blob: 036ec7d8a83a9693044d4024b8676ab1a6296e26 [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>
Jason Wang5203cd22013-03-26 23:11:21 +000035#include <net/flow_keys.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070036
Herbert Xu60476372007-04-09 11:59:39 -070037/* Don't change this without changing skb_csum_unnecessary! */
Linus Torvalds1da177e2005-04-16 15:20:36 -070038#define CHECKSUM_NONE 0
Herbert Xu60476372007-04-09 11:59:39 -070039#define CHECKSUM_UNNECESSARY 1
40#define CHECKSUM_COMPLETE 2
41#define CHECKSUM_PARTIAL 3
Linus Torvalds1da177e2005-04-16 15:20:36 -070042
43#define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
44 ~(SMP_CACHE_BYTES - 1))
David S. Millerfc910a22007-03-25 20:27:59 -070045#define SKB_WITH_OVERHEAD(X) \
Herbert Xudeea84b2007-10-21 16:27:46 -070046 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
David S. Millerfc910a22007-03-25 20:27:59 -070047#define SKB_MAX_ORDER(X, ORDER) \
48 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
Linus Torvalds1da177e2005-04-16 15:20:36 -070049#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
50#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
51
Eric Dumazet87fb4b72011-10-13 07:28:54 +000052/* return minimum truesize of one skb containing X bytes of data */
53#define SKB_TRUESIZE(X) ((X) + \
54 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
55 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
56
Linus Torvalds1da177e2005-04-16 15:20:36 -070057/* A. Checksumming of received packets by device.
58 *
59 * NONE: device failed to checksum this packet.
60 * skb->csum is undefined.
61 *
62 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
63 * skb->csum is undefined.
64 * It is bad option, but, unfortunately, many of vendors do this.
65 * Apparently with secret goal to sell you new device, when you
66 * will add new protocol to your host. F.e. IPv6. 8)
67 *
Patrick McHardy84fa7932006-08-29 16:44:56 -070068 * COMPLETE: the most generic way. Device supplied checksum of _all_
Linus Torvalds1da177e2005-04-16 15:20:36 -070069 * the packet as seen by netif_rx in skb->csum.
70 * NOTE: Even if device supports only some protocols, but
Patrick McHardy84fa7932006-08-29 16:44:56 -070071 * is able to produce some skb->csum, it MUST use COMPLETE,
Linus Torvalds1da177e2005-04-16 15:20:36 -070072 * not UNNECESSARY.
73 *
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070074 * PARTIAL: identical to the case for output below. This may occur
75 * on a packet received directly from another Linux OS, e.g.,
76 * a virtualised Linux kernel on the same host. The packet can
77 * be treated in the same way as UNNECESSARY except that on
78 * output (i.e., forwarding) the checksum must be filled in
79 * by the OS or the hardware.
80 *
Linus Torvalds1da177e2005-04-16 15:20:36 -070081 * B. Checksumming on output.
82 *
83 * NONE: skb is checksummed by protocol or csum is not required.
84 *
Patrick McHardy84fa7932006-08-29 16:44:56 -070085 * PARTIAL: device is required to csum packet as seen by hard_start_xmit
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070086 * from skb->csum_start to the end and to record the checksum
87 * at skb->csum_start + skb->csum_offset.
Linus Torvalds1da177e2005-04-16 15:20:36 -070088 *
89 * Device must show its capabilities in dev->features, set
90 * at device setup time.
91 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
92 * everything.
Linus Torvalds1da177e2005-04-16 15:20:36 -070093 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
94 * TCP/UDP over IPv4. Sigh. Vendors like this
95 * way by an unknown reason. Though, see comment above
96 * about CHECKSUM_UNNECESSARY. 8)
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070097 * NETIF_F_IPV6_CSUM about as dumb as the last one but does IPv6 instead.
Linus Torvalds1da177e2005-04-16 15:20:36 -070098 *
Yi Zou3af79302012-03-19 11:12:41 +000099 * UNNECESSARY: device will do per protocol specific csum. Protocol drivers
100 * that do not want net to perform the checksum calculation should use
101 * this flag in their outgoing skbs.
102 * NETIF_F_FCOE_CRC this indicates the device can do FCoE FC CRC
103 * offload. Correspondingly, the FCoE protocol driver
104 * stack should use CHECKSUM_UNNECESSARY.
105 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700106 * Any questions? No questions, good. --ANK
107 */
108
Linus Torvalds1da177e2005-04-16 15:20:36 -0700109struct net_device;
David Howells716ea3a2007-04-02 20:19:53 -0700110struct scatterlist;
Jens Axboe9c55e012007-11-06 23:30:13 -0800111struct pipe_inode_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700112
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700113#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114struct nf_conntrack {
115 atomic_t use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700116};
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700117#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700118
119#ifdef CONFIG_BRIDGE_NETFILTER
120struct nf_bridge_info {
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000121 atomic_t use;
122 unsigned int mask;
123 struct net_device *physindev;
124 struct net_device *physoutdev;
125 unsigned long data[32 / sizeof(unsigned long)];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700126};
127#endif
128
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129struct sk_buff_head {
130 /* These two members must be first. */
131 struct sk_buff *next;
132 struct sk_buff *prev;
133
134 __u32 qlen;
135 spinlock_t lock;
136};
137
138struct sk_buff;
139
Ian Campbell9d4dde52011-12-22 23:39:14 +0000140/* To allow 64K frame to be packed as single skb without frag_list we
141 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
142 * buffers which do not start on a page boundary.
143 *
144 * Since GRO uses frags we allocate at least 16 regardless of page
145 * size.
Anton Blancharda715dea2011-03-27 14:57:26 +0000146 */
Ian Campbell9d4dde52011-12-22 23:39:14 +0000147#if (65536/PAGE_SIZE + 1) < 16
David S. Millereec00952011-03-29 23:34:08 -0700148#define MAX_SKB_FRAGS 16UL
Anton Blancharda715dea2011-03-27 14:57:26 +0000149#else
Ian Campbell9d4dde52011-12-22 23:39:14 +0000150#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
Anton Blancharda715dea2011-03-27 14:57:26 +0000151#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152
153typedef struct skb_frag_struct skb_frag_t;
154
155struct skb_frag_struct {
Ian Campbella8605c62011-10-19 23:01:49 +0000156 struct {
157 struct page *p;
158 } page;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000159#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
David S. Millera309bb02007-07-30 18:47:03 -0700160 __u32 page_offset;
161 __u32 size;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000162#else
163 __u16 page_offset;
164 __u16 size;
165#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166};
167
Eric Dumazet9e903e02011-10-18 21:00:24 +0000168static inline unsigned int skb_frag_size(const skb_frag_t *frag)
169{
170 return frag->size;
171}
172
173static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
174{
175 frag->size = size;
176}
177
178static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
179{
180 frag->size += delta;
181}
182
183static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
184{
185 frag->size -= delta;
186}
187
Patrick Ohlyac45f602009-02-12 05:03:37 +0000188#define HAVE_HW_TIME_STAMP
189
190/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800191 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000192 * @hwtstamp: hardware time stamp transformed into duration
193 * since arbitrary point in time
194 * @syststamp: hwtstamp transformed to system time base
195 *
196 * Software time stamps generated by ktime_get_real() are stored in
197 * skb->tstamp. The relation between the different kinds of time
198 * stamps is as follows:
199 *
200 * syststamp and tstamp can be compared against each other in
201 * arbitrary combinations. The accuracy of a
202 * syststamp/tstamp/"syststamp from other device" comparison is
203 * limited by the accuracy of the transformation into system time
204 * base. This depends on the device driver and its underlying
205 * hardware.
206 *
207 * hwtstamps can only be compared against other hwtstamps from
208 * the same device.
209 *
210 * This structure is attached to packets as part of the
211 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
212 */
213struct skb_shared_hwtstamps {
214 ktime_t hwtstamp;
215 ktime_t syststamp;
216};
217
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000218/* Definitions for tx_flags in struct skb_shared_info */
219enum {
220 /* generate hardware time stamp */
221 SKBTX_HW_TSTAMP = 1 << 0,
222
223 /* generate software time stamp */
224 SKBTX_SW_TSTAMP = 1 << 1,
225
226 /* device driver is going to provide hardware time stamp */
227 SKBTX_IN_PROGRESS = 1 << 2,
228
Shirley Maa6686f22011-07-06 12:22:12 +0000229 /* device driver supports TX zero-copy buffers */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000230 SKBTX_DEV_ZEROCOPY = 1 << 3,
Johannes Berg6e3e9392011-11-09 10:15:42 +0100231
232 /* generate wifi status information (where possible) */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000233 SKBTX_WIFI_STATUS = 1 << 4,
Pravin B Shelarc9af6db2013-02-11 09:27:41 +0000234
235 /* This indicates at least one fragment might be overwritten
236 * (as in vmsplice(), sendfile() ...)
237 * If we need to compute a TX checksum, we'll need to copy
238 * all frags to avoid possible bad checksum
239 */
240 SKBTX_SHARED_FRAG = 1 << 5,
Shirley Maa6686f22011-07-06 12:22:12 +0000241};
242
243/*
244 * The callback notifies userspace to release buffers when skb DMA is done in
245 * lower device, the skb last reference should be 0 when calling this.
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000246 * The zerocopy_success argument is true if zero copy transmit occurred,
247 * false on data copy or out of memory error caused by data copy attempt.
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000248 * The ctx field is used to track device context.
249 * The desc field is used to track userspace buffer index.
Shirley Maa6686f22011-07-06 12:22:12 +0000250 */
251struct ubuf_info {
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000252 void (*callback)(struct ubuf_info *, bool zerocopy_success);
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000253 void *ctx;
Shirley Maa6686f22011-07-06 12:22:12 +0000254 unsigned long desc;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000255};
256
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257/* This data is invariant across clones and lives at
258 * the end of the header data, ie. at skb->end.
259 */
260struct skb_shared_info {
Ian Campbell9f42f122012-01-05 07:13:39 +0000261 unsigned char nr_frags;
262 __u8 tx_flags;
Herbert Xu79671682006-06-22 02:40:14 -0700263 unsigned short gso_size;
264 /* Warning: this field is not always filled in (UFO)! */
265 unsigned short gso_segs;
266 unsigned short gso_type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700267 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000268 struct skb_shared_hwtstamps hwtstamps;
Ian Campbell9f42f122012-01-05 07:13:39 +0000269 __be32 ip6_frag_id;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700270
271 /*
272 * Warning : all fields before dataref are cleared in __alloc_skb()
273 */
274 atomic_t dataref;
275
Johann Baudy69e3c752009-05-18 22:11:22 -0700276 /* Intermediate layers must ensure that destructor_arg
277 * remains valid until skb destructor */
278 void * destructor_arg;
Shirley Maa6686f22011-07-06 12:22:12 +0000279
Eric Dumazetfed66382010-07-22 19:09:08 +0000280 /* must be last field, see pskb_expand_head() */
281 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282};
283
284/* We divide dataref into two halves. The higher 16 bits hold references
285 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700286 * the entire skb->data. A clone of a headerless skb holds the length of
287 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288 *
289 * All users must obey the rule that the skb->data reference count must be
290 * greater than or equal to the payload reference count.
291 *
292 * Holding a reference to the payload part means that the user does not
293 * care about modifications to the header part of skb->data.
294 */
295#define SKB_DATAREF_SHIFT 16
296#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
297
David S. Millerd179cd12005-08-17 14:57:30 -0700298
299enum {
300 SKB_FCLONE_UNAVAILABLE,
301 SKB_FCLONE_ORIG,
302 SKB_FCLONE_CLONE,
303};
304
Herbert Xu79671682006-06-22 02:40:14 -0700305enum {
306 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700307 SKB_GSO_UDP = 1 << 1,
Herbert Xu576a30e2006-06-27 13:22:38 -0700308
309 /* This indicates the skb is from an untrusted source. */
310 SKB_GSO_DODGY = 1 << 2,
Michael Chanb0da85372006-06-29 12:30:00 -0700311
312 /* This indicates the tcp segment has CWR set. */
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700313 SKB_GSO_TCP_ECN = 1 << 3,
314
315 SKB_GSO_TCPV6 = 1 << 4,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800316
317 SKB_GSO_FCOE = 1 << 5,
Pravin B Shelar68c33162013-02-14 14:02:41 +0000318
319 SKB_GSO_GRE = 1 << 6,
Pravin B Shelar73136262013-03-07 13:21:51 +0000320
Eric Dumazetcb32f512013-10-19 11:42:57 -0700321 SKB_GSO_IPIP = 1 << 7,
Simon Horman0d89d202013-05-23 21:02:52 +0000322
Eric Dumazet61c1db72013-10-20 20:47:30 -0700323 SKB_GSO_SIT = 1 << 8,
Eric Dumazetcb32f512013-10-19 11:42:57 -0700324
Eric Dumazet61c1db72013-10-20 20:47:30 -0700325 SKB_GSO_UDP_TUNNEL = 1 << 9,
326
327 SKB_GSO_MPLS = 1 << 10,
Herbert Xu79671682006-06-22 02:40:14 -0700328};
329
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700330#if BITS_PER_LONG > 32
331#define NET_SKBUFF_DATA_USES_OFFSET 1
332#endif
333
334#ifdef NET_SKBUFF_DATA_USES_OFFSET
335typedef unsigned int sk_buff_data_t;
336#else
337typedef unsigned char *sk_buff_data_t;
338#endif
339
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100340#if defined(CONFIG_NF_DEFRAG_IPV4) || defined(CONFIG_NF_DEFRAG_IPV4_MODULE) || \
341 defined(CONFIG_NF_DEFRAG_IPV6) || defined(CONFIG_NF_DEFRAG_IPV6_MODULE)
342#define NET_SKBUFF_NF_DEFRAG_NEEDED 1
343#endif
344
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345/**
346 * struct sk_buff - socket buffer
347 * @next: Next buffer in list
348 * @prev: Previous buffer in list
Herbert Xu325ed822005-10-03 13:57:23 -0700349 * @tstamp: Time we arrived
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700350 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700351 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700352 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000353 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700354 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700355 * @len: Length of actual data
356 * @data_len: Data length
357 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700358 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700359 * @csum: Checksum (must include start/offset pair)
360 * @csum_start: Offset from skb->head where checksumming should start
361 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700362 * @priority: Packet queueing priority
Martin Waitz67be2dd2005-05-01 08:59:26 -0700363 * @local_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700364 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700365 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700366 * @nohdr: Payload reference only, must not modify header
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700367 * @nfctinfo: Relationship of this skb to the connection
Linus Torvalds1da177e2005-04-16 15:20:36 -0700368 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700369 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700370 * @ipvs_property: skbuff is owned by ipvs
Randy Dunlap31729362008-02-18 20:52:13 -0800371 * @peeked: this packet has been seen already, so stats have been
372 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700373 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700374 * @protocol: Packet protocol from driver
375 * @destructor: Destruct function
376 * @nfct: Associated connection, if any
Randy Dunlap461ddf32005-11-20 21:25:15 -0800377 * @nfct_reasm: netfilter conntrack re-assembly pointer
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800379 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380 * @tc_index: Traffic control index
381 * @tc_verd: traffic control verdict
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700382 * @rxhash: the packet hash computed on receive
383 * @queue_mapping: Queue mapping for multiqueue devices
Randy Dunlap553a5672008-04-20 10:51:01 -0700384 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700385 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Changli Gao4ca24622011-08-19 07:26:44 -0700386 * @l4_rxhash: indicate rxhash is a canonical 4-tuple hash over transport
387 * ports.
Johannes Berg6e3e9392011-11-09 10:15:42 +0100388 * @wifi_acked_valid: wifi_acked was set
389 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000390 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Randy Dunlapf4b8ea72006-06-22 16:00:11 -0700391 * @dma_cookie: a cookie to one of several possible DMA operations
392 * done by skb DMA functions
Eliezer Tamir06021292013-06-10 11:39:50 +0300393 * @napi_id: id of the NAPI struct this skb came from
James Morris984bc162006-06-09 00:29:17 -0700394 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700395 * @mark: Generic packet mark
396 * @dropcount: total number of sk_receive_queue overflows
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000397 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b02008-07-14 22:49:06 -0700398 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000399 * @inner_protocol: Protocol (encapsulation)
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000400 * @inner_transport_header: Inner transport layer header (encapsulation)
401 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000402 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700403 * @transport_header: Transport layer header
404 * @network_header: Network layer header
405 * @mac_header: Link layer header
406 * @tail: Tail pointer
407 * @end: End pointer
408 * @head: Head of buffer
409 * @data: Data head pointer
410 * @truesize: Buffer size
411 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412 */
413
414struct sk_buff {
415 /* These two members must be first. */
416 struct sk_buff *next;
417 struct sk_buff *prev;
418
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700419 ktime_t tstamp;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000420
421 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424 /*
425 * This is the control buffer. It is free to use for every
426 * layer. Please put your private variables there. If you
427 * want to keep them across layers you have to do a skb_clone()
428 * first. This is owned by whoever has the skb queued ATM.
429 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000430 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431
Eric Dumazet7fee2262010-05-11 23:19:48 +0000432 unsigned long _skb_refdst;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000433#ifdef CONFIG_XFRM
434 struct sec_path *sp;
435#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700437 data_len;
438 __u16 mac_len,
439 hdr_len;
Al Viroff1dcad2006-11-20 18:07:29 -0800440 union {
441 __wsum csum;
Herbert Xu663ead32007-04-09 11:59:07 -0700442 struct {
443 __u16 csum_start;
444 __u16 csum_offset;
445 };
Al Viroff1dcad2006-11-20 18:07:29 -0800446 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 __u32 priority;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200448 kmemcheck_bitfield_begin(flags1);
Thomas Graf1cbb3382005-07-05 14:13:41 -0700449 __u8 local_df:1,
450 cloned:1,
451 ip_summed:2,
Harald Welte6869c4d2005-08-09 19:24:19 -0700452 nohdr:1,
453 nfctinfo:3;
David S. Millerd179cd12005-08-17 14:57:30 -0700454 __u8 pkt_type:3,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800455 fclone:2,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700456 ipvs_property:1,
Herbert Xua59322b2007-12-05 01:53:40 -0800457 peeked:1,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700458 nf_trace:1;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200459 kmemcheck_bitfield_end(flags1);
Eric Dumazet4ab408d2010-03-01 03:09:26 +0000460 __be16 protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461
462 void (*destructor)(struct sk_buff *skb);
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800463#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700464 struct nf_conntrack *nfct;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100465#endif
466#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800467 struct sk_buff *nfct_reasm;
468#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700469#ifdef CONFIG_BRIDGE_NETFILTER
470 struct nf_bridge_info *nf_bridge;
471#endif
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700472
Eric Dumazet8964be42009-11-20 15:35:04 -0800473 int skb_iif;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000474
475 __u32 rxhash;
476
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000477 __be16 vlan_proto;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000478 __u16 vlan_tci;
479
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700481 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700483 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200486
Eric Dumazet0a148422011-04-20 09:27:32 +0000487 __u16 queue_mapping;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200488 kmemcheck_bitfield_begin(flags2);
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400489#ifdef CONFIG_IPV6_NDISC_NODETYPE
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000490 __u8 ndisc_nodetype:2;
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400491#endif
Mel Gormanc93bdd02012-07-31 16:44:19 -0700492 __u8 pfmemalloc:1;
Tom Herbert3853b582010-11-21 13:17:29 +0000493 __u8 ooo_okay:1;
Tom Herbertbdeab992011-08-14 19:45:55 +0000494 __u8 l4_rxhash:1;
Johannes Berg6e3e9392011-11-09 10:15:42 +0100495 __u8 wifi_acked_valid:1;
496 __u8 wifi_acked:1;
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000497 __u8 no_fcs:1;
Eric Dumazetd3836f22012-04-27 00:33:38 +0000498 __u8 head_frag:1;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000499 /* Encapsulation protocol and NIC drivers should use
500 * this flag to indicate to each other if the skb contains
501 * encapsulated packet or not and maybe use the inner packet
502 * headers if needed
503 */
504 __u8 encapsulation:1;
Nicolas Dichtel45906722013-09-30 14:16:41 +0200505 /* 6/8 bit hole (depending on ndisc_nodetype presence) */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200506 kmemcheck_bitfield_end(flags2);
507
Cong Wange0d10952013-08-01 11:10:25 +0800508#if defined CONFIG_NET_DMA || defined CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir06021292013-06-10 11:39:50 +0300509 union {
510 unsigned int napi_id;
511 dma_cookie_t dma_cookie;
512 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700513#endif
James Morris984bc162006-06-09 00:29:17 -0700514#ifdef CONFIG_NETWORK_SECMARK
515 __u32 secmark;
516#endif
Neil Horman3b885782009-10-12 13:26:31 -0700517 union {
518 __u32 mark;
519 __u32 dropcount;
Eric Dumazet16fad692013-03-14 05:40:32 +0000520 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700521 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522
Simon Horman0d89d202013-05-23 21:02:52 +0000523 __be16 inner_protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000524 __u16 inner_transport_header;
525 __u16 inner_network_header;
526 __u16 inner_mac_header;
527 __u16 transport_header;
528 __u16 network_header;
529 __u16 mac_header;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700530 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700531 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700532 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700534 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700535 unsigned int truesize;
536 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537};
538
539#ifdef __KERNEL__
540/*
541 * Handling routines are only of interest to the kernel
542 */
543#include <linux/slab.h>
544
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545
Mel Gormanc93bdd02012-07-31 16:44:19 -0700546#define SKB_ALLOC_FCLONE 0x01
547#define SKB_ALLOC_RX 0x02
548
549/* Returns true if the skb was allocated from PFMEMALLOC reserves */
550static inline bool skb_pfmemalloc(const struct sk_buff *skb)
551{
552 return unlikely(skb->pfmemalloc);
553}
554
Eric Dumazet7fee2262010-05-11 23:19:48 +0000555/*
556 * skb might have a dst pointer attached, refcounted or not.
557 * _skb_refdst low order bit is set if refcount was _not_ taken
558 */
559#define SKB_DST_NOREF 1UL
560#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
561
562/**
563 * skb_dst - returns skb dst_entry
564 * @skb: buffer
565 *
566 * Returns skb dst_entry, regardless of reference taken or not.
567 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000568static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
569{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000570 /* If refdst was not refcounted, check we still are in a
571 * rcu_read_lock section
572 */
573 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
574 !rcu_read_lock_held() &&
575 !rcu_read_lock_bh_held());
576 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000577}
578
Eric Dumazet7fee2262010-05-11 23:19:48 +0000579/**
580 * skb_dst_set - sets skb dst
581 * @skb: buffer
582 * @dst: dst entry
583 *
584 * Sets skb dst, assuming a reference was taken on dst and should
585 * be released by skb_dst_drop()
586 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000587static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
588{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000589 skb->_skb_refdst = (unsigned long)dst;
590}
591
Joe Perches7965bd42013-09-26 14:48:15 -0700592void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
593 bool force);
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200594
595/**
596 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
597 * @skb: buffer
598 * @dst: dst entry
599 *
600 * Sets skb dst, assuming a reference was not taken on dst.
601 * If dst entry is cached, we do not take reference and dst_release
602 * will be avoided by refdst_drop. If dst entry is not cached, we take
603 * reference, so that last dst_release can destroy the dst immediately.
604 */
605static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
606{
607 __skb_dst_set_noref(skb, dst, false);
608}
609
610/**
611 * skb_dst_set_noref_force - sets skb dst, without taking reference
612 * @skb: buffer
613 * @dst: dst entry
614 *
615 * Sets skb dst, assuming a reference was not taken on dst.
616 * No reference is taken and no dst_release will be called. While for
617 * cached dsts deferred reclaim is a basic feature, for entries that are
618 * not cached it is caller's job to guarantee that last dst_release for
619 * provided dst happens when nobody uses it, eg. after a RCU grace period.
620 */
621static inline void skb_dst_set_noref_force(struct sk_buff *skb,
622 struct dst_entry *dst)
623{
624 __skb_dst_set_noref(skb, dst, true);
625}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000626
627/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300628 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000629 * @skb: buffer
630 */
631static inline bool skb_dst_is_noref(const struct sk_buff *skb)
632{
633 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000634}
635
Eric Dumazet511c3f92009-06-02 05:14:27 +0000636static inline struct rtable *skb_rtable(const struct sk_buff *skb)
637{
Eric Dumazetadf30902009-06-02 05:19:30 +0000638 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000639}
640
Joe Perches7965bd42013-09-26 14:48:15 -0700641void kfree_skb(struct sk_buff *skb);
642void kfree_skb_list(struct sk_buff *segs);
643void skb_tx_error(struct sk_buff *skb);
644void consume_skb(struct sk_buff *skb);
645void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000646extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000647
Joe Perches7965bd42013-09-26 14:48:15 -0700648void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
649bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
650 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000651
Joe Perches7965bd42013-09-26 14:48:15 -0700652struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
653 int node);
654struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700655static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100656 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700657{
Eric Dumazet564824b2010-10-11 19:05:25 +0000658 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700659}
660
661static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100662 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700663{
Mel Gormanc93bdd02012-07-31 16:44:19 -0700664 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700665}
666
Joe Perches7965bd42013-09-26 14:48:15 -0700667struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
Patrick McHardy0ebd0ac2013-04-17 06:46:58 +0000668static inline struct sk_buff *alloc_skb_head(gfp_t priority)
669{
670 return __alloc_skb_head(priority, -1);
671}
672
Joe Perches7965bd42013-09-26 14:48:15 -0700673struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
674int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
675struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
676struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
677struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom, gfp_t gfp_mask);
Eric Dumazet117632e2011-12-03 21:39:53 +0000678
Joe Perches7965bd42013-09-26 14:48:15 -0700679int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
680struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
681 unsigned int headroom);
682struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
683 int newtailroom, gfp_t priority);
684int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
685 int len);
686int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
687int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000688#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700689
Joe Perches7965bd42013-09-26 14:48:15 -0700690int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
691 int getfrag(void *from, char *to, int offset,
692 int len, int odd, struct sk_buff *skb),
693 void *from, int length);
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700694
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800695struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700696 __u32 lower_offset;
697 __u32 upper_offset;
698 __u32 frag_idx;
699 __u32 stepped_offset;
700 struct sk_buff *root_skb;
701 struct sk_buff *cur_skb;
702 __u8 *frag_data;
703};
704
Joe Perches7965bd42013-09-26 14:48:15 -0700705void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
706 unsigned int to, struct skb_seq_state *st);
707unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
708 struct skb_seq_state *st);
709void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -0700710
Joe Perches7965bd42013-09-26 14:48:15 -0700711unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
712 unsigned int to, struct ts_config *config,
713 struct ts_state *state);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700714
Joe Perches7965bd42013-09-26 14:48:15 -0700715void __skb_get_rxhash(struct sk_buff *skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000716static inline __u32 skb_get_rxhash(struct sk_buff *skb)
717{
Willem de Bruijnecd5cf52012-10-26 11:52:08 +0000718 if (!skb->l4_rxhash)
Tom Herbertbdeab992011-08-14 19:45:55 +0000719 __skb_get_rxhash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000720
721 return skb->rxhash;
722}
723
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700724#ifdef NET_SKBUFF_DATA_USES_OFFSET
725static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
726{
727 return skb->head + skb->end;
728}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000729
730static inline unsigned int skb_end_offset(const struct sk_buff *skb)
731{
732 return skb->end;
733}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700734#else
735static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
736{
737 return skb->end;
738}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000739
740static inline unsigned int skb_end_offset(const struct sk_buff *skb)
741{
742 return skb->end - skb->head;
743}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700744#endif
745
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700747#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748
Patrick Ohlyac45f602009-02-12 05:03:37 +0000749static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
750{
751 return &skb_shinfo(skb)->hwtstamps;
752}
753
Linus Torvalds1da177e2005-04-16 15:20:36 -0700754/**
755 * skb_queue_empty - check if a queue is empty
756 * @list: queue head
757 *
758 * Returns true if the queue is empty, false otherwise.
759 */
760static inline int skb_queue_empty(const struct sk_buff_head *list)
761{
762 return list->next == (struct sk_buff *)list;
763}
764
765/**
David S. Millerfc7ebb22008-09-23 00:34:07 -0700766 * skb_queue_is_last - check if skb is the last entry in the queue
767 * @list: queue head
768 * @skb: buffer
769 *
770 * Returns true if @skb is the last buffer on the list.
771 */
772static inline bool skb_queue_is_last(const struct sk_buff_head *list,
773 const struct sk_buff *skb)
774{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000775 return skb->next == (struct sk_buff *)list;
David S. Millerfc7ebb22008-09-23 00:34:07 -0700776}
777
778/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800779 * skb_queue_is_first - check if skb is the first entry in the queue
780 * @list: queue head
781 * @skb: buffer
782 *
783 * Returns true if @skb is the first buffer on the list.
784 */
785static inline bool skb_queue_is_first(const struct sk_buff_head *list,
786 const struct sk_buff *skb)
787{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000788 return skb->prev == (struct sk_buff *)list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800789}
790
791/**
David S. Miller249c8b42008-09-23 00:44:42 -0700792 * skb_queue_next - return the next packet in the queue
793 * @list: queue head
794 * @skb: current buffer
795 *
796 * Return the next packet in @list after @skb. It is only valid to
797 * call this if skb_queue_is_last() evaluates to false.
798 */
799static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
800 const struct sk_buff *skb)
801{
802 /* This BUG_ON may seem severe, but if we just return then we
803 * are going to dereference garbage.
804 */
805 BUG_ON(skb_queue_is_last(list, skb));
806 return skb->next;
807}
808
809/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800810 * skb_queue_prev - return the prev packet in the queue
811 * @list: queue head
812 * @skb: current buffer
813 *
814 * Return the prev packet in @list before @skb. It is only valid to
815 * call this if skb_queue_is_first() evaluates to false.
816 */
817static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
818 const struct sk_buff *skb)
819{
820 /* This BUG_ON may seem severe, but if we just return then we
821 * are going to dereference garbage.
822 */
823 BUG_ON(skb_queue_is_first(list, skb));
824 return skb->prev;
825}
826
827/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828 * skb_get - reference buffer
829 * @skb: buffer to reference
830 *
831 * Makes another reference to a socket buffer and returns a pointer
832 * to the buffer.
833 */
834static inline struct sk_buff *skb_get(struct sk_buff *skb)
835{
836 atomic_inc(&skb->users);
837 return skb;
838}
839
840/*
841 * If users == 1, we are the only owner and are can avoid redundant
842 * atomic change.
843 */
844
845/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700846 * skb_cloned - is the buffer a clone
847 * @skb: buffer to check
848 *
849 * Returns true if the buffer was generated with skb_clone() and is
850 * one of multiple shared copies of the buffer. Cloned buffers are
851 * shared data so must not be written to under normal circumstances.
852 */
853static inline int skb_cloned(const struct sk_buff *skb)
854{
855 return skb->cloned &&
856 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
857}
858
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +0000859static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
860{
861 might_sleep_if(pri & __GFP_WAIT);
862
863 if (skb_cloned(skb))
864 return pskb_expand_head(skb, 0, 0, pri);
865
866 return 0;
867}
868
Linus Torvalds1da177e2005-04-16 15:20:36 -0700869/**
870 * skb_header_cloned - is the header a clone
871 * @skb: buffer to check
872 *
873 * Returns true if modifying the header part of the buffer requires
874 * the data to be copied.
875 */
876static inline int skb_header_cloned(const struct sk_buff *skb)
877{
878 int dataref;
879
880 if (!skb->cloned)
881 return 0;
882
883 dataref = atomic_read(&skb_shinfo(skb)->dataref);
884 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
885 return dataref != 1;
886}
887
888/**
889 * skb_header_release - release reference to header
890 * @skb: buffer to operate on
891 *
892 * Drop a reference to the header part of the buffer. This is done
893 * by acquiring a payload reference. You must not read from the header
894 * part of skb->data after this.
895 */
896static inline void skb_header_release(struct sk_buff *skb)
897{
898 BUG_ON(skb->nohdr);
899 skb->nohdr = 1;
900 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
901}
902
903/**
904 * skb_shared - is the buffer shared
905 * @skb: buffer to check
906 *
907 * Returns true if more than one person has a reference to this
908 * buffer.
909 */
910static inline int skb_shared(const struct sk_buff *skb)
911{
912 return atomic_read(&skb->users) != 1;
913}
914
915/**
916 * skb_share_check - check if buffer is shared and if so clone it
917 * @skb: buffer to check
918 * @pri: priority for memory allocation
919 *
920 * If the buffer is shared the buffer is cloned and the old copy
921 * drops a reference. A new clone with a single reference is returned.
922 * If the buffer is not shared the original buffer is returned. When
923 * being called from interrupt status or with spinlocks held pri must
924 * be GFP_ATOMIC.
925 *
926 * NULL is returned on a memory allocation failure.
927 */
Eric Dumazet47061bc2012-08-03 20:54:15 +0000928static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700929{
930 might_sleep_if(pri & __GFP_WAIT);
931 if (skb_shared(skb)) {
932 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +0000933
934 if (likely(nskb))
935 consume_skb(skb);
936 else
937 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700938 skb = nskb;
939 }
940 return skb;
941}
942
943/*
944 * Copy shared buffers into a new sk_buff. We effectively do COW on
945 * packets to handle cases where we have a local reader and forward
946 * and a couple of other messy ones. The normal one is tcpdumping
947 * a packet thats being forwarded.
948 */
949
950/**
951 * skb_unshare - make a copy of a shared buffer
952 * @skb: buffer to check
953 * @pri: priority for memory allocation
954 *
955 * If the socket buffer is a clone then this function creates a new
956 * copy of the data, drops a reference count on the old copy and returns
957 * the new copy with the reference count at 1. If the buffer is not a clone
958 * the original buffer is returned. When called with a spinlock held or
959 * from interrupt state @pri must be %GFP_ATOMIC
960 *
961 * %NULL is returned on a memory allocation failure.
962 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -0700963static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100964 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700965{
966 might_sleep_if(pri & __GFP_WAIT);
967 if (skb_cloned(skb)) {
968 struct sk_buff *nskb = skb_copy(skb, pri);
969 kfree_skb(skb); /* Free our shared copy */
970 skb = nskb;
971 }
972 return skb;
973}
974
975/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -0800976 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977 * @list_: list to peek at
978 *
979 * Peek an &sk_buff. Unlike most other operations you _MUST_
980 * be careful with this one. A peek leaves the buffer on the
981 * list and someone else may run off with it. You must hold
982 * the appropriate locks or have a private queue to do this.
983 *
984 * Returns %NULL for an empty list or a pointer to the head element.
985 * The reference count is not incremented and the reference is therefore
986 * volatile. Use with caution.
987 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -0400988static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989{
Eric Dumazet18d07002012-04-30 16:31:46 +0000990 struct sk_buff *skb = list_->next;
991
992 if (skb == (struct sk_buff *)list_)
993 skb = NULL;
994 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700995}
996
997/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +0000998 * skb_peek_next - peek skb following the given one from a queue
999 * @skb: skb to start from
1000 * @list_: list to peek at
1001 *
1002 * Returns %NULL when the end of the list is met or a pointer to the
1003 * next element. The reference count is not incremented and the
1004 * reference is therefore volatile. Use with caution.
1005 */
1006static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1007 const struct sk_buff_head *list_)
1008{
1009 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001010
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001011 if (next == (struct sk_buff *)list_)
1012 next = NULL;
1013 return next;
1014}
1015
1016/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001017 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001018 * @list_: list to peek at
1019 *
1020 * Peek an &sk_buff. Unlike most other operations you _MUST_
1021 * be careful with this one. A peek leaves the buffer on the
1022 * list and someone else may run off with it. You must hold
1023 * the appropriate locks or have a private queue to do this.
1024 *
1025 * Returns %NULL for an empty list or a pointer to the tail element.
1026 * The reference count is not incremented and the reference is therefore
1027 * volatile. Use with caution.
1028 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001029static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001030{
Eric Dumazet18d07002012-04-30 16:31:46 +00001031 struct sk_buff *skb = list_->prev;
1032
1033 if (skb == (struct sk_buff *)list_)
1034 skb = NULL;
1035 return skb;
1036
Linus Torvalds1da177e2005-04-16 15:20:36 -07001037}
1038
1039/**
1040 * skb_queue_len - get queue length
1041 * @list_: list to measure
1042 *
1043 * Return the length of an &sk_buff queue.
1044 */
1045static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1046{
1047 return list_->qlen;
1048}
1049
David S. Miller67fed452008-09-21 22:36:24 -07001050/**
1051 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1052 * @list: queue to initialize
1053 *
1054 * This initializes only the list and queue length aspects of
1055 * an sk_buff_head object. This allows to initialize the list
1056 * aspects of an sk_buff_head without reinitializing things like
1057 * the spinlock. It can also be used for on-stack sk_buff_head
1058 * objects where the spinlock is known to not be used.
1059 */
1060static inline void __skb_queue_head_init(struct sk_buff_head *list)
1061{
1062 list->prev = list->next = (struct sk_buff *)list;
1063 list->qlen = 0;
1064}
1065
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001066/*
1067 * This function creates a split out lock class for each invocation;
1068 * this is needed for now since a whole lot of users of the skb-queue
1069 * infrastructure in drivers have different locking usage (in hardirq)
1070 * than the networking core (in softirq only). In the long run either the
1071 * network layer or drivers should need annotation to consolidate the
1072 * main types of usage into 3 classes.
1073 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074static inline void skb_queue_head_init(struct sk_buff_head *list)
1075{
1076 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001077 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001078}
1079
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001080static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1081 struct lock_class_key *class)
1082{
1083 skb_queue_head_init(list);
1084 lockdep_set_class(&list->lock, class);
1085}
1086
Linus Torvalds1da177e2005-04-16 15:20:36 -07001087/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001088 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001089 *
1090 * The "__skb_xxxx()" functions are the non-atomic ones that
1091 * can only be called with interrupts disabled.
1092 */
Joe Perches7965bd42013-09-26 14:48:15 -07001093void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1094 struct sk_buff_head *list);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001095static inline void __skb_insert(struct sk_buff *newsk,
1096 struct sk_buff *prev, struct sk_buff *next,
1097 struct sk_buff_head *list)
1098{
1099 newsk->next = next;
1100 newsk->prev = prev;
1101 next->prev = prev->next = newsk;
1102 list->qlen++;
1103}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001104
David S. Miller67fed452008-09-21 22:36:24 -07001105static inline void __skb_queue_splice(const struct sk_buff_head *list,
1106 struct sk_buff *prev,
1107 struct sk_buff *next)
1108{
1109 struct sk_buff *first = list->next;
1110 struct sk_buff *last = list->prev;
1111
1112 first->prev = prev;
1113 prev->next = first;
1114
1115 last->next = next;
1116 next->prev = last;
1117}
1118
1119/**
1120 * skb_queue_splice - join two skb lists, this is designed for stacks
1121 * @list: the new list to add
1122 * @head: the place to add it in the first list
1123 */
1124static inline void skb_queue_splice(const struct sk_buff_head *list,
1125 struct sk_buff_head *head)
1126{
1127 if (!skb_queue_empty(list)) {
1128 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001129 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001130 }
1131}
1132
1133/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001134 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001135 * @list: the new list to add
1136 * @head: the place to add it in the first list
1137 *
1138 * The list at @list is reinitialised
1139 */
1140static inline void skb_queue_splice_init(struct sk_buff_head *list,
1141 struct sk_buff_head *head)
1142{
1143 if (!skb_queue_empty(list)) {
1144 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001145 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001146 __skb_queue_head_init(list);
1147 }
1148}
1149
1150/**
1151 * skb_queue_splice_tail - join two skb lists, each list being a queue
1152 * @list: the new list to add
1153 * @head: the place to add it in the first list
1154 */
1155static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1156 struct sk_buff_head *head)
1157{
1158 if (!skb_queue_empty(list)) {
1159 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001160 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001161 }
1162}
1163
1164/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001165 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001166 * @list: the new list to add
1167 * @head: the place to add it in the first list
1168 *
1169 * Each of the lists is a queue.
1170 * The list at @list is reinitialised
1171 */
1172static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1173 struct sk_buff_head *head)
1174{
1175 if (!skb_queue_empty(list)) {
1176 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001177 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001178 __skb_queue_head_init(list);
1179 }
1180}
1181
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001183 * __skb_queue_after - queue a buffer at the list head
1184 * @list: list to use
1185 * @prev: place after this buffer
1186 * @newsk: buffer to queue
1187 *
1188 * Queue a buffer int the middle of a list. This function takes no locks
1189 * and you must therefore hold required locks before calling it.
1190 *
1191 * A buffer cannot be placed on two lists at the same time.
1192 */
1193static inline void __skb_queue_after(struct sk_buff_head *list,
1194 struct sk_buff *prev,
1195 struct sk_buff *newsk)
1196{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001197 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001198}
1199
Joe Perches7965bd42013-09-26 14:48:15 -07001200void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1201 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001202
Gerrit Renkerf5572852008-04-14 00:05:28 -07001203static inline void __skb_queue_before(struct sk_buff_head *list,
1204 struct sk_buff *next,
1205 struct sk_buff *newsk)
1206{
1207 __skb_insert(newsk, next->prev, next, list);
1208}
1209
Stephen Hemminger300ce172005-10-30 13:47:34 -08001210/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211 * __skb_queue_head - queue a buffer at the list head
1212 * @list: list to use
1213 * @newsk: buffer to queue
1214 *
1215 * Queue a buffer at the start of a list. This function takes no locks
1216 * and you must therefore hold required locks before calling it.
1217 *
1218 * A buffer cannot be placed on two lists at the same time.
1219 */
Joe Perches7965bd42013-09-26 14:48:15 -07001220void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001221static inline void __skb_queue_head(struct sk_buff_head *list,
1222 struct sk_buff *newsk)
1223{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001224 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001225}
1226
1227/**
1228 * __skb_queue_tail - queue a buffer at the list tail
1229 * @list: list to use
1230 * @newsk: buffer to queue
1231 *
1232 * Queue a buffer at the end of a list. This function takes no locks
1233 * and you must therefore hold required locks before calling it.
1234 *
1235 * A buffer cannot be placed on two lists at the same time.
1236 */
Joe Perches7965bd42013-09-26 14:48:15 -07001237void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001238static inline void __skb_queue_tail(struct sk_buff_head *list,
1239 struct sk_buff *newsk)
1240{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001241 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001242}
1243
Linus Torvalds1da177e2005-04-16 15:20:36 -07001244/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001245 * remove sk_buff from list. _Must_ be called atomically, and with
1246 * the list known..
1247 */
Joe Perches7965bd42013-09-26 14:48:15 -07001248void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001249static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1250{
1251 struct sk_buff *next, *prev;
1252
1253 list->qlen--;
1254 next = skb->next;
1255 prev = skb->prev;
1256 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001257 next->prev = prev;
1258 prev->next = next;
1259}
1260
Gerrit Renkerf525c062008-04-14 00:04:12 -07001261/**
1262 * __skb_dequeue - remove from the head of the queue
1263 * @list: list to dequeue from
1264 *
1265 * Remove the head of the list. This function does not take any locks
1266 * so must be used with appropriate locks held only. The head item is
1267 * returned or %NULL if the list is empty.
1268 */
Joe Perches7965bd42013-09-26 14:48:15 -07001269struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Gerrit Renkerf525c062008-04-14 00:04:12 -07001270static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1271{
1272 struct sk_buff *skb = skb_peek(list);
1273 if (skb)
1274 __skb_unlink(skb, list);
1275 return skb;
1276}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001277
1278/**
1279 * __skb_dequeue_tail - remove from the tail of the queue
1280 * @list: list to dequeue from
1281 *
1282 * Remove the tail of the list. This function does not take any locks
1283 * so must be used with appropriate locks held only. The tail item is
1284 * returned or %NULL if the list is empty.
1285 */
Joe Perches7965bd42013-09-26 14:48:15 -07001286struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001287static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1288{
1289 struct sk_buff *skb = skb_peek_tail(list);
1290 if (skb)
1291 __skb_unlink(skb, list);
1292 return skb;
1293}
1294
1295
David S. Millerbdcc0922012-03-07 20:53:36 -05001296static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001297{
1298 return skb->data_len;
1299}
1300
1301static inline unsigned int skb_headlen(const struct sk_buff *skb)
1302{
1303 return skb->len - skb->data_len;
1304}
1305
1306static inline int skb_pagelen(const struct sk_buff *skb)
1307{
1308 int i, len = 0;
1309
1310 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001311 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001312 return len + skb_headlen(skb);
1313}
1314
Ian Campbell131ea662011-08-19 06:25:00 +00001315/**
1316 * __skb_fill_page_desc - initialise a paged fragment in an skb
1317 * @skb: buffer containing fragment to be initialised
1318 * @i: paged fragment index to initialise
1319 * @page: the page to use for this fragment
1320 * @off: the offset to the data with @page
1321 * @size: the length of the data
1322 *
1323 * Initialises the @i'th fragment of @skb to point to &size bytes at
1324 * offset @off within @page.
1325 *
1326 * Does not take any additional reference on the fragment.
1327 */
1328static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1329 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001330{
1331 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1332
Mel Gormanc48a11c2012-07-31 16:44:23 -07001333 /*
1334 * Propagate page->pfmemalloc to the skb if we can. The problem is
1335 * that not all callers have unique ownership of the page. If
1336 * pfmemalloc is set, we check the mapping as a mapping implies
1337 * page->index is set (index and pfmemalloc share space).
1338 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1339 * do not lose pfmemalloc information as the pages would not be
1340 * allocated using __GFP_MEMALLOC.
1341 */
Ian Campbella8605c62011-10-19 23:01:49 +00001342 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001344 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001345
1346 page = compound_head(page);
1347 if (page->pfmemalloc && !page->mapping)
1348 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00001349}
1350
1351/**
1352 * skb_fill_page_desc - initialise a paged fragment in an skb
1353 * @skb: buffer containing fragment to be initialised
1354 * @i: paged fragment index to initialise
1355 * @page: the page to use for this fragment
1356 * @off: the offset to the data with @page
1357 * @size: the length of the data
1358 *
1359 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01001360 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00001361 * addition updates @skb such that @i is the last fragment.
1362 *
1363 * Does not take any additional reference on the fragment.
1364 */
1365static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1366 struct page *page, int off, int size)
1367{
1368 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001369 skb_shinfo(skb)->nr_frags = i + 1;
1370}
1371
Joe Perches7965bd42013-09-26 14:48:15 -07001372void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1373 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001374
Jason Wangf8e617e2013-11-01 14:07:47 +08001375void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1376 unsigned int truesize);
1377
Linus Torvalds1da177e2005-04-16 15:20:36 -07001378#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001379#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001380#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1381
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001382#ifdef NET_SKBUFF_DATA_USES_OFFSET
1383static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1384{
1385 return skb->head + skb->tail;
1386}
1387
1388static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1389{
1390 skb->tail = skb->data - skb->head;
1391}
1392
1393static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1394{
1395 skb_reset_tail_pointer(skb);
1396 skb->tail += offset;
1397}
Simon Horman7cc46192013-05-28 20:34:29 +00001398
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001399#else /* NET_SKBUFF_DATA_USES_OFFSET */
1400static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1401{
1402 return skb->tail;
1403}
1404
1405static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1406{
1407 skb->tail = skb->data;
1408}
1409
1410static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1411{
1412 skb->tail = skb->data + offset;
1413}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001414
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001415#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1416
Linus Torvalds1da177e2005-04-16 15:20:36 -07001417/*
1418 * Add data to an sk_buff
1419 */
Mathias Krause0c7ddf32013-11-07 14:18:24 +01001420unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001421unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001422static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1423{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001424 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425 SKB_LINEAR_ASSERT(skb);
1426 skb->tail += len;
1427 skb->len += len;
1428 return tmp;
1429}
1430
Joe Perches7965bd42013-09-26 14:48:15 -07001431unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001432static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1433{
1434 skb->data -= len;
1435 skb->len += len;
1436 return skb->data;
1437}
1438
Joe Perches7965bd42013-09-26 14:48:15 -07001439unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001440static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1441{
1442 skb->len -= len;
1443 BUG_ON(skb->len < skb->data_len);
1444 return skb->data += len;
1445}
1446
David S. Miller47d29642010-05-02 02:21:44 -07001447static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1448{
1449 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1450}
1451
Joe Perches7965bd42013-09-26 14:48:15 -07001452unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001453
1454static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1455{
1456 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001457 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001458 return NULL;
1459 skb->len -= len;
1460 return skb->data += len;
1461}
1462
1463static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1464{
1465 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1466}
1467
1468static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1469{
1470 if (likely(len <= skb_headlen(skb)))
1471 return 1;
1472 if (unlikely(len > skb->len))
1473 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001474 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001475}
1476
1477/**
1478 * skb_headroom - bytes at buffer head
1479 * @skb: buffer to check
1480 *
1481 * Return the number of bytes of free space at the head of an &sk_buff.
1482 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001483static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001484{
1485 return skb->data - skb->head;
1486}
1487
1488/**
1489 * skb_tailroom - bytes at buffer end
1490 * @skb: buffer to check
1491 *
1492 * Return the number of bytes of free space at the tail of an sk_buff
1493 */
1494static inline int skb_tailroom(const struct sk_buff *skb)
1495{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001496 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001497}
1498
1499/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00001500 * skb_availroom - bytes at buffer end
1501 * @skb: buffer to check
1502 *
1503 * Return the number of bytes of free space at the tail of an sk_buff
1504 * allocated by sk_stream_alloc()
1505 */
1506static inline int skb_availroom(const struct sk_buff *skb)
1507{
Eric Dumazet16fad692013-03-14 05:40:32 +00001508 if (skb_is_nonlinear(skb))
1509 return 0;
1510
1511 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00001512}
1513
1514/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001515 * skb_reserve - adjust headroom
1516 * @skb: buffer to alter
1517 * @len: bytes to move
1518 *
1519 * Increase the headroom of an empty &sk_buff by reducing the tail
1520 * room. This is only allowed for an empty buffer.
1521 */
David S. Miller8243126c2006-01-17 02:54:21 -08001522static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523{
1524 skb->data += len;
1525 skb->tail += len;
1526}
1527
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001528static inline void skb_reset_inner_headers(struct sk_buff *skb)
1529{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001530 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001531 skb->inner_network_header = skb->network_header;
1532 skb->inner_transport_header = skb->transport_header;
1533}
1534
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00001535static inline void skb_reset_mac_len(struct sk_buff *skb)
1536{
1537 skb->mac_len = skb->network_header - skb->mac_header;
1538}
1539
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001540static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1541 *skb)
1542{
1543 return skb->head + skb->inner_transport_header;
1544}
1545
1546static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1547{
1548 skb->inner_transport_header = skb->data - skb->head;
1549}
1550
1551static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1552 const int offset)
1553{
1554 skb_reset_inner_transport_header(skb);
1555 skb->inner_transport_header += offset;
1556}
1557
1558static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1559{
1560 return skb->head + skb->inner_network_header;
1561}
1562
1563static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1564{
1565 skb->inner_network_header = skb->data - skb->head;
1566}
1567
1568static inline void skb_set_inner_network_header(struct sk_buff *skb,
1569 const int offset)
1570{
1571 skb_reset_inner_network_header(skb);
1572 skb->inner_network_header += offset;
1573}
1574
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001575static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1576{
1577 return skb->head + skb->inner_mac_header;
1578}
1579
1580static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1581{
1582 skb->inner_mac_header = skb->data - skb->head;
1583}
1584
1585static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1586 const int offset)
1587{
1588 skb_reset_inner_mac_header(skb);
1589 skb->inner_mac_header += offset;
1590}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001591static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1592{
Cong Wang35d04612013-05-29 15:16:05 +08001593 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001594}
1595
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001596static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1597{
1598 return skb->head + skb->transport_header;
1599}
1600
1601static inline void skb_reset_transport_header(struct sk_buff *skb)
1602{
1603 skb->transport_header = skb->data - skb->head;
1604}
1605
1606static inline void skb_set_transport_header(struct sk_buff *skb,
1607 const int offset)
1608{
1609 skb_reset_transport_header(skb);
1610 skb->transport_header += offset;
1611}
1612
1613static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1614{
1615 return skb->head + skb->network_header;
1616}
1617
1618static inline void skb_reset_network_header(struct sk_buff *skb)
1619{
1620 skb->network_header = skb->data - skb->head;
1621}
1622
1623static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1624{
1625 skb_reset_network_header(skb);
1626 skb->network_header += offset;
1627}
1628
1629static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1630{
1631 return skb->head + skb->mac_header;
1632}
1633
1634static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1635{
Cong Wang35d04612013-05-29 15:16:05 +08001636 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001637}
1638
1639static inline void skb_reset_mac_header(struct sk_buff *skb)
1640{
1641 skb->mac_header = skb->data - skb->head;
1642}
1643
1644static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1645{
1646 skb_reset_mac_header(skb);
1647 skb->mac_header += offset;
1648}
1649
Ying Xuefbbdb8f2013-03-27 16:46:06 +00001650static inline void skb_probe_transport_header(struct sk_buff *skb,
1651 const int offset_hint)
1652{
1653 struct flow_keys keys;
1654
1655 if (skb_transport_header_was_set(skb))
1656 return;
1657 else if (skb_flow_dissect(skb, &keys))
1658 skb_set_transport_header(skb, keys.thoff);
1659 else
1660 skb_set_transport_header(skb, offset_hint);
1661}
1662
Eric Dumazet03606892012-02-23 10:55:02 +00001663static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1664{
1665 if (skb_mac_header_was_set(skb)) {
1666 const unsigned char *old_mac = skb_mac_header(skb);
1667
1668 skb_set_mac_header(skb, -skb->mac_len);
1669 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1670 }
1671}
1672
Michał Mirosław04fb4512010-12-14 15:24:08 +00001673static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1674{
1675 return skb->csum_start - skb_headroom(skb);
1676}
1677
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001678static inline int skb_transport_offset(const struct sk_buff *skb)
1679{
1680 return skb_transport_header(skb) - skb->data;
1681}
1682
1683static inline u32 skb_network_header_len(const struct sk_buff *skb)
1684{
1685 return skb->transport_header - skb->network_header;
1686}
1687
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001688static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1689{
1690 return skb->inner_transport_header - skb->inner_network_header;
1691}
1692
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001693static inline int skb_network_offset(const struct sk_buff *skb)
1694{
1695 return skb_network_header(skb) - skb->data;
1696}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001697
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001698static inline int skb_inner_network_offset(const struct sk_buff *skb)
1699{
1700 return skb_inner_network_header(skb) - skb->data;
1701}
1702
Changli Gaof9599ce2010-08-04 04:43:44 +00001703static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1704{
1705 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1706}
1707
Linus Torvalds1da177e2005-04-16 15:20:36 -07001708/*
1709 * CPUs often take a performance hit when accessing unaligned memory
1710 * locations. The actual performance hit varies, it can be small if the
1711 * hardware handles it or large if we have to take an exception and fix it
1712 * in software.
1713 *
1714 * Since an ethernet header is 14 bytes network drivers often end up with
1715 * the IP header at an unaligned offset. The IP header can be aligned by
1716 * shifting the start of the packet by 2 bytes. Drivers should do this
1717 * with:
1718 *
Tobias Klauser8660c122009-07-13 22:48:16 +00001719 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001720 *
1721 * The downside to this alignment of the IP header is that the DMA is now
1722 * unaligned. On some architectures the cost of an unaligned DMA is high
1723 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00001724 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001725 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1726 * to be overridden.
1727 */
1728#ifndef NET_IP_ALIGN
1729#define NET_IP_ALIGN 2
1730#endif
1731
Anton Blanchard025be812006-03-31 02:27:06 -08001732/*
1733 * The networking layer reserves some headroom in skb data (via
1734 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1735 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08001736 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08001737 *
1738 * Unfortunately this headroom changes the DMA alignment of the resulting
1739 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1740 * on some architectures. An architecture can override this value,
1741 * perhaps setting it to a cacheline in size (since that will maintain
1742 * cacheline alignment of the DMA). It must be a power of 2.
1743 *
David S. Millerd6301d32009-02-08 19:24:13 -08001744 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08001745 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07001746 *
1747 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1748 * to reduce average number of cache lines per packet.
1749 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07001750 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08001751 */
1752#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07001753#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08001754#endif
1755
Joe Perches7965bd42013-09-26 14:48:15 -07001756int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001757
1758static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1759{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00001760 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07001761 WARN_ON(1);
1762 return;
1763 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001764 skb->len = len;
1765 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001766}
1767
Joe Perches7965bd42013-09-26 14:48:15 -07001768void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001769
1770static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1771{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001772 if (skb->data_len)
1773 return ___pskb_trim(skb, len);
1774 __skb_trim(skb, len);
1775 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001776}
1777
1778static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1779{
1780 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1781}
1782
1783/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07001784 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1785 * @skb: buffer to alter
1786 * @len: new length
1787 *
1788 * This is identical to pskb_trim except that the caller knows that
1789 * the skb is not cloned so we should never get an error due to out-
1790 * of-memory.
1791 */
1792static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1793{
1794 int err = pskb_trim(skb, len);
1795 BUG_ON(err);
1796}
1797
1798/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001799 * skb_orphan - orphan a buffer
1800 * @skb: buffer to orphan
1801 *
1802 * If a buffer currently has an owner then we call the owner's
1803 * destructor function and make the @skb unowned. The buffer continues
1804 * to exist but is no longer charged to its former owner.
1805 */
1806static inline void skb_orphan(struct sk_buff *skb)
1807{
Eric Dumazetc34a7612013-07-30 16:11:15 -07001808 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001809 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07001810 skb->destructor = NULL;
1811 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07001812 } else {
1813 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07001814 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001815}
1816
1817/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00001818 * skb_orphan_frags - orphan the frags contained in a buffer
1819 * @skb: buffer to orphan frags from
1820 * @gfp_mask: allocation mask for replacement pages
1821 *
1822 * For each frag in the SKB which needs a destructor (i.e. has an
1823 * owner) create a copy of that frag and release the original
1824 * page by calling the destructor.
1825 */
1826static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1827{
1828 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1829 return 0;
1830 return skb_copy_ubufs(skb, gfp_mask);
1831}
1832
1833/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001834 * __skb_queue_purge - empty a list
1835 * @list: list to empty
1836 *
1837 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1838 * the list and one reference dropped. This function does not take the
1839 * list lock and the caller must hold the relevant locks to use it.
1840 */
Joe Perches7965bd42013-09-26 14:48:15 -07001841void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001842static inline void __skb_queue_purge(struct sk_buff_head *list)
1843{
1844 struct sk_buff *skb;
1845 while ((skb = __skb_dequeue(list)) != NULL)
1846 kfree_skb(skb);
1847}
1848
Alexander Duycke5e67302013-02-08 10:17:15 +00001849#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
1850#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
1851#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
1852
Joe Perches7965bd42013-09-26 14:48:15 -07001853void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001854
Joe Perches7965bd42013-09-26 14:48:15 -07001855struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
1856 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07001857
1858/**
1859 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1860 * @dev: network device to receive on
1861 * @length: length to allocate
1862 *
1863 * Allocate a new &sk_buff and assign it a usage count of one. The
1864 * buffer has unspecified headroom built in. Users should allocate
1865 * the headroom they think they need without accounting for the
1866 * built in space. The built in space is used for optimisations.
1867 *
1868 * %NULL is returned if there is no free memory. Although this function
1869 * allocates memory it can be called from an interrupt.
1870 */
1871static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00001872 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07001873{
1874 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1875}
1876
Eric Dumazet6f532612012-05-18 05:12:12 +00001877/* legacy helper around __netdev_alloc_skb() */
1878static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1879 gfp_t gfp_mask)
1880{
1881 return __netdev_alloc_skb(NULL, length, gfp_mask);
1882}
1883
1884/* legacy helper around netdev_alloc_skb() */
1885static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1886{
1887 return netdev_alloc_skb(NULL, length);
1888}
1889
1890
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001891static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
1892 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00001893{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001894 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00001895
1896 if (NET_IP_ALIGN && skb)
1897 skb_reserve(skb, NET_IP_ALIGN);
1898 return skb;
1899}
1900
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001901static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1902 unsigned int length)
1903{
1904 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
1905}
1906
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01001907/**
1908 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
Mel Gorman06140022012-07-31 16:44:24 -07001909 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1910 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1911 * @order: size of the allocation
1912 *
1913 * Allocate a new page.
1914 *
1915 * %NULL is returned if there is no free memory.
1916*/
1917static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
1918 struct sk_buff *skb,
1919 unsigned int order)
1920{
1921 struct page *page;
1922
1923 gfp_mask |= __GFP_COLD;
1924
1925 if (!(gfp_mask & __GFP_NOMEMALLOC))
1926 gfp_mask |= __GFP_MEMALLOC;
1927
1928 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
1929 if (skb && page && page->pfmemalloc)
1930 skb->pfmemalloc = true;
1931
1932 return page;
1933}
1934
1935/**
1936 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
1937 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1938 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1939 *
1940 * Allocate a new page.
1941 *
1942 * %NULL is returned if there is no free memory.
1943 */
1944static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
1945 struct sk_buff *skb)
1946{
1947 return __skb_alloc_pages(gfp_mask, skb, 0);
1948}
1949
1950/**
1951 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
1952 * @page: The page that was allocated from skb_alloc_page
1953 * @skb: The skb that may need pfmemalloc set
1954 */
1955static inline void skb_propagate_pfmemalloc(struct page *page,
1956 struct sk_buff *skb)
1957{
1958 if (page && page->pfmemalloc)
1959 skb->pfmemalloc = true;
1960}
1961
Eric Dumazet564824b2010-10-11 19:05:25 +00001962/**
Ian Campbell131ea662011-08-19 06:25:00 +00001963 * skb_frag_page - retrieve the page refered to by a paged fragment
1964 * @frag: the paged fragment
1965 *
1966 * Returns the &struct page associated with @frag.
1967 */
1968static inline struct page *skb_frag_page(const skb_frag_t *frag)
1969{
Ian Campbella8605c62011-10-19 23:01:49 +00001970 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00001971}
1972
1973/**
1974 * __skb_frag_ref - take an addition reference on a paged fragment.
1975 * @frag: the paged fragment
1976 *
1977 * Takes an additional reference on the paged fragment @frag.
1978 */
1979static inline void __skb_frag_ref(skb_frag_t *frag)
1980{
1981 get_page(skb_frag_page(frag));
1982}
1983
1984/**
1985 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
1986 * @skb: the buffer
1987 * @f: the fragment offset.
1988 *
1989 * Takes an additional reference on the @f'th paged fragment of @skb.
1990 */
1991static inline void skb_frag_ref(struct sk_buff *skb, int f)
1992{
1993 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
1994}
1995
1996/**
1997 * __skb_frag_unref - release a reference on a paged fragment.
1998 * @frag: the paged fragment
1999 *
2000 * Releases a reference on the paged fragment @frag.
2001 */
2002static inline void __skb_frag_unref(skb_frag_t *frag)
2003{
2004 put_page(skb_frag_page(frag));
2005}
2006
2007/**
2008 * skb_frag_unref - release a reference on a paged fragment of an skb.
2009 * @skb: the buffer
2010 * @f: the fragment offset
2011 *
2012 * Releases a reference on the @f'th paged fragment of @skb.
2013 */
2014static inline void skb_frag_unref(struct sk_buff *skb, int f)
2015{
2016 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2017}
2018
2019/**
2020 * skb_frag_address - gets the address of the data contained in a paged fragment
2021 * @frag: the paged fragment buffer
2022 *
2023 * Returns the address of the data within @frag. The page must already
2024 * be mapped.
2025 */
2026static inline void *skb_frag_address(const skb_frag_t *frag)
2027{
2028 return page_address(skb_frag_page(frag)) + frag->page_offset;
2029}
2030
2031/**
2032 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2033 * @frag: the paged fragment buffer
2034 *
2035 * Returns the address of the data within @frag. Checks that the page
2036 * is mapped and returns %NULL otherwise.
2037 */
2038static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2039{
2040 void *ptr = page_address(skb_frag_page(frag));
2041 if (unlikely(!ptr))
2042 return NULL;
2043
2044 return ptr + frag->page_offset;
2045}
2046
2047/**
2048 * __skb_frag_set_page - sets the page contained in a paged fragment
2049 * @frag: the paged fragment
2050 * @page: the page to set
2051 *
2052 * Sets the fragment @frag to contain @page.
2053 */
2054static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2055{
Ian Campbella8605c62011-10-19 23:01:49 +00002056 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002057}
2058
2059/**
2060 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2061 * @skb: the buffer
2062 * @f: the fragment offset
2063 * @page: the page to set
2064 *
2065 * Sets the @f'th fragment of @skb to contain @page.
2066 */
2067static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2068 struct page *page)
2069{
2070 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2071}
2072
Eric Dumazet400dfd32013-10-17 16:27:07 -07002073bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2074
Ian Campbell131ea662011-08-19 06:25:00 +00002075/**
2076 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002077 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002078 * @frag: the paged fragment to map
2079 * @offset: the offset within the fragment (starting at the
2080 * fragment's own offset)
2081 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002082 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00002083 *
2084 * Maps the page associated with @frag to @device.
2085 */
2086static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2087 const skb_frag_t *frag,
2088 size_t offset, size_t size,
2089 enum dma_data_direction dir)
2090{
2091 return dma_map_page(dev, skb_frag_page(frag),
2092 frag->page_offset + offset, size, dir);
2093}
2094
Eric Dumazet117632e2011-12-03 21:39:53 +00002095static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2096 gfp_t gfp_mask)
2097{
2098 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2099}
2100
Ian Campbell131ea662011-08-19 06:25:00 +00002101/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002102 * skb_clone_writable - is the header of a clone writable
2103 * @skb: buffer to check
2104 * @len: length up to which to write
2105 *
2106 * Returns true if modifying the header part of the cloned buffer
2107 * does not requires the data to be copied.
2108 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002109static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002110{
2111 return !skb_header_cloned(skb) &&
2112 skb_headroom(skb) + len <= skb->hdr_len;
2113}
2114
Herbert Xud9cc2042007-09-16 16:21:16 -07002115static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2116 int cloned)
2117{
2118 int delta = 0;
2119
Herbert Xud9cc2042007-09-16 16:21:16 -07002120 if (headroom > skb_headroom(skb))
2121 delta = headroom - skb_headroom(skb);
2122
2123 if (delta || cloned)
2124 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2125 GFP_ATOMIC);
2126 return 0;
2127}
2128
Patrick McHardy334a8132007-06-25 04:35:20 -07002129/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002130 * skb_cow - copy header of skb when it is required
2131 * @skb: buffer to cow
2132 * @headroom: needed headroom
2133 *
2134 * If the skb passed lacks sufficient headroom or its data part
2135 * is shared, data is reallocated. If reallocation fails, an error
2136 * is returned and original skb is not changed.
2137 *
2138 * The result is skb with writable area skb->head...skb->tail
2139 * and at least @headroom of space at head.
2140 */
2141static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2142{
Herbert Xud9cc2042007-09-16 16:21:16 -07002143 return __skb_cow(skb, headroom, skb_cloned(skb));
2144}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002145
Herbert Xud9cc2042007-09-16 16:21:16 -07002146/**
2147 * skb_cow_head - skb_cow but only making the head writable
2148 * @skb: buffer to cow
2149 * @headroom: needed headroom
2150 *
2151 * This function is identical to skb_cow except that we replace the
2152 * skb_cloned check by skb_header_cloned. It should be used when
2153 * you only need to push on some header and do not need to modify
2154 * the data.
2155 */
2156static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2157{
2158 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002159}
2160
2161/**
2162 * skb_padto - pad an skbuff up to a minimal size
2163 * @skb: buffer to pad
2164 * @len: minimal length
2165 *
2166 * Pads up a buffer to ensure the trailing bytes exist and are
2167 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002168 * is untouched. Otherwise it is extended. Returns zero on
2169 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002170 */
2171
Herbert Xu5b057c62006-06-23 02:06:41 -07002172static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002173{
2174 unsigned int size = skb->len;
2175 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002176 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002177 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002178}
2179
2180static inline int skb_add_data(struct sk_buff *skb,
2181 char __user *from, int copy)
2182{
2183 const int off = skb->len;
2184
2185 if (skb->ip_summed == CHECKSUM_NONE) {
2186 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08002187 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
Linus Torvalds1da177e2005-04-16 15:20:36 -07002188 copy, 0, &err);
2189 if (!err) {
2190 skb->csum = csum_block_add(skb->csum, csum, off);
2191 return 0;
2192 }
2193 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2194 return 0;
2195
2196 __skb_trim(skb, off);
2197 return -EFAULT;
2198}
2199
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002200static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2201 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002202{
2203 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002204 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002205
Ian Campbellea2ab692011-08-22 23:44:58 +00002206 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002207 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002208 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002209 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002210}
2211
Herbert Xu364c6ba2006-06-09 16:10:40 -07002212static inline int __skb_linearize(struct sk_buff *skb)
2213{
2214 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2215}
2216
Linus Torvalds1da177e2005-04-16 15:20:36 -07002217/**
2218 * skb_linearize - convert paged skb to linear one
2219 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002220 *
2221 * If there is no free memory -ENOMEM is returned, otherwise zero
2222 * is returned and the old skb data released.
2223 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002224static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002225{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002226 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2227}
2228
2229/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00002230 * skb_has_shared_frag - can any frag be overwritten
2231 * @skb: buffer to test
2232 *
2233 * Return true if the skb has at least one frag that might be modified
2234 * by an external entity (as in vmsplice()/sendfile())
2235 */
2236static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2237{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00002238 return skb_is_nonlinear(skb) &&
2239 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00002240}
2241
2242/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07002243 * skb_linearize_cow - make sure skb is linear and writable
2244 * @skb: buffer to process
2245 *
2246 * If there is no free memory -ENOMEM is returned, otherwise zero
2247 * is returned and the old skb data released.
2248 */
2249static inline int skb_linearize_cow(struct sk_buff *skb)
2250{
2251 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2252 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002253}
2254
2255/**
2256 * skb_postpull_rcsum - update checksum for received skb after pull
2257 * @skb: buffer to update
2258 * @start: start of data before pull
2259 * @len: length of data pulled
2260 *
2261 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002262 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2263 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002264 */
2265
2266static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002267 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002268{
Patrick McHardy84fa7932006-08-29 16:44:56 -07002269 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002270 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2271}
2272
Herbert Xucbb042f2006-03-20 22:43:56 -08002273unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2274
Linus Torvalds1da177e2005-04-16 15:20:36 -07002275/**
2276 * pskb_trim_rcsum - trim received skb and update checksum
2277 * @skb: buffer to trim
2278 * @len: new length
2279 *
2280 * This is exactly the same as pskb_trim except that it ensures the
2281 * checksum of received packets are still valid after the operation.
2282 */
2283
2284static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2285{
Stephen Hemminger0e4e4222005-09-08 12:32:03 -07002286 if (likely(len >= skb->len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002287 return 0;
Patrick McHardy84fa7932006-08-29 16:44:56 -07002288 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002289 skb->ip_summed = CHECKSUM_NONE;
2290 return __pskb_trim(skb, len);
2291}
2292
Linus Torvalds1da177e2005-04-16 15:20:36 -07002293#define skb_queue_walk(queue, skb) \
2294 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002295 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002296 skb = skb->next)
2297
James Chapman46f89142007-04-30 00:07:31 -07002298#define skb_queue_walk_safe(queue, skb, tmp) \
2299 for (skb = (queue)->next, tmp = skb->next; \
2300 skb != (struct sk_buff *)(queue); \
2301 skb = tmp, tmp = skb->next)
2302
David S. Miller1164f522008-09-23 00:49:44 -07002303#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002304 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002305 skb = skb->next)
2306
2307#define skb_queue_walk_from_safe(queue, skb, tmp) \
2308 for (tmp = skb->next; \
2309 skb != (struct sk_buff *)(queue); \
2310 skb = tmp, tmp = skb->next)
2311
Stephen Hemminger300ce172005-10-30 13:47:34 -08002312#define skb_queue_reverse_walk(queue, skb) \
2313 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002314 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08002315 skb = skb->prev)
2316
David S. Miller686a2952011-01-20 22:47:32 -08002317#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2318 for (skb = (queue)->prev, tmp = skb->prev; \
2319 skb != (struct sk_buff *)(queue); \
2320 skb = tmp, tmp = skb->prev)
2321
2322#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2323 for (tmp = skb->prev; \
2324 skb != (struct sk_buff *)(queue); \
2325 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002326
David S. Miller21dc3302010-08-23 00:13:46 -07002327static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07002328{
2329 return skb_shinfo(skb)->frag_list != NULL;
2330}
2331
2332static inline void skb_frag_list_init(struct sk_buff *skb)
2333{
2334 skb_shinfo(skb)->frag_list = NULL;
2335}
2336
2337static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2338{
2339 frag->next = skb_shinfo(skb)->frag_list;
2340 skb_shinfo(skb)->frag_list = frag;
2341}
2342
2343#define skb_walk_frags(skb, iter) \
2344 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2345
Joe Perches7965bd42013-09-26 14:48:15 -07002346struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2347 int *peeked, int *off, int *err);
2348struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2349 int *err);
2350unsigned int datagram_poll(struct file *file, struct socket *sock,
2351 struct poll_table_struct *wait);
2352int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2353 struct iovec *to, int size);
2354int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2355 struct iovec *iov);
2356int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2357 const struct iovec *from, int from_offset,
2358 int len);
2359int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2360 int offset, size_t count);
2361int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2362 const struct iovec *to, int to_offset,
2363 int size);
2364void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2365void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2366int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07002367int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2368int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2369__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2370 int len, __wsum csum);
2371int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2372 struct pipe_inode_info *pipe, unsigned int len,
2373 unsigned int flags);
2374void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
2375void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2376int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2377void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Joe Perches7965bd42013-09-26 14:48:15 -07002378struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002379
Daniel Borkmann2817a332013-10-30 11:50:51 +01002380struct skb_checksum_ops {
2381 __wsum (*update)(const void *mem, int len, __wsum wsum);
2382 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2383};
2384
2385__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2386 __wsum csum, const struct skb_checksum_ops *ops);
2387__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2388 __wsum csum);
2389
Linus Torvalds1da177e2005-04-16 15:20:36 -07002390static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2391 int len, void *buffer)
2392{
2393 int hlen = skb_headlen(skb);
2394
Patrick McHardy55820ee2005-07-05 14:08:10 -07002395 if (hlen - offset >= len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002396 return skb->data + offset;
2397
2398 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2399 return NULL;
2400
2401 return buffer;
2402}
2403
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03002404static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2405 void *to,
2406 const unsigned int len)
2407{
2408 memcpy(to, skb->data, len);
2409}
2410
2411static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2412 const int offset, void *to,
2413 const unsigned int len)
2414{
2415 memcpy(to, skb->data + offset, len);
2416}
2417
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03002418static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2419 const void *from,
2420 const unsigned int len)
2421{
2422 memcpy(skb->data, from, len);
2423}
2424
2425static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2426 const int offset,
2427 const void *from,
2428 const unsigned int len)
2429{
2430 memcpy(skb->data + offset, from, len);
2431}
2432
Joe Perches7965bd42013-09-26 14:48:15 -07002433void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002434
Patrick Ohlyac45f602009-02-12 05:03:37 +00002435static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2436{
2437 return skb->tstamp;
2438}
2439
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002440/**
2441 * skb_get_timestamp - get timestamp from a skb
2442 * @skb: skb to get stamp from
2443 * @stamp: pointer to struct timeval to store stamp in
2444 *
2445 * Timestamps are stored in the skb as offsets to a base timestamp.
2446 * This function converts the offset back to a struct timeval and stores
2447 * it in stamp.
2448 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00002449static inline void skb_get_timestamp(const struct sk_buff *skb,
2450 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002451{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002452 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002453}
2454
Patrick Ohlyac45f602009-02-12 05:03:37 +00002455static inline void skb_get_timestampns(const struct sk_buff *skb,
2456 struct timespec *stamp)
2457{
2458 *stamp = ktime_to_timespec(skb->tstamp);
2459}
2460
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002461static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002462{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002463 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002464}
2465
Stephen Hemminger164891a2007-04-23 22:26:16 -07002466static inline ktime_t net_timedelta(ktime_t t)
2467{
2468 return ktime_sub(ktime_get_real(), t);
2469}
2470
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07002471static inline ktime_t net_invalid_timestamp(void)
2472{
2473 return ktime_set(0, 0);
2474}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002475
Joe Perches7965bd42013-09-26 14:48:15 -07002476void skb_timestamping_init(void);
Richard Cochranc1f19b52010-07-17 08:49:36 +00002477
2478#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2479
Joe Perches7965bd42013-09-26 14:48:15 -07002480void skb_clone_tx_timestamp(struct sk_buff *skb);
2481bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00002482
2483#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2484
2485static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2486{
2487}
2488
2489static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2490{
2491 return false;
2492}
2493
2494#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2495
2496/**
2497 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2498 *
Richard Cochranda92b192011-10-21 00:49:15 +00002499 * PHY drivers may accept clones of transmitted packets for
2500 * timestamping via their phy_driver.txtstamp method. These drivers
2501 * must call this function to return the skb back to the stack, with
2502 * or without a timestamp.
2503 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00002504 * @skb: clone of the the original outgoing packet
Richard Cochranda92b192011-10-21 00:49:15 +00002505 * @hwtstamps: hardware time stamps, may be NULL if not available
Richard Cochranc1f19b52010-07-17 08:49:36 +00002506 *
2507 */
2508void skb_complete_tx_timestamp(struct sk_buff *skb,
2509 struct skb_shared_hwtstamps *hwtstamps);
2510
Patrick Ohlyac45f602009-02-12 05:03:37 +00002511/**
2512 * skb_tstamp_tx - queue clone of skb with send time stamps
2513 * @orig_skb: the original outgoing packet
2514 * @hwtstamps: hardware time stamps, may be NULL if not available
2515 *
2516 * If the skb has a socket associated, then this function clones the
2517 * skb (thus sharing the actual data and optional structures), stores
2518 * the optional hardware time stamping information (if non NULL) or
2519 * generates a software time stamp (otherwise), then queues the clone
2520 * to the error queue of the socket. Errors are silently ignored.
2521 */
Joe Perches7965bd42013-09-26 14:48:15 -07002522void skb_tstamp_tx(struct sk_buff *orig_skb,
2523 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00002524
Richard Cochran4507a712010-07-17 08:48:28 +00002525static inline void sw_tx_timestamp(struct sk_buff *skb)
2526{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002527 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2528 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00002529 skb_tstamp_tx(skb, NULL);
2530}
2531
2532/**
2533 * skb_tx_timestamp() - Driver hook for transmit timestamping
2534 *
2535 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00002536 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00002537 *
2538 * @skb: A socket buffer.
2539 */
2540static inline void skb_tx_timestamp(struct sk_buff *skb)
2541{
Richard Cochranc1f19b52010-07-17 08:49:36 +00002542 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002543 sw_tx_timestamp(skb);
2544}
2545
Johannes Berg6e3e9392011-11-09 10:15:42 +01002546/**
2547 * skb_complete_wifi_ack - deliver skb with wifi status
2548 *
2549 * @skb: the original outgoing packet
2550 * @acked: ack status
2551 *
2552 */
2553void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2554
Joe Perches7965bd42013-09-26 14:48:15 -07002555__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2556__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002557
Herbert Xu60476372007-04-09 11:59:39 -07002558static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2559{
2560 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2561}
2562
Herbert Xufb286bb2005-11-10 13:01:24 -08002563/**
2564 * skb_checksum_complete - Calculate checksum of an entire packet
2565 * @skb: packet to process
2566 *
2567 * This function calculates the checksum over the entire packet plus
2568 * the value of skb->csum. The latter can be used to supply the
2569 * checksum of a pseudo header as used by TCP/UDP. It returns the
2570 * checksum.
2571 *
2572 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2573 * this function can be used to verify that checksum on received
2574 * packets. In that case the function should return zero if the
2575 * checksum is correct. In particular, this function will return zero
2576 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2577 * hardware has already verified the correctness of the checksum.
2578 */
Al Viro4381ca32007-07-15 21:00:11 +01002579static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002580{
Herbert Xu60476372007-04-09 11:59:39 -07002581 return skb_csum_unnecessary(skb) ?
2582 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002583}
2584
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002585#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07002586void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002587static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2588{
2589 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002590 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002591}
2592static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2593{
2594 if (nfct)
2595 atomic_inc(&nfct->use);
2596}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002597#endif
2598#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -08002599static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
2600{
2601 if (skb)
2602 atomic_inc(&skb->users);
2603}
2604static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
2605{
2606 if (skb)
2607 kfree_skb(skb);
2608}
2609#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002610#ifdef CONFIG_BRIDGE_NETFILTER
2611static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2612{
2613 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2614 kfree(nf_bridge);
2615}
2616static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2617{
2618 if (nf_bridge)
2619 atomic_inc(&nf_bridge->use);
2620}
2621#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08002622static inline void nf_reset(struct sk_buff *skb)
2623{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002624#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08002625 nf_conntrack_put(skb->nfct);
2626 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002627#endif
2628#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Patrick McHardya193a4a2006-03-20 19:23:05 -08002629 nf_conntrack_put_reasm(skb->nfct_reasm);
2630 skb->nfct_reasm = NULL;
2631#endif
2632#ifdef CONFIG_BRIDGE_NETFILTER
2633 nf_bridge_put(skb->nf_bridge);
2634 skb->nf_bridge = NULL;
2635#endif
2636}
2637
Patrick McHardy124dff02013-04-05 20:42:05 +02002638static inline void nf_reset_trace(struct sk_buff *skb)
2639{
Gao feng130549fe2013-03-21 19:48:41 +00002640#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
2641 skb->nf_trace = 0;
2642#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002643}
2644
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002645/* Note: This doesn't put any conntrack and bridge info in dst. */
2646static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2647{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002648#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002649 dst->nfct = src->nfct;
2650 nf_conntrack_get(src->nfct);
2651 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002652#endif
2653#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002654 dst->nfct_reasm = src->nfct_reasm;
2655 nf_conntrack_get_reasm(src->nfct_reasm);
2656#endif
2657#ifdef CONFIG_BRIDGE_NETFILTER
2658 dst->nf_bridge = src->nf_bridge;
2659 nf_bridge_get(src->nf_bridge);
2660#endif
2661}
2662
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002663static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2664{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002665#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002666 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002667#endif
2668#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002669 nf_conntrack_put_reasm(dst->nfct_reasm);
2670#endif
2671#ifdef CONFIG_BRIDGE_NETFILTER
2672 nf_bridge_put(dst->nf_bridge);
2673#endif
2674 __nf_copy(dst, src);
2675}
2676
James Morris984bc162006-06-09 00:29:17 -07002677#ifdef CONFIG_NETWORK_SECMARK
2678static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2679{
2680 to->secmark = from->secmark;
2681}
2682
2683static inline void skb_init_secmark(struct sk_buff *skb)
2684{
2685 skb->secmark = 0;
2686}
2687#else
2688static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2689{ }
2690
2691static inline void skb_init_secmark(struct sk_buff *skb)
2692{ }
2693#endif
2694
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002695static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2696{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002697 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002698}
2699
Stephen Hemminger92477442009-03-21 13:39:26 -07002700static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002701{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002702 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002703}
2704
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002705static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2706{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002707 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002708}
2709
David S. Millerd5a9e242009-01-27 16:22:11 -08002710static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2711{
2712 skb->queue_mapping = rx_queue + 1;
2713}
2714
Stephen Hemminger92477442009-03-21 13:39:26 -07002715static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002716{
2717 return skb->queue_mapping - 1;
2718}
2719
Stephen Hemminger92477442009-03-21 13:39:26 -07002720static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002721{
Eric Dumazeta02cec22010-09-22 20:43:57 +00002722 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08002723}
2724
Joe Perches7965bd42013-09-26 14:48:15 -07002725u16 __skb_tx_hash(const struct net_device *dev, const struct sk_buff *skb,
2726 unsigned int num_tx_queues);
Stephen Hemminger92477442009-03-21 13:39:26 -07002727
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04002728static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2729{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002730#ifdef CONFIG_XFRM
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002731 return skb->sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002732#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002733 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002734#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04002735}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002736
Pravin B Shelar68c33162013-02-14 14:02:41 +00002737/* Keeps track of mac header offset relative to skb->head.
2738 * It is useful for TSO of Tunneling protocol. e.g. GRE.
2739 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07002740 * tunnel skb it points to outer mac header.
2741 * Keeps track of level of encapsulation of network headers.
2742 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00002743struct skb_gso_cb {
Eric Dumazet3347c962013-10-19 11:42:56 -07002744 int mac_offset;
2745 int encap_level;
Pravin B Shelar68c33162013-02-14 14:02:41 +00002746};
2747#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
2748
2749static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
2750{
2751 return (skb_mac_header(inner_skb) - inner_skb->head) -
2752 SKB_GSO_CB(inner_skb)->mac_offset;
2753}
2754
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00002755static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
2756{
2757 int new_headroom, headroom;
2758 int ret;
2759
2760 headroom = skb_headroom(skb);
2761 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
2762 if (ret)
2763 return ret;
2764
2765 new_headroom = skb_headroom(skb);
2766 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
2767 return 0;
2768}
2769
David S. Millerbdcc0922012-03-07 20:53:36 -05002770static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07002771{
2772 return skb_shinfo(skb)->gso_size;
2773}
2774
Eric Dumazet36a8f392013-09-29 01:21:32 -07002775/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05002776static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02002777{
2778 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2779}
2780
Joe Perches7965bd42013-09-26 14:48:15 -07002781void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07002782
2783static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2784{
2785 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2786 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002787 const struct skb_shared_info *shinfo = skb_shinfo(skb);
2788
Alexander Duyckb78462e2010-06-02 12:24:37 +00002789 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2790 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07002791 __skb_warn_lro_forwarding(skb);
2792 return true;
2793 }
2794 return false;
2795}
2796
Herbert Xu35fc92a2007-03-26 23:22:20 -07002797static inline void skb_forward_csum(struct sk_buff *skb)
2798{
2799 /* Unfortunately we don't support this one. Any brave souls? */
2800 if (skb->ip_summed == CHECKSUM_COMPLETE)
2801 skb->ip_summed = CHECKSUM_NONE;
2802}
2803
Eric Dumazetbc8acf22010-09-02 13:07:41 -07002804/**
2805 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2806 * @skb: skb to check
2807 *
2808 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2809 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2810 * use this helper, to document places where we make this assertion.
2811 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002812static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07002813{
2814#ifdef DEBUG
2815 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2816#endif
2817}
2818
Rusty Russellf35d9d82008-02-04 23:49:54 -05002819bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00002820
Daniel Borkmannf77668d2013-03-19 06:39:30 +00002821u32 __skb_get_poff(const struct sk_buff *skb);
2822
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00002823/**
2824 * skb_head_is_locked - Determine if the skb->head is locked down
2825 * @skb: skb to check
2826 *
2827 * The head on skbs build around a head frag can be removed if they are
2828 * not cloned. This function returns true if the skb head is locked down
2829 * due to either being allocated via kmalloc, or by being a clone with
2830 * multiple references to the head.
2831 */
2832static inline bool skb_head_is_locked(const struct sk_buff *skb)
2833{
2834 return !skb->head_frag || skb_cloned(skb);
2835}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002836#endif /* __KERNEL__ */
2837#endif /* _LINUX_SKBUFF_H */