(struct sk_buff_head *list,
1108 struct sk_buff *newsk)
1109{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001110 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001111}
1112
1113/**
1114 * __skb_queue_tail - queue a buffer at the list tail
1115 * @list: list to use
1116 * @newsk: buffer to queue
1117 *
1118 * Queue a buffer at the end of a list. This function takes no locks
1119 * and you must therefore hold required locks before calling it.
1120 *
1121 * A buffer cannot be placed on two lists at the same time.
1122 */
1123extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1124static inline void __skb_queue_tail(struct sk_buff_head *list,
1125 struct sk_buff *newsk)
1126{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001127 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128}
1129
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131 * remove sk_buff from list. _Must_ be called atomically, and with
1132 * the list known..
1133 */
David S. Miller8728b832005-08-09 19:25:21 -07001134extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001135static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1136{
1137 struct sk_buff *next, *prev;
1138
1139 list->qlen--;
1140 next = skb->next;
1141 prev = skb->prev;
1142 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001143 next->prev = prev;
1144 prev->next = next;
1145}
1146
Gerrit Renkerf525c062008-04-14 00:04:12 -07001147/**
1148 * __skb_dequeue - remove from the head of the queue
1149 * @list: list to dequeue from
1150 *
1151 * Remove the head of the list. This function does not take any locks
1152 * so must be used with appropriate locks held only. The head item is
1153 * returned or %NULL if the list is empty.
1154 */
1155extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1156static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1157{
1158 struct sk_buff *skb = skb_peek(list);
1159 if (skb)
1160 __skb_unlink(skb, list);
1161 return skb;
1162}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163
1164/**
1165 * __skb_dequeue_tail - remove from the tail of the queue
1166 * @list: list to dequeue from
1167 *
1168 * Remove the tail of the list. This function does not take any locks
1169 * so must be used with appropriate locks held only. The tail item is
1170 * returned or %NULL if the list is empty.
1171 */
1172extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1173static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1174{
1175 struct sk_buff *skb = skb_peek_tail(list);
1176 if (skb)
1177 __skb_unlink(skb, list);
1178 return skb;
1179}
1180
1181
David S. Millerbdcc0922012-03-07 20:53:36 -05001182static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183{
1184 return skb->data_len;
1185}
1186
1187static inline unsigned int skb_headlen(const struct sk_buff *skb)
1188{
1189 return skb->len - skb->data_len;
1190}
1191
1192static inline int skb_pagelen(const struct sk_buff *skb)
1193{
1194 int i, len = 0;
1195
1196 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001197 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001198 return len + skb_headlen(skb);
1199}
1200
Ian Campbell131ea662011-08-19 06:25:00 +00001201/**
1202 * __skb_fill_page_desc - initialise a paged fragment in an skb
1203 * @skb: buffer containing fragment to be initialised
1204 * @i: paged fragment index to initialise
1205 * @page: the page to use for this fragment
1206 * @off: the offset to the data with @page
1207 * @size: the length of the data
1208 *
1209 * Initialises the @i'th fragment of @skb to point to &size bytes at
1210 * offset @off within @page.
1211 *
1212 * Does not take any additional reference on the fragment.
1213 */
1214static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1215 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001216{
1217 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1218
Ian Campbella8605c62011-10-19 23:01:49 +00001219 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001220 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001221 skb_frag_size_set(frag, size);
Ian Campbell131ea662011-08-19 06:25:00 +00001222}
1223
1224/**
1225 * skb_fill_page_desc - initialise a paged fragment in an skb
1226 * @skb: buffer containing fragment to be initialised
1227 * @i: paged fragment index to initialise
1228 * @page: the page to use for this fragment
1229 * @off: the offset to the data with @page
1230 * @size: the length of the data
1231 *
1232 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1233 * @skb to point to &size bytes at offset @off within @page. In
1234 * addition updates @skb such that @i is the last fragment.
1235 *
1236 * Does not take any additional reference on the fragment.
1237 */
1238static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1239 struct page *page, int off, int size)
1240{
1241 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001242 skb_shinfo(skb)->nr_frags = i + 1;
1243}
1244
Peter Zijlstra654bed12008-10-07 14:22:33 -07001245extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
Eric Dumazet50269e12012-03-23 23:59:33 +00001246 int off, int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001247
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001249#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1251
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001252#ifdef NET_SKBUFF_DATA_USES_OFFSET
1253static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1254{
1255 return skb->head + skb->tail;
1256}
1257
1258static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1259{
1260 skb->tail = skb->data - skb->head;
1261}
1262
1263static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1264{
1265 skb_reset_tail_pointer(skb);
1266 skb->tail += offset;
1267}
1268#else /* NET_SKBUFF_DATA_USES_OFFSET */
1269static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1270{
1271 return skb->tail;
1272}
1273
1274static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1275{
1276 skb->tail = skb->data;
1277}
1278
1279static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1280{
1281 skb->tail = skb->data + offset;
1282}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001283
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001284#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1285
Linus Torvalds1da177e2005-04-16 15:20:36 -07001286/*
1287 * Add data to an sk_buff
1288 */
Ilpo Järvinen0dde3e12008-03-27 17:43:41 -07001289extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001290static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1291{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001292 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001293 SKB_LINEAR_ASSERT(skb);
1294 skb->tail += len;
1295 skb->len += len;
1296 return tmp;
1297}
1298
Ilpo Järvinenc2aa2702008-03-27 17:52:40 -07001299extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001300static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1301{
1302 skb->data -= len;
1303 skb->len += len;
1304 return skb->data;
1305}
1306
Ilpo Järvinen6be8ac22008-03-27 17:47:24 -07001307extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001308static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1309{
1310 skb->len -= len;
1311 BUG_ON(skb->len < skb->data_len);
1312 return skb->data += len;
1313}
1314
David S. Miller47d29642010-05-02 02:21:44 -07001315static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1316{
1317 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1318}
1319
Linus Torvalds1da177e2005-04-16 15:20:36 -07001320extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1321
1322static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1323{
1324 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001325 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001326 return NULL;
1327 skb->len -= len;
1328 return skb->data += len;
1329}
1330
1331static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1332{
1333 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1334}
1335
1336static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1337{
1338 if (likely(len <= skb_headlen(skb)))
1339 return 1;
1340 if (unlikely(len > skb->len))
1341 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001342 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343}
1344
1345/**
1346 * skb_headroom - bytes at buffer head
1347 * @skb: buffer to check
1348 *
1349 * Return the number of bytes of free space at the head of an &sk_buff.
1350 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001351static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352{
1353 return skb->data - skb->head;
1354}
1355
1356/**
1357 * skb_tailroom - bytes at buffer end
1358 * @skb: buffer to check
1359 *
1360 * Return the number of bytes of free space at the tail of an sk_buff
1361 */
1362static inline int skb_tailroom(const struct sk_buff *skb)
1363{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001364 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365}
1366
1367/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00001368 * skb_availroom - bytes at buffer end
1369 * @skb: buffer to check
1370 *
1371 * Return the number of bytes of free space at the tail of an sk_buff
1372 * allocated by sk_stream_alloc()
1373 */
1374static inline int skb_availroom(const struct sk_buff *skb)
1375{
Eric Dumazetb6da5782013-03-14 05:40:32 +00001376 if (skb_is_nonlinear(skb))
1377 return 0;
1378
1379 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00001380}
1381
1382/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383 * skb_reserve - adjust headroom
1384 * @skb: buffer to alter
1385 * @len: bytes to move
1386 *
1387 * Increase the headroom of an empty &sk_buff by reducing the tail
1388 * room. This is only allowed for an empty buffer.
1389 */
David S. Miller82431262006-01-17 02:54:21 -08001390static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001391{
1392 skb->data += len;
1393 skb->tail += len;
1394}
1395
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00001396static inline void skb_reset_mac_len(struct sk_buff *skb)
1397{
1398 skb->mac_len = skb->network_header - skb->mac_header;
1399}
1400
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001401#ifdef NET_SKBUFF_DATA_USES_OFFSET
1402static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1403{
1404 return skb->head + skb->transport_header;
1405}
1406
1407static inline void skb_reset_transport_header(struct sk_buff *skb)
1408{
1409 skb->transport_header = skb->data - skb->head;
1410}
1411
1412static inline void skb_set_transport_header(struct sk_buff *skb,
1413 const int offset)
1414{
1415 skb_reset_transport_header(skb);
1416 skb->transport_header += offset;
1417}
1418
1419static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1420{
1421 return skb->head + skb->network_header;
1422}
1423
1424static inline void skb_reset_network_header(struct sk_buff *skb)
1425{
1426 skb->network_header = skb->data - skb->head;
1427}
1428
1429static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1430{
1431 skb_reset_network_header(skb);
1432 skb->network_header += offset;
1433}
1434
1435static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1436{
1437 return skb->head + skb->mac_header;
1438}
1439
1440static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1441{
1442 return skb->mac_header != ~0U;
1443}
1444
1445static inline void skb_reset_mac_header(struct sk_buff *skb)
1446{
1447 skb->mac_header = skb->data - skb->head;
1448}
1449
1450static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1451{
1452 skb_reset_mac_header(skb);
1453 skb->mac_header += offset;
1454}
1455
1456#else /* NET_SKBUFF_DATA_USES_OFFSET */
1457
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001458static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1459{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001460 return skb->transport_header;
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001461}
1462
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001463static inline void skb_reset_transport_header(struct sk_buff *skb)
1464{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001465 skb->transport_header = skb->data;
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001466}
1467
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001468static inline void skb_set_transport_header(struct sk_buff *skb,
1469 const int offset)
1470{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001471 skb->transport_header = skb->data + offset;
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001472}
1473
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001474static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1475{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001476 return skb->network_header;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001477}
1478
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001479static inline void skb_reset_network_header(struct sk_buff *skb)
1480{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001481 skb->network_header = skb->data;
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001482}
1483
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001484static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1485{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001486 skb->network_header = skb->data + offset;
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001487}
1488
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001489static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1490{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001491 return skb->mac_header;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001492}
1493
1494static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1495{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001496 return skb->mac_header != NULL;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001497}
1498
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001499static inline void skb_reset_mac_header(struct sk_buff *skb)
1500{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001501 skb->mac_header = skb->data;
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001502}
1503
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001504static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1505{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001506 skb->mac_header = skb->data + offset;
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001507}
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001508#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1509
Eric Dumazet03606892012-02-23 10:55:02 +00001510static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1511{
1512 if (skb_mac_header_was_set(skb)) {
1513 const unsigned char *old_mac = skb_mac_header(skb);
1514
1515 skb_set_mac_header(skb, -skb->mac_len);
1516 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1517 }
1518}
1519
Michał Mirosław04fb4512010-12-14 15:24:08 +00001520static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1521{
1522 return skb->csum_start - skb_headroom(skb);
1523}
1524
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001525static inline int skb_transport_offset(const struct sk_buff *skb)
1526{
1527 return skb_transport_header(skb) - skb->data;
1528}
1529
1530static inline u32 skb_network_header_len(const struct sk_buff *skb)
1531{
1532 return skb->transport_header - skb->network_header;
1533}
1534
1535static inline int skb_network_offset(const struct sk_buff *skb)
1536{
1537 return skb_network_header(skb) - skb->data;
1538}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001539
Changli Gaof9599ce2010-08-04 04:43:44 +00001540static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1541{
1542 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1543}
1544
Linus Torvalds1da177e2005-04-16 15:20:36 -07001545/*
1546 * CPUs often take a performance hit when accessing unaligned memory
1547 * locations. The actual performance hit varies, it can be small if the
1548 * hardware handles it or large if we have to take an exception and fix it
1549 * in software.
1550 *
1551 * Since an ethernet header is 14 bytes network drivers often end up with
1552 * the IP header at an unaligned offset. The IP header can be aligned by
1553 * shifting the start of the packet by 2 bytes. Drivers should do this
1554 * with:
1555 *
Tobias Klauser8660c122009-07-13 22:48:16 +00001556 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001557 *
1558 * The downside to this alignment of the IP header is that the DMA is now
1559 * unaligned. On some architectures the cost of an unaligned DMA is high
1560 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00001561 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001562 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1563 * to be overridden.
1564 */
1565#ifndef NET_IP_ALIGN
1566#define NET_IP_ALIGN 2
1567#endif
1568
Anton Blanchard025be812006-03-31 02:27:06 -08001569/*
1570 * The networking layer reserves some headroom in skb data (via
1571 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1572 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08001573 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08001574 *
1575 * Unfortunately this headroom changes the DMA alignment of the resulting
1576 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1577 * on some architectures. An architecture can override this value,
1578 * perhaps setting it to a cacheline in size (since that will maintain
1579 * cacheline alignment of the DMA). It must be a power of 2.
1580 *
David S. Millerd6301d32009-02-08 19:24:13 -08001581 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08001582 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07001583 *
1584 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1585 * to reduce average number of cache lines per packet.
1586 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07001587 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08001588 */
1589#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07001590#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08001591#endif
1592
Herbert Xu3cc0e872006-06-09 16:13:38 -07001593extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001594
1595static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1596{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00001597 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07001598 WARN_ON(1);
1599 return;
1600 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001601 skb->len = len;
1602 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001603}
1604
Ilpo Järvinen419ae742008-03-27 17:54:01 -07001605extern void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001606
1607static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1608{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001609 if (skb->data_len)
1610 return ___pskb_trim(skb, len);
1611 __skb_trim(skb, len);
1612 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001613}
1614
1615static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1616{
1617 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1618}
1619
1620/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07001621 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1622 * @skb: buffer to alter
1623 * @len: new length
1624 *
1625 * This is identical to pskb_trim except that the caller knows that
1626 * the skb is not cloned so we should never get an error due to out-
1627 * of-memory.
1628 */
1629static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1630{
1631 int err = pskb_trim(skb, len);
1632 BUG_ON(err);
1633}
1634
1635/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001636 * skb_orphan - orphan a buffer
1637 * @skb: buffer to orphan
1638 *
1639 * If a buffer currently has an owner then we call the owner's
1640 * destructor function and make the @skb unowned. The buffer continues
1641 * to exist but is no longer charged to its former owner.
1642 */
1643static inline void skb_orphan(struct sk_buff *skb)
1644{
1645 if (skb->destructor)
1646 skb->destructor(skb);
1647 skb->destructor = NULL;
1648 skb->sk = NULL;
1649}
1650
1651/**
1652 * __skb_queue_purge - empty a list
1653 * @list: list to empty
1654 *
1655 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1656 * the list and one reference dropped. This function does not take the
1657 * list lock and the caller must hold the relevant locks to use it.
1658 */
1659extern void skb_queue_purge(struct sk_buff_head *list);
1660static inline void __skb_queue_purge(struct sk_buff_head *list)
1661{
1662 struct sk_buff *skb;
1663 while ((skb = __skb_dequeue(list)) != NULL)
1664 kfree_skb(skb);
1665}
1666
1667/**
Christoph Hellwigb4e54de2006-07-24 15:31:14 -07001668 * __dev_alloc_skb - allocate an skbuff for receiving
Linus Torvalds1da177e2005-04-16 15:20:36 -07001669 * @length: length to allocate
1670 * @gfp_mask: get_free_pages mask, passed to alloc_skb
1671 *
1672 * Allocate a new &sk_buff and assign it a usage count of one. The
1673 * buffer has unspecified headroom built in. Users should allocate
1674 * the headroom they think they need without accounting for the
1675 * built in space. The built in space is used for optimisations.
1676 *
Christoph Hellwig766ea8c2006-08-07 15:49:53 -07001677 * %NULL is returned if there is no free memory.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001678 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001679static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
Al Virodd0fc662005-10-07 07:46:04 +01001680 gfp_t gfp_mask)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001681{
Anton Blanchard025be812006-03-31 02:27:06 -08001682 struct sk_buff *skb = alloc_skb(length + NET_SKB_PAD, gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001683 if (likely(skb))
Anton Blanchard025be812006-03-31 02:27:06 -08001684 skb_reserve(skb, NET_SKB_PAD);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001685 return skb;
1686}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687
Ilpo Järvinenf58518e2008-03-27 17:51:31 -07001688extern struct sk_buff *dev_alloc_skb(unsigned int length);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001689
Christoph Hellwig8af27452006-07-31 22:35:23 -07001690extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
1691 unsigned int length, gfp_t gfp_mask);
1692
1693/**
1694 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1695 * @dev: network device to receive on
1696 * @length: length to allocate
1697 *
1698 * Allocate a new &sk_buff and assign it a usage count of one. The
1699 * buffer has unspecified headroom built in. Users should allocate
1700 * the headroom they think they need without accounting for the
1701 * built in space. The built in space is used for optimisations.
1702 *
1703 * %NULL is returned if there is no free memory. Although this function
1704 * allocates memory it can be called from an interrupt.
1705 */
1706static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
1707 unsigned int length)
1708{
1709 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1710}
1711
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001712static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
1713 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00001714{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001715 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00001716
1717 if (NET_IP_ALIGN && skb)
1718 skb_reserve(skb, NET_IP_ALIGN);
1719 return skb;
1720}
1721
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001722static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1723 unsigned int length)
1724{
1725 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
1726}
1727
Eric Dumazet564824b2010-10-11 19:05:25 +00001728/**
Ian Campbell131ea662011-08-19 06:25:00 +00001729 * skb_frag_page - retrieve the page refered to by a paged fragment
1730 * @frag: the paged fragment
1731 *
1732 * Returns the &struct page associated with @frag.
1733 */
1734static inline struct page *skb_frag_page(const skb_frag_t *frag)
1735{
Ian Campbella8605c62011-10-19 23:01:49 +00001736 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00001737}
1738
1739/**
1740 * __skb_frag_ref - take an addition reference on a paged fragment.
1741 * @frag: the paged fragment
1742 *
1743 * Takes an additional reference on the paged fragment @frag.
1744 */
1745static inline void __skb_frag_ref(skb_frag_t *frag)
1746{
1747 get_page(skb_frag_page(frag));
1748}
1749
1750/**
1751 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
1752 * @skb: the buffer
1753 * @f: the fragment offset.
1754 *
1755 * Takes an additional reference on the @f'th paged fragment of @skb.
1756 */
1757static inline void skb_frag_ref(struct sk_buff *skb, int f)
1758{
1759 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
1760}
1761
1762/**
1763 * __skb_frag_unref - release a reference on a paged fragment.
1764 * @frag: the paged fragment
1765 *
1766 * Releases a reference on the paged fragment @frag.
1767 */
1768static inline void __skb_frag_unref(skb_frag_t *frag)
1769{
1770 put_page(skb_frag_page(frag));
1771}
1772
1773/**
1774 * skb_frag_unref - release a reference on a paged fragment of an skb.
1775 * @skb: the buffer
1776 * @f: the fragment offset
1777 *
1778 * Releases a reference on the @f'th paged fragment of @skb.
1779 */
1780static inline void skb_frag_unref(struct sk_buff *skb, int f)
1781{
1782 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
1783}
1784
1785/**
1786 * skb_frag_address - gets the address of the data contained in a paged fragment
1787 * @frag: the paged fragment buffer
1788 *
1789 * Returns the address of the data within @frag. The page must already
1790 * be mapped.
1791 */
1792static inline void *skb_frag_address(const skb_frag_t *frag)
1793{
1794 return page_address(skb_frag_page(frag)) + frag->page_offset;
1795}
1796
1797/**
1798 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
1799 * @frag: the paged fragment buffer
1800 *
1801 * Returns the address of the data within @frag. Checks that the page
1802 * is mapped and returns %NULL otherwise.
1803 */
1804static inline void *skb_frag_address_safe(const skb_frag_t *frag)
1805{
1806 void *ptr = page_address(skb_frag_page(frag));
1807 if (unlikely(!ptr))
1808 return NULL;
1809
1810 return ptr + frag->page_offset;
1811}
1812
1813/**
1814 * __skb_frag_set_page - sets the page contained in a paged fragment
1815 * @frag: the paged fragment
1816 * @page: the page to set
1817 *
1818 * Sets the fragment @frag to contain @page.
1819 */
1820static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
1821{
Ian Campbella8605c62011-10-19 23:01:49 +00001822 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00001823}
1824
1825/**
1826 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
1827 * @skb: the buffer
1828 * @f: the fragment offset
1829 * @page: the page to set
1830 *
1831 * Sets the @f'th fragment of @skb to contain @page.
1832 */
1833static inline void skb_frag_set_page(struct sk_buff *skb, int f,
1834 struct page *page)
1835{
1836 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
1837}
1838
1839/**
1840 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00001841 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00001842 * @frag: the paged fragment to map
1843 * @offset: the offset within the fragment (starting at the
1844 * fragment's own offset)
1845 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00001846 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00001847 *
1848 * Maps the page associated with @frag to @device.
1849 */
1850static inline dma_addr_t skb_frag_dma_map(struct device *dev,
1851 const skb_frag_t *frag,
1852 size_t offset, size_t size,
1853 enum dma_data_direction dir)
1854{
1855 return dma_map_page(dev, skb_frag_page(frag),
1856 frag->page_offset + offset, size, dir);
1857}
1858
Eric Dumazet117632e2011-12-03 21:39:53 +00001859static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
1860 gfp_t gfp_mask)
1861{
1862 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
1863}
1864
Ian Campbell131ea662011-08-19 06:25:00 +00001865/**
Patrick McHardy334a8132007-06-25 04:35:20 -07001866 * skb_clone_writable - is the header of a clone writable
1867 * @skb: buffer to check
1868 * @len: length up to which to write
1869 *
1870 * Returns true if modifying the header part of the cloned buffer
1871 * does not requires the data to be copied.
1872 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001873static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07001874{
1875 return !skb_header_cloned(skb) &&
1876 skb_headroom(skb) + len <= skb->hdr_len;
1877}
1878
Herbert Xud9cc2042007-09-16 16:21:16 -07001879static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
1880 int cloned)
1881{
1882 int delta = 0;
1883
Herbert Xud9cc2042007-09-16 16:21:16 -07001884 if (headroom > skb_headroom(skb))
1885 delta = headroom - skb_headroom(skb);
1886
1887 if (delta || cloned)
1888 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
1889 GFP_ATOMIC);
1890 return 0;
1891}
1892
Patrick McHardy334a8132007-06-25 04:35:20 -07001893/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001894 * skb_cow - copy header of skb when it is required
1895 * @skb: buffer to cow
1896 * @headroom: needed headroom
1897 *
1898 * If the skb passed lacks sufficient headroom or its data part
1899 * is shared, data is reallocated. If reallocation fails, an error
1900 * is returned and original skb is not changed.
1901 *
1902 * The result is skb with writable area skb->head...skb->tail
1903 * and at least @headroom of space at head.
1904 */
1905static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
1906{
Herbert Xud9cc2042007-09-16 16:21:16 -07001907 return __skb_cow(skb, headroom, skb_cloned(skb));
1908}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001909
Herbert Xud9cc2042007-09-16 16:21:16 -07001910/**
1911 * skb_cow_head - skb_cow but only making the head writable
1912 * @skb: buffer to cow
1913 * @headroom: needed headroom
1914 *
1915 * This function is identical to skb_cow except that we replace the
1916 * skb_cloned check by skb_header_cloned. It should be used when
1917 * you only need to push on some header and do not need to modify
1918 * the data.
1919 */
1920static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
1921{
1922 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001923}
1924
1925/**
1926 * skb_padto - pad an skbuff up to a minimal size
1927 * @skb: buffer to pad
1928 * @len: minimal length
1929 *
1930 * Pads up a buffer to ensure the trailing bytes exist and are
1931 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07001932 * is untouched. Otherwise it is extended. Returns zero on
1933 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001934 */
1935
Herbert Xu5b057c62006-06-23 02:06:41 -07001936static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001937{
1938 unsigned int size = skb->len;
1939 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07001940 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001941 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001942}
1943
1944static inline int skb_add_data(struct sk_buff *skb,
1945 char __user *from, int copy)
1946{
1947 const int off = skb->len;
1948
1949 if (skb->ip_summed == CHECKSUM_NONE) {
1950 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08001951 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
Linus Torvalds1da177e2005-04-16 15:20:36 -07001952 copy, 0, &err);
1953 if (!err) {
1954 skb->csum = csum_block_add(skb->csum, csum, off);
1955 return 0;
1956 }
1957 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
1958 return 0;
1959
1960 __skb_trim(skb, off);
1961 return -EFAULT;
1962}
1963
1964static inline int skb_can_coalesce(struct sk_buff *skb, int i,
Ian Campbellea2ab692011-08-22 23:44:58 +00001965 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001966{
1967 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00001968 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07001969
Ian Campbellea2ab692011-08-22 23:44:58 +00001970 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00001971 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001972 }
1973 return 0;
1974}
1975
Herbert Xu364c6ba2006-06-09 16:10:40 -07001976static inline int __skb_linearize(struct sk_buff *skb)
1977{
1978 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
1979}
1980
Linus Torvalds1da177e2005-04-16 15:20:36 -07001981/**
1982 * skb_linearize - convert paged skb to linear one
1983 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07001984 *
1985 * If there is no free memory -ENOMEM is returned, otherwise zero
1986 * is returned and the old skb data released.
1987 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07001988static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001989{
Herbert Xu364c6ba2006-06-09 16:10:40 -07001990 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
1991}
1992
1993/**
1994 * skb_linearize_cow - make sure skb is linear and writable
1995 * @skb: buffer to process
1996 *
1997 * If there is no free memory -ENOMEM is returned, otherwise zero
1998 * is returned and the old skb data released.
1999 */
2000static inline int skb_linearize_cow(struct sk_buff *skb)
2001{
2002 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2003 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002004}
2005
2006/**
2007 * skb_postpull_rcsum - update checksum for received skb after pull
2008 * @skb: buffer to update
2009 * @start: start of data before pull
2010 * @len: length of data pulled
2011 *
2012 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002013 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2014 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002015 */
2016
2017static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002018 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002019{
Patrick McHardy84fa7932006-08-29 16:44:56 -07002020 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2022}
2023
Herbert Xucbb042f2006-03-20 22:43:56 -08002024unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2025
Linus Torvalds1da177e2005-04-16 15:20:36 -07002026/**
2027 * pskb_trim_rcsum - trim received skb and update checksum
2028 * @skb: buffer to trim
2029 * @len: new length
2030 *
2031 * This is exactly the same as pskb_trim except that it ensures the
2032 * checksum of received packets are still valid after the operation.
2033 */
2034
2035static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2036{
Stephen Hemminger0e4e4222005-09-08 12:32:03 -07002037 if (likely(len >= skb->len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002038 return 0;
Patrick McHardy84fa7932006-08-29 16:44:56 -07002039 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002040 skb->ip_summed = CHECKSUM_NONE;
2041 return __pskb_trim(skb, len);
2042}
2043
Linus Torvalds1da177e2005-04-16 15:20:36 -07002044#define skb_queue_walk(queue, skb) \
2045 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002046 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002047 skb = skb->next)
2048
James Chapman46f89142007-04-30 00:07:31 -07002049#define skb_queue_walk_safe(queue, skb, tmp) \
2050 for (skb = (queue)->next, tmp = skb->next; \
2051 skb != (struct sk_buff *)(queue); \
2052 skb = tmp, tmp = skb->next)
2053
David S. Miller1164f522008-09-23 00:49:44 -07002054#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002055 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002056 skb = skb->next)
2057
2058#define skb_queue_walk_from_safe(queue, skb, tmp) \
2059 for (tmp = skb->next; \
2060 skb != (struct sk_buff *)(queue); \
2061 skb = tmp, tmp = skb->next)
2062
Stephen Hemminger300ce172005-10-30 13:47:34 -08002063#define skb_queue_reverse_walk(queue, skb) \
2064 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002065 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08002066 skb = skb->prev)
2067
David S. Miller686a2952011-01-20 22:47:32 -08002068#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2069 for (skb = (queue)->prev, tmp = skb->prev; \
2070 skb != (struct sk_buff *)(queue); \
2071 skb = tmp, tmp = skb->prev)
2072
2073#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2074 for (tmp = skb->prev; \
2075 skb != (struct sk_buff *)(queue); \
2076 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002077
David S. Miller21dc3302010-08-23 00:13:46 -07002078static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07002079{
2080 return skb_shinfo(skb)->frag_list != NULL;
2081}
2082
2083static inline void skb_frag_list_init(struct sk_buff *skb)
2084{
2085 skb_shinfo(skb)->frag_list = NULL;
2086}
2087
2088static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2089{
2090 frag->next = skb_shinfo(skb)->frag_list;
2091 skb_shinfo(skb)->frag_list = frag;
2092}
2093
2094#define skb_walk_frags(skb, iter) \
2095 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2096
Herbert Xua59322b2007-12-05 01:53:40 -08002097extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
Pavel Emelyanov3f518bf2012-02-21 07:30:58 +00002098 int *peeked, int *off, int *err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002099extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
2100 int noblock, int *err);
2101extern unsigned int datagram_poll(struct file *file, struct socket *sock,
2102 struct poll_table_struct *wait);
2103extern int skb_copy_datagram_iovec(const struct sk_buff *from,
2104 int offset, struct iovec *to,
2105 int size);
Herbert Xufb286bb2005-11-10 13:01:24 -08002106extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002107 int hlen,
2108 struct iovec *iov);
Rusty Russelldb543c12008-08-15 15:13:53 -07002109extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
2110 int offset,
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +00002111 const struct iovec *from,
2112 int from_offset,
Rusty Russelldb543c12008-08-15 15:13:53 -07002113 int len);
Michael S. Tsirkin0a1ec072009-04-20 01:25:46 +00002114extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
2115 int offset,
2116 const struct iovec *to,
2117 int to_offset,
2118 int size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002119extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
Eric Dumazet9d410c72009-10-30 05:03:53 +00002120extern void skb_free_datagram_locked(struct sock *sk,
2121 struct sk_buff *skb);
Herbert Xu27ab2562007-12-05 01:51:58 -08002122extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
Herbert Xu3305b802005-12-13 23:16:37 -08002123 unsigned int flags);
Al Viro2bbbc862006-11-14 21:37:14 -08002124extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
2125 int len, __wsum csum);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002126extern int skb_copy_bits(const struct sk_buff *skb, int offset,
2127 void *to, int len);
Stephen Hemminger0c6fcc82007-04-20 16:40:01 -07002128extern int skb_store_bits(struct sk_buff *skb, int offset,
2129 const void *from, int len);
Al Viro81d77662006-11-14 21:37:33 -08002130extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002131 int offset, u8 *to, int len,
Al Viro81d77662006-11-14 21:37:33 -08002132 __wsum csum);
Jens Axboe9c55e012007-11-06 23:30:13 -08002133extern int skb_splice_bits(struct sk_buff *skb,
2134 unsigned int offset,
2135 struct pipe_inode_info *pipe,
2136 unsigned int len,
2137 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002138extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
2139extern void skb_split(struct sk_buff *skb,
2140 struct sk_buff *skb1, const u32 len);
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08002141extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
2142 int shiftlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002143
Michał Mirosławc8f44af2011-11-15 15:29:55 +00002144extern struct sk_buff *skb_segment(struct sk_buff *skb,
2145 netdev_features_t features);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002146
Linus Torvalds1da177e2005-04-16 15:20:36 -07002147static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2148 int len, void *buffer)
2149{
2150 int hlen = skb_headlen(skb);
2151
Patrick McHardy55820ee2005-07-05 14:08:10 -07002152 if (hlen - offset >= len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002153 return skb->data + offset;
2154
2155 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2156 return NULL;
2157
2158 return buffer;
2159}
2160
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03002161static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2162 void *to,
2163 const unsigned int len)
2164{
2165 memcpy(to, skb->data, len);
2166}
2167
2168static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2169 const int offset, void *to,
2170 const unsigned int len)
2171{
2172 memcpy(to, skb->data + offset, len);
2173}
2174
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03002175static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2176 const void *from,
2177 const unsigned int len)
2178{
2179 memcpy(skb->data, from, len);
2180}
2181
2182static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2183 const int offset,
2184 const void *from,
2185 const unsigned int len)
2186{
2187 memcpy(skb->data + offset, from, len);
2188}
2189
Linus Torvalds1da177e2005-04-16 15:20:36 -07002190extern void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002191
Patrick Ohlyac45f602009-02-12 05:03:37 +00002192static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2193{
2194 return skb->tstamp;
2195}
2196
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002197/**
2198 * skb_get_timestamp - get timestamp from a skb
2199 * @skb: skb to get stamp from
2200 * @stamp: pointer to struct timeval to store stamp in
2201 *
2202 * Timestamps are stored in the skb as offsets to a base timestamp.
2203 * This function converts the offset back to a struct timeval and stores
2204 * it in stamp.
2205 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00002206static inline void skb_get_timestamp(const struct sk_buff *skb,
2207 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002208{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002209 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002210}
2211
Patrick Ohlyac45f602009-02-12 05:03:37 +00002212static inline void skb_get_timestampns(const struct sk_buff *skb,
2213 struct timespec *stamp)
2214{
2215 *stamp = ktime_to_timespec(skb->tstamp);
2216}
2217
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002218static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002219{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002220 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002221}
2222
Stephen Hemminger164891a2007-04-23 22:26:16 -07002223static inline ktime_t net_timedelta(ktime_t t)
2224{
2225 return ktime_sub(ktime_get_real(), t);
2226}
2227
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07002228static inline ktime_t net_invalid_timestamp(void)
2229{
2230 return ktime_set(0, 0);
2231}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002232
Richard Cochranc1f19b52010-07-17 08:49:36 +00002233extern void skb_timestamping_init(void);
2234
2235#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2236
2237extern void skb_clone_tx_timestamp(struct sk_buff *skb);
2238extern bool skb_defer_rx_timestamp(struct sk_buff *skb);
2239
2240#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2241
2242static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2243{
2244}
2245
2246static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2247{
2248 return false;
2249}
2250
2251#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2252
2253/**
2254 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2255 *
Richard Cochranda92b192011-10-21 00:49:15 +00002256 * PHY drivers may accept clones of transmitted packets for
2257 * timestamping via their phy_driver.txtstamp method. These drivers
2258 * must call this function to return the skb back to the stack, with
2259 * or without a timestamp.
2260 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00002261 * @skb: clone of the the original outgoing packet
Richard Cochranda92b192011-10-21 00:49:15 +00002262 * @hwtstamps: hardware time stamps, may be NULL if not available
Richard Cochranc1f19b52010-07-17 08:49:36 +00002263 *
2264 */
2265void skb_complete_tx_timestamp(struct sk_buff *skb,
2266 struct skb_shared_hwtstamps *hwtstamps);
2267
Patrick Ohlyac45f602009-02-12 05:03:37 +00002268/**
2269 * skb_tstamp_tx - queue clone of skb with send time stamps
2270 * @orig_skb: the original outgoing packet
2271 * @hwtstamps: hardware time stamps, may be NULL if not available
2272 *
2273 * If the skb has a socket associated, then this function clones the
2274 * skb (thus sharing the actual data and optional structures), stores
2275 * the optional hardware time stamping information (if non NULL) or
2276 * generates a software time stamp (otherwise), then queues the clone
2277 * to the error queue of the socket. Errors are silently ignored.
2278 */
2279extern void skb_tstamp_tx(struct sk_buff *orig_skb,
2280 struct skb_shared_hwtstamps *hwtstamps);
2281
Richard Cochran4507a712010-07-17 08:48:28 +00002282static inline void sw_tx_timestamp(struct sk_buff *skb)
2283{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002284 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2285 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00002286 skb_tstamp_tx(skb, NULL);
2287}
2288
2289/**
2290 * skb_tx_timestamp() - Driver hook for transmit timestamping
2291 *
2292 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00002293 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00002294 *
2295 * @skb: A socket buffer.
2296 */
2297static inline void skb_tx_timestamp(struct sk_buff *skb)
2298{
Richard Cochranc1f19b52010-07-17 08:49:36 +00002299 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002300 sw_tx_timestamp(skb);
2301}
2302
Johannes Berg6e3e9392011-11-09 10:15:42 +01002303/**
2304 * skb_complete_wifi_ack - deliver skb with wifi status
2305 *
2306 * @skb: the original outgoing packet
2307 * @acked: ack status
2308 *
2309 */
2310void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2311
Herbert Xu759e5d02007-03-25 20:10:56 -07002312extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
Al Virob51655b2006-11-14 21:40:42 -08002313extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002314
Herbert Xu60476372007-04-09 11:59:39 -07002315static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2316{
2317 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2318}
2319
Herbert Xufb286bb2005-11-10 13:01:24 -08002320/**
2321 * skb_checksum_complete - Calculate checksum of an entire packet
2322 * @skb: packet to process
2323 *
2324 * This function calculates the checksum over the entire packet plus
2325 * the value of skb->csum. The latter can be used to supply the
2326 * checksum of a pseudo header as used by TCP/UDP. It returns the
2327 * checksum.
2328 *
2329 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2330 * this function can be used to verify that checksum on received
2331 * packets. In that case the function should return zero if the
2332 * checksum is correct. In particular, this function will return zero
2333 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2334 * hardware has already verified the correctness of the checksum.
2335 */
Al Viro4381ca32007-07-15 21:00:11 +01002336static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002337{
Herbert Xu60476372007-04-09 11:59:39 -07002338 return skb_csum_unnecessary(skb) ?
2339 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002340}
2341
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002342#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002343extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002344static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2345{
2346 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002347 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002348}
2349static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2350{
2351 if (nfct)
2352 atomic_inc(&nfct->use);
2353}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002354#endif
2355#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -08002356static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
2357{
2358 if (skb)
2359 atomic_inc(&skb->users);
2360}
2361static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
2362{
2363 if (skb)
2364 kfree_skb(skb);
2365}
2366#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002367#ifdef CONFIG_BRIDGE_NETFILTER
2368static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2369{
2370 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2371 kfree(nf_bridge);
2372}
2373static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2374{
2375 if (nf_bridge)
2376 atomic_inc(&nf_bridge->use);
2377}
2378#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08002379static inline void nf_reset(struct sk_buff *skb)
2380{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002381#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08002382 nf_conntrack_put(skb->nfct);
2383 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002384#endif
2385#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Patrick McHardya193a4a2006-03-20 19:23:05 -08002386 nf_conntrack_put_reasm(skb->nfct_reasm);
2387 skb->nfct_reasm = NULL;
2388#endif
2389#ifdef CONFIG_BRIDGE_NETFILTER
2390 nf_bridge_put(skb->nf_bridge);
2391 skb->nf_bridge = NULL;
2392#endif
2393}
2394
Patrick McHardy1ab6b2a2013-04-05 20:42:05 +02002395static inline void nf_reset_trace(struct sk_buff *skb)
2396{
2397#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
2398 skb->nf_trace = 0;
2399#endif
2400}
2401
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002402/* Note: This doesn't put any conntrack and bridge info in dst. */
2403static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2404{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002405#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002406 dst->nfct = src->nfct;
2407 nf_conntrack_get(src->nfct);
2408 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002409#endif
2410#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002411 dst->nfct_reasm = src->nfct_reasm;
2412 nf_conntrack_get_reasm(src->nfct_reasm);
2413#endif
2414#ifdef CONFIG_BRIDGE_NETFILTER
2415 dst->nf_bridge = src->nf_bridge;
2416 nf_bridge_get(src->nf_bridge);
2417#endif
2418}
2419
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002420static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2421{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002422#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002423 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002424#endif
2425#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002426 nf_conntrack_put_reasm(dst->nfct_reasm);
2427#endif
2428#ifdef CONFIG_BRIDGE_NETFILTER
2429 nf_bridge_put(dst->nf_bridge);
2430#endif
2431 __nf_copy(dst, src);
2432}
2433
James Morris984bc162006-06-09 00:29:17 -07002434#ifdef CONFIG_NETWORK_SECMARK
2435static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2436{
2437 to->secmark = from->secmark;
2438}
2439
2440static inline void skb_init_secmark(struct sk_buff *skb)
2441{
2442 skb->secmark = 0;
2443}
2444#else
2445static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2446{ }
2447
2448static inline void skb_init_secmark(struct sk_buff *skb)
2449{ }
2450#endif
2451
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002452static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2453{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002454 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002455}
2456
Stephen Hemminger92477442009-03-21 13:39:26 -07002457static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002458{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002459 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002460}
2461
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002462static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2463{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002464 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002465}
2466
David S. Millerd5a9e242009-01-27 16:22:11 -08002467static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2468{
2469 skb->queue_mapping = rx_queue + 1;
2470}
2471
Stephen Hemminger92477442009-03-21 13:39:26 -07002472static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002473{
2474 return skb->queue_mapping - 1;
2475}
2476
Stephen Hemminger92477442009-03-21 13:39:26 -07002477static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002478{
Eric Dumazeta02cec22010-09-22 20:43:57 +00002479 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08002480}
2481
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00002482extern u16 __skb_tx_hash(const struct net_device *dev,
2483 const struct sk_buff *skb,
2484 unsigned int num_tx_queues);
Stephen Hemminger92477442009-03-21 13:39:26 -07002485
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002486#ifdef CONFIG_XFRM
2487static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2488{
2489 return skb->sp;
2490}
2491#else
2492static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2493{
2494 return NULL;
2495}
2496#endif
2497
David S. Millerbdcc0922012-03-07 20:53:36 -05002498static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07002499{
2500 return skb_shinfo(skb)->gso_size;
2501}
2502
David S. Millerbdcc0922012-03-07 20:53:36 -05002503static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02002504{
2505 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2506}
2507
Ben Hutchings4497b072008-06-19 16:22:28 -07002508extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2509
2510static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2511{
2512 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2513 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002514 const struct skb_shared_info *shinfo = skb_shinfo(skb);
2515
Alexander Duyckb78462e2010-06-02 12:24:37 +00002516 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2517 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07002518 __skb_warn_lro_forwarding(skb);
2519 return true;
2520 }
2521 return false;
2522}
2523
Herbert Xu35fc92a2007-03-26 23:22:20 -07002524static inline void skb_forward_csum(struct sk_buff *skb)
2525{
2526 /* Unfortunately we don't support this one. Any brave souls? */
2527 if (skb->ip_summed == CHECKSUM_COMPLETE)
2528 skb->ip_summed = CHECKSUM_NONE;
2529}
2530
Eric Dumazetbc8acf22010-09-02 13:07:41 -07002531/**
2532 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2533 * @skb: skb to check
2534 *
2535 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2536 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2537 * use this helper, to document places where we make this assertion.
2538 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002539static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07002540{
2541#ifdef DEBUG
2542 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2543#endif
2544}
2545
Rusty Russellf35d9d82008-02-04 23:49:54 -05002546bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00002547
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002548static inline bool skb_is_recycleable(const struct sk_buff *skb, int skb_size)
Andy Fleming3d153a72011-10-13 04:33:54 +00002549{
2550 if (irqs_disabled())
2551 return false;
2552
2553 if (skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)
2554 return false;
2555
2556 if (skb_is_nonlinear(skb) || skb->fclone != SKB_FCLONE_UNAVAILABLE)
2557 return false;
2558
2559 skb_size = SKB_DATA_ALIGN(skb_size + NET_SKB_PAD);
2560 if (skb_end_pointer(skb) - skb->head < skb_size)
2561 return false;
2562
2563 if (skb_shared(skb) || skb_cloned(skb))
2564 return false;
2565
2566 return true;
2567}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002568#endif /* __KERNEL__ */
2569#endif /* _LINUX_SKBUFF_H */