blob: ca6ee7d93edb4b325bf6a4a85545b93c497597ee [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,
Pravin B Shelarc9af6db2013-02-11 09:27:41 +0000233
234 /* This indicates at least one fragment might be overwritten
235 * (as in vmsplice(), sendfile() ...)
236 * If we need to compute a TX checksum, we'll need to copy
237 * all frags to avoid possible bad checksum
238 */
239 SKBTX_SHARED_FRAG = 1 << 5,
Shirley Maa6686f22011-07-06 12:22:12 +0000240};
241
242/*
243 * The callback notifies userspace to release buffers when skb DMA is done in
244 * lower device, the skb last reference should be 0 when calling this.
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000245 * The zerocopy_success argument is true if zero copy transmit occurred,
246 * false on data copy or out of memory error caused by data copy attempt.
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000247 * The ctx field is used to track device context.
248 * The desc field is used to track userspace buffer index.
Shirley Maa6686f22011-07-06 12:22:12 +0000249 */
250struct ubuf_info {
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000251 void (*callback)(struct ubuf_info *, bool zerocopy_success);
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000252 void *ctx;
Shirley Maa6686f22011-07-06 12:22:12 +0000253 unsigned long desc;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000254};
255
Linus Torvalds1da177e2005-04-16 15:20:36 -0700256/* This data is invariant across clones and lives at
257 * the end of the header data, ie. at skb->end.
258 */
259struct skb_shared_info {
Ian Campbell9f42f122012-01-05 07:13:39 +0000260 unsigned char nr_frags;
261 __u8 tx_flags;
Herbert Xu79671682006-06-22 02:40:14 -0700262 unsigned short gso_size;
263 /* Warning: this field is not always filled in (UFO)! */
264 unsigned short gso_segs;
265 unsigned short gso_type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700266 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000267 struct skb_shared_hwtstamps hwtstamps;
Ian Campbell9f42f122012-01-05 07:13:39 +0000268 __be32 ip6_frag_id;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700269
270 /*
271 * Warning : all fields before dataref are cleared in __alloc_skb()
272 */
273 atomic_t dataref;
274
Johann Baudy69e3c752009-05-18 22:11:22 -0700275 /* Intermediate layers must ensure that destructor_arg
276 * remains valid until skb destructor */
277 void * destructor_arg;
Shirley Maa6686f22011-07-06 12:22:12 +0000278
Eric Dumazetfed66382010-07-22 19:09:08 +0000279 /* must be last field, see pskb_expand_head() */
280 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281};
282
283/* We divide dataref into two halves. The higher 16 bits hold references
284 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700285 * the entire skb->data. A clone of a headerless skb holds the length of
286 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287 *
288 * All users must obey the rule that the skb->data reference count must be
289 * greater than or equal to the payload reference count.
290 *
291 * Holding a reference to the payload part means that the user does not
292 * care about modifications to the header part of skb->data.
293 */
294#define SKB_DATAREF_SHIFT 16
295#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
296
David S. Millerd179cd12005-08-17 14:57:30 -0700297
298enum {
299 SKB_FCLONE_UNAVAILABLE,
300 SKB_FCLONE_ORIG,
301 SKB_FCLONE_CLONE,
302};
303
Herbert Xu79671682006-06-22 02:40:14 -0700304enum {
305 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700306 SKB_GSO_UDP = 1 << 1,
Herbert Xu576a30e2006-06-27 13:22:38 -0700307
308 /* This indicates the skb is from an untrusted source. */
309 SKB_GSO_DODGY = 1 << 2,
Michael Chanb0da85372006-06-29 12:30:00 -0700310
311 /* This indicates the tcp segment has CWR set. */
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700312 SKB_GSO_TCP_ECN = 1 << 3,
313
314 SKB_GSO_TCPV6 = 1 << 4,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800315
316 SKB_GSO_FCOE = 1 << 5,
Herbert Xu79671682006-06-22 02:40:14 -0700317};
318
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700319#if BITS_PER_LONG > 32
320#define NET_SKBUFF_DATA_USES_OFFSET 1
321#endif
322
323#ifdef NET_SKBUFF_DATA_USES_OFFSET
324typedef unsigned int sk_buff_data_t;
325#else
326typedef unsigned char *sk_buff_data_t;
327#endif
328
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100329#if defined(CONFIG_NF_DEFRAG_IPV4) || defined(CONFIG_NF_DEFRAG_IPV4_MODULE) || \
330 defined(CONFIG_NF_DEFRAG_IPV6) || defined(CONFIG_NF_DEFRAG_IPV6_MODULE)
331#define NET_SKBUFF_NF_DEFRAG_NEEDED 1
332#endif
333
Linus Torvalds1da177e2005-04-16 15:20:36 -0700334/**
335 * struct sk_buff - socket buffer
336 * @next: Next buffer in list
337 * @prev: Previous buffer in list
Herbert Xu325ed822005-10-03 13:57:23 -0700338 * @tstamp: Time we arrived
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700339 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700341 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000342 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700343 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344 * @len: Length of actual data
345 * @data_len: Data length
346 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700347 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700348 * @csum: Checksum (must include start/offset pair)
349 * @csum_start: Offset from skb->head where checksumming should start
350 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700351 * @priority: Packet queueing priority
Martin Waitz67be2dd2005-05-01 08:59:26 -0700352 * @local_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700354 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355 * @nohdr: Payload reference only, must not modify header
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700356 * @nfctinfo: Relationship of this skb to the connection
Linus Torvalds1da177e2005-04-16 15:20:36 -0700357 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700358 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700359 * @ipvs_property: skbuff is owned by ipvs
Randy Dunlap31729362008-02-18 20:52:13 -0800360 * @peeked: this packet has been seen already, so stats have been
361 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700362 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700363 * @protocol: Packet protocol from driver
364 * @destructor: Destruct function
365 * @nfct: Associated connection, if any
Randy Dunlap461ddf32005-11-20 21:25:15 -0800366 * @nfct_reasm: netfilter conntrack re-assembly pointer
Linus Torvalds1da177e2005-04-16 15:20:36 -0700367 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800368 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369 * @tc_index: Traffic control index
370 * @tc_verd: traffic control verdict
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700371 * @rxhash: the packet hash computed on receive
372 * @queue_mapping: Queue mapping for multiqueue devices
Randy Dunlap553a5672008-04-20 10:51:01 -0700373 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700374 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Changli Gao4ca24622011-08-19 07:26:44 -0700375 * @l4_rxhash: indicate rxhash is a canonical 4-tuple hash over transport
376 * ports.
Johannes Berg6e3e9392011-11-09 10:15:42 +0100377 * @wifi_acked_valid: wifi_acked was set
378 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000379 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Randy Dunlapf4b8ea72006-06-22 16:00:11 -0700380 * @dma_cookie: a cookie to one of several possible DMA operations
381 * done by skb DMA functions
James Morris984bc162006-06-09 00:29:17 -0700382 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700383 * @mark: Generic packet mark
384 * @dropcount: total number of sk_receive_queue overflows
Patrick McHardy6aa895b02008-07-14 22:49:06 -0700385 * @vlan_tci: vlan tag control information
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000386 * @inner_transport_header: Inner transport layer header (encapsulation)
387 * @inner_network_header: Network layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700388 * @transport_header: Transport layer header
389 * @network_header: Network layer header
390 * @mac_header: Link layer header
391 * @tail: Tail pointer
392 * @end: End pointer
393 * @head: Head of buffer
394 * @data: Data head pointer
395 * @truesize: Buffer size
396 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397 */
398
399struct sk_buff {
400 /* These two members must be first. */
401 struct sk_buff *next;
402 struct sk_buff *prev;
403
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700404 ktime_t tstamp;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000405
406 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700407 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700408
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409 /*
410 * This is the control buffer. It is free to use for every
411 * layer. Please put your private variables there. If you
412 * want to keep them across layers you have to do a skb_clone()
413 * first. This is owned by whoever has the skb queued ATM.
414 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000415 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416
Eric Dumazet7fee2262010-05-11 23:19:48 +0000417 unsigned long _skb_refdst;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000418#ifdef CONFIG_XFRM
419 struct sec_path *sp;
420#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700422 data_len;
423 __u16 mac_len,
424 hdr_len;
Al Viroff1dcad2006-11-20 18:07:29 -0800425 union {
426 __wsum csum;
Herbert Xu663ead32007-04-09 11:59:07 -0700427 struct {
428 __u16 csum_start;
429 __u16 csum_offset;
430 };
Al Viroff1dcad2006-11-20 18:07:29 -0800431 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432 __u32 priority;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200433 kmemcheck_bitfield_begin(flags1);
Thomas Graf1cbb3382005-07-05 14:13:41 -0700434 __u8 local_df:1,
435 cloned:1,
436 ip_summed:2,
Harald Welte6869c4d2005-08-09 19:24:19 -0700437 nohdr:1,
438 nfctinfo:3;
David S. Millerd179cd12005-08-17 14:57:30 -0700439 __u8 pkt_type:3,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800440 fclone:2,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700441 ipvs_property:1,
Herbert Xua59322b2007-12-05 01:53:40 -0800442 peeked:1,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700443 nf_trace:1;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200444 kmemcheck_bitfield_end(flags1);
Eric Dumazet4ab408d2010-03-01 03:09:26 +0000445 __be16 protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446
447 void (*destructor)(struct sk_buff *skb);
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800448#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700449 struct nf_conntrack *nfct;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100450#endif
451#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800452 struct sk_buff *nfct_reasm;
453#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700454#ifdef CONFIG_BRIDGE_NETFILTER
455 struct nf_bridge_info *nf_bridge;
456#endif
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700457
Eric Dumazet8964be42009-11-20 15:35:04 -0800458 int skb_iif;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000459
460 __u32 rxhash;
461
462 __u16 vlan_tci;
463
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700465 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700467 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700468#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200470
Eric Dumazet0a148422011-04-20 09:27:32 +0000471 __u16 queue_mapping;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200472 kmemcheck_bitfield_begin(flags2);
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400473#ifdef CONFIG_IPV6_NDISC_NODETYPE
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000474 __u8 ndisc_nodetype:2;
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400475#endif
Mel Gormanc93bdd02012-07-31 16:44:19 -0700476 __u8 pfmemalloc:1;
Tom Herbert3853b582010-11-21 13:17:29 +0000477 __u8 ooo_okay:1;
Tom Herbertbdeab992011-08-14 19:45:55 +0000478 __u8 l4_rxhash:1;
Johannes Berg6e3e9392011-11-09 10:15:42 +0100479 __u8 wifi_acked_valid:1;
480 __u8 wifi_acked:1;
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000481 __u8 no_fcs:1;
Eric Dumazetd3836f22012-04-27 00:33:38 +0000482 __u8 head_frag:1;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000483 /* Encapsulation protocol and NIC drivers should use
484 * this flag to indicate to each other if the skb contains
485 * encapsulated packet or not and maybe use the inner packet
486 * headers if needed
487 */
488 __u8 encapsulation:1;
489 /* 7/9 bit hole (depending on ndisc_nodetype presence) */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200490 kmemcheck_bitfield_end(flags2);
491
Chris Leech97fc2f02006-05-23 17:55:33 -0700492#ifdef CONFIG_NET_DMA
493 dma_cookie_t dma_cookie;
494#endif
James Morris984bc162006-06-09 00:29:17 -0700495#ifdef CONFIG_NETWORK_SECMARK
496 __u32 secmark;
497#endif
Neil Horman3b885782009-10-12 13:26:31 -0700498 union {
499 __u32 mark;
500 __u32 dropcount;
Eric Dumazeta21d4572012-04-10 20:30:48 +0000501 __u32 avail_size;
Neil Horman3b885782009-10-12 13:26:31 -0700502 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000504 sk_buff_data_t inner_transport_header;
505 sk_buff_data_t inner_network_header;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700506 sk_buff_data_t transport_header;
507 sk_buff_data_t network_header;
508 sk_buff_data_t mac_header;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700509 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700510 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700511 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700513 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700514 unsigned int truesize;
515 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516};
517
518#ifdef __KERNEL__
519/*
520 * Handling routines are only of interest to the kernel
521 */
522#include <linux/slab.h>
523
Linus Torvalds1da177e2005-04-16 15:20:36 -0700524
Mel Gormanc93bdd02012-07-31 16:44:19 -0700525#define SKB_ALLOC_FCLONE 0x01
526#define SKB_ALLOC_RX 0x02
527
528/* Returns true if the skb was allocated from PFMEMALLOC reserves */
529static inline bool skb_pfmemalloc(const struct sk_buff *skb)
530{
531 return unlikely(skb->pfmemalloc);
532}
533
Eric Dumazet7fee2262010-05-11 23:19:48 +0000534/*
535 * skb might have a dst pointer attached, refcounted or not.
536 * _skb_refdst low order bit is set if refcount was _not_ taken
537 */
538#define SKB_DST_NOREF 1UL
539#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
540
541/**
542 * skb_dst - returns skb dst_entry
543 * @skb: buffer
544 *
545 * Returns skb dst_entry, regardless of reference taken or not.
546 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000547static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
548{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000549 /* If refdst was not refcounted, check we still are in a
550 * rcu_read_lock section
551 */
552 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
553 !rcu_read_lock_held() &&
554 !rcu_read_lock_bh_held());
555 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000556}
557
Eric Dumazet7fee2262010-05-11 23:19:48 +0000558/**
559 * skb_dst_set - sets skb dst
560 * @skb: buffer
561 * @dst: dst entry
562 *
563 * Sets skb dst, assuming a reference was taken on dst and should
564 * be released by skb_dst_drop()
565 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000566static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
567{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000568 skb->_skb_refdst = (unsigned long)dst;
569}
570
Eric Dumazet27b75c92010-10-15 05:44:11 +0000571extern void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst);
Eric Dumazet7fee2262010-05-11 23:19:48 +0000572
573/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300574 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000575 * @skb: buffer
576 */
577static inline bool skb_dst_is_noref(const struct sk_buff *skb)
578{
579 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000580}
581
Eric Dumazet511c3f92009-06-02 05:14:27 +0000582static inline struct rtable *skb_rtable(const struct sk_buff *skb)
583{
Eric Dumazetadf30902009-06-02 05:19:30 +0000584 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000585}
586
Jörn Engel231d06a2006-03-20 21:28:35 -0800587extern void kfree_skb(struct sk_buff *skb);
Michael S. Tsirkin25121172012-11-01 09:16:28 +0000588extern void skb_tx_error(struct sk_buff *skb);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000589extern void consume_skb(struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590extern void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000591extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000592
593extern void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
594extern bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
595 bool *fragstolen, int *delta_truesize);
596
David S. Millerd179cd12005-08-17 14:57:30 -0700597extern struct sk_buff *__alloc_skb(unsigned int size,
Mel Gormanc93bdd02012-07-31 16:44:19 -0700598 gfp_t priority, int flags, int node);
Eric Dumazetd3836f22012-04-27 00:33:38 +0000599extern struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700600static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100601 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700602{
Eric Dumazet564824b2010-10-11 19:05:25 +0000603 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700604}
605
606static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100607 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700608{
Mel Gormanc93bdd02012-07-31 16:44:19 -0700609 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700610}
611
Herbert Xue0053ec2007-10-14 00:37:52 -0700612extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
Michael S. Tsirkin48c83012011-08-31 08:03:29 +0000613extern int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700614extern struct sk_buff *skb_clone(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100615 gfp_t priority);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700616extern struct sk_buff *skb_copy(const struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100617 gfp_t priority);
Eric Dumazet117632e2011-12-03 21:39:53 +0000618extern struct sk_buff *__pskb_copy(struct sk_buff *skb,
619 int headroom, gfp_t gfp_mask);
620
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621extern int pskb_expand_head(struct sk_buff *skb,
Victor Fusco86a76ca2005-07-08 14:57:47 -0700622 int nhead, int ntail,
Al Virodd0fc662005-10-07 07:46:04 +0100623 gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
625 unsigned int headroom);
626extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
627 int newheadroom, int newtailroom,
Al Virodd0fc662005-10-07 07:46:04 +0100628 gfp_t priority);
David Howells716ea3a2007-04-02 20:19:53 -0700629extern int skb_to_sgvec(struct sk_buff *skb,
630 struct scatterlist *sg, int offset,
631 int len);
632extern int skb_cow_data(struct sk_buff *skb, int tailbits,
633 struct sk_buff **trailer);
Herbert Xu5b057c62006-06-23 02:06:41 -0700634extern int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000635#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700637extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
638 int getfrag(void *from, char *to, int offset,
639 int len,int odd, struct sk_buff *skb),
640 void *from, int length);
641
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800642struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700643 __u32 lower_offset;
644 __u32 upper_offset;
645 __u32 frag_idx;
646 __u32 stepped_offset;
647 struct sk_buff *root_skb;
648 struct sk_buff *cur_skb;
649 __u8 *frag_data;
650};
651
652extern void skb_prepare_seq_read(struct sk_buff *skb,
653 unsigned int from, unsigned int to,
654 struct skb_seq_state *st);
655extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
656 struct skb_seq_state *st);
657extern void skb_abort_seq_read(struct skb_seq_state *st);
658
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700659extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
660 unsigned int to, struct ts_config *config,
661 struct ts_state *state);
662
Tom Herbertbdeab992011-08-14 19:45:55 +0000663extern void __skb_get_rxhash(struct sk_buff *skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000664static inline __u32 skb_get_rxhash(struct sk_buff *skb)
665{
Willem de Bruijnecd5cf52012-10-26 11:52:08 +0000666 if (!skb->l4_rxhash)
Tom Herbertbdeab992011-08-14 19:45:55 +0000667 __skb_get_rxhash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000668
669 return skb->rxhash;
670}
671
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700672#ifdef NET_SKBUFF_DATA_USES_OFFSET
673static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
674{
675 return skb->head + skb->end;
676}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000677
678static inline unsigned int skb_end_offset(const struct sk_buff *skb)
679{
680 return skb->end;
681}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700682#else
683static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
684{
685 return skb->end;
686}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000687
688static inline unsigned int skb_end_offset(const struct sk_buff *skb)
689{
690 return skb->end - skb->head;
691}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700692#endif
693
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700695#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700696
Patrick Ohlyac45f602009-02-12 05:03:37 +0000697static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
698{
699 return &skb_shinfo(skb)->hwtstamps;
700}
701
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702/**
703 * skb_queue_empty - check if a queue is empty
704 * @list: queue head
705 *
706 * Returns true if the queue is empty, false otherwise.
707 */
708static inline int skb_queue_empty(const struct sk_buff_head *list)
709{
710 return list->next == (struct sk_buff *)list;
711}
712
713/**
David S. Millerfc7ebb22008-09-23 00:34:07 -0700714 * skb_queue_is_last - check if skb is the last entry in the queue
715 * @list: queue head
716 * @skb: buffer
717 *
718 * Returns true if @skb is the last buffer on the list.
719 */
720static inline bool skb_queue_is_last(const struct sk_buff_head *list,
721 const struct sk_buff *skb)
722{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000723 return skb->next == (struct sk_buff *)list;
David S. Millerfc7ebb22008-09-23 00:34:07 -0700724}
725
726/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800727 * skb_queue_is_first - check if skb is the first entry in the queue
728 * @list: queue head
729 * @skb: buffer
730 *
731 * Returns true if @skb is the first buffer on the list.
732 */
733static inline bool skb_queue_is_first(const struct sk_buff_head *list,
734 const struct sk_buff *skb)
735{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000736 return skb->prev == (struct sk_buff *)list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800737}
738
739/**
David S. Miller249c8b42008-09-23 00:44:42 -0700740 * skb_queue_next - return the next packet in the queue
741 * @list: queue head
742 * @skb: current buffer
743 *
744 * Return the next packet in @list after @skb. It is only valid to
745 * call this if skb_queue_is_last() evaluates to false.
746 */
747static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
748 const struct sk_buff *skb)
749{
750 /* This BUG_ON may seem severe, but if we just return then we
751 * are going to dereference garbage.
752 */
753 BUG_ON(skb_queue_is_last(list, skb));
754 return skb->next;
755}
756
757/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800758 * skb_queue_prev - return the prev packet in the queue
759 * @list: queue head
760 * @skb: current buffer
761 *
762 * Return the prev packet in @list before @skb. It is only valid to
763 * call this if skb_queue_is_first() evaluates to false.
764 */
765static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
766 const struct sk_buff *skb)
767{
768 /* This BUG_ON may seem severe, but if we just return then we
769 * are going to dereference garbage.
770 */
771 BUG_ON(skb_queue_is_first(list, skb));
772 return skb->prev;
773}
774
775/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700776 * skb_get - reference buffer
777 * @skb: buffer to reference
778 *
779 * Makes another reference to a socket buffer and returns a pointer
780 * to the buffer.
781 */
782static inline struct sk_buff *skb_get(struct sk_buff *skb)
783{
784 atomic_inc(&skb->users);
785 return skb;
786}
787
788/*
789 * If users == 1, we are the only owner and are can avoid redundant
790 * atomic change.
791 */
792
793/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700794 * skb_cloned - is the buffer a clone
795 * @skb: buffer to check
796 *
797 * Returns true if the buffer was generated with skb_clone() and is
798 * one of multiple shared copies of the buffer. Cloned buffers are
799 * shared data so must not be written to under normal circumstances.
800 */
801static inline int skb_cloned(const struct sk_buff *skb)
802{
803 return skb->cloned &&
804 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
805}
806
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +0000807static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
808{
809 might_sleep_if(pri & __GFP_WAIT);
810
811 if (skb_cloned(skb))
812 return pskb_expand_head(skb, 0, 0, pri);
813
814 return 0;
815}
816
Linus Torvalds1da177e2005-04-16 15:20:36 -0700817/**
818 * skb_header_cloned - is the header a clone
819 * @skb: buffer to check
820 *
821 * Returns true if modifying the header part of the buffer requires
822 * the data to be copied.
823 */
824static inline int skb_header_cloned(const struct sk_buff *skb)
825{
826 int dataref;
827
828 if (!skb->cloned)
829 return 0;
830
831 dataref = atomic_read(&skb_shinfo(skb)->dataref);
832 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
833 return dataref != 1;
834}
835
836/**
837 * skb_header_release - release reference to header
838 * @skb: buffer to operate on
839 *
840 * Drop a reference to the header part of the buffer. This is done
841 * by acquiring a payload reference. You must not read from the header
842 * part of skb->data after this.
843 */
844static inline void skb_header_release(struct sk_buff *skb)
845{
846 BUG_ON(skb->nohdr);
847 skb->nohdr = 1;
848 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
849}
850
851/**
852 * skb_shared - is the buffer shared
853 * @skb: buffer to check
854 *
855 * Returns true if more than one person has a reference to this
856 * buffer.
857 */
858static inline int skb_shared(const struct sk_buff *skb)
859{
860 return atomic_read(&skb->users) != 1;
861}
862
863/**
864 * skb_share_check - check if buffer is shared and if so clone it
865 * @skb: buffer to check
866 * @pri: priority for memory allocation
867 *
868 * If the buffer is shared the buffer is cloned and the old copy
869 * drops a reference. A new clone with a single reference is returned.
870 * If the buffer is not shared the original buffer is returned. When
871 * being called from interrupt status or with spinlocks held pri must
872 * be GFP_ATOMIC.
873 *
874 * NULL is returned on a memory allocation failure.
875 */
Eric Dumazet47061bc2012-08-03 20:54:15 +0000876static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877{
878 might_sleep_if(pri & __GFP_WAIT);
879 if (skb_shared(skb)) {
880 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +0000881
882 if (likely(nskb))
883 consume_skb(skb);
884 else
885 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700886 skb = nskb;
887 }
888 return skb;
889}
890
891/*
892 * Copy shared buffers into a new sk_buff. We effectively do COW on
893 * packets to handle cases where we have a local reader and forward
894 * and a couple of other messy ones. The normal one is tcpdumping
895 * a packet thats being forwarded.
896 */
897
898/**
899 * skb_unshare - make a copy of a shared buffer
900 * @skb: buffer to check
901 * @pri: priority for memory allocation
902 *
903 * If the socket buffer is a clone then this function creates a new
904 * copy of the data, drops a reference count on the old copy and returns
905 * the new copy with the reference count at 1. If the buffer is not a clone
906 * the original buffer is returned. When called with a spinlock held or
907 * from interrupt state @pri must be %GFP_ATOMIC
908 *
909 * %NULL is returned on a memory allocation failure.
910 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -0700911static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100912 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700913{
914 might_sleep_if(pri & __GFP_WAIT);
915 if (skb_cloned(skb)) {
916 struct sk_buff *nskb = skb_copy(skb, pri);
917 kfree_skb(skb); /* Free our shared copy */
918 skb = nskb;
919 }
920 return skb;
921}
922
923/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -0800924 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -0700925 * @list_: list to peek at
926 *
927 * Peek an &sk_buff. Unlike most other operations you _MUST_
928 * be careful with this one. A peek leaves the buffer on the
929 * list and someone else may run off with it. You must hold
930 * the appropriate locks or have a private queue to do this.
931 *
932 * Returns %NULL for an empty list or a pointer to the head element.
933 * The reference count is not incremented and the reference is therefore
934 * volatile. Use with caution.
935 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -0400936static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700937{
Eric Dumazet18d07002012-04-30 16:31:46 +0000938 struct sk_buff *skb = list_->next;
939
940 if (skb == (struct sk_buff *)list_)
941 skb = NULL;
942 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700943}
944
945/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +0000946 * skb_peek_next - peek skb following the given one from a queue
947 * @skb: skb to start from
948 * @list_: list to peek at
949 *
950 * Returns %NULL when the end of the list is met or a pointer to the
951 * next element. The reference count is not incremented and the
952 * reference is therefore volatile. Use with caution.
953 */
954static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
955 const struct sk_buff_head *list_)
956{
957 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +0000958
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +0000959 if (next == (struct sk_buff *)list_)
960 next = NULL;
961 return next;
962}
963
964/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -0800965 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -0700966 * @list_: list to peek at
967 *
968 * Peek an &sk_buff. Unlike most other operations you _MUST_
969 * be careful with this one. A peek leaves the buffer on the
970 * list and someone else may run off with it. You must hold
971 * the appropriate locks or have a private queue to do this.
972 *
973 * Returns %NULL for an empty list or a pointer to the tail element.
974 * The reference count is not incremented and the reference is therefore
975 * volatile. Use with caution.
976 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -0400977static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700978{
Eric Dumazet18d07002012-04-30 16:31:46 +0000979 struct sk_buff *skb = list_->prev;
980
981 if (skb == (struct sk_buff *)list_)
982 skb = NULL;
983 return skb;
984
Linus Torvalds1da177e2005-04-16 15:20:36 -0700985}
986
987/**
988 * skb_queue_len - get queue length
989 * @list_: list to measure
990 *
991 * Return the length of an &sk_buff queue.
992 */
993static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
994{
995 return list_->qlen;
996}
997
David S. Miller67fed452008-09-21 22:36:24 -0700998/**
999 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1000 * @list: queue to initialize
1001 *
1002 * This initializes only the list and queue length aspects of
1003 * an sk_buff_head object. This allows to initialize the list
1004 * aspects of an sk_buff_head without reinitializing things like
1005 * the spinlock. It can also be used for on-stack sk_buff_head
1006 * objects where the spinlock is known to not be used.
1007 */
1008static inline void __skb_queue_head_init(struct sk_buff_head *list)
1009{
1010 list->prev = list->next = (struct sk_buff *)list;
1011 list->qlen = 0;
1012}
1013
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001014/*
1015 * This function creates a split out lock class for each invocation;
1016 * this is needed for now since a whole lot of users of the skb-queue
1017 * infrastructure in drivers have different locking usage (in hardirq)
1018 * than the networking core (in softirq only). In the long run either the
1019 * network layer or drivers should need annotation to consolidate the
1020 * main types of usage into 3 classes.
1021 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001022static inline void skb_queue_head_init(struct sk_buff_head *list)
1023{
1024 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001025 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001026}
1027
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001028static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1029 struct lock_class_key *class)
1030{
1031 skb_queue_head_init(list);
1032 lockdep_set_class(&list->lock, class);
1033}
1034
Linus Torvalds1da177e2005-04-16 15:20:36 -07001035/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001036 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037 *
1038 * The "__skb_xxxx()" functions are the non-atomic ones that
1039 * can only be called with interrupts disabled.
1040 */
Gerrit Renkerbf299272008-04-14 00:04:51 -07001041extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
1042static inline void __skb_insert(struct sk_buff *newsk,
1043 struct sk_buff *prev, struct sk_buff *next,
1044 struct sk_buff_head *list)
1045{
1046 newsk->next = next;
1047 newsk->prev = prev;
1048 next->prev = prev->next = newsk;
1049 list->qlen++;
1050}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001051
David S. Miller67fed452008-09-21 22:36:24 -07001052static inline void __skb_queue_splice(const struct sk_buff_head *list,
1053 struct sk_buff *prev,
1054 struct sk_buff *next)
1055{
1056 struct sk_buff *first = list->next;
1057 struct sk_buff *last = list->prev;
1058
1059 first->prev = prev;
1060 prev->next = first;
1061
1062 last->next = next;
1063 next->prev = last;
1064}
1065
1066/**
1067 * skb_queue_splice - join two skb lists, this is designed for stacks
1068 * @list: the new list to add
1069 * @head: the place to add it in the first list
1070 */
1071static inline void skb_queue_splice(const struct sk_buff_head *list,
1072 struct sk_buff_head *head)
1073{
1074 if (!skb_queue_empty(list)) {
1075 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001076 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001077 }
1078}
1079
1080/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001081 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001082 * @list: the new list to add
1083 * @head: the place to add it in the first list
1084 *
1085 * The list at @list is reinitialised
1086 */
1087static inline void skb_queue_splice_init(struct sk_buff_head *list,
1088 struct sk_buff_head *head)
1089{
1090 if (!skb_queue_empty(list)) {
1091 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001092 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001093 __skb_queue_head_init(list);
1094 }
1095}
1096
1097/**
1098 * skb_queue_splice_tail - join two skb lists, each list being a queue
1099 * @list: the new list to add
1100 * @head: the place to add it in the first list
1101 */
1102static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1103 struct sk_buff_head *head)
1104{
1105 if (!skb_queue_empty(list)) {
1106 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001107 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001108 }
1109}
1110
1111/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001112 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001113 * @list: the new list to add
1114 * @head: the place to add it in the first list
1115 *
1116 * Each of the lists is a queue.
1117 * The list at @list is reinitialised
1118 */
1119static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1120 struct sk_buff_head *head)
1121{
1122 if (!skb_queue_empty(list)) {
1123 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001124 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001125 __skb_queue_head_init(list);
1126 }
1127}
1128
Linus Torvalds1da177e2005-04-16 15:20:36 -07001129/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001130 * __skb_queue_after - queue a buffer at the list head
1131 * @list: list to use
1132 * @prev: place after this buffer
1133 * @newsk: buffer to queue
1134 *
1135 * Queue a buffer int the middle 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 */
1140static inline void __skb_queue_after(struct sk_buff_head *list,
1141 struct sk_buff *prev,
1142 struct sk_buff *newsk)
1143{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001144 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001145}
1146
Gerrit Renker7de6c032008-04-14 00:05:09 -07001147extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1148 struct sk_buff_head *list);
1149
Gerrit Renkerf5572852008-04-14 00:05:28 -07001150static inline void __skb_queue_before(struct sk_buff_head *list,
1151 struct sk_buff *next,
1152 struct sk_buff *newsk)
1153{
1154 __skb_insert(newsk, next->prev, next, list);
1155}
1156
Stephen Hemminger300ce172005-10-30 13:47:34 -08001157/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001158 * __skb_queue_head - queue a buffer at the list head
1159 * @list: list to use
1160 * @newsk: buffer to queue
1161 *
1162 * Queue a buffer at the start of a list. This function takes no locks
1163 * and you must therefore hold required locks before calling it.
1164 *
1165 * A buffer cannot be placed on two lists at the same time.
1166 */
1167extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1168static inline void __skb_queue_head(struct sk_buff_head *list,
1169 struct sk_buff *newsk)
1170{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001171 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172}
1173
1174/**
1175 * __skb_queue_tail - queue a buffer at the list tail
1176 * @list: list to use
1177 * @newsk: buffer to queue
1178 *
1179 * Queue a buffer at the end of a list. This function takes no locks
1180 * and you must therefore hold required locks before calling it.
1181 *
1182 * A buffer cannot be placed on two lists at the same time.
1183 */
1184extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1185static inline void __skb_queue_tail(struct sk_buff_head *list,
1186 struct sk_buff *newsk)
1187{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001188 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001189}
1190
Linus Torvalds1da177e2005-04-16 15:20:36 -07001191/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001192 * remove sk_buff from list. _Must_ be called atomically, and with
1193 * the list known..
1194 */
David S. Miller8728b832005-08-09 19:25:21 -07001195extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1197{
1198 struct sk_buff *next, *prev;
1199
1200 list->qlen--;
1201 next = skb->next;
1202 prev = skb->prev;
1203 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001204 next->prev = prev;
1205 prev->next = next;
1206}
1207
Gerrit Renkerf525c062008-04-14 00:04:12 -07001208/**
1209 * __skb_dequeue - remove from the head of the queue
1210 * @list: list to dequeue from
1211 *
1212 * Remove the head of the list. This function does not take any locks
1213 * so must be used with appropriate locks held only. The head item is
1214 * returned or %NULL if the list is empty.
1215 */
1216extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1217static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1218{
1219 struct sk_buff *skb = skb_peek(list);
1220 if (skb)
1221 __skb_unlink(skb, list);
1222 return skb;
1223}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001224
1225/**
1226 * __skb_dequeue_tail - remove from the tail of the queue
1227 * @list: list to dequeue from
1228 *
1229 * Remove the tail of the list. This function does not take any locks
1230 * so must be used with appropriate locks held only. The tail item is
1231 * returned or %NULL if the list is empty.
1232 */
1233extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1234static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1235{
1236 struct sk_buff *skb = skb_peek_tail(list);
1237 if (skb)
1238 __skb_unlink(skb, list);
1239 return skb;
1240}
1241
1242
David S. Millerbdcc0922012-03-07 20:53:36 -05001243static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001244{
1245 return skb->data_len;
1246}
1247
1248static inline unsigned int skb_headlen(const struct sk_buff *skb)
1249{
1250 return skb->len - skb->data_len;
1251}
1252
1253static inline int skb_pagelen(const struct sk_buff *skb)
1254{
1255 int i, len = 0;
1256
1257 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001258 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001259 return len + skb_headlen(skb);
1260}
1261
Ian Campbell131ea662011-08-19 06:25:00 +00001262/**
1263 * __skb_fill_page_desc - initialise a paged fragment in an skb
1264 * @skb: buffer containing fragment to be initialised
1265 * @i: paged fragment index to initialise
1266 * @page: the page to use for this fragment
1267 * @off: the offset to the data with @page
1268 * @size: the length of the data
1269 *
1270 * Initialises the @i'th fragment of @skb to point to &size bytes at
1271 * offset @off within @page.
1272 *
1273 * Does not take any additional reference on the fragment.
1274 */
1275static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1276 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001277{
1278 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1279
Mel Gormanc48a11c2012-07-31 16:44:23 -07001280 /*
1281 * Propagate page->pfmemalloc to the skb if we can. The problem is
1282 * that not all callers have unique ownership of the page. If
1283 * pfmemalloc is set, we check the mapping as a mapping implies
1284 * page->index is set (index and pfmemalloc share space).
1285 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1286 * do not lose pfmemalloc information as the pages would not be
1287 * allocated using __GFP_MEMALLOC.
1288 */
1289 if (page->pfmemalloc && !page->mapping)
1290 skb->pfmemalloc = true;
Ian Campbella8605c62011-10-19 23:01:49 +00001291 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001293 skb_frag_size_set(frag, size);
Ian Campbell131ea662011-08-19 06:25:00 +00001294}
1295
1296/**
1297 * skb_fill_page_desc - initialise a paged fragment in an skb
1298 * @skb: buffer containing fragment to be initialised
1299 * @i: paged fragment index to initialise
1300 * @page: the page to use for this fragment
1301 * @off: the offset to the data with @page
1302 * @size: the length of the data
1303 *
1304 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1305 * @skb to point to &size bytes at offset @off within @page. In
1306 * addition updates @skb such that @i is the last fragment.
1307 *
1308 * Does not take any additional reference on the fragment.
1309 */
1310static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1311 struct page *page, int off, int size)
1312{
1313 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001314 skb_shinfo(skb)->nr_frags = i + 1;
1315}
1316
Peter Zijlstra654bed12008-10-07 14:22:33 -07001317extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
Eric Dumazet50269e12012-03-23 23:59:33 +00001318 int off, int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001319
Linus Torvalds1da177e2005-04-16 15:20:36 -07001320#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001321#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001322#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1323
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001324#ifdef NET_SKBUFF_DATA_USES_OFFSET
1325static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1326{
1327 return skb->head + skb->tail;
1328}
1329
1330static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1331{
1332 skb->tail = skb->data - skb->head;
1333}
1334
1335static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1336{
1337 skb_reset_tail_pointer(skb);
1338 skb->tail += offset;
1339}
1340#else /* NET_SKBUFF_DATA_USES_OFFSET */
1341static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1342{
1343 return skb->tail;
1344}
1345
1346static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1347{
1348 skb->tail = skb->data;
1349}
1350
1351static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1352{
1353 skb->tail = skb->data + offset;
1354}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001355
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001356#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1357
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358/*
1359 * Add data to an sk_buff
1360 */
Ilpo Järvinen0dde3e12008-03-27 17:43:41 -07001361extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1363{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001364 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365 SKB_LINEAR_ASSERT(skb);
1366 skb->tail += len;
1367 skb->len += len;
1368 return tmp;
1369}
1370
Ilpo Järvinenc2aa2702008-03-27 17:52:40 -07001371extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1373{
1374 skb->data -= len;
1375 skb->len += len;
1376 return skb->data;
1377}
1378
Ilpo Järvinen6be8ac22008-03-27 17:47:24 -07001379extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001380static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1381{
1382 skb->len -= len;
1383 BUG_ON(skb->len < skb->data_len);
1384 return skb->data += len;
1385}
1386
David S. Miller47d29642010-05-02 02:21:44 -07001387static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1388{
1389 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1390}
1391
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1393
1394static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1395{
1396 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001397 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398 return NULL;
1399 skb->len -= len;
1400 return skb->data += len;
1401}
1402
1403static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1404{
1405 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1406}
1407
1408static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1409{
1410 if (likely(len <= skb_headlen(skb)))
1411 return 1;
1412 if (unlikely(len > skb->len))
1413 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001414 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001415}
1416
1417/**
1418 * skb_headroom - bytes at buffer head
1419 * @skb: buffer to check
1420 *
1421 * Return the number of bytes of free space at the head of an &sk_buff.
1422 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001423static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001424{
1425 return skb->data - skb->head;
1426}
1427
1428/**
1429 * skb_tailroom - bytes at buffer end
1430 * @skb: buffer to check
1431 *
1432 * Return the number of bytes of free space at the tail of an sk_buff
1433 */
1434static inline int skb_tailroom(const struct sk_buff *skb)
1435{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001436 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437}
1438
1439/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00001440 * skb_availroom - bytes at buffer end
1441 * @skb: buffer to check
1442 *
1443 * Return the number of bytes of free space at the tail of an sk_buff
1444 * allocated by sk_stream_alloc()
1445 */
1446static inline int skb_availroom(const struct sk_buff *skb)
1447{
1448 return skb_is_nonlinear(skb) ? 0 : skb->avail_size - skb->len;
1449}
1450
1451/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001452 * skb_reserve - adjust headroom
1453 * @skb: buffer to alter
1454 * @len: bytes to move
1455 *
1456 * Increase the headroom of an empty &sk_buff by reducing the tail
1457 * room. This is only allowed for an empty buffer.
1458 */
David S. Miller8243126c2006-01-17 02:54:21 -08001459static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001460{
1461 skb->data += len;
1462 skb->tail += len;
1463}
1464
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001465static inline void skb_reset_inner_headers(struct sk_buff *skb)
1466{
1467 skb->inner_network_header = skb->network_header;
1468 skb->inner_transport_header = skb->transport_header;
1469}
1470
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00001471static inline void skb_reset_mac_len(struct sk_buff *skb)
1472{
1473 skb->mac_len = skb->network_header - skb->mac_header;
1474}
1475
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001476#ifdef NET_SKBUFF_DATA_USES_OFFSET
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001477static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1478 *skb)
1479{
1480 return skb->head + skb->inner_transport_header;
1481}
1482
1483static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1484{
1485 skb->inner_transport_header = skb->data - skb->head;
1486}
1487
1488static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1489 const int offset)
1490{
1491 skb_reset_inner_transport_header(skb);
1492 skb->inner_transport_header += offset;
1493}
1494
1495static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1496{
1497 return skb->head + skb->inner_network_header;
1498}
1499
1500static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1501{
1502 skb->inner_network_header = skb->data - skb->head;
1503}
1504
1505static inline void skb_set_inner_network_header(struct sk_buff *skb,
1506 const int offset)
1507{
1508 skb_reset_inner_network_header(skb);
1509 skb->inner_network_header += offset;
1510}
1511
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001512static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1513{
1514 return skb->transport_header != ~0U;
1515}
1516
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001517static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1518{
1519 return skb->head + skb->transport_header;
1520}
1521
1522static inline void skb_reset_transport_header(struct sk_buff *skb)
1523{
1524 skb->transport_header = skb->data - skb->head;
1525}
1526
1527static inline void skb_set_transport_header(struct sk_buff *skb,
1528 const int offset)
1529{
1530 skb_reset_transport_header(skb);
1531 skb->transport_header += offset;
1532}
1533
1534static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1535{
1536 return skb->head + skb->network_header;
1537}
1538
1539static inline void skb_reset_network_header(struct sk_buff *skb)
1540{
1541 skb->network_header = skb->data - skb->head;
1542}
1543
1544static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1545{
1546 skb_reset_network_header(skb);
1547 skb->network_header += offset;
1548}
1549
1550static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1551{
1552 return skb->head + skb->mac_header;
1553}
1554
1555static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1556{
1557 return skb->mac_header != ~0U;
1558}
1559
1560static inline void skb_reset_mac_header(struct sk_buff *skb)
1561{
1562 skb->mac_header = skb->data - skb->head;
1563}
1564
1565static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1566{
1567 skb_reset_mac_header(skb);
1568 skb->mac_header += offset;
1569}
1570
1571#else /* NET_SKBUFF_DATA_USES_OFFSET */
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001572static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1573 *skb)
1574{
1575 return skb->inner_transport_header;
1576}
1577
1578static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1579{
1580 skb->inner_transport_header = skb->data;
1581}
1582
1583static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1584 const int offset)
1585{
1586 skb->inner_transport_header = skb->data + offset;
1587}
1588
1589static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1590{
1591 return skb->inner_network_header;
1592}
1593
1594static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1595{
1596 skb->inner_network_header = skb->data;
1597}
1598
1599static inline void skb_set_inner_network_header(struct sk_buff *skb,
1600 const int offset)
1601{
1602 skb->inner_network_header = skb->data + offset;
1603}
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001604
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001605static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1606{
1607 return skb->transport_header != NULL;
1608}
1609
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001610static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1611{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001612 return skb->transport_header;
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001613}
1614
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001615static inline void skb_reset_transport_header(struct sk_buff *skb)
1616{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001617 skb->transport_header = skb->data;
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001618}
1619
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001620static inline void skb_set_transport_header(struct sk_buff *skb,
1621 const int offset)
1622{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001623 skb->transport_header = skb->data + offset;
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001624}
1625
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001626static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1627{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001628 return skb->network_header;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001629}
1630
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001631static inline void skb_reset_network_header(struct sk_buff *skb)
1632{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001633 skb->network_header = skb->data;
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001634}
1635
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001636static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1637{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001638 skb->network_header = skb->data + offset;
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001639}
1640
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001641static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1642{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001643 return skb->mac_header;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001644}
1645
1646static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1647{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001648 return skb->mac_header != NULL;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001649}
1650
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001651static inline void skb_reset_mac_header(struct sk_buff *skb)
1652{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001653 skb->mac_header = skb->data;
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001654}
1655
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001656static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1657{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001658 skb->mac_header = skb->data + offset;
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001659}
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001660#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1661
Eric Dumazet03606892012-02-23 10:55:02 +00001662static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1663{
1664 if (skb_mac_header_was_set(skb)) {
1665 const unsigned char *old_mac = skb_mac_header(skb);
1666
1667 skb_set_mac_header(skb, -skb->mac_len);
1668 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1669 }
1670}
1671
Michał Mirosław04fb4512010-12-14 15:24:08 +00001672static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1673{
1674 return skb->csum_start - skb_headroom(skb);
1675}
1676
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001677static inline int skb_transport_offset(const struct sk_buff *skb)
1678{
1679 return skb_transport_header(skb) - skb->data;
1680}
1681
1682static inline u32 skb_network_header_len(const struct sk_buff *skb)
1683{
1684 return skb->transport_header - skb->network_header;
1685}
1686
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001687static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1688{
1689 return skb->inner_transport_header - skb->inner_network_header;
1690}
1691
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001692static inline int skb_network_offset(const struct sk_buff *skb)
1693{
1694 return skb_network_header(skb) - skb->data;
1695}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001696
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001697static inline int skb_inner_network_offset(const struct sk_buff *skb)
1698{
1699 return skb_inner_network_header(skb) - skb->data;
1700}
1701
Changli Gaof9599ce2010-08-04 04:43:44 +00001702static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1703{
1704 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1705}
1706
Linus Torvalds1da177e2005-04-16 15:20:36 -07001707/*
1708 * CPUs often take a performance hit when accessing unaligned memory
1709 * locations. The actual performance hit varies, it can be small if the
1710 * hardware handles it or large if we have to take an exception and fix it
1711 * in software.
1712 *
1713 * Since an ethernet header is 14 bytes network drivers often end up with
1714 * the IP header at an unaligned offset. The IP header can be aligned by
1715 * shifting the start of the packet by 2 bytes. Drivers should do this
1716 * with:
1717 *
Tobias Klauser8660c122009-07-13 22:48:16 +00001718 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001719 *
1720 * The downside to this alignment of the IP header is that the DMA is now
1721 * unaligned. On some architectures the cost of an unaligned DMA is high
1722 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00001723 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1725 * to be overridden.
1726 */
1727#ifndef NET_IP_ALIGN
1728#define NET_IP_ALIGN 2
1729#endif
1730
Anton Blanchard025be812006-03-31 02:27:06 -08001731/*
1732 * The networking layer reserves some headroom in skb data (via
1733 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1734 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08001735 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08001736 *
1737 * Unfortunately this headroom changes the DMA alignment of the resulting
1738 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1739 * on some architectures. An architecture can override this value,
1740 * perhaps setting it to a cacheline in size (since that will maintain
1741 * cacheline alignment of the DMA). It must be a power of 2.
1742 *
David S. Millerd6301d32009-02-08 19:24:13 -08001743 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08001744 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07001745 *
1746 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1747 * to reduce average number of cache lines per packet.
1748 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07001749 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08001750 */
1751#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07001752#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08001753#endif
1754
Herbert Xu3cc0e872006-06-09 16:13:38 -07001755extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001756
1757static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1758{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00001759 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07001760 WARN_ON(1);
1761 return;
1762 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001763 skb->len = len;
1764 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001765}
1766
Ilpo Järvinen419ae742008-03-27 17:54:01 -07001767extern void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001768
1769static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1770{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001771 if (skb->data_len)
1772 return ___pskb_trim(skb, len);
1773 __skb_trim(skb, len);
1774 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001775}
1776
1777static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1778{
1779 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1780}
1781
1782/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07001783 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1784 * @skb: buffer to alter
1785 * @len: new length
1786 *
1787 * This is identical to pskb_trim except that the caller knows that
1788 * the skb is not cloned so we should never get an error due to out-
1789 * of-memory.
1790 */
1791static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1792{
1793 int err = pskb_trim(skb, len);
1794 BUG_ON(err);
1795}
1796
1797/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001798 * skb_orphan - orphan a buffer
1799 * @skb: buffer to orphan
1800 *
1801 * If a buffer currently has an owner then we call the owner's
1802 * destructor function and make the @skb unowned. The buffer continues
1803 * to exist but is no longer charged to its former owner.
1804 */
1805static inline void skb_orphan(struct sk_buff *skb)
1806{
1807 if (skb->destructor)
1808 skb->destructor(skb);
1809 skb->destructor = NULL;
1810 skb->sk = NULL;
1811}
1812
1813/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00001814 * skb_orphan_frags - orphan the frags contained in a buffer
1815 * @skb: buffer to orphan frags from
1816 * @gfp_mask: allocation mask for replacement pages
1817 *
1818 * For each frag in the SKB which needs a destructor (i.e. has an
1819 * owner) create a copy of that frag and release the original
1820 * page by calling the destructor.
1821 */
1822static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1823{
1824 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1825 return 0;
1826 return skb_copy_ubufs(skb, gfp_mask);
1827}
1828
1829/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001830 * __skb_queue_purge - empty a list
1831 * @list: list to empty
1832 *
1833 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1834 * the list and one reference dropped. This function does not take the
1835 * list lock and the caller must hold the relevant locks to use it.
1836 */
1837extern void skb_queue_purge(struct sk_buff_head *list);
1838static inline void __skb_queue_purge(struct sk_buff_head *list)
1839{
1840 struct sk_buff *skb;
1841 while ((skb = __skb_dequeue(list)) != NULL)
1842 kfree_skb(skb);
1843}
1844
Alexander Duycke5e67302013-02-08 10:17:15 +00001845#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
1846#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
1847#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
1848
Eric Dumazet6f532612012-05-18 05:12:12 +00001849extern void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001850
Christoph Hellwig8af27452006-07-31 22:35:23 -07001851extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00001852 unsigned int length,
1853 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07001854
1855/**
1856 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1857 * @dev: network device to receive on
1858 * @length: length to allocate
1859 *
1860 * Allocate a new &sk_buff and assign it a usage count of one. The
1861 * buffer has unspecified headroom built in. Users should allocate
1862 * the headroom they think they need without accounting for the
1863 * built in space. The built in space is used for optimisations.
1864 *
1865 * %NULL is returned if there is no free memory. Although this function
1866 * allocates memory it can be called from an interrupt.
1867 */
1868static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00001869 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07001870{
1871 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1872}
1873
Eric Dumazet6f532612012-05-18 05:12:12 +00001874/* legacy helper around __netdev_alloc_skb() */
1875static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1876 gfp_t gfp_mask)
1877{
1878 return __netdev_alloc_skb(NULL, length, gfp_mask);
1879}
1880
1881/* legacy helper around netdev_alloc_skb() */
1882static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1883{
1884 return netdev_alloc_skb(NULL, length);
1885}
1886
1887
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001888static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
1889 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00001890{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001891 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00001892
1893 if (NET_IP_ALIGN && skb)
1894 skb_reserve(skb, NET_IP_ALIGN);
1895 return skb;
1896}
1897
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001898static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1899 unsigned int length)
1900{
1901 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
1902}
1903
Mel Gorman06140022012-07-31 16:44:24 -07001904/*
1905 * __skb_alloc_page - allocate pages for ps-rx on a skb and preserve pfmemalloc data
1906 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1907 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1908 * @order: size of the allocation
1909 *
1910 * Allocate a new page.
1911 *
1912 * %NULL is returned if there is no free memory.
1913*/
1914static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
1915 struct sk_buff *skb,
1916 unsigned int order)
1917{
1918 struct page *page;
1919
1920 gfp_mask |= __GFP_COLD;
1921
1922 if (!(gfp_mask & __GFP_NOMEMALLOC))
1923 gfp_mask |= __GFP_MEMALLOC;
1924
1925 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
1926 if (skb && page && page->pfmemalloc)
1927 skb->pfmemalloc = true;
1928
1929 return page;
1930}
1931
1932/**
1933 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
1934 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1935 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1936 *
1937 * Allocate a new page.
1938 *
1939 * %NULL is returned if there is no free memory.
1940 */
1941static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
1942 struct sk_buff *skb)
1943{
1944 return __skb_alloc_pages(gfp_mask, skb, 0);
1945}
1946
1947/**
1948 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
1949 * @page: The page that was allocated from skb_alloc_page
1950 * @skb: The skb that may need pfmemalloc set
1951 */
1952static inline void skb_propagate_pfmemalloc(struct page *page,
1953 struct sk_buff *skb)
1954{
1955 if (page && page->pfmemalloc)
1956 skb->pfmemalloc = true;
1957}
1958
Eric Dumazet564824b2010-10-11 19:05:25 +00001959/**
Ian Campbell131ea662011-08-19 06:25:00 +00001960 * skb_frag_page - retrieve the page refered to by a paged fragment
1961 * @frag: the paged fragment
1962 *
1963 * Returns the &struct page associated with @frag.
1964 */
1965static inline struct page *skb_frag_page(const skb_frag_t *frag)
1966{
Ian Campbella8605c62011-10-19 23:01:49 +00001967 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00001968}
1969
1970/**
1971 * __skb_frag_ref - take an addition reference on a paged fragment.
1972 * @frag: the paged fragment
1973 *
1974 * Takes an additional reference on the paged fragment @frag.
1975 */
1976static inline void __skb_frag_ref(skb_frag_t *frag)
1977{
1978 get_page(skb_frag_page(frag));
1979}
1980
1981/**
1982 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
1983 * @skb: the buffer
1984 * @f: the fragment offset.
1985 *
1986 * Takes an additional reference on the @f'th paged fragment of @skb.
1987 */
1988static inline void skb_frag_ref(struct sk_buff *skb, int f)
1989{
1990 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
1991}
1992
1993/**
1994 * __skb_frag_unref - release a reference on a paged fragment.
1995 * @frag: the paged fragment
1996 *
1997 * Releases a reference on the paged fragment @frag.
1998 */
1999static inline void __skb_frag_unref(skb_frag_t *frag)
2000{
2001 put_page(skb_frag_page(frag));
2002}
2003
2004/**
2005 * skb_frag_unref - release a reference on a paged fragment of an skb.
2006 * @skb: the buffer
2007 * @f: the fragment offset
2008 *
2009 * Releases a reference on the @f'th paged fragment of @skb.
2010 */
2011static inline void skb_frag_unref(struct sk_buff *skb, int f)
2012{
2013 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2014}
2015
2016/**
2017 * skb_frag_address - gets the address of the data contained in a paged fragment
2018 * @frag: the paged fragment buffer
2019 *
2020 * Returns the address of the data within @frag. The page must already
2021 * be mapped.
2022 */
2023static inline void *skb_frag_address(const skb_frag_t *frag)
2024{
2025 return page_address(skb_frag_page(frag)) + frag->page_offset;
2026}
2027
2028/**
2029 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2030 * @frag: the paged fragment buffer
2031 *
2032 * Returns the address of the data within @frag. Checks that the page
2033 * is mapped and returns %NULL otherwise.
2034 */
2035static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2036{
2037 void *ptr = page_address(skb_frag_page(frag));
2038 if (unlikely(!ptr))
2039 return NULL;
2040
2041 return ptr + frag->page_offset;
2042}
2043
2044/**
2045 * __skb_frag_set_page - sets the page contained in a paged fragment
2046 * @frag: the paged fragment
2047 * @page: the page to set
2048 *
2049 * Sets the fragment @frag to contain @page.
2050 */
2051static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2052{
Ian Campbella8605c62011-10-19 23:01:49 +00002053 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002054}
2055
2056/**
2057 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2058 * @skb: the buffer
2059 * @f: the fragment offset
2060 * @page: the page to set
2061 *
2062 * Sets the @f'th fragment of @skb to contain @page.
2063 */
2064static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2065 struct page *page)
2066{
2067 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2068}
2069
2070/**
2071 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002072 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002073 * @frag: the paged fragment to map
2074 * @offset: the offset within the fragment (starting at the
2075 * fragment's own offset)
2076 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002077 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00002078 *
2079 * Maps the page associated with @frag to @device.
2080 */
2081static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2082 const skb_frag_t *frag,
2083 size_t offset, size_t size,
2084 enum dma_data_direction dir)
2085{
2086 return dma_map_page(dev, skb_frag_page(frag),
2087 frag->page_offset + offset, size, dir);
2088}
2089
Eric Dumazet117632e2011-12-03 21:39:53 +00002090static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2091 gfp_t gfp_mask)
2092{
2093 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2094}
2095
Ian Campbell131ea662011-08-19 06:25:00 +00002096/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002097 * skb_clone_writable - is the header of a clone writable
2098 * @skb: buffer to check
2099 * @len: length up to which to write
2100 *
2101 * Returns true if modifying the header part of the cloned buffer
2102 * does not requires the data to be copied.
2103 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002104static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002105{
2106 return !skb_header_cloned(skb) &&
2107 skb_headroom(skb) + len <= skb->hdr_len;
2108}
2109
Herbert Xud9cc2042007-09-16 16:21:16 -07002110static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2111 int cloned)
2112{
2113 int delta = 0;
2114
Herbert Xud9cc2042007-09-16 16:21:16 -07002115 if (headroom > skb_headroom(skb))
2116 delta = headroom - skb_headroom(skb);
2117
2118 if (delta || cloned)
2119 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2120 GFP_ATOMIC);
2121 return 0;
2122}
2123
Patrick McHardy334a8132007-06-25 04:35:20 -07002124/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002125 * skb_cow - copy header of skb when it is required
2126 * @skb: buffer to cow
2127 * @headroom: needed headroom
2128 *
2129 * If the skb passed lacks sufficient headroom or its data part
2130 * is shared, data is reallocated. If reallocation fails, an error
2131 * is returned and original skb is not changed.
2132 *
2133 * The result is skb with writable area skb->head...skb->tail
2134 * and at least @headroom of space at head.
2135 */
2136static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2137{
Herbert Xud9cc2042007-09-16 16:21:16 -07002138 return __skb_cow(skb, headroom, skb_cloned(skb));
2139}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002140
Herbert Xud9cc2042007-09-16 16:21:16 -07002141/**
2142 * skb_cow_head - skb_cow but only making the head writable
2143 * @skb: buffer to cow
2144 * @headroom: needed headroom
2145 *
2146 * This function is identical to skb_cow except that we replace the
2147 * skb_cloned check by skb_header_cloned. It should be used when
2148 * you only need to push on some header and do not need to modify
2149 * the data.
2150 */
2151static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2152{
2153 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002154}
2155
2156/**
2157 * skb_padto - pad an skbuff up to a minimal size
2158 * @skb: buffer to pad
2159 * @len: minimal length
2160 *
2161 * Pads up a buffer to ensure the trailing bytes exist and are
2162 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002163 * is untouched. Otherwise it is extended. Returns zero on
2164 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002165 */
2166
Herbert Xu5b057c62006-06-23 02:06:41 -07002167static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002168{
2169 unsigned int size = skb->len;
2170 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002171 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002172 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002173}
2174
2175static inline int skb_add_data(struct sk_buff *skb,
2176 char __user *from, int copy)
2177{
2178 const int off = skb->len;
2179
2180 if (skb->ip_summed == CHECKSUM_NONE) {
2181 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08002182 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
Linus Torvalds1da177e2005-04-16 15:20:36 -07002183 copy, 0, &err);
2184 if (!err) {
2185 skb->csum = csum_block_add(skb->csum, csum, off);
2186 return 0;
2187 }
2188 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2189 return 0;
2190
2191 __skb_trim(skb, off);
2192 return -EFAULT;
2193}
2194
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002195static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2196 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002197{
2198 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002199 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002200
Ian Campbellea2ab692011-08-22 23:44:58 +00002201 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002202 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002203 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002204 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002205}
2206
Herbert Xu364c6ba2006-06-09 16:10:40 -07002207static inline int __skb_linearize(struct sk_buff *skb)
2208{
2209 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2210}
2211
Linus Torvalds1da177e2005-04-16 15:20:36 -07002212/**
2213 * skb_linearize - convert paged skb to linear one
2214 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002215 *
2216 * If there is no free memory -ENOMEM is returned, otherwise zero
2217 * is returned and the old skb data released.
2218 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002219static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002220{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002221 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2222}
2223
2224/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00002225 * skb_has_shared_frag - can any frag be overwritten
2226 * @skb: buffer to test
2227 *
2228 * Return true if the skb has at least one frag that might be modified
2229 * by an external entity (as in vmsplice()/sendfile())
2230 */
2231static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2232{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00002233 return skb_is_nonlinear(skb) &&
2234 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00002235}
2236
2237/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07002238 * skb_linearize_cow - make sure skb is linear and writable
2239 * @skb: buffer to process
2240 *
2241 * If there is no free memory -ENOMEM is returned, otherwise zero
2242 * is returned and the old skb data released.
2243 */
2244static inline int skb_linearize_cow(struct sk_buff *skb)
2245{
2246 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2247 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002248}
2249
2250/**
2251 * skb_postpull_rcsum - update checksum for received skb after pull
2252 * @skb: buffer to update
2253 * @start: start of data before pull
2254 * @len: length of data pulled
2255 *
2256 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002257 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2258 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002259 */
2260
2261static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002262 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002263{
Patrick McHardy84fa7932006-08-29 16:44:56 -07002264 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002265 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2266}
2267
Herbert Xucbb042f2006-03-20 22:43:56 -08002268unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2269
Linus Torvalds1da177e2005-04-16 15:20:36 -07002270/**
2271 * pskb_trim_rcsum - trim received skb and update checksum
2272 * @skb: buffer to trim
2273 * @len: new length
2274 *
2275 * This is exactly the same as pskb_trim except that it ensures the
2276 * checksum of received packets are still valid after the operation.
2277 */
2278
2279static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2280{
Stephen Hemminger0e4e4222005-09-08 12:32:03 -07002281 if (likely(len >= skb->len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002282 return 0;
Patrick McHardy84fa7932006-08-29 16:44:56 -07002283 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002284 skb->ip_summed = CHECKSUM_NONE;
2285 return __pskb_trim(skb, len);
2286}
2287
Linus Torvalds1da177e2005-04-16 15:20:36 -07002288#define skb_queue_walk(queue, skb) \
2289 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002290 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002291 skb = skb->next)
2292
James Chapman46f89142007-04-30 00:07:31 -07002293#define skb_queue_walk_safe(queue, skb, tmp) \
2294 for (skb = (queue)->next, tmp = skb->next; \
2295 skb != (struct sk_buff *)(queue); \
2296 skb = tmp, tmp = skb->next)
2297
David S. Miller1164f522008-09-23 00:49:44 -07002298#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002299 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002300 skb = skb->next)
2301
2302#define skb_queue_walk_from_safe(queue, skb, tmp) \
2303 for (tmp = skb->next; \
2304 skb != (struct sk_buff *)(queue); \
2305 skb = tmp, tmp = skb->next)
2306
Stephen Hemminger300ce172005-10-30 13:47:34 -08002307#define skb_queue_reverse_walk(queue, skb) \
2308 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002309 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08002310 skb = skb->prev)
2311
David S. Miller686a2952011-01-20 22:47:32 -08002312#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2313 for (skb = (queue)->prev, tmp = skb->prev; \
2314 skb != (struct sk_buff *)(queue); \
2315 skb = tmp, tmp = skb->prev)
2316
2317#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2318 for (tmp = skb->prev; \
2319 skb != (struct sk_buff *)(queue); \
2320 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002321
David S. Miller21dc3302010-08-23 00:13:46 -07002322static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07002323{
2324 return skb_shinfo(skb)->frag_list != NULL;
2325}
2326
2327static inline void skb_frag_list_init(struct sk_buff *skb)
2328{
2329 skb_shinfo(skb)->frag_list = NULL;
2330}
2331
2332static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2333{
2334 frag->next = skb_shinfo(skb)->frag_list;
2335 skb_shinfo(skb)->frag_list = frag;
2336}
2337
2338#define skb_walk_frags(skb, iter) \
2339 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2340
Herbert Xua59322b2007-12-05 01:53:40 -08002341extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
Pavel Emelyanov3f518bf2012-02-21 07:30:58 +00002342 int *peeked, int *off, int *err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002343extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
2344 int noblock, int *err);
2345extern unsigned int datagram_poll(struct file *file, struct socket *sock,
2346 struct poll_table_struct *wait);
2347extern int skb_copy_datagram_iovec(const struct sk_buff *from,
2348 int offset, struct iovec *to,
2349 int size);
Herbert Xufb286bb2005-11-10 13:01:24 -08002350extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002351 int hlen,
2352 struct iovec *iov);
Rusty Russelldb543c12008-08-15 15:13:53 -07002353extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
2354 int offset,
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +00002355 const struct iovec *from,
2356 int from_offset,
Rusty Russelldb543c12008-08-15 15:13:53 -07002357 int len);
Michael S. Tsirkin0a1ec072009-04-20 01:25:46 +00002358extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
2359 int offset,
2360 const struct iovec *to,
2361 int to_offset,
2362 int size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002363extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
Eric Dumazet9d410c72009-10-30 05:03:53 +00002364extern void skb_free_datagram_locked(struct sock *sk,
2365 struct sk_buff *skb);
Herbert Xu27ab2562007-12-05 01:51:58 -08002366extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
Herbert Xu3305b802005-12-13 23:16:37 -08002367 unsigned int flags);
Al Viro2bbbc862006-11-14 21:37:14 -08002368extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
2369 int len, __wsum csum);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002370extern int skb_copy_bits(const struct sk_buff *skb, int offset,
2371 void *to, int len);
Stephen Hemminger0c6fcc82007-04-20 16:40:01 -07002372extern int skb_store_bits(struct sk_buff *skb, int offset,
2373 const void *from, int len);
Al Viro81d77662006-11-14 21:37:33 -08002374extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002375 int offset, u8 *to, int len,
Al Viro81d77662006-11-14 21:37:33 -08002376 __wsum csum);
Jens Axboe9c55e012007-11-06 23:30:13 -08002377extern int skb_splice_bits(struct sk_buff *skb,
2378 unsigned int offset,
2379 struct pipe_inode_info *pipe,
2380 unsigned int len,
2381 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002382extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
2383extern void skb_split(struct sk_buff *skb,
2384 struct sk_buff *skb1, const u32 len);
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08002385extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
2386 int shiftlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002387
Michał Mirosławc8f44af2011-11-15 15:29:55 +00002388extern struct sk_buff *skb_segment(struct sk_buff *skb,
2389 netdev_features_t features);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002390
Linus Torvalds1da177e2005-04-16 15:20:36 -07002391static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2392 int len, void *buffer)
2393{
2394 int hlen = skb_headlen(skb);
2395
Patrick McHardy55820ee2005-07-05 14:08:10 -07002396 if (hlen - offset >= len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002397 return skb->data + offset;
2398
2399 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2400 return NULL;
2401
2402 return buffer;
2403}
2404
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03002405static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2406 void *to,
2407 const unsigned int len)
2408{
2409 memcpy(to, skb->data, len);
2410}
2411
2412static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2413 const int offset, void *to,
2414 const unsigned int len)
2415{
2416 memcpy(to, skb->data + offset, len);
2417}
2418
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03002419static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2420 const void *from,
2421 const unsigned int len)
2422{
2423 memcpy(skb->data, from, len);
2424}
2425
2426static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2427 const int offset,
2428 const void *from,
2429 const unsigned int len)
2430{
2431 memcpy(skb->data + offset, from, len);
2432}
2433
Linus Torvalds1da177e2005-04-16 15:20:36 -07002434extern void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002435
Patrick Ohlyac45f602009-02-12 05:03:37 +00002436static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2437{
2438 return skb->tstamp;
2439}
2440
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002441/**
2442 * skb_get_timestamp - get timestamp from a skb
2443 * @skb: skb to get stamp from
2444 * @stamp: pointer to struct timeval to store stamp in
2445 *
2446 * Timestamps are stored in the skb as offsets to a base timestamp.
2447 * This function converts the offset back to a struct timeval and stores
2448 * it in stamp.
2449 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00002450static inline void skb_get_timestamp(const struct sk_buff *skb,
2451 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002452{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002453 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002454}
2455
Patrick Ohlyac45f602009-02-12 05:03:37 +00002456static inline void skb_get_timestampns(const struct sk_buff *skb,
2457 struct timespec *stamp)
2458{
2459 *stamp = ktime_to_timespec(skb->tstamp);
2460}
2461
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002462static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002463{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002464 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002465}
2466
Stephen Hemminger164891a2007-04-23 22:26:16 -07002467static inline ktime_t net_timedelta(ktime_t t)
2468{
2469 return ktime_sub(ktime_get_real(), t);
2470}
2471
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07002472static inline ktime_t net_invalid_timestamp(void)
2473{
2474 return ktime_set(0, 0);
2475}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002476
Richard Cochranc1f19b52010-07-17 08:49:36 +00002477extern void skb_timestamping_init(void);
2478
2479#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2480
2481extern void skb_clone_tx_timestamp(struct sk_buff *skb);
2482extern bool skb_defer_rx_timestamp(struct sk_buff *skb);
2483
2484#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2485
2486static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2487{
2488}
2489
2490static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2491{
2492 return false;
2493}
2494
2495#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2496
2497/**
2498 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2499 *
Richard Cochranda92b192011-10-21 00:49:15 +00002500 * PHY drivers may accept clones of transmitted packets for
2501 * timestamping via their phy_driver.txtstamp method. These drivers
2502 * must call this function to return the skb back to the stack, with
2503 * or without a timestamp.
2504 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00002505 * @skb: clone of the the original outgoing packet
Richard Cochranda92b192011-10-21 00:49:15 +00002506 * @hwtstamps: hardware time stamps, may be NULL if not available
Richard Cochranc1f19b52010-07-17 08:49:36 +00002507 *
2508 */
2509void skb_complete_tx_timestamp(struct sk_buff *skb,
2510 struct skb_shared_hwtstamps *hwtstamps);
2511
Patrick Ohlyac45f602009-02-12 05:03:37 +00002512/**
2513 * skb_tstamp_tx - queue clone of skb with send time stamps
2514 * @orig_skb: the original outgoing packet
2515 * @hwtstamps: hardware time stamps, may be NULL if not available
2516 *
2517 * If the skb has a socket associated, then this function clones the
2518 * skb (thus sharing the actual data and optional structures), stores
2519 * the optional hardware time stamping information (if non NULL) or
2520 * generates a software time stamp (otherwise), then queues the clone
2521 * to the error queue of the socket. Errors are silently ignored.
2522 */
2523extern void skb_tstamp_tx(struct sk_buff *orig_skb,
2524 struct skb_shared_hwtstamps *hwtstamps);
2525
Richard Cochran4507a712010-07-17 08:48:28 +00002526static inline void sw_tx_timestamp(struct sk_buff *skb)
2527{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002528 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2529 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00002530 skb_tstamp_tx(skb, NULL);
2531}
2532
2533/**
2534 * skb_tx_timestamp() - Driver hook for transmit timestamping
2535 *
2536 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00002537 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00002538 *
2539 * @skb: A socket buffer.
2540 */
2541static inline void skb_tx_timestamp(struct sk_buff *skb)
2542{
Richard Cochranc1f19b52010-07-17 08:49:36 +00002543 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002544 sw_tx_timestamp(skb);
2545}
2546
Johannes Berg6e3e9392011-11-09 10:15:42 +01002547/**
2548 * skb_complete_wifi_ack - deliver skb with wifi status
2549 *
2550 * @skb: the original outgoing packet
2551 * @acked: ack status
2552 *
2553 */
2554void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2555
Herbert Xu759e5d02007-03-25 20:10:56 -07002556extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
Al Virob51655b2006-11-14 21:40:42 -08002557extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002558
Herbert Xu60476372007-04-09 11:59:39 -07002559static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2560{
2561 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2562}
2563
Herbert Xufb286bb2005-11-10 13:01:24 -08002564/**
2565 * skb_checksum_complete - Calculate checksum of an entire packet
2566 * @skb: packet to process
2567 *
2568 * This function calculates the checksum over the entire packet plus
2569 * the value of skb->csum. The latter can be used to supply the
2570 * checksum of a pseudo header as used by TCP/UDP. It returns the
2571 * checksum.
2572 *
2573 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2574 * this function can be used to verify that checksum on received
2575 * packets. In that case the function should return zero if the
2576 * checksum is correct. In particular, this function will return zero
2577 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2578 * hardware has already verified the correctness of the checksum.
2579 */
Al Viro4381ca32007-07-15 21:00:11 +01002580static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002581{
Herbert Xu60476372007-04-09 11:59:39 -07002582 return skb_csum_unnecessary(skb) ?
2583 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002584}
2585
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002586#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002587extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002588static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2589{
2590 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002591 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002592}
2593static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2594{
2595 if (nfct)
2596 atomic_inc(&nfct->use);
2597}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002598#endif
2599#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -08002600static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
2601{
2602 if (skb)
2603 atomic_inc(&skb->users);
2604}
2605static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
2606{
2607 if (skb)
2608 kfree_skb(skb);
2609}
2610#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002611#ifdef CONFIG_BRIDGE_NETFILTER
2612static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2613{
2614 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2615 kfree(nf_bridge);
2616}
2617static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2618{
2619 if (nf_bridge)
2620 atomic_inc(&nf_bridge->use);
2621}
2622#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08002623static inline void nf_reset(struct sk_buff *skb)
2624{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002625#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08002626 nf_conntrack_put(skb->nfct);
2627 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002628#endif
2629#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Patrick McHardya193a4a2006-03-20 19:23:05 -08002630 nf_conntrack_put_reasm(skb->nfct_reasm);
2631 skb->nfct_reasm = NULL;
2632#endif
2633#ifdef CONFIG_BRIDGE_NETFILTER
2634 nf_bridge_put(skb->nf_bridge);
2635 skb->nf_bridge = NULL;
2636#endif
2637}
2638
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002639/* Note: This doesn't put any conntrack and bridge info in dst. */
2640static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2641{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002642#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002643 dst->nfct = src->nfct;
2644 nf_conntrack_get(src->nfct);
2645 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002646#endif
2647#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002648 dst->nfct_reasm = src->nfct_reasm;
2649 nf_conntrack_get_reasm(src->nfct_reasm);
2650#endif
2651#ifdef CONFIG_BRIDGE_NETFILTER
2652 dst->nf_bridge = src->nf_bridge;
2653 nf_bridge_get(src->nf_bridge);
2654#endif
2655}
2656
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002657static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2658{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002659#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002660 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002661#endif
2662#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002663 nf_conntrack_put_reasm(dst->nfct_reasm);
2664#endif
2665#ifdef CONFIG_BRIDGE_NETFILTER
2666 nf_bridge_put(dst->nf_bridge);
2667#endif
2668 __nf_copy(dst, src);
2669}
2670
James Morris984bc162006-06-09 00:29:17 -07002671#ifdef CONFIG_NETWORK_SECMARK
2672static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2673{
2674 to->secmark = from->secmark;
2675}
2676
2677static inline void skb_init_secmark(struct sk_buff *skb)
2678{
2679 skb->secmark = 0;
2680}
2681#else
2682static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2683{ }
2684
2685static inline void skb_init_secmark(struct sk_buff *skb)
2686{ }
2687#endif
2688
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002689static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2690{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002691 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002692}
2693
Stephen Hemminger92477442009-03-21 13:39:26 -07002694static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002695{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002696 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002697}
2698
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002699static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2700{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002701 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002702}
2703
David S. Millerd5a9e242009-01-27 16:22:11 -08002704static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2705{
2706 skb->queue_mapping = rx_queue + 1;
2707}
2708
Stephen Hemminger92477442009-03-21 13:39:26 -07002709static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002710{
2711 return skb->queue_mapping - 1;
2712}
2713
Stephen Hemminger92477442009-03-21 13:39:26 -07002714static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002715{
Eric Dumazeta02cec22010-09-22 20:43:57 +00002716 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08002717}
2718
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00002719extern u16 __skb_tx_hash(const struct net_device *dev,
2720 const struct sk_buff *skb,
2721 unsigned int num_tx_queues);
Stephen Hemminger92477442009-03-21 13:39:26 -07002722
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002723#ifdef CONFIG_XFRM
2724static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2725{
2726 return skb->sp;
2727}
2728#else
2729static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2730{
2731 return NULL;
2732}
2733#endif
2734
David S. Millerbdcc0922012-03-07 20:53:36 -05002735static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07002736{
2737 return skb_shinfo(skb)->gso_size;
2738}
2739
David S. Millerbdcc0922012-03-07 20:53:36 -05002740static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02002741{
2742 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2743}
2744
Ben Hutchings4497b072008-06-19 16:22:28 -07002745extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2746
2747static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2748{
2749 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2750 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002751 const struct skb_shared_info *shinfo = skb_shinfo(skb);
2752
Alexander Duyckb78462e2010-06-02 12:24:37 +00002753 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2754 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07002755 __skb_warn_lro_forwarding(skb);
2756 return true;
2757 }
2758 return false;
2759}
2760
Herbert Xu35fc92a2007-03-26 23:22:20 -07002761static inline void skb_forward_csum(struct sk_buff *skb)
2762{
2763 /* Unfortunately we don't support this one. Any brave souls? */
2764 if (skb->ip_summed == CHECKSUM_COMPLETE)
2765 skb->ip_summed = CHECKSUM_NONE;
2766}
2767
Eric Dumazetbc8acf22010-09-02 13:07:41 -07002768/**
2769 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2770 * @skb: skb to check
2771 *
2772 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2773 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2774 * use this helper, to document places where we make this assertion.
2775 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002776static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07002777{
2778#ifdef DEBUG
2779 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2780#endif
2781}
2782
Rusty Russellf35d9d82008-02-04 23:49:54 -05002783bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00002784
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00002785/**
2786 * skb_head_is_locked - Determine if the skb->head is locked down
2787 * @skb: skb to check
2788 *
2789 * The head on skbs build around a head frag can be removed if they are
2790 * not cloned. This function returns true if the skb head is locked down
2791 * due to either being allocated via kmalloc, or by being a clone with
2792 * multiple references to the head.
2793 */
2794static inline bool skb_head_is_locked(const struct sk_buff *skb)
2795{
2796 return !skb->head_frag || skb_cloned(skb);
2797}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002798#endif /* __KERNEL__ */
2799#endif /* _LINUX_SKBUFF_H */