blob: b33a3a1f205e45c747c9125a5b5920365ca2f8be [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * Definitions for the 'struct sk_buff' memory handlers.
3 *
4 * Authors:
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14#ifndef _LINUX_SKBUFF_H
15#define _LINUX_SKBUFF_H
16
Linus Torvalds1da177e2005-04-16 15:20:36 -070017#include <linux/kernel.h>
Vegard Nossumfe55f6d2008-08-30 12:16:35 +020018#include <linux/kmemcheck.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070019#include <linux/compiler.h>
20#include <linux/time.h>
Paul Gortmaker187f1882011-11-23 20:12:59 -050021#include <linux/bug.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070022#include <linux/cache.h>
23
Arun Sharma600634972011-07-26 16:09:06 -070024#include <linux/atomic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <asm/types.h>
26#include <linux/spinlock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <linux/net.h>
Thomas Graf3fc7e8a2005-06-23 21:00:17 -070028#include <linux/textsearch.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070029#include <net/checksum.h>
Al Viroa80958f2006-12-04 20:41:19 +000030#include <linux/rcupdate.h>
Chris Leech97fc2f02006-05-23 17:55:33 -070031#include <linux/dmaengine.h>
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -070032#include <linux/hrtimer.h>
Ian Campbell131ea662011-08-19 06:25:00 +000033#include <linux/dma-mapping.h>
Michał Mirosławc8f44af2011-11-15 15:29:55 +000034#include <linux/netdev_features.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070035
Herbert Xu60476372007-04-09 11:59:39 -070036/* Don't change this without changing skb_csum_unnecessary! */
Linus Torvalds1da177e2005-04-16 15:20:36 -070037#define CHECKSUM_NONE 0
Herbert Xu60476372007-04-09 11:59:39 -070038#define CHECKSUM_UNNECESSARY 1
39#define CHECKSUM_COMPLETE 2
40#define CHECKSUM_PARTIAL 3
Linus Torvalds1da177e2005-04-16 15:20:36 -070041
42#define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
43 ~(SMP_CACHE_BYTES - 1))
David S. Millerfc910a22007-03-25 20:27:59 -070044#define SKB_WITH_OVERHEAD(X) \
Herbert Xudeea84b2007-10-21 16:27:46 -070045 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
David S. Millerfc910a22007-03-25 20:27:59 -070046#define SKB_MAX_ORDER(X, ORDER) \
47 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
Linus Torvalds1da177e2005-04-16 15:20:36 -070048#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
49#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
50
Eric Dumazet87fb4b72011-10-13 07:28:54 +000051/* return minimum truesize of one skb containing X bytes of data */
52#define SKB_TRUESIZE(X) ((X) + \
53 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
54 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
55
Linus Torvalds1da177e2005-04-16 15:20:36 -070056/* A. Checksumming of received packets by device.
57 *
58 * NONE: device failed to checksum this packet.
59 * skb->csum is undefined.
60 *
61 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
62 * skb->csum is undefined.
63 * It is bad option, but, unfortunately, many of vendors do this.
64 * Apparently with secret goal to sell you new device, when you
65 * will add new protocol to your host. F.e. IPv6. 8)
66 *
Patrick McHardy84fa7932006-08-29 16:44:56 -070067 * COMPLETE: the most generic way. Device supplied checksum of _all_
Linus Torvalds1da177e2005-04-16 15:20:36 -070068 * the packet as seen by netif_rx in skb->csum.
69 * NOTE: Even if device supports only some protocols, but
Patrick McHardy84fa7932006-08-29 16:44:56 -070070 * is able to produce some skb->csum, it MUST use COMPLETE,
Linus Torvalds1da177e2005-04-16 15:20:36 -070071 * not UNNECESSARY.
72 *
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070073 * PARTIAL: identical to the case for output below. This may occur
74 * on a packet received directly from another Linux OS, e.g.,
75 * a virtualised Linux kernel on the same host. The packet can
76 * be treated in the same way as UNNECESSARY except that on
77 * output (i.e., forwarding) the checksum must be filled in
78 * by the OS or the hardware.
79 *
Linus Torvalds1da177e2005-04-16 15:20:36 -070080 * B. Checksumming on output.
81 *
82 * NONE: skb is checksummed by protocol or csum is not required.
83 *
Patrick McHardy84fa7932006-08-29 16:44:56 -070084 * PARTIAL: device is required to csum packet as seen by hard_start_xmit
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070085 * from skb->csum_start to the end and to record the checksum
86 * at skb->csum_start + skb->csum_offset.
Linus Torvalds1da177e2005-04-16 15:20:36 -070087 *
88 * Device must show its capabilities in dev->features, set
89 * at device setup time.
90 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
91 * everything.
Linus Torvalds1da177e2005-04-16 15:20:36 -070092 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
93 * TCP/UDP over IPv4. Sigh. Vendors like this
94 * way by an unknown reason. Though, see comment above
95 * about CHECKSUM_UNNECESSARY. 8)
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070096 * NETIF_F_IPV6_CSUM about as dumb as the last one but does IPv6 instead.
Linus Torvalds1da177e2005-04-16 15:20:36 -070097 *
Yi Zou3af79302012-03-19 11:12:41 +000098 * UNNECESSARY: device will do per protocol specific csum. Protocol drivers
99 * that do not want net to perform the checksum calculation should use
100 * this flag in their outgoing skbs.
101 * NETIF_F_FCOE_CRC this indicates the device can do FCoE FC CRC
102 * offload. Correspondingly, the FCoE protocol driver
103 * stack should use CHECKSUM_UNNECESSARY.
104 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105 * Any questions? No questions, good. --ANK
106 */
107
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108struct net_device;
David Howells716ea3a2007-04-02 20:19:53 -0700109struct scatterlist;
Jens Axboe9c55e012007-11-06 23:30:13 -0800110struct pipe_inode_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700112#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113struct nf_conntrack {
114 atomic_t use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115};
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700116#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117
118#ifdef CONFIG_BRIDGE_NETFILTER
119struct nf_bridge_info {
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000120 atomic_t use;
121 unsigned int mask;
122 struct net_device *physindev;
123 struct net_device *physoutdev;
124 unsigned long data[32 / sizeof(unsigned long)];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125};
126#endif
127
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128struct sk_buff_head {
129 /* These two members must be first. */
130 struct sk_buff *next;
131 struct sk_buff *prev;
132
133 __u32 qlen;
134 spinlock_t lock;
135};
136
137struct sk_buff;
138
Ian Campbell9d4dde52011-12-22 23:39:14 +0000139/* To allow 64K frame to be packed as single skb without frag_list we
140 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
141 * buffers which do not start on a page boundary.
142 *
143 * Since GRO uses frags we allocate at least 16 regardless of page
144 * size.
Anton Blancharda715dea2011-03-27 14:57:26 +0000145 */
Ian Campbell9d4dde52011-12-22 23:39:14 +0000146#if (65536/PAGE_SIZE + 1) < 16
David S. Millereec00952011-03-29 23:34:08 -0700147#define MAX_SKB_FRAGS 16UL
Anton Blancharda715dea2011-03-27 14:57:26 +0000148#else
Ian Campbell9d4dde52011-12-22 23:39:14 +0000149#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
Anton Blancharda715dea2011-03-27 14:57:26 +0000150#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151
152typedef struct skb_frag_struct skb_frag_t;
153
154struct skb_frag_struct {
Ian Campbella8605c62011-10-19 23:01:49 +0000155 struct {
156 struct page *p;
157 } page;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000158#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
David S. Millera309bb02007-07-30 18:47:03 -0700159 __u32 page_offset;
160 __u32 size;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000161#else
162 __u16 page_offset;
163 __u16 size;
164#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165};
166
Eric Dumazet9e903e02011-10-18 21:00:24 +0000167static inline unsigned int skb_frag_size(const skb_frag_t *frag)
168{
169 return frag->size;
170}
171
172static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
173{
174 frag->size = size;
175}
176
177static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
178{
179 frag->size += delta;
180}
181
182static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
183{
184 frag->size -= delta;
185}
186
Patrick Ohlyac45f602009-02-12 05:03:37 +0000187#define HAVE_HW_TIME_STAMP
188
189/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800190 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000191 * @hwtstamp: hardware time stamp transformed into duration
192 * since arbitrary point in time
193 * @syststamp: hwtstamp transformed to system time base
194 *
195 * Software time stamps generated by ktime_get_real() are stored in
196 * skb->tstamp. The relation between the different kinds of time
197 * stamps is as follows:
198 *
199 * syststamp and tstamp can be compared against each other in
200 * arbitrary combinations. The accuracy of a
201 * syststamp/tstamp/"syststamp from other device" comparison is
202 * limited by the accuracy of the transformation into system time
203 * base. This depends on the device driver and its underlying
204 * hardware.
205 *
206 * hwtstamps can only be compared against other hwtstamps from
207 * the same device.
208 *
209 * This structure is attached to packets as part of the
210 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
211 */
212struct skb_shared_hwtstamps {
213 ktime_t hwtstamp;
214 ktime_t syststamp;
215};
216
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000217/* Definitions for tx_flags in struct skb_shared_info */
218enum {
219 /* generate hardware time stamp */
220 SKBTX_HW_TSTAMP = 1 << 0,
221
222 /* generate software time stamp */
223 SKBTX_SW_TSTAMP = 1 << 1,
224
225 /* device driver is going to provide hardware time stamp */
226 SKBTX_IN_PROGRESS = 1 << 2,
227
Shirley Maa6686f22011-07-06 12:22:12 +0000228 /* device driver supports TX zero-copy buffers */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000229 SKBTX_DEV_ZEROCOPY = 1 << 3,
Johannes Berg6e3e9392011-11-09 10:15:42 +0100230
231 /* generate wifi status information (where possible) */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000232 SKBTX_WIFI_STATUS = 1 << 4,
Shirley Maa6686f22011-07-06 12:22:12 +0000233};
234
235/*
236 * The callback notifies userspace to release buffers when skb DMA is done in
237 * lower device, the skb last reference should be 0 when calling this.
Michael S. 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 McHardy6aa895b02008-07-14 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
Mel Gormanc93bdd02012-07-31 16:44:19 -0700465 __u8 pfmemalloc:1;
Tom Herbert3853b582010-11-21 13:17:29 +0000466 __u8 ooo_okay:1;
Tom Herbertbdeab992011-08-14 19:45:55 +0000467 __u8 l4_rxhash:1;
Johannes Berg6e3e9392011-11-09 10:15:42 +0100468 __u8 wifi_acked_valid:1;
469 __u8 wifi_acked:1;
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000470 __u8 no_fcs:1;
Eric Dumazetd3836f22012-04-27 00:33:38 +0000471 __u8 head_frag:1;
472 /* 8/10 bit hole (depending on ndisc_nodetype presence) */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200473 kmemcheck_bitfield_end(flags2);
474
Chris Leech97fc2f02006-05-23 17:55:33 -0700475#ifdef CONFIG_NET_DMA
476 dma_cookie_t dma_cookie;
477#endif
James Morris984bc162006-06-09 00:29:17 -0700478#ifdef CONFIG_NETWORK_SECMARK
479 __u32 secmark;
480#endif
Neil Horman3b885782009-10-12 13:26:31 -0700481 union {
482 __u32 mark;
483 __u32 dropcount;
Eric Dumazeta21d4572012-04-10 20:30:48 +0000484 __u32 avail_size;
Neil Horman3b885782009-10-12 13:26:31 -0700485 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700487 sk_buff_data_t transport_header;
488 sk_buff_data_t network_header;
489 sk_buff_data_t mac_header;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700491 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700492 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700494 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700495 unsigned int truesize;
496 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497};
498
499#ifdef __KERNEL__
500/*
501 * Handling routines are only of interest to the kernel
502 */
503#include <linux/slab.h>
504
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505
Mel Gormanc93bdd02012-07-31 16:44:19 -0700506#define SKB_ALLOC_FCLONE 0x01
507#define SKB_ALLOC_RX 0x02
508
509/* Returns true if the skb was allocated from PFMEMALLOC reserves */
510static inline bool skb_pfmemalloc(const struct sk_buff *skb)
511{
512 return unlikely(skb->pfmemalloc);
513}
514
Eric Dumazet7fee2262010-05-11 23:19:48 +0000515/*
516 * skb might have a dst pointer attached, refcounted or not.
517 * _skb_refdst low order bit is set if refcount was _not_ taken
518 */
519#define SKB_DST_NOREF 1UL
520#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
521
522/**
523 * skb_dst - returns skb dst_entry
524 * @skb: buffer
525 *
526 * Returns skb dst_entry, regardless of reference taken or not.
527 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000528static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
529{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000530 /* If refdst was not refcounted, check we still are in a
531 * rcu_read_lock section
532 */
533 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
534 !rcu_read_lock_held() &&
535 !rcu_read_lock_bh_held());
536 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000537}
538
Eric Dumazet7fee2262010-05-11 23:19:48 +0000539/**
540 * skb_dst_set - sets skb dst
541 * @skb: buffer
542 * @dst: dst entry
543 *
544 * Sets skb dst, assuming a reference was taken on dst and should
545 * be released by skb_dst_drop()
546 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000547static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
548{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000549 skb->_skb_refdst = (unsigned long)dst;
550}
551
Eric Dumazet27b75c92010-10-15 05:44:11 +0000552extern void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst);
Eric Dumazet7fee2262010-05-11 23:19:48 +0000553
554/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300555 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000556 * @skb: buffer
557 */
558static inline bool skb_dst_is_noref(const struct sk_buff *skb)
559{
560 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000561}
562
Eric Dumazet511c3f92009-06-02 05:14:27 +0000563static inline struct rtable *skb_rtable(const struct sk_buff *skb)
564{
Eric Dumazetadf30902009-06-02 05:19:30 +0000565 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000566}
567
Jörn Engel231d06a2006-03-20 21:28:35 -0800568extern void kfree_skb(struct sk_buff *skb);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000569extern void consume_skb(struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700570extern void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000571extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000572
573extern void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
574extern bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
575 bool *fragstolen, int *delta_truesize);
576
David S. Millerd179cd12005-08-17 14:57:30 -0700577extern struct sk_buff *__alloc_skb(unsigned int size,
Mel Gormanc93bdd02012-07-31 16:44:19 -0700578 gfp_t priority, int flags, int node);
Eric Dumazetd3836f22012-04-27 00:33:38 +0000579extern struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700580static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100581 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700582{
Eric Dumazet564824b2010-10-11 19:05:25 +0000583 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700584}
585
586static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100587 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700588{
Mel Gormanc93bdd02012-07-31 16:44:19 -0700589 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700590}
591
Andy Fleming3d153a72011-10-13 04:33:54 +0000592extern void skb_recycle(struct sk_buff *skb);
Changli Gao5b0daa32010-05-29 00:12:13 -0700593extern bool skb_recycle_check(struct sk_buff *skb, int skb_size);
Lennert Buytenhek04a4bb52008-10-01 02:33:12 -0700594
Herbert Xue0053ec2007-10-14 00:37:52 -0700595extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
Michael S. Tsirkin48c83012011-08-31 08:03:29 +0000596extern int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700597extern struct sk_buff *skb_clone(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100598 gfp_t priority);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700599extern struct sk_buff *skb_copy(const struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100600 gfp_t priority);
Eric Dumazet117632e2011-12-03 21:39:53 +0000601extern struct sk_buff *__pskb_copy(struct sk_buff *skb,
602 int headroom, gfp_t gfp_mask);
603
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604extern int pskb_expand_head(struct sk_buff *skb,
Victor Fusco86a76ca2005-07-08 14:57:47 -0700605 int nhead, int ntail,
Al Virodd0fc662005-10-07 07:46:04 +0100606 gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
608 unsigned int headroom);
609extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
610 int newheadroom, int newtailroom,
Al Virodd0fc662005-10-07 07:46:04 +0100611 gfp_t priority);
David Howells716ea3a2007-04-02 20:19:53 -0700612extern int skb_to_sgvec(struct sk_buff *skb,
613 struct scatterlist *sg, int offset,
614 int len);
615extern int skb_cow_data(struct sk_buff *skb, int tailbits,
616 struct sk_buff **trailer);
Herbert Xu5b057c62006-06-23 02:06:41 -0700617extern int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000618#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700620extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
621 int getfrag(void *from, char *to, int offset,
622 int len,int odd, struct sk_buff *skb),
623 void *from, int length);
624
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800625struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700626 __u32 lower_offset;
627 __u32 upper_offset;
628 __u32 frag_idx;
629 __u32 stepped_offset;
630 struct sk_buff *root_skb;
631 struct sk_buff *cur_skb;
632 __u8 *frag_data;
633};
634
635extern void skb_prepare_seq_read(struct sk_buff *skb,
636 unsigned int from, unsigned int to,
637 struct skb_seq_state *st);
638extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
639 struct skb_seq_state *st);
640extern void skb_abort_seq_read(struct skb_seq_state *st);
641
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700642extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
643 unsigned int to, struct ts_config *config,
644 struct ts_state *state);
645
Tom Herbertbdeab992011-08-14 19:45:55 +0000646extern void __skb_get_rxhash(struct sk_buff *skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000647static inline __u32 skb_get_rxhash(struct sk_buff *skb)
648{
649 if (!skb->rxhash)
Tom Herbertbdeab992011-08-14 19:45:55 +0000650 __skb_get_rxhash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000651
652 return skb->rxhash;
653}
654
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700655#ifdef NET_SKBUFF_DATA_USES_OFFSET
656static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
657{
658 return skb->head + skb->end;
659}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000660
661static inline unsigned int skb_end_offset(const struct sk_buff *skb)
662{
663 return skb->end;
664}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700665#else
666static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
667{
668 return skb->end;
669}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000670
671static inline unsigned int skb_end_offset(const struct sk_buff *skb)
672{
673 return skb->end - skb->head;
674}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700675#endif
676
Linus Torvalds1da177e2005-04-16 15:20:36 -0700677/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700678#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679
Patrick Ohlyac45f602009-02-12 05:03:37 +0000680static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
681{
682 return &skb_shinfo(skb)->hwtstamps;
683}
684
Linus Torvalds1da177e2005-04-16 15:20:36 -0700685/**
686 * skb_queue_empty - check if a queue is empty
687 * @list: queue head
688 *
689 * Returns true if the queue is empty, false otherwise.
690 */
691static inline int skb_queue_empty(const struct sk_buff_head *list)
692{
693 return list->next == (struct sk_buff *)list;
694}
695
696/**
David S. Millerfc7ebb22008-09-23 00:34:07 -0700697 * skb_queue_is_last - check if skb is the last entry in the queue
698 * @list: queue head
699 * @skb: buffer
700 *
701 * Returns true if @skb is the last buffer on the list.
702 */
703static inline bool skb_queue_is_last(const struct sk_buff_head *list,
704 const struct sk_buff *skb)
705{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000706 return skb->next == (struct sk_buff *)list;
David S. Millerfc7ebb22008-09-23 00:34:07 -0700707}
708
709/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800710 * skb_queue_is_first - check if skb is the first entry in the queue
711 * @list: queue head
712 * @skb: buffer
713 *
714 * Returns true if @skb is the first buffer on the list.
715 */
716static inline bool skb_queue_is_first(const struct sk_buff_head *list,
717 const struct sk_buff *skb)
718{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000719 return skb->prev == (struct sk_buff *)list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800720}
721
722/**
David S. Miller249c8b42008-09-23 00:44:42 -0700723 * skb_queue_next - return the next packet in the queue
724 * @list: queue head
725 * @skb: current buffer
726 *
727 * Return the next packet in @list after @skb. It is only valid to
728 * call this if skb_queue_is_last() evaluates to false.
729 */
730static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
731 const struct sk_buff *skb)
732{
733 /* This BUG_ON may seem severe, but if we just return then we
734 * are going to dereference garbage.
735 */
736 BUG_ON(skb_queue_is_last(list, skb));
737 return skb->next;
738}
739
740/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800741 * skb_queue_prev - return the prev packet in the queue
742 * @list: queue head
743 * @skb: current buffer
744 *
745 * Return the prev packet in @list before @skb. It is only valid to
746 * call this if skb_queue_is_first() evaluates to false.
747 */
748static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
749 const struct sk_buff *skb)
750{
751 /* This BUG_ON may seem severe, but if we just return then we
752 * are going to dereference garbage.
753 */
754 BUG_ON(skb_queue_is_first(list, skb));
755 return skb->prev;
756}
757
758/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759 * skb_get - reference buffer
760 * @skb: buffer to reference
761 *
762 * Makes another reference to a socket buffer and returns a pointer
763 * to the buffer.
764 */
765static inline struct sk_buff *skb_get(struct sk_buff *skb)
766{
767 atomic_inc(&skb->users);
768 return skb;
769}
770
771/*
772 * If users == 1, we are the only owner and are can avoid redundant
773 * atomic change.
774 */
775
776/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700777 * skb_cloned - is the buffer a clone
778 * @skb: buffer to check
779 *
780 * Returns true if the buffer was generated with skb_clone() and is
781 * one of multiple shared copies of the buffer. Cloned buffers are
782 * shared data so must not be written to under normal circumstances.
783 */
784static inline int skb_cloned(const struct sk_buff *skb)
785{
786 return skb->cloned &&
787 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
788}
789
790/**
791 * skb_header_cloned - is the header a clone
792 * @skb: buffer to check
793 *
794 * Returns true if modifying the header part of the buffer requires
795 * the data to be copied.
796 */
797static inline int skb_header_cloned(const struct sk_buff *skb)
798{
799 int dataref;
800
801 if (!skb->cloned)
802 return 0;
803
804 dataref = atomic_read(&skb_shinfo(skb)->dataref);
805 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
806 return dataref != 1;
807}
808
809/**
810 * skb_header_release - release reference to header
811 * @skb: buffer to operate on
812 *
813 * Drop a reference to the header part of the buffer. This is done
814 * by acquiring a payload reference. You must not read from the header
815 * part of skb->data after this.
816 */
817static inline void skb_header_release(struct sk_buff *skb)
818{
819 BUG_ON(skb->nohdr);
820 skb->nohdr = 1;
821 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
822}
823
824/**
825 * skb_shared - is the buffer shared
826 * @skb: buffer to check
827 *
828 * Returns true if more than one person has a reference to this
829 * buffer.
830 */
831static inline int skb_shared(const struct sk_buff *skb)
832{
833 return atomic_read(&skb->users) != 1;
834}
835
836/**
837 * skb_share_check - check if buffer is shared and if so clone it
838 * @skb: buffer to check
839 * @pri: priority for memory allocation
840 *
841 * If the buffer is shared the buffer is cloned and the old copy
842 * drops a reference. A new clone with a single reference is returned.
843 * If the buffer is not shared the original buffer is returned. When
844 * being called from interrupt status or with spinlocks held pri must
845 * be GFP_ATOMIC.
846 *
847 * NULL is returned on a memory allocation failure.
848 */
Eric Dumazet47061bc2012-08-03 20:54:15 +0000849static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700850{
851 might_sleep_if(pri & __GFP_WAIT);
852 if (skb_shared(skb)) {
853 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +0000854
855 if (likely(nskb))
856 consume_skb(skb);
857 else
858 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700859 skb = nskb;
860 }
861 return skb;
862}
863
864/*
865 * Copy shared buffers into a new sk_buff. We effectively do COW on
866 * packets to handle cases where we have a local reader and forward
867 * and a couple of other messy ones. The normal one is tcpdumping
868 * a packet thats being forwarded.
869 */
870
871/**
872 * skb_unshare - make a copy of a shared buffer
873 * @skb: buffer to check
874 * @pri: priority for memory allocation
875 *
876 * If the socket buffer is a clone then this function creates a new
877 * copy of the data, drops a reference count on the old copy and returns
878 * the new copy with the reference count at 1. If the buffer is not a clone
879 * the original buffer is returned. When called with a spinlock held or
880 * from interrupt state @pri must be %GFP_ATOMIC
881 *
882 * %NULL is returned on a memory allocation failure.
883 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -0700884static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100885 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886{
887 might_sleep_if(pri & __GFP_WAIT);
888 if (skb_cloned(skb)) {
889 struct sk_buff *nskb = skb_copy(skb, pri);
890 kfree_skb(skb); /* Free our shared copy */
891 skb = nskb;
892 }
893 return skb;
894}
895
896/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -0800897 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898 * @list_: list to peek at
899 *
900 * Peek an &sk_buff. Unlike most other operations you _MUST_
901 * be careful with this one. A peek leaves the buffer on the
902 * list and someone else may run off with it. You must hold
903 * the appropriate locks or have a private queue to do this.
904 *
905 * Returns %NULL for an empty list or a pointer to the head element.
906 * The reference count is not incremented and the reference is therefore
907 * volatile. Use with caution.
908 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -0400909static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910{
Eric Dumazet18d07002012-04-30 16:31:46 +0000911 struct sk_buff *skb = list_->next;
912
913 if (skb == (struct sk_buff *)list_)
914 skb = NULL;
915 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700916}
917
918/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +0000919 * skb_peek_next - peek skb following the given one from a queue
920 * @skb: skb to start from
921 * @list_: list to peek at
922 *
923 * Returns %NULL when the end of the list is met or a pointer to the
924 * next element. The reference count is not incremented and the
925 * reference is therefore volatile. Use with caution.
926 */
927static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
928 const struct sk_buff_head *list_)
929{
930 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +0000931
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +0000932 if (next == (struct sk_buff *)list_)
933 next = NULL;
934 return next;
935}
936
937/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -0800938 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -0700939 * @list_: list to peek at
940 *
941 * Peek an &sk_buff. Unlike most other operations you _MUST_
942 * be careful with this one. A peek leaves the buffer on the
943 * list and someone else may run off with it. You must hold
944 * the appropriate locks or have a private queue to do this.
945 *
946 * Returns %NULL for an empty list or a pointer to the tail element.
947 * The reference count is not incremented and the reference is therefore
948 * volatile. Use with caution.
949 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -0400950static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700951{
Eric Dumazet18d07002012-04-30 16:31:46 +0000952 struct sk_buff *skb = list_->prev;
953
954 if (skb == (struct sk_buff *)list_)
955 skb = NULL;
956 return skb;
957
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958}
959
960/**
961 * skb_queue_len - get queue length
962 * @list_: list to measure
963 *
964 * Return the length of an &sk_buff queue.
965 */
966static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
967{
968 return list_->qlen;
969}
970
David S. Miller67fed452008-09-21 22:36:24 -0700971/**
972 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
973 * @list: queue to initialize
974 *
975 * This initializes only the list and queue length aspects of
976 * an sk_buff_head object. This allows to initialize the list
977 * aspects of an sk_buff_head without reinitializing things like
978 * the spinlock. It can also be used for on-stack sk_buff_head
979 * objects where the spinlock is known to not be used.
980 */
981static inline void __skb_queue_head_init(struct sk_buff_head *list)
982{
983 list->prev = list->next = (struct sk_buff *)list;
984 list->qlen = 0;
985}
986
Arjan van de Ven76f10ad2006-08-02 14:06:55 -0700987/*
988 * This function creates a split out lock class for each invocation;
989 * this is needed for now since a whole lot of users of the skb-queue
990 * infrastructure in drivers have different locking usage (in hardirq)
991 * than the networking core (in softirq only). In the long run either the
992 * network layer or drivers should need annotation to consolidate the
993 * main types of usage into 3 classes.
994 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995static inline void skb_queue_head_init(struct sk_buff_head *list)
996{
997 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -0700998 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999}
1000
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001001static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1002 struct lock_class_key *class)
1003{
1004 skb_queue_head_init(list);
1005 lockdep_set_class(&list->lock, class);
1006}
1007
Linus Torvalds1da177e2005-04-16 15:20:36 -07001008/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001009 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010 *
1011 * The "__skb_xxxx()" functions are the non-atomic ones that
1012 * can only be called with interrupts disabled.
1013 */
Gerrit Renkerbf299272008-04-14 00:04:51 -07001014extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
1015static inline void __skb_insert(struct sk_buff *newsk,
1016 struct sk_buff *prev, struct sk_buff *next,
1017 struct sk_buff_head *list)
1018{
1019 newsk->next = next;
1020 newsk->prev = prev;
1021 next->prev = prev->next = newsk;
1022 list->qlen++;
1023}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001024
David S. Miller67fed452008-09-21 22:36:24 -07001025static inline void __skb_queue_splice(const struct sk_buff_head *list,
1026 struct sk_buff *prev,
1027 struct sk_buff *next)
1028{
1029 struct sk_buff *first = list->next;
1030 struct sk_buff *last = list->prev;
1031
1032 first->prev = prev;
1033 prev->next = first;
1034
1035 last->next = next;
1036 next->prev = last;
1037}
1038
1039/**
1040 * skb_queue_splice - join two skb lists, this is designed for stacks
1041 * @list: the new list to add
1042 * @head: the place to add it in the first list
1043 */
1044static inline void skb_queue_splice(const struct sk_buff_head *list,
1045 struct sk_buff_head *head)
1046{
1047 if (!skb_queue_empty(list)) {
1048 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001049 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001050 }
1051}
1052
1053/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001054 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001055 * @list: the new list to add
1056 * @head: the place to add it in the first list
1057 *
1058 * The list at @list is reinitialised
1059 */
1060static inline void skb_queue_splice_init(struct sk_buff_head *list,
1061 struct sk_buff_head *head)
1062{
1063 if (!skb_queue_empty(list)) {
1064 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001065 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001066 __skb_queue_head_init(list);
1067 }
1068}
1069
1070/**
1071 * skb_queue_splice_tail - join two skb lists, each list being a queue
1072 * @list: the new list to add
1073 * @head: the place to add it in the first list
1074 */
1075static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1076 struct sk_buff_head *head)
1077{
1078 if (!skb_queue_empty(list)) {
1079 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001080 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001081 }
1082}
1083
1084/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001085 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001086 * @list: the new list to add
1087 * @head: the place to add it in the first list
1088 *
1089 * Each of the lists is a queue.
1090 * The list at @list is reinitialised
1091 */
1092static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1093 struct sk_buff_head *head)
1094{
1095 if (!skb_queue_empty(list)) {
1096 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001097 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001098 __skb_queue_head_init(list);
1099 }
1100}
1101
Linus Torvalds1da177e2005-04-16 15:20:36 -07001102/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001103 * __skb_queue_after - queue a buffer at the list head
1104 * @list: list to use
1105 * @prev: place after this buffer
1106 * @newsk: buffer to queue
1107 *
1108 * Queue a buffer int the middle of a list. This function takes no locks
1109 * and you must therefore hold required locks before calling it.
1110 *
1111 * A buffer cannot be placed on two lists at the same time.
1112 */
1113static inline void __skb_queue_after(struct sk_buff_head *list,
1114 struct sk_buff *prev,
1115 struct sk_buff *newsk)
1116{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001117 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001118}
1119
Gerrit Renker7de6c032008-04-14 00:05:09 -07001120extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1121 struct sk_buff_head *list);
1122
Gerrit Renkerf5572852008-04-14 00:05:28 -07001123static inline void __skb_queue_before(struct sk_buff_head *list,
1124 struct sk_buff *next,
1125 struct sk_buff *newsk)
1126{
1127 __skb_insert(newsk, next->prev, next, list);
1128}
1129
Stephen Hemminger300ce172005-10-30 13:47:34 -08001130/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131 * __skb_queue_head - queue a buffer at the list head
1132 * @list: list to use
1133 * @newsk: buffer to queue
1134 *
1135 * Queue a buffer at the start of a list. This function takes no locks
1136 * and you must therefore hold required locks before calling it.
1137 *
1138 * A buffer cannot be placed on two lists at the same time.
1139 */
1140extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1141static inline void __skb_queue_head(struct sk_buff_head *list,
1142 struct sk_buff *newsk)
1143{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001144 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001145}
1146
1147/**
1148 * __skb_queue_tail - queue a buffer at the list tail
1149 * @list: list to use
1150 * @newsk: buffer to queue
1151 *
1152 * Queue a buffer at the end of a list. This function takes no locks
1153 * and you must therefore hold required locks before calling it.
1154 *
1155 * A buffer cannot be placed on two lists at the same time.
1156 */
1157extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1158static inline void __skb_queue_tail(struct sk_buff_head *list,
1159 struct sk_buff *newsk)
1160{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001161 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162}
1163
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001165 * remove sk_buff from list. _Must_ be called atomically, and with
1166 * the list known..
1167 */
David S. Miller8728b832005-08-09 19:25:21 -07001168extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001169static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1170{
1171 struct sk_buff *next, *prev;
1172
1173 list->qlen--;
1174 next = skb->next;
1175 prev = skb->prev;
1176 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001177 next->prev = prev;
1178 prev->next = next;
1179}
1180
Gerrit Renkerf525c062008-04-14 00:04:12 -07001181/**
1182 * __skb_dequeue - remove from the head of the queue
1183 * @list: list to dequeue from
1184 *
1185 * Remove the head of the list. This function does not take any locks
1186 * so must be used with appropriate locks held only. The head item is
1187 * returned or %NULL if the list is empty.
1188 */
1189extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1190static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1191{
1192 struct sk_buff *skb = skb_peek(list);
1193 if (skb)
1194 __skb_unlink(skb, list);
1195 return skb;
1196}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001197
1198/**
1199 * __skb_dequeue_tail - remove from the tail of the queue
1200 * @list: list to dequeue from
1201 *
1202 * Remove the tail of the list. This function does not take any locks
1203 * so must be used with appropriate locks held only. The tail item is
1204 * returned or %NULL if the list is empty.
1205 */
1206extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1207static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1208{
1209 struct sk_buff *skb = skb_peek_tail(list);
1210 if (skb)
1211 __skb_unlink(skb, list);
1212 return skb;
1213}
1214
1215
David S. Millerbdcc0922012-03-07 20:53:36 -05001216static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001217{
1218 return skb->data_len;
1219}
1220
1221static inline unsigned int skb_headlen(const struct sk_buff *skb)
1222{
1223 return skb->len - skb->data_len;
1224}
1225
1226static inline int skb_pagelen(const struct sk_buff *skb)
1227{
1228 int i, len = 0;
1229
1230 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001231 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001232 return len + skb_headlen(skb);
1233}
1234
Ian Campbell131ea662011-08-19 06:25:00 +00001235/**
1236 * __skb_fill_page_desc - initialise a paged fragment in an skb
1237 * @skb: buffer containing fragment to be initialised
1238 * @i: paged fragment index to initialise
1239 * @page: the page to use for this fragment
1240 * @off: the offset to the data with @page
1241 * @size: the length of the data
1242 *
1243 * Initialises the @i'th fragment of @skb to point to &size bytes at
1244 * offset @off within @page.
1245 *
1246 * Does not take any additional reference on the fragment.
1247 */
1248static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1249 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250{
1251 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1252
Mel Gormanc48a11c2012-07-31 16:44:23 -07001253 /*
1254 * Propagate page->pfmemalloc to the skb if we can. The problem is
1255 * that not all callers have unique ownership of the page. If
1256 * pfmemalloc is set, we check the mapping as a mapping implies
1257 * page->index is set (index and pfmemalloc share space).
1258 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1259 * do not lose pfmemalloc information as the pages would not be
1260 * allocated using __GFP_MEMALLOC.
1261 */
1262 if (page->pfmemalloc && !page->mapping)
1263 skb->pfmemalloc = true;
Ian Campbella8605c62011-10-19 23:01:49 +00001264 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001265 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001266 skb_frag_size_set(frag, size);
Ian Campbell131ea662011-08-19 06:25:00 +00001267}
1268
1269/**
1270 * skb_fill_page_desc - initialise a paged fragment in an skb
1271 * @skb: buffer containing fragment to be initialised
1272 * @i: paged fragment index to initialise
1273 * @page: the page to use for this fragment
1274 * @off: the offset to the data with @page
1275 * @size: the length of the data
1276 *
1277 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1278 * @skb to point to &size bytes at offset @off within @page. In
1279 * addition updates @skb such that @i is the last fragment.
1280 *
1281 * Does not take any additional reference on the fragment.
1282 */
1283static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1284 struct page *page, int off, int size)
1285{
1286 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001287 skb_shinfo(skb)->nr_frags = i + 1;
1288}
1289
Peter Zijlstra654bed12008-10-07 14:22:33 -07001290extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
Eric Dumazet50269e12012-03-23 23:59:33 +00001291 int off, int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001292
Linus Torvalds1da177e2005-04-16 15:20:36 -07001293#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001294#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001295#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1296
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001297#ifdef NET_SKBUFF_DATA_USES_OFFSET
1298static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1299{
1300 return skb->head + skb->tail;
1301}
1302
1303static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1304{
1305 skb->tail = skb->data - skb->head;
1306}
1307
1308static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1309{
1310 skb_reset_tail_pointer(skb);
1311 skb->tail += offset;
1312}
1313#else /* NET_SKBUFF_DATA_USES_OFFSET */
1314static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1315{
1316 return skb->tail;
1317}
1318
1319static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1320{
1321 skb->tail = skb->data;
1322}
1323
1324static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1325{
1326 skb->tail = skb->data + offset;
1327}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001328
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001329#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1330
Linus Torvalds1da177e2005-04-16 15:20:36 -07001331/*
1332 * Add data to an sk_buff
1333 */
Ilpo Järvinen0dde3e12008-03-27 17:43:41 -07001334extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1336{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001337 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001338 SKB_LINEAR_ASSERT(skb);
1339 skb->tail += len;
1340 skb->len += len;
1341 return tmp;
1342}
1343
Ilpo Järvinenc2aa2702008-03-27 17:52:40 -07001344extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1346{
1347 skb->data -= len;
1348 skb->len += len;
1349 return skb->data;
1350}
1351
Ilpo Järvinen6be8ac22008-03-27 17:47:24 -07001352extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001353static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1354{
1355 skb->len -= len;
1356 BUG_ON(skb->len < skb->data_len);
1357 return skb->data += len;
1358}
1359
David S. Miller47d29642010-05-02 02:21:44 -07001360static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1361{
1362 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1363}
1364
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1366
1367static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1368{
1369 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001370 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001371 return NULL;
1372 skb->len -= len;
1373 return skb->data += len;
1374}
1375
1376static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1377{
1378 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1379}
1380
1381static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1382{
1383 if (likely(len <= skb_headlen(skb)))
1384 return 1;
1385 if (unlikely(len > skb->len))
1386 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001387 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001388}
1389
1390/**
1391 * skb_headroom - bytes at buffer head
1392 * @skb: buffer to check
1393 *
1394 * Return the number of bytes of free space at the head of an &sk_buff.
1395 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001396static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001397{
1398 return skb->data - skb->head;
1399}
1400
1401/**
1402 * skb_tailroom - bytes at buffer end
1403 * @skb: buffer to check
1404 *
1405 * Return the number of bytes of free space at the tail of an sk_buff
1406 */
1407static inline int skb_tailroom(const struct sk_buff *skb)
1408{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001409 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001410}
1411
1412/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00001413 * skb_availroom - bytes at buffer end
1414 * @skb: buffer to check
1415 *
1416 * Return the number of bytes of free space at the tail of an sk_buff
1417 * allocated by sk_stream_alloc()
1418 */
1419static inline int skb_availroom(const struct sk_buff *skb)
1420{
1421 return skb_is_nonlinear(skb) ? 0 : skb->avail_size - skb->len;
1422}
1423
1424/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425 * skb_reserve - adjust headroom
1426 * @skb: buffer to alter
1427 * @len: bytes to move
1428 *
1429 * Increase the headroom of an empty &sk_buff by reducing the tail
1430 * room. This is only allowed for an empty buffer.
1431 */
David S. Miller8243126c2006-01-17 02:54:21 -08001432static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001433{
1434 skb->data += len;
1435 skb->tail += len;
1436}
1437
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00001438static inline void skb_reset_mac_len(struct sk_buff *skb)
1439{
1440 skb->mac_len = skb->network_header - skb->mac_header;
1441}
1442
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001443#ifdef NET_SKBUFF_DATA_USES_OFFSET
1444static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1445{
1446 return skb->head + skb->transport_header;
1447}
1448
1449static inline void skb_reset_transport_header(struct sk_buff *skb)
1450{
1451 skb->transport_header = skb->data - skb->head;
1452}
1453
1454static inline void skb_set_transport_header(struct sk_buff *skb,
1455 const int offset)
1456{
1457 skb_reset_transport_header(skb);
1458 skb->transport_header += offset;
1459}
1460
1461static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1462{
1463 return skb->head + skb->network_header;
1464}
1465
1466static inline void skb_reset_network_header(struct sk_buff *skb)
1467{
1468 skb->network_header = skb->data - skb->head;
1469}
1470
1471static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1472{
1473 skb_reset_network_header(skb);
1474 skb->network_header += offset;
1475}
1476
1477static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1478{
1479 return skb->head + skb->mac_header;
1480}
1481
1482static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1483{
1484 return skb->mac_header != ~0U;
1485}
1486
1487static inline void skb_reset_mac_header(struct sk_buff *skb)
1488{
1489 skb->mac_header = skb->data - skb->head;
1490}
1491
1492static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1493{
1494 skb_reset_mac_header(skb);
1495 skb->mac_header += offset;
1496}
1497
1498#else /* NET_SKBUFF_DATA_USES_OFFSET */
1499
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001500static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1501{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001502 return skb->transport_header;
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001503}
1504
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001505static inline void skb_reset_transport_header(struct sk_buff *skb)
1506{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001507 skb->transport_header = skb->data;
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001508}
1509
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001510static inline void skb_set_transport_header(struct sk_buff *skb,
1511 const int offset)
1512{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001513 skb->transport_header = skb->data + offset;
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001514}
1515
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001516static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1517{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001518 return skb->network_header;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001519}
1520
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001521static inline void skb_reset_network_header(struct sk_buff *skb)
1522{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001523 skb->network_header = skb->data;
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001524}
1525
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001526static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1527{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001528 skb->network_header = skb->data + offset;
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001529}
1530
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001531static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1532{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001533 return skb->mac_header;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001534}
1535
1536static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1537{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001538 return skb->mac_header != NULL;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001539}
1540
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001541static inline void skb_reset_mac_header(struct sk_buff *skb)
1542{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001543 skb->mac_header = skb->data;
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001544}
1545
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001546static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1547{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001548 skb->mac_header = skb->data + offset;
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001549}
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001550#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1551
Eric Dumazet03606892012-02-23 10:55:02 +00001552static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1553{
1554 if (skb_mac_header_was_set(skb)) {
1555 const unsigned char *old_mac = skb_mac_header(skb);
1556
1557 skb_set_mac_header(skb, -skb->mac_len);
1558 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1559 }
1560}
1561
Michał Mirosław04fb4512010-12-14 15:24:08 +00001562static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1563{
1564 return skb->csum_start - skb_headroom(skb);
1565}
1566
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001567static inline int skb_transport_offset(const struct sk_buff *skb)
1568{
1569 return skb_transport_header(skb) - skb->data;
1570}
1571
1572static inline u32 skb_network_header_len(const struct sk_buff *skb)
1573{
1574 return skb->transport_header - skb->network_header;
1575}
1576
1577static inline int skb_network_offset(const struct sk_buff *skb)
1578{
1579 return skb_network_header(skb) - skb->data;
1580}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001581
Changli Gaof9599ce2010-08-04 04:43:44 +00001582static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1583{
1584 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1585}
1586
Linus Torvalds1da177e2005-04-16 15:20:36 -07001587/*
1588 * CPUs often take a performance hit when accessing unaligned memory
1589 * locations. The actual performance hit varies, it can be small if the
1590 * hardware handles it or large if we have to take an exception and fix it
1591 * in software.
1592 *
1593 * Since an ethernet header is 14 bytes network drivers often end up with
1594 * the IP header at an unaligned offset. The IP header can be aligned by
1595 * shifting the start of the packet by 2 bytes. Drivers should do this
1596 * with:
1597 *
Tobias Klauser8660c122009-07-13 22:48:16 +00001598 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001599 *
1600 * The downside to this alignment of the IP header is that the DMA is now
1601 * unaligned. On some architectures the cost of an unaligned DMA is high
1602 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00001603 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001604 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1605 * to be overridden.
1606 */
1607#ifndef NET_IP_ALIGN
1608#define NET_IP_ALIGN 2
1609#endif
1610
Anton Blanchard025be812006-03-31 02:27:06 -08001611/*
1612 * The networking layer reserves some headroom in skb data (via
1613 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1614 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08001615 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08001616 *
1617 * Unfortunately this headroom changes the DMA alignment of the resulting
1618 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1619 * on some architectures. An architecture can override this value,
1620 * perhaps setting it to a cacheline in size (since that will maintain
1621 * cacheline alignment of the DMA). It must be a power of 2.
1622 *
David S. Millerd6301d32009-02-08 19:24:13 -08001623 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08001624 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07001625 *
1626 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1627 * to reduce average number of cache lines per packet.
1628 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07001629 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08001630 */
1631#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07001632#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08001633#endif
1634
Herbert Xu3cc0e872006-06-09 16:13:38 -07001635extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001636
1637static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1638{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00001639 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07001640 WARN_ON(1);
1641 return;
1642 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001643 skb->len = len;
1644 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001645}
1646
Ilpo Järvinen419ae742008-03-27 17:54:01 -07001647extern void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001648
1649static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1650{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001651 if (skb->data_len)
1652 return ___pskb_trim(skb, len);
1653 __skb_trim(skb, len);
1654 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001655}
1656
1657static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1658{
1659 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1660}
1661
1662/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07001663 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1664 * @skb: buffer to alter
1665 * @len: new length
1666 *
1667 * This is identical to pskb_trim except that the caller knows that
1668 * the skb is not cloned so we should never get an error due to out-
1669 * of-memory.
1670 */
1671static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1672{
1673 int err = pskb_trim(skb, len);
1674 BUG_ON(err);
1675}
1676
1677/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001678 * skb_orphan - orphan a buffer
1679 * @skb: buffer to orphan
1680 *
1681 * If a buffer currently has an owner then we call the owner's
1682 * destructor function and make the @skb unowned. The buffer continues
1683 * to exist but is no longer charged to its former owner.
1684 */
1685static inline void skb_orphan(struct sk_buff *skb)
1686{
1687 if (skb->destructor)
1688 skb->destructor(skb);
1689 skb->destructor = NULL;
1690 skb->sk = NULL;
1691}
1692
1693/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00001694 * skb_orphan_frags - orphan the frags contained in a buffer
1695 * @skb: buffer to orphan frags from
1696 * @gfp_mask: allocation mask for replacement pages
1697 *
1698 * For each frag in the SKB which needs a destructor (i.e. has an
1699 * owner) create a copy of that frag and release the original
1700 * page by calling the destructor.
1701 */
1702static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1703{
1704 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1705 return 0;
1706 return skb_copy_ubufs(skb, gfp_mask);
1707}
1708
1709/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710 * __skb_queue_purge - empty a list
1711 * @list: list to empty
1712 *
1713 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1714 * the list and one reference dropped. This function does not take the
1715 * list lock and the caller must hold the relevant locks to use it.
1716 */
1717extern void skb_queue_purge(struct sk_buff_head *list);
1718static inline void __skb_queue_purge(struct sk_buff_head *list)
1719{
1720 struct sk_buff *skb;
1721 while ((skb = __skb_dequeue(list)) != NULL)
1722 kfree_skb(skb);
1723}
1724
Eric Dumazet6f532612012-05-18 05:12:12 +00001725extern void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001726
Christoph Hellwig8af27452006-07-31 22:35:23 -07001727extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00001728 unsigned int length,
1729 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07001730
1731/**
1732 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1733 * @dev: network device to receive on
1734 * @length: length to allocate
1735 *
1736 * Allocate a new &sk_buff and assign it a usage count of one. The
1737 * buffer has unspecified headroom built in. Users should allocate
1738 * the headroom they think they need without accounting for the
1739 * built in space. The built in space is used for optimisations.
1740 *
1741 * %NULL is returned if there is no free memory. Although this function
1742 * allocates memory it can be called from an interrupt.
1743 */
1744static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00001745 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07001746{
1747 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1748}
1749
Eric Dumazet6f532612012-05-18 05:12:12 +00001750/* legacy helper around __netdev_alloc_skb() */
1751static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1752 gfp_t gfp_mask)
1753{
1754 return __netdev_alloc_skb(NULL, length, gfp_mask);
1755}
1756
1757/* legacy helper around netdev_alloc_skb() */
1758static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1759{
1760 return netdev_alloc_skb(NULL, length);
1761}
1762
1763
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001764static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
1765 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00001766{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001767 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00001768
1769 if (NET_IP_ALIGN && skb)
1770 skb_reserve(skb, NET_IP_ALIGN);
1771 return skb;
1772}
1773
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001774static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1775 unsigned int length)
1776{
1777 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
1778}
1779
Mel Gorman06140022012-07-31 16:44:24 -07001780/*
1781 * __skb_alloc_page - allocate pages for ps-rx on a skb and preserve pfmemalloc data
1782 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1783 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1784 * @order: size of the allocation
1785 *
1786 * Allocate a new page.
1787 *
1788 * %NULL is returned if there is no free memory.
1789*/
1790static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
1791 struct sk_buff *skb,
1792 unsigned int order)
1793{
1794 struct page *page;
1795
1796 gfp_mask |= __GFP_COLD;
1797
1798 if (!(gfp_mask & __GFP_NOMEMALLOC))
1799 gfp_mask |= __GFP_MEMALLOC;
1800
1801 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
1802 if (skb && page && page->pfmemalloc)
1803 skb->pfmemalloc = true;
1804
1805 return page;
1806}
1807
1808/**
1809 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
1810 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1811 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1812 *
1813 * Allocate a new page.
1814 *
1815 * %NULL is returned if there is no free memory.
1816 */
1817static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
1818 struct sk_buff *skb)
1819{
1820 return __skb_alloc_pages(gfp_mask, skb, 0);
1821}
1822
1823/**
1824 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
1825 * @page: The page that was allocated from skb_alloc_page
1826 * @skb: The skb that may need pfmemalloc set
1827 */
1828static inline void skb_propagate_pfmemalloc(struct page *page,
1829 struct sk_buff *skb)
1830{
1831 if (page && page->pfmemalloc)
1832 skb->pfmemalloc = true;
1833}
1834
Eric Dumazet564824b2010-10-11 19:05:25 +00001835/**
Ian Campbell131ea662011-08-19 06:25:00 +00001836 * skb_frag_page - retrieve the page refered to by a paged fragment
1837 * @frag: the paged fragment
1838 *
1839 * Returns the &struct page associated with @frag.
1840 */
1841static inline struct page *skb_frag_page(const skb_frag_t *frag)
1842{
Ian Campbella8605c62011-10-19 23:01:49 +00001843 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00001844}
1845
1846/**
1847 * __skb_frag_ref - take an addition reference on a paged fragment.
1848 * @frag: the paged fragment
1849 *
1850 * Takes an additional reference on the paged fragment @frag.
1851 */
1852static inline void __skb_frag_ref(skb_frag_t *frag)
1853{
1854 get_page(skb_frag_page(frag));
1855}
1856
1857/**
1858 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
1859 * @skb: the buffer
1860 * @f: the fragment offset.
1861 *
1862 * Takes an additional reference on the @f'th paged fragment of @skb.
1863 */
1864static inline void skb_frag_ref(struct sk_buff *skb, int f)
1865{
1866 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
1867}
1868
1869/**
1870 * __skb_frag_unref - release a reference on a paged fragment.
1871 * @frag: the paged fragment
1872 *
1873 * Releases a reference on the paged fragment @frag.
1874 */
1875static inline void __skb_frag_unref(skb_frag_t *frag)
1876{
1877 put_page(skb_frag_page(frag));
1878}
1879
1880/**
1881 * skb_frag_unref - release a reference on a paged fragment of an skb.
1882 * @skb: the buffer
1883 * @f: the fragment offset
1884 *
1885 * Releases a reference on the @f'th paged fragment of @skb.
1886 */
1887static inline void skb_frag_unref(struct sk_buff *skb, int f)
1888{
1889 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
1890}
1891
1892/**
1893 * skb_frag_address - gets the address of the data contained in a paged fragment
1894 * @frag: the paged fragment buffer
1895 *
1896 * Returns the address of the data within @frag. The page must already
1897 * be mapped.
1898 */
1899static inline void *skb_frag_address(const skb_frag_t *frag)
1900{
1901 return page_address(skb_frag_page(frag)) + frag->page_offset;
1902}
1903
1904/**
1905 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
1906 * @frag: the paged fragment buffer
1907 *
1908 * Returns the address of the data within @frag. Checks that the page
1909 * is mapped and returns %NULL otherwise.
1910 */
1911static inline void *skb_frag_address_safe(const skb_frag_t *frag)
1912{
1913 void *ptr = page_address(skb_frag_page(frag));
1914 if (unlikely(!ptr))
1915 return NULL;
1916
1917 return ptr + frag->page_offset;
1918}
1919
1920/**
1921 * __skb_frag_set_page - sets the page contained in a paged fragment
1922 * @frag: the paged fragment
1923 * @page: the page to set
1924 *
1925 * Sets the fragment @frag to contain @page.
1926 */
1927static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
1928{
Ian Campbella8605c62011-10-19 23:01:49 +00001929 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00001930}
1931
1932/**
1933 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
1934 * @skb: the buffer
1935 * @f: the fragment offset
1936 * @page: the page to set
1937 *
1938 * Sets the @f'th fragment of @skb to contain @page.
1939 */
1940static inline void skb_frag_set_page(struct sk_buff *skb, int f,
1941 struct page *page)
1942{
1943 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
1944}
1945
1946/**
1947 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00001948 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00001949 * @frag: the paged fragment to map
1950 * @offset: the offset within the fragment (starting at the
1951 * fragment's own offset)
1952 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00001953 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00001954 *
1955 * Maps the page associated with @frag to @device.
1956 */
1957static inline dma_addr_t skb_frag_dma_map(struct device *dev,
1958 const skb_frag_t *frag,
1959 size_t offset, size_t size,
1960 enum dma_data_direction dir)
1961{
1962 return dma_map_page(dev, skb_frag_page(frag),
1963 frag->page_offset + offset, size, dir);
1964}
1965
Eric Dumazet117632e2011-12-03 21:39:53 +00001966static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
1967 gfp_t gfp_mask)
1968{
1969 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
1970}
1971
Ian Campbell131ea662011-08-19 06:25:00 +00001972/**
Patrick McHardy334a8132007-06-25 04:35:20 -07001973 * skb_clone_writable - is the header of a clone writable
1974 * @skb: buffer to check
1975 * @len: length up to which to write
1976 *
1977 * Returns true if modifying the header part of the cloned buffer
1978 * does not requires the data to be copied.
1979 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001980static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07001981{
1982 return !skb_header_cloned(skb) &&
1983 skb_headroom(skb) + len <= skb->hdr_len;
1984}
1985
Herbert Xud9cc2042007-09-16 16:21:16 -07001986static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
1987 int cloned)
1988{
1989 int delta = 0;
1990
Herbert Xud9cc2042007-09-16 16:21:16 -07001991 if (headroom > skb_headroom(skb))
1992 delta = headroom - skb_headroom(skb);
1993
1994 if (delta || cloned)
1995 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
1996 GFP_ATOMIC);
1997 return 0;
1998}
1999
Patrick McHardy334a8132007-06-25 04:35:20 -07002000/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002001 * skb_cow - copy header of skb when it is required
2002 * @skb: buffer to cow
2003 * @headroom: needed headroom
2004 *
2005 * If the skb passed lacks sufficient headroom or its data part
2006 * is shared, data is reallocated. If reallocation fails, an error
2007 * is returned and original skb is not changed.
2008 *
2009 * The result is skb with writable area skb->head...skb->tail
2010 * and at least @headroom of space at head.
2011 */
2012static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2013{
Herbert Xud9cc2042007-09-16 16:21:16 -07002014 return __skb_cow(skb, headroom, skb_cloned(skb));
2015}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002016
Herbert Xud9cc2042007-09-16 16:21:16 -07002017/**
2018 * skb_cow_head - skb_cow but only making the head writable
2019 * @skb: buffer to cow
2020 * @headroom: needed headroom
2021 *
2022 * This function is identical to skb_cow except that we replace the
2023 * skb_cloned check by skb_header_cloned. It should be used when
2024 * you only need to push on some header and do not need to modify
2025 * the data.
2026 */
2027static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2028{
2029 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002030}
2031
2032/**
2033 * skb_padto - pad an skbuff up to a minimal size
2034 * @skb: buffer to pad
2035 * @len: minimal length
2036 *
2037 * Pads up a buffer to ensure the trailing bytes exist and are
2038 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002039 * is untouched. Otherwise it is extended. Returns zero on
2040 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002041 */
2042
Herbert Xu5b057c62006-06-23 02:06:41 -07002043static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002044{
2045 unsigned int size = skb->len;
2046 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002047 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002048 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002049}
2050
2051static inline int skb_add_data(struct sk_buff *skb,
2052 char __user *from, int copy)
2053{
2054 const int off = skb->len;
2055
2056 if (skb->ip_summed == CHECKSUM_NONE) {
2057 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08002058 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
Linus Torvalds1da177e2005-04-16 15:20:36 -07002059 copy, 0, &err);
2060 if (!err) {
2061 skb->csum = csum_block_add(skb->csum, csum, off);
2062 return 0;
2063 }
2064 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2065 return 0;
2066
2067 __skb_trim(skb, off);
2068 return -EFAULT;
2069}
2070
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002071static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2072 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002073{
2074 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002075 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002076
Ian Campbellea2ab692011-08-22 23:44:58 +00002077 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002078 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002079 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002080 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002081}
2082
Herbert Xu364c6ba2006-06-09 16:10:40 -07002083static inline int __skb_linearize(struct sk_buff *skb)
2084{
2085 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2086}
2087
Linus Torvalds1da177e2005-04-16 15:20:36 -07002088/**
2089 * skb_linearize - convert paged skb to linear one
2090 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002091 *
2092 * If there is no free memory -ENOMEM is returned, otherwise zero
2093 * is returned and the old skb data released.
2094 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002095static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002096{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002097 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2098}
2099
2100/**
2101 * skb_linearize_cow - make sure skb is linear and writable
2102 * @skb: buffer to process
2103 *
2104 * If there is no free memory -ENOMEM is returned, otherwise zero
2105 * is returned and the old skb data released.
2106 */
2107static inline int skb_linearize_cow(struct sk_buff *skb)
2108{
2109 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2110 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002111}
2112
2113/**
2114 * skb_postpull_rcsum - update checksum for received skb after pull
2115 * @skb: buffer to update
2116 * @start: start of data before pull
2117 * @len: length of data pulled
2118 *
2119 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002120 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2121 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002122 */
2123
2124static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002125 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002126{
Patrick McHardy84fa7932006-08-29 16:44:56 -07002127 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002128 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2129}
2130
Herbert Xucbb042f2006-03-20 22:43:56 -08002131unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2132
Linus Torvalds1da177e2005-04-16 15:20:36 -07002133/**
2134 * pskb_trim_rcsum - trim received skb and update checksum
2135 * @skb: buffer to trim
2136 * @len: new length
2137 *
2138 * This is exactly the same as pskb_trim except that it ensures the
2139 * checksum of received packets are still valid after the operation.
2140 */
2141
2142static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2143{
Stephen Hemminger0e4e4222005-09-08 12:32:03 -07002144 if (likely(len >= skb->len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002145 return 0;
Patrick McHardy84fa7932006-08-29 16:44:56 -07002146 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002147 skb->ip_summed = CHECKSUM_NONE;
2148 return __pskb_trim(skb, len);
2149}
2150
Linus Torvalds1da177e2005-04-16 15:20:36 -07002151#define skb_queue_walk(queue, skb) \
2152 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002153 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002154 skb = skb->next)
2155
James Chapman46f89142007-04-30 00:07:31 -07002156#define skb_queue_walk_safe(queue, skb, tmp) \
2157 for (skb = (queue)->next, tmp = skb->next; \
2158 skb != (struct sk_buff *)(queue); \
2159 skb = tmp, tmp = skb->next)
2160
David S. Miller1164f522008-09-23 00:49:44 -07002161#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002162 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002163 skb = skb->next)
2164
2165#define skb_queue_walk_from_safe(queue, skb, tmp) \
2166 for (tmp = skb->next; \
2167 skb != (struct sk_buff *)(queue); \
2168 skb = tmp, tmp = skb->next)
2169
Stephen Hemminger300ce172005-10-30 13:47:34 -08002170#define skb_queue_reverse_walk(queue, skb) \
2171 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002172 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08002173 skb = skb->prev)
2174
David S. Miller686a2952011-01-20 22:47:32 -08002175#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2176 for (skb = (queue)->prev, tmp = skb->prev; \
2177 skb != (struct sk_buff *)(queue); \
2178 skb = tmp, tmp = skb->prev)
2179
2180#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2181 for (tmp = skb->prev; \
2182 skb != (struct sk_buff *)(queue); \
2183 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002184
David S. Miller21dc3302010-08-23 00:13:46 -07002185static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07002186{
2187 return skb_shinfo(skb)->frag_list != NULL;
2188}
2189
2190static inline void skb_frag_list_init(struct sk_buff *skb)
2191{
2192 skb_shinfo(skb)->frag_list = NULL;
2193}
2194
2195static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2196{
2197 frag->next = skb_shinfo(skb)->frag_list;
2198 skb_shinfo(skb)->frag_list = frag;
2199}
2200
2201#define skb_walk_frags(skb, iter) \
2202 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2203
Herbert Xua59322b2007-12-05 01:53:40 -08002204extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
Pavel Emelyanov3f518bf2012-02-21 07:30:58 +00002205 int *peeked, int *off, int *err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002206extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
2207 int noblock, int *err);
2208extern unsigned int datagram_poll(struct file *file, struct socket *sock,
2209 struct poll_table_struct *wait);
2210extern int skb_copy_datagram_iovec(const struct sk_buff *from,
2211 int offset, struct iovec *to,
2212 int size);
Herbert Xufb286bb2005-11-10 13:01:24 -08002213extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002214 int hlen,
2215 struct iovec *iov);
Rusty Russelldb543c12008-08-15 15:13:53 -07002216extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
2217 int offset,
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +00002218 const struct iovec *from,
2219 int from_offset,
Rusty Russelldb543c12008-08-15 15:13:53 -07002220 int len);
Michael S. Tsirkin0a1ec072009-04-20 01:25:46 +00002221extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
2222 int offset,
2223 const struct iovec *to,
2224 int to_offset,
2225 int size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002226extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
Eric Dumazet9d410c72009-10-30 05:03:53 +00002227extern void skb_free_datagram_locked(struct sock *sk,
2228 struct sk_buff *skb);
Herbert Xu27ab2562007-12-05 01:51:58 -08002229extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
Herbert Xu3305b802005-12-13 23:16:37 -08002230 unsigned int flags);
Al Viro2bbbc862006-11-14 21:37:14 -08002231extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
2232 int len, __wsum csum);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002233extern int skb_copy_bits(const struct sk_buff *skb, int offset,
2234 void *to, int len);
Stephen Hemminger0c6fcc82007-04-20 16:40:01 -07002235extern int skb_store_bits(struct sk_buff *skb, int offset,
2236 const void *from, int len);
Al Viro81d77662006-11-14 21:37:33 -08002237extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002238 int offset, u8 *to, int len,
Al Viro81d77662006-11-14 21:37:33 -08002239 __wsum csum);
Jens Axboe9c55e012007-11-06 23:30:13 -08002240extern int skb_splice_bits(struct sk_buff *skb,
2241 unsigned int offset,
2242 struct pipe_inode_info *pipe,
2243 unsigned int len,
2244 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002245extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
2246extern void skb_split(struct sk_buff *skb,
2247 struct sk_buff *skb1, const u32 len);
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08002248extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
2249 int shiftlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002250
Michał Mirosławc8f44af2011-11-15 15:29:55 +00002251extern struct sk_buff *skb_segment(struct sk_buff *skb,
2252 netdev_features_t features);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002253
Linus Torvalds1da177e2005-04-16 15:20:36 -07002254static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2255 int len, void *buffer)
2256{
2257 int hlen = skb_headlen(skb);
2258
Patrick McHardy55820ee2005-07-05 14:08:10 -07002259 if (hlen - offset >= len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002260 return skb->data + offset;
2261
2262 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2263 return NULL;
2264
2265 return buffer;
2266}
2267
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03002268static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2269 void *to,
2270 const unsigned int len)
2271{
2272 memcpy(to, skb->data, len);
2273}
2274
2275static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2276 const int offset, void *to,
2277 const unsigned int len)
2278{
2279 memcpy(to, skb->data + offset, len);
2280}
2281
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03002282static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2283 const void *from,
2284 const unsigned int len)
2285{
2286 memcpy(skb->data, from, len);
2287}
2288
2289static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2290 const int offset,
2291 const void *from,
2292 const unsigned int len)
2293{
2294 memcpy(skb->data + offset, from, len);
2295}
2296
Linus Torvalds1da177e2005-04-16 15:20:36 -07002297extern void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002298
Patrick Ohlyac45f602009-02-12 05:03:37 +00002299static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2300{
2301 return skb->tstamp;
2302}
2303
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002304/**
2305 * skb_get_timestamp - get timestamp from a skb
2306 * @skb: skb to get stamp from
2307 * @stamp: pointer to struct timeval to store stamp in
2308 *
2309 * Timestamps are stored in the skb as offsets to a base timestamp.
2310 * This function converts the offset back to a struct timeval and stores
2311 * it in stamp.
2312 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00002313static inline void skb_get_timestamp(const struct sk_buff *skb,
2314 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002315{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002316 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002317}
2318
Patrick Ohlyac45f602009-02-12 05:03:37 +00002319static inline void skb_get_timestampns(const struct sk_buff *skb,
2320 struct timespec *stamp)
2321{
2322 *stamp = ktime_to_timespec(skb->tstamp);
2323}
2324
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002325static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002326{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002327 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002328}
2329
Stephen Hemminger164891a2007-04-23 22:26:16 -07002330static inline ktime_t net_timedelta(ktime_t t)
2331{
2332 return ktime_sub(ktime_get_real(), t);
2333}
2334
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07002335static inline ktime_t net_invalid_timestamp(void)
2336{
2337 return ktime_set(0, 0);
2338}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002339
Richard Cochranc1f19b52010-07-17 08:49:36 +00002340extern void skb_timestamping_init(void);
2341
2342#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2343
2344extern void skb_clone_tx_timestamp(struct sk_buff *skb);
2345extern bool skb_defer_rx_timestamp(struct sk_buff *skb);
2346
2347#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2348
2349static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2350{
2351}
2352
2353static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2354{
2355 return false;
2356}
2357
2358#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2359
2360/**
2361 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2362 *
Richard Cochranda92b192011-10-21 00:49:15 +00002363 * PHY drivers may accept clones of transmitted packets for
2364 * timestamping via their phy_driver.txtstamp method. These drivers
2365 * must call this function to return the skb back to the stack, with
2366 * or without a timestamp.
2367 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00002368 * @skb: clone of the the original outgoing packet
Richard Cochranda92b192011-10-21 00:49:15 +00002369 * @hwtstamps: hardware time stamps, may be NULL if not available
Richard Cochranc1f19b52010-07-17 08:49:36 +00002370 *
2371 */
2372void skb_complete_tx_timestamp(struct sk_buff *skb,
2373 struct skb_shared_hwtstamps *hwtstamps);
2374
Patrick Ohlyac45f602009-02-12 05:03:37 +00002375/**
2376 * skb_tstamp_tx - queue clone of skb with send time stamps
2377 * @orig_skb: the original outgoing packet
2378 * @hwtstamps: hardware time stamps, may be NULL if not available
2379 *
2380 * If the skb has a socket associated, then this function clones the
2381 * skb (thus sharing the actual data and optional structures), stores
2382 * the optional hardware time stamping information (if non NULL) or
2383 * generates a software time stamp (otherwise), then queues the clone
2384 * to the error queue of the socket. Errors are silently ignored.
2385 */
2386extern void skb_tstamp_tx(struct sk_buff *orig_skb,
2387 struct skb_shared_hwtstamps *hwtstamps);
2388
Richard Cochran4507a712010-07-17 08:48:28 +00002389static inline void sw_tx_timestamp(struct sk_buff *skb)
2390{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002391 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2392 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00002393 skb_tstamp_tx(skb, NULL);
2394}
2395
2396/**
2397 * skb_tx_timestamp() - Driver hook for transmit timestamping
2398 *
2399 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00002400 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00002401 *
2402 * @skb: A socket buffer.
2403 */
2404static inline void skb_tx_timestamp(struct sk_buff *skb)
2405{
Richard Cochranc1f19b52010-07-17 08:49:36 +00002406 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002407 sw_tx_timestamp(skb);
2408}
2409
Johannes Berg6e3e9392011-11-09 10:15:42 +01002410/**
2411 * skb_complete_wifi_ack - deliver skb with wifi status
2412 *
2413 * @skb: the original outgoing packet
2414 * @acked: ack status
2415 *
2416 */
2417void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2418
Herbert Xu759e5d02007-03-25 20:10:56 -07002419extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
Al Virob51655b2006-11-14 21:40:42 -08002420extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002421
Herbert Xu60476372007-04-09 11:59:39 -07002422static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2423{
2424 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2425}
2426
Herbert Xufb286bb2005-11-10 13:01:24 -08002427/**
2428 * skb_checksum_complete - Calculate checksum of an entire packet
2429 * @skb: packet to process
2430 *
2431 * This function calculates the checksum over the entire packet plus
2432 * the value of skb->csum. The latter can be used to supply the
2433 * checksum of a pseudo header as used by TCP/UDP. It returns the
2434 * checksum.
2435 *
2436 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2437 * this function can be used to verify that checksum on received
2438 * packets. In that case the function should return zero if the
2439 * checksum is correct. In particular, this function will return zero
2440 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2441 * hardware has already verified the correctness of the checksum.
2442 */
Al Viro4381ca32007-07-15 21:00:11 +01002443static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002444{
Herbert Xu60476372007-04-09 11:59:39 -07002445 return skb_csum_unnecessary(skb) ?
2446 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002447}
2448
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002449#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002450extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002451static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2452{
2453 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002454 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002455}
2456static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2457{
2458 if (nfct)
2459 atomic_inc(&nfct->use);
2460}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002461#endif
2462#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -08002463static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
2464{
2465 if (skb)
2466 atomic_inc(&skb->users);
2467}
2468static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
2469{
2470 if (skb)
2471 kfree_skb(skb);
2472}
2473#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002474#ifdef CONFIG_BRIDGE_NETFILTER
2475static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2476{
2477 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2478 kfree(nf_bridge);
2479}
2480static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2481{
2482 if (nf_bridge)
2483 atomic_inc(&nf_bridge->use);
2484}
2485#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08002486static inline void nf_reset(struct sk_buff *skb)
2487{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002488#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08002489 nf_conntrack_put(skb->nfct);
2490 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002491#endif
2492#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Patrick McHardya193a4a2006-03-20 19:23:05 -08002493 nf_conntrack_put_reasm(skb->nfct_reasm);
2494 skb->nfct_reasm = NULL;
2495#endif
2496#ifdef CONFIG_BRIDGE_NETFILTER
2497 nf_bridge_put(skb->nf_bridge);
2498 skb->nf_bridge = NULL;
2499#endif
2500}
2501
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002502/* Note: This doesn't put any conntrack and bridge info in dst. */
2503static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2504{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002505#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002506 dst->nfct = src->nfct;
2507 nf_conntrack_get(src->nfct);
2508 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002509#endif
2510#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002511 dst->nfct_reasm = src->nfct_reasm;
2512 nf_conntrack_get_reasm(src->nfct_reasm);
2513#endif
2514#ifdef CONFIG_BRIDGE_NETFILTER
2515 dst->nf_bridge = src->nf_bridge;
2516 nf_bridge_get(src->nf_bridge);
2517#endif
2518}
2519
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002520static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2521{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002522#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002523 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002524#endif
2525#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002526 nf_conntrack_put_reasm(dst->nfct_reasm);
2527#endif
2528#ifdef CONFIG_BRIDGE_NETFILTER
2529 nf_bridge_put(dst->nf_bridge);
2530#endif
2531 __nf_copy(dst, src);
2532}
2533
James Morris984bc162006-06-09 00:29:17 -07002534#ifdef CONFIG_NETWORK_SECMARK
2535static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2536{
2537 to->secmark = from->secmark;
2538}
2539
2540static inline void skb_init_secmark(struct sk_buff *skb)
2541{
2542 skb->secmark = 0;
2543}
2544#else
2545static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2546{ }
2547
2548static inline void skb_init_secmark(struct sk_buff *skb)
2549{ }
2550#endif
2551
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002552static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2553{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002554 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002555}
2556
Stephen Hemminger92477442009-03-21 13:39:26 -07002557static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002558{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002559 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002560}
2561
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002562static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2563{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002564 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002565}
2566
David S. Millerd5a9e242009-01-27 16:22:11 -08002567static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2568{
2569 skb->queue_mapping = rx_queue + 1;
2570}
2571
Stephen Hemminger92477442009-03-21 13:39:26 -07002572static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002573{
2574 return skb->queue_mapping - 1;
2575}
2576
Stephen Hemminger92477442009-03-21 13:39:26 -07002577static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002578{
Eric Dumazeta02cec22010-09-22 20:43:57 +00002579 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08002580}
2581
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00002582extern u16 __skb_tx_hash(const struct net_device *dev,
2583 const struct sk_buff *skb,
2584 unsigned int num_tx_queues);
Stephen Hemminger92477442009-03-21 13:39:26 -07002585
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002586#ifdef CONFIG_XFRM
2587static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2588{
2589 return skb->sp;
2590}
2591#else
2592static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2593{
2594 return NULL;
2595}
2596#endif
2597
David S. Millerbdcc0922012-03-07 20:53:36 -05002598static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07002599{
2600 return skb_shinfo(skb)->gso_size;
2601}
2602
David S. Millerbdcc0922012-03-07 20:53:36 -05002603static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02002604{
2605 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2606}
2607
Ben Hutchings4497b072008-06-19 16:22:28 -07002608extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2609
2610static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2611{
2612 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2613 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002614 const struct skb_shared_info *shinfo = skb_shinfo(skb);
2615
Alexander Duyckb78462e2010-06-02 12:24:37 +00002616 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2617 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07002618 __skb_warn_lro_forwarding(skb);
2619 return true;
2620 }
2621 return false;
2622}
2623
Herbert Xu35fc92a2007-03-26 23:22:20 -07002624static inline void skb_forward_csum(struct sk_buff *skb)
2625{
2626 /* Unfortunately we don't support this one. Any brave souls? */
2627 if (skb->ip_summed == CHECKSUM_COMPLETE)
2628 skb->ip_summed = CHECKSUM_NONE;
2629}
2630
Eric Dumazetbc8acf22010-09-02 13:07:41 -07002631/**
2632 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2633 * @skb: skb to check
2634 *
2635 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2636 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2637 * use this helper, to document places where we make this assertion.
2638 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002639static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07002640{
2641#ifdef DEBUG
2642 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2643#endif
2644}
2645
Rusty Russellf35d9d82008-02-04 23:49:54 -05002646bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00002647
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002648static inline bool skb_is_recycleable(const struct sk_buff *skb, int skb_size)
Andy Fleming3d153a72011-10-13 04:33:54 +00002649{
2650 if (irqs_disabled())
2651 return false;
2652
2653 if (skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)
2654 return false;
2655
2656 if (skb_is_nonlinear(skb) || skb->fclone != SKB_FCLONE_UNAVAILABLE)
2657 return false;
2658
2659 skb_size = SKB_DATA_ALIGN(skb_size + NET_SKB_PAD);
Alexander Duyckec47ea82012-05-04 14:26:56 +00002660 if (skb_end_offset(skb) < skb_size)
Andy Fleming3d153a72011-10-13 04:33:54 +00002661 return false;
2662
2663 if (skb_shared(skb) || skb_cloned(skb))
2664 return false;
2665
2666 return true;
2667}
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00002668
2669/**
2670 * skb_head_is_locked - Determine if the skb->head is locked down
2671 * @skb: skb to check
2672 *
2673 * The head on skbs build around a head frag can be removed if they are
2674 * not cloned. This function returns true if the skb head is locked down
2675 * due to either being allocated via kmalloc, or by being a clone with
2676 * multiple references to the head.
2677 */
2678static inline bool skb_head_is_locked(const struct sk_buff *skb)
2679{
2680 return !skb->head_frag || skb_cloned(skb);
2681}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002682#endif /* __KERNEL__ */
2683#endif /* _LINUX_SKBUFF_H */