blob: c93b5859a772b9dd58eab291892af34e1da8fd07 [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>
Eric Dumazet363ec39232014-02-26 14:02:11 -080035#include <linux/sched.h>
Jason Wang5203cd22013-03-26 23:11:21 +000036#include <net/flow_keys.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070037
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010038/* A. Checksumming of received packets by device.
39 *
40 * CHECKSUM_NONE:
41 *
42 * Device failed to checksum this packet e.g. due to lack of capabilities.
43 * The packet contains full (though not verified) checksum in packet but
44 * not in skb->csum. Thus, skb->csum is undefined in this case.
45 *
46 * CHECKSUM_UNNECESSARY:
47 *
48 * The hardware you're dealing with doesn't calculate the full checksum
49 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
Tom Herbert77cffe22014-08-27 21:26:46 -070050 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
51 * if their checksums are okay. skb->csum is still undefined in this case
52 * though. It is a bad option, but, unfortunately, nowadays most vendors do
53 * this. Apparently with the secret goal to sell you new devices, when you
54 * will add new protocol to your host, f.e. IPv6 8)
55 *
56 * CHECKSUM_UNNECESSARY is applicable to following protocols:
57 * TCP: IPv6 and IPv4.
58 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
59 * zero UDP checksum for either IPv4 or IPv6, the networking stack
60 * may perform further validation in this case.
61 * GRE: only if the checksum is present in the header.
62 * SCTP: indicates the CRC in SCTP header has been validated.
63 *
64 * skb->csum_level indicates the number of consecutive checksums found in
65 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
66 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
67 * and a device is able to verify the checksums for UDP (possibly zero),
68 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
69 * two. If the device were only able to verify the UDP checksum and not
70 * GRE, either because it doesn't support GRE checksum of because GRE
71 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
72 * not considered in this case).
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010073 *
74 * CHECKSUM_COMPLETE:
75 *
76 * This is the most generic way. The device supplied checksum of the _whole_
77 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
78 * hardware doesn't need to parse L3/L4 headers to implement this.
79 *
80 * Note: Even if device supports only some protocols, but is able to produce
81 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
82 *
83 * CHECKSUM_PARTIAL:
84 *
85 * This is identical to the case for output below. This may occur on a packet
86 * received directly from another Linux OS, e.g., a virtualized Linux kernel
87 * on the same host. The packet can be treated in the same way as
88 * CHECKSUM_UNNECESSARY, except that on output (i.e., forwarding) the
89 * checksum must be filled in by the OS or the hardware.
90 *
91 * B. Checksumming on output.
92 *
93 * CHECKSUM_NONE:
94 *
95 * The skb was already checksummed by the protocol, or a checksum is not
96 * required.
97 *
98 * CHECKSUM_PARTIAL:
99 *
100 * The device is required to checksum the packet as seen by hard_start_xmit()
101 * from skb->csum_start up to the end, and to record/write the checksum at
102 * offset skb->csum_start + skb->csum_offset.
103 *
104 * The device must show its capabilities in dev->features, set up at device
105 * setup time, e.g. netdev_features.h:
106 *
107 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
108 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
109 * IPv4. Sigh. Vendors like this way for an unknown reason.
110 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
111 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
112 * NETIF_F_... - Well, you get the picture.
113 *
114 * CHECKSUM_UNNECESSARY:
115 *
116 * Normally, the device will do per protocol specific checksumming. Protocol
117 * implementations that do not want the NIC to perform the checksum
118 * calculation should use this flag in their outgoing skbs.
119 *
120 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
121 * offload. Correspondingly, the FCoE protocol driver
122 * stack should use CHECKSUM_UNNECESSARY.
123 *
124 * Any questions? No questions, good. --ANK
125 */
126
Herbert Xu60476372007-04-09 11:59:39 -0700127/* Don't change this without changing skb_csum_unnecessary! */
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100128#define CHECKSUM_NONE 0
129#define CHECKSUM_UNNECESSARY 1
130#define CHECKSUM_COMPLETE 2
131#define CHECKSUM_PARTIAL 3
Linus Torvalds1da177e2005-04-16 15:20:36 -0700132
Tom Herbert77cffe22014-08-27 21:26:46 -0700133/* Maximum value in skb->csum_level */
134#define SKB_MAX_CSUM_LEVEL 3
135
Tobias Klauser0bec8c82014-07-22 12:06:23 +0200136#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
David S. Millerfc910a22007-03-25 20:27:59 -0700137#define SKB_WITH_OVERHEAD(X) \
Herbert Xudeea84b2007-10-21 16:27:46 -0700138 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
David S. Millerfc910a22007-03-25 20:27:59 -0700139#define SKB_MAX_ORDER(X, ORDER) \
140 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700141#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
142#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
143
Eric Dumazet87fb4b72011-10-13 07:28:54 +0000144/* return minimum truesize of one skb containing X bytes of data */
145#define SKB_TRUESIZE(X) ((X) + \
146 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
147 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
148
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149struct net_device;
David Howells716ea3a2007-04-02 20:19:53 -0700150struct scatterlist;
Jens Axboe9c55e012007-11-06 23:30:13 -0800151struct pipe_inode_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700152
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700153#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700154struct nf_conntrack {
155 atomic_t use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156};
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700157#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700158
159#ifdef CONFIG_BRIDGE_NETFILTER
160struct nf_bridge_info {
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000161 atomic_t use;
162 unsigned int mask;
163 struct net_device *physindev;
164 struct net_device *physoutdev;
165 unsigned long data[32 / sizeof(unsigned long)];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166};
167#endif
168
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169struct sk_buff_head {
170 /* These two members must be first. */
171 struct sk_buff *next;
172 struct sk_buff *prev;
173
174 __u32 qlen;
175 spinlock_t lock;
176};
177
178struct sk_buff;
179
Ian Campbell9d4dde52011-12-22 23:39:14 +0000180/* To allow 64K frame to be packed as single skb without frag_list we
181 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
182 * buffers which do not start on a page boundary.
183 *
184 * Since GRO uses frags we allocate at least 16 regardless of page
185 * size.
Anton Blancharda715dea2011-03-27 14:57:26 +0000186 */
Ian Campbell9d4dde52011-12-22 23:39:14 +0000187#if (65536/PAGE_SIZE + 1) < 16
David S. Millereec00952011-03-29 23:34:08 -0700188#define MAX_SKB_FRAGS 16UL
Anton Blancharda715dea2011-03-27 14:57:26 +0000189#else
Ian Campbell9d4dde52011-12-22 23:39:14 +0000190#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
Anton Blancharda715dea2011-03-27 14:57:26 +0000191#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192
193typedef struct skb_frag_struct skb_frag_t;
194
195struct skb_frag_struct {
Ian Campbella8605c62011-10-19 23:01:49 +0000196 struct {
197 struct page *p;
198 } page;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000199#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
David S. Millera309bb02007-07-30 18:47:03 -0700200 __u32 page_offset;
201 __u32 size;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000202#else
203 __u16 page_offset;
204 __u16 size;
205#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700206};
207
Eric Dumazet9e903e02011-10-18 21:00:24 +0000208static inline unsigned int skb_frag_size(const skb_frag_t *frag)
209{
210 return frag->size;
211}
212
213static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
214{
215 frag->size = size;
216}
217
218static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
219{
220 frag->size += delta;
221}
222
223static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
224{
225 frag->size -= delta;
226}
227
Patrick Ohlyac45f602009-02-12 05:03:37 +0000228#define HAVE_HW_TIME_STAMP
229
230/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800231 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000232 * @hwtstamp: hardware time stamp transformed into duration
233 * since arbitrary point in time
Patrick Ohlyac45f602009-02-12 05:03:37 +0000234 *
235 * Software time stamps generated by ktime_get_real() are stored in
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400236 * skb->tstamp.
Patrick Ohlyac45f602009-02-12 05:03:37 +0000237 *
238 * hwtstamps can only be compared against other hwtstamps from
239 * the same device.
240 *
241 * This structure is attached to packets as part of the
242 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
243 */
244struct skb_shared_hwtstamps {
245 ktime_t hwtstamp;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000246};
247
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000248/* Definitions for tx_flags in struct skb_shared_info */
249enum {
250 /* generate hardware time stamp */
251 SKBTX_HW_TSTAMP = 1 << 0,
252
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400253 /* generate software time stamp when queueing packet to NIC */
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000254 SKBTX_SW_TSTAMP = 1 << 1,
255
256 /* device driver is going to provide hardware time stamp */
257 SKBTX_IN_PROGRESS = 1 << 2,
258
Shirley Maa6686f22011-07-06 12:22:12 +0000259 /* device driver supports TX zero-copy buffers */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000260 SKBTX_DEV_ZEROCOPY = 1 << 3,
Johannes Berg6e3e9392011-11-09 10:15:42 +0100261
262 /* generate wifi status information (where possible) */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000263 SKBTX_WIFI_STATUS = 1 << 4,
Pravin B Shelarc9af6db2013-02-11 09:27:41 +0000264
265 /* This indicates at least one fragment might be overwritten
266 * (as in vmsplice(), sendfile() ...)
267 * If we need to compute a TX checksum, we'll need to copy
268 * all frags to avoid possible bad checksum
269 */
270 SKBTX_SHARED_FRAG = 1 << 5,
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400271
272 /* generate software time stamp when entering packet scheduling */
273 SKBTX_SCHED_TSTAMP = 1 << 6,
Willem de Bruijne1c8a602014-08-04 22:11:50 -0400274
275 /* generate software timestamp on peer data acknowledgment */
276 SKBTX_ACK_TSTAMP = 1 << 7,
Shirley Maa6686f22011-07-06 12:22:12 +0000277};
278
Willem de Bruijne1c8a602014-08-04 22:11:50 -0400279#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
280 SKBTX_SCHED_TSTAMP | \
281 SKBTX_ACK_TSTAMP)
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400282#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
283
Shirley Maa6686f22011-07-06 12:22:12 +0000284/*
285 * The callback notifies userspace to release buffers when skb DMA is done in
286 * lower device, the skb last reference should be 0 when calling this.
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000287 * The zerocopy_success argument is true if zero copy transmit occurred,
288 * false on data copy or out of memory error caused by data copy attempt.
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000289 * The ctx field is used to track device context.
290 * The desc field is used to track userspace buffer index.
Shirley Maa6686f22011-07-06 12:22:12 +0000291 */
292struct ubuf_info {
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000293 void (*callback)(struct ubuf_info *, bool zerocopy_success);
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000294 void *ctx;
Shirley Maa6686f22011-07-06 12:22:12 +0000295 unsigned long desc;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000296};
297
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298/* This data is invariant across clones and lives at
299 * the end of the header data, ie. at skb->end.
300 */
301struct skb_shared_info {
Ian Campbell9f42f122012-01-05 07:13:39 +0000302 unsigned char nr_frags;
303 __u8 tx_flags;
Herbert Xu79671682006-06-22 02:40:14 -0700304 unsigned short gso_size;
305 /* Warning: this field is not always filled in (UFO)! */
306 unsigned short gso_segs;
307 unsigned short gso_type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000309 struct skb_shared_hwtstamps hwtstamps;
Willem de Bruijn09c2d252014-08-04 22:11:47 -0400310 u32 tskey;
Ian Campbell9f42f122012-01-05 07:13:39 +0000311 __be32 ip6_frag_id;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700312
313 /*
314 * Warning : all fields before dataref are cleared in __alloc_skb()
315 */
316 atomic_t dataref;
317
Johann Baudy69e3c752009-05-18 22:11:22 -0700318 /* Intermediate layers must ensure that destructor_arg
319 * remains valid until skb destructor */
320 void * destructor_arg;
Shirley Maa6686f22011-07-06 12:22:12 +0000321
Eric Dumazetfed66382010-07-22 19:09:08 +0000322 /* must be last field, see pskb_expand_head() */
323 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324};
325
326/* We divide dataref into two halves. The higher 16 bits hold references
327 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700328 * the entire skb->data. A clone of a headerless skb holds the length of
329 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700330 *
331 * All users must obey the rule that the skb->data reference count must be
332 * greater than or equal to the payload reference count.
333 *
334 * Holding a reference to the payload part means that the user does not
335 * care about modifications to the header part of skb->data.
336 */
337#define SKB_DATAREF_SHIFT 16
338#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
339
David S. Millerd179cd12005-08-17 14:57:30 -0700340
341enum {
342 SKB_FCLONE_UNAVAILABLE,
343 SKB_FCLONE_ORIG,
344 SKB_FCLONE_CLONE,
345};
346
Herbert Xu79671682006-06-22 02:40:14 -0700347enum {
348 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700349 SKB_GSO_UDP = 1 << 1,
Herbert Xu576a30e2006-06-27 13:22:38 -0700350
351 /* This indicates the skb is from an untrusted source. */
352 SKB_GSO_DODGY = 1 << 2,
Michael Chanb0da85372006-06-29 12:30:00 -0700353
354 /* This indicates the tcp segment has CWR set. */
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700355 SKB_GSO_TCP_ECN = 1 << 3,
356
357 SKB_GSO_TCPV6 = 1 << 4,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800358
359 SKB_GSO_FCOE = 1 << 5,
Pravin B Shelar68c33162013-02-14 14:02:41 +0000360
361 SKB_GSO_GRE = 1 << 6,
Pravin B Shelar73136262013-03-07 13:21:51 +0000362
Tom Herbert4b282522014-06-14 23:23:52 -0700363 SKB_GSO_GRE_CSUM = 1 << 7,
Simon Horman0d89d202013-05-23 21:02:52 +0000364
Tom Herbert4b282522014-06-14 23:23:52 -0700365 SKB_GSO_IPIP = 1 << 8,
Eric Dumazetcb32f512013-10-19 11:42:57 -0700366
Tom Herbert4b282522014-06-14 23:23:52 -0700367 SKB_GSO_SIT = 1 << 9,
Eric Dumazet61c1db72013-10-20 20:47:30 -0700368
Tom Herbert4b282522014-06-14 23:23:52 -0700369 SKB_GSO_UDP_TUNNEL = 1 << 10,
Tom Herbert0f4f4ff2014-06-04 17:20:16 -0700370
371 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
Tom Herbert4749c092014-06-04 17:20:23 -0700372
Tom Herbert4b282522014-06-14 23:23:52 -0700373 SKB_GSO_MPLS = 1 << 12,
374
Herbert Xu79671682006-06-22 02:40:14 -0700375};
376
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700377#if BITS_PER_LONG > 32
378#define NET_SKBUFF_DATA_USES_OFFSET 1
379#endif
380
381#ifdef NET_SKBUFF_DATA_USES_OFFSET
382typedef unsigned int sk_buff_data_t;
383#else
384typedef unsigned char *sk_buff_data_t;
385#endif
386
Eric Dumazet363ec39232014-02-26 14:02:11 -0800387/**
388 * struct skb_mstamp - multi resolution time stamps
389 * @stamp_us: timestamp in us resolution
390 * @stamp_jiffies: timestamp in jiffies
391 */
392struct skb_mstamp {
393 union {
394 u64 v64;
395 struct {
396 u32 stamp_us;
397 u32 stamp_jiffies;
398 };
399 };
400};
401
402/**
403 * skb_mstamp_get - get current timestamp
404 * @cl: place to store timestamps
405 */
406static inline void skb_mstamp_get(struct skb_mstamp *cl)
407{
408 u64 val = local_clock();
409
410 do_div(val, NSEC_PER_USEC);
411 cl->stamp_us = (u32)val;
412 cl->stamp_jiffies = (u32)jiffies;
413}
414
415/**
416 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
417 * @t1: pointer to newest sample
418 * @t0: pointer to oldest sample
419 */
420static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
421 const struct skb_mstamp *t0)
422{
423 s32 delta_us = t1->stamp_us - t0->stamp_us;
424 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
425
426 /* If delta_us is negative, this might be because interval is too big,
427 * or local_clock() drift is too big : fallback using jiffies.
428 */
429 if (delta_us <= 0 ||
430 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
431
432 delta_us = jiffies_to_usecs(delta_jiffies);
433
434 return delta_us;
435}
436
437
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438/**
439 * struct sk_buff - socket buffer
440 * @next: Next buffer in list
441 * @prev: Previous buffer in list
Eric Dumazet363ec39232014-02-26 14:02:11 -0800442 * @tstamp: Time we arrived/left
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700443 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700445 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000446 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700447 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448 * @len: Length of actual data
449 * @data_len: Data length
450 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700451 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700452 * @csum: Checksum (must include start/offset pair)
453 * @csum_start: Offset from skb->head where checksumming should start
454 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700455 * @priority: Packet queueing priority
WANG Cong60ff7462014-05-04 16:39:18 -0700456 * @ignore_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700458 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459 * @nohdr: Payload reference only, must not modify header
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700460 * @nfctinfo: Relationship of this skb to the connection
Linus Torvalds1da177e2005-04-16 15:20:36 -0700461 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700462 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700463 * @ipvs_property: skbuff is owned by ipvs
Randy Dunlap31729362008-02-18 20:52:13 -0800464 * @peeked: this packet has been seen already, so stats have been
465 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700466 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700467 * @protocol: Packet protocol from driver
468 * @destructor: Destruct function
469 * @nfct: Associated connection, if any
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800471 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472 * @tc_index: Traffic control index
473 * @tc_verd: traffic control verdict
Tom Herbert61b905d2014-03-24 15:34:47 -0700474 * @hash: the packet hash
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700475 * @queue_mapping: Queue mapping for multiqueue devices
David S. Miller0b725a22014-08-25 15:51:53 -0700476 * @xmit_more: More SKBs are pending for this queue
Randy Dunlap553a5672008-04-20 10:51:01 -0700477 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700478 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Tom Herbert61b905d2014-03-24 15:34:47 -0700479 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
Changli Gao4ca24622011-08-19 07:26:44 -0700480 * ports.
Tom Herberta3b18dd2014-07-01 21:33:17 -0700481 * @sw_hash: indicates hash was computed in software stack
Johannes Berg6e3e9392011-11-09 10:15:42 +0100482 * @wifi_acked_valid: wifi_acked was set
483 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000484 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Randy Dunlapf4b8ea72006-06-22 16:00:11 -0700485 * @dma_cookie: a cookie to one of several possible DMA operations
486 * done by skb DMA functions
Eliezer Tamir06021292013-06-10 11:39:50 +0300487 * @napi_id: id of the NAPI struct this skb came from
James Morris984bc162006-06-09 00:29:17 -0700488 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700489 * @mark: Generic packet mark
490 * @dropcount: total number of sk_receive_queue overflows
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000491 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b02008-07-14 22:49:06 -0700492 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000493 * @inner_protocol: Protocol (encapsulation)
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000494 * @inner_transport_header: Inner transport layer header (encapsulation)
495 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000496 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700497 * @transport_header: Transport layer header
498 * @network_header: Network layer header
499 * @mac_header: Link layer header
500 * @tail: Tail pointer
501 * @end: End pointer
502 * @head: Head of buffer
503 * @data: Data head pointer
504 * @truesize: Buffer size
505 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506 */
507
508struct sk_buff {
509 /* These two members must be first. */
510 struct sk_buff *next;
511 struct sk_buff *prev;
512
Eric Dumazet363ec39232014-02-26 14:02:11 -0800513 union {
514 ktime_t tstamp;
515 struct skb_mstamp skb_mstamp;
516 };
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000517
518 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520
Linus Torvalds1da177e2005-04-16 15:20:36 -0700521 /*
522 * This is the control buffer. It is free to use for every
523 * layer. Please put your private variables there. If you
524 * want to keep them across layers you have to do a skb_clone()
525 * first. This is owned by whoever has the skb queued ATM.
526 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000527 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528
Eric Dumazet7fee2262010-05-11 23:19:48 +0000529 unsigned long _skb_refdst;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000530#ifdef CONFIG_XFRM
531 struct sec_path *sp;
532#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700534 data_len;
535 __u16 mac_len,
536 hdr_len;
Al Viroff1dcad2006-11-20 18:07:29 -0800537 union {
538 __wsum csum;
Herbert Xu663ead32007-04-09 11:59:07 -0700539 struct {
540 __u16 csum_start;
541 __u16 csum_offset;
542 };
Al Viroff1dcad2006-11-20 18:07:29 -0800543 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544 __u32 priority;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200545 kmemcheck_bitfield_begin(flags1);
WANG Cong60ff7462014-05-04 16:39:18 -0700546 __u8 ignore_df:1,
Thomas Graf1cbb3382005-07-05 14:13:41 -0700547 cloned:1,
548 ip_summed:2,
Harald Welte6869c4d2005-08-09 19:24:19 -0700549 nohdr:1,
550 nfctinfo:3;
David S. Millerd179cd12005-08-17 14:57:30 -0700551 __u8 pkt_type:3,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800552 fclone:2,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700553 ipvs_property:1,
Herbert Xua59322b2007-12-05 01:53:40 -0800554 peeked:1,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700555 nf_trace:1;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200556 kmemcheck_bitfield_end(flags1);
Eric Dumazet4ab408d2010-03-01 03:09:26 +0000557 __be16 protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558
559 void (*destructor)(struct sk_buff *skb);
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800560#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700561 struct nf_conntrack *nfct;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100562#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700563#ifdef CONFIG_BRIDGE_NETFILTER
564 struct nf_bridge_info *nf_bridge;
565#endif
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700566
Eric Dumazet8964be42009-11-20 15:35:04 -0800567 int skb_iif;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000568
Tom Herbert61b905d2014-03-24 15:34:47 -0700569 __u32 hash;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000570
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000571 __be16 vlan_proto;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000572 __u16 vlan_tci;
573
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700575 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700577 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700579#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200580
Eric Dumazet0a148422011-04-20 09:27:32 +0000581 __u16 queue_mapping;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200582 kmemcheck_bitfield_begin(flags2);
David S. Miller0b725a22014-08-25 15:51:53 -0700583 __u8 xmit_more:1;
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400584#ifdef CONFIG_IPV6_NDISC_NODETYPE
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000585 __u8 ndisc_nodetype:2;
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400586#endif
Mel Gormanc93bdd02012-07-31 16:44:19 -0700587 __u8 pfmemalloc:1;
Tom Herbert3853b582010-11-21 13:17:29 +0000588 __u8 ooo_okay:1;
Tom Herbert61b905d2014-03-24 15:34:47 -0700589 __u8 l4_hash:1;
Tom Herberta3b18dd2014-07-01 21:33:17 -0700590 __u8 sw_hash:1;
Johannes Berg6e3e9392011-11-09 10:15:42 +0100591 __u8 wifi_acked_valid:1;
592 __u8 wifi_acked:1;
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000593 __u8 no_fcs:1;
Eric Dumazetd3836f22012-04-27 00:33:38 +0000594 __u8 head_frag:1;
Tom Herbert77cffe22014-08-27 21:26:46 -0700595 /* Indicates the inner headers are valid in the skbuff. */
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000596 __u8 encapsulation:1;
Tom Herbert7e2b10c2014-06-04 17:20:02 -0700597 __u8 encap_hdr_csum:1;
Tom Herbert5d0c2b92014-06-10 18:54:13 -0700598 __u8 csum_valid:1;
Tom Herbert7e3cead2014-06-10 18:54:19 -0700599 __u8 csum_complete_sw:1;
David S. Miller5c214032014-08-27 14:39:04 -0700600 /* 1/3 bit hole (depending on ndisc_nodetype presence) */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200601 kmemcheck_bitfield_end(flags2);
602
Cong Wange0d10952013-08-01 11:10:25 +0800603#if defined CONFIG_NET_DMA || defined CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir06021292013-06-10 11:39:50 +0300604 union {
605 unsigned int napi_id;
606 dma_cookie_t dma_cookie;
607 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700608#endif
James Morris984bc162006-06-09 00:29:17 -0700609#ifdef CONFIG_NETWORK_SECMARK
610 __u32 secmark;
611#endif
Neil Horman3b885782009-10-12 13:26:31 -0700612 union {
613 __u32 mark;
614 __u32 dropcount;
Eric Dumazet16fad692013-03-14 05:40:32 +0000615 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700616 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617
Tom Herbertde20fe82014-08-27 21:26:35 -0700618 kmemcheck_bitfield_begin(flags3);
Tom Herbert77cffe22014-08-27 21:26:46 -0700619 __u8 csum_level:2;
620 /* 14 bit hole */
Tom Herbertde20fe82014-08-27 21:26:35 -0700621 kmemcheck_bitfield_end(flags3);
622
Simon Horman0d89d202013-05-23 21:02:52 +0000623 __be16 inner_protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000624 __u16 inner_transport_header;
625 __u16 inner_network_header;
626 __u16 inner_mac_header;
627 __u16 transport_header;
628 __u16 network_header;
629 __u16 mac_header;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700631 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700632 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700634 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700635 unsigned int truesize;
636 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700637};
638
639#ifdef __KERNEL__
640/*
641 * Handling routines are only of interest to the kernel
642 */
643#include <linux/slab.h>
644
Linus Torvalds1da177e2005-04-16 15:20:36 -0700645
Mel Gormanc93bdd02012-07-31 16:44:19 -0700646#define SKB_ALLOC_FCLONE 0x01
647#define SKB_ALLOC_RX 0x02
648
649/* Returns true if the skb was allocated from PFMEMALLOC reserves */
650static inline bool skb_pfmemalloc(const struct sk_buff *skb)
651{
652 return unlikely(skb->pfmemalloc);
653}
654
Eric Dumazet7fee2262010-05-11 23:19:48 +0000655/*
656 * skb might have a dst pointer attached, refcounted or not.
657 * _skb_refdst low order bit is set if refcount was _not_ taken
658 */
659#define SKB_DST_NOREF 1UL
660#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
661
662/**
663 * skb_dst - returns skb dst_entry
664 * @skb: buffer
665 *
666 * Returns skb dst_entry, regardless of reference taken or not.
667 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000668static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
669{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000670 /* If refdst was not refcounted, check we still are in a
671 * rcu_read_lock section
672 */
673 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
674 !rcu_read_lock_held() &&
675 !rcu_read_lock_bh_held());
676 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000677}
678
Eric Dumazet7fee2262010-05-11 23:19:48 +0000679/**
680 * skb_dst_set - sets skb dst
681 * @skb: buffer
682 * @dst: dst entry
683 *
684 * Sets skb dst, assuming a reference was taken on dst and should
685 * be released by skb_dst_drop()
686 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000687static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
688{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000689 skb->_skb_refdst = (unsigned long)dst;
690}
691
Joe Perches7965bd42013-09-26 14:48:15 -0700692void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
693 bool force);
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200694
695/**
696 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
697 * @skb: buffer
698 * @dst: dst entry
699 *
700 * Sets skb dst, assuming a reference was not taken on dst.
701 * If dst entry is cached, we do not take reference and dst_release
702 * will be avoided by refdst_drop. If dst entry is not cached, we take
703 * reference, so that last dst_release can destroy the dst immediately.
704 */
705static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
706{
707 __skb_dst_set_noref(skb, dst, false);
708}
709
710/**
711 * skb_dst_set_noref_force - sets skb dst, without taking reference
712 * @skb: buffer
713 * @dst: dst entry
714 *
715 * Sets skb dst, assuming a reference was not taken on dst.
716 * No reference is taken and no dst_release will be called. While for
717 * cached dsts deferred reclaim is a basic feature, for entries that are
718 * not cached it is caller's job to guarantee that last dst_release for
719 * provided dst happens when nobody uses it, eg. after a RCU grace period.
720 */
721static inline void skb_dst_set_noref_force(struct sk_buff *skb,
722 struct dst_entry *dst)
723{
724 __skb_dst_set_noref(skb, dst, true);
725}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000726
727/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300728 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000729 * @skb: buffer
730 */
731static inline bool skb_dst_is_noref(const struct sk_buff *skb)
732{
733 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000734}
735
Eric Dumazet511c3f92009-06-02 05:14:27 +0000736static inline struct rtable *skb_rtable(const struct sk_buff *skb)
737{
Eric Dumazetadf30902009-06-02 05:19:30 +0000738 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000739}
740
Joe Perches7965bd42013-09-26 14:48:15 -0700741void kfree_skb(struct sk_buff *skb);
742void kfree_skb_list(struct sk_buff *segs);
743void skb_tx_error(struct sk_buff *skb);
744void consume_skb(struct sk_buff *skb);
745void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000746extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000747
Joe Perches7965bd42013-09-26 14:48:15 -0700748void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
749bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
750 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000751
Joe Perches7965bd42013-09-26 14:48:15 -0700752struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
753 int node);
754struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700755static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100756 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700757{
Eric Dumazet564824b2010-10-11 19:05:25 +0000758 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700759}
760
761static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100762 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700763{
Mel Gormanc93bdd02012-07-31 16:44:19 -0700764 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700765}
766
Joe Perches7965bd42013-09-26 14:48:15 -0700767struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
Patrick McHardy0ebd0ac2013-04-17 06:46:58 +0000768static inline struct sk_buff *alloc_skb_head(gfp_t priority)
769{
770 return __alloc_skb_head(priority, -1);
771}
772
Joe Perches7965bd42013-09-26 14:48:15 -0700773struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
774int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
775struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
776struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
Octavian Purdilabad93e92014-06-12 01:36:26 +0300777struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
778 gfp_t gfp_mask, bool fclone);
779static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
780 gfp_t gfp_mask)
781{
782 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
783}
Eric Dumazet117632e2011-12-03 21:39:53 +0000784
Joe Perches7965bd42013-09-26 14:48:15 -0700785int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
786struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
787 unsigned int headroom);
788struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
789 int newtailroom, gfp_t priority);
Fan Du25a91d82014-01-18 09:54:23 +0800790int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
791 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -0700792int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
793 int len);
794int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
795int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000796#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700797
Joe Perches7965bd42013-09-26 14:48:15 -0700798int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
799 int getfrag(void *from, char *to, int offset,
800 int len, int odd, struct sk_buff *skb),
801 void *from, int length);
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700802
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800803struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700804 __u32 lower_offset;
805 __u32 upper_offset;
806 __u32 frag_idx;
807 __u32 stepped_offset;
808 struct sk_buff *root_skb;
809 struct sk_buff *cur_skb;
810 __u8 *frag_data;
811};
812
Joe Perches7965bd42013-09-26 14:48:15 -0700813void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
814 unsigned int to, struct skb_seq_state *st);
815unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
816 struct skb_seq_state *st);
817void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -0700818
Joe Perches7965bd42013-09-26 14:48:15 -0700819unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
820 unsigned int to, struct ts_config *config,
821 struct ts_state *state);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700822
Tom Herbert09323cc2013-12-15 22:16:19 -0800823/*
824 * Packet hash types specify the type of hash in skb_set_hash.
825 *
826 * Hash types refer to the protocol layer addresses which are used to
827 * construct a packet's hash. The hashes are used to differentiate or identify
828 * flows of the protocol layer for the hash type. Hash types are either
829 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
830 *
831 * Properties of hashes:
832 *
833 * 1) Two packets in different flows have different hash values
834 * 2) Two packets in the same flow should have the same hash value
835 *
836 * A hash at a higher layer is considered to be more specific. A driver should
837 * set the most specific hash possible.
838 *
839 * A driver cannot indicate a more specific hash than the layer at which a hash
840 * was computed. For instance an L3 hash cannot be set as an L4 hash.
841 *
842 * A driver may indicate a hash level which is less specific than the
843 * actual layer the hash was computed on. For instance, a hash computed
844 * at L4 may be considered an L3 hash. This should only be done if the
845 * driver can't unambiguously determine that the HW computed the hash at
846 * the higher layer. Note that the "should" in the second property above
847 * permits this.
848 */
849enum pkt_hash_types {
850 PKT_HASH_TYPE_NONE, /* Undefined type */
851 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
852 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
853 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
854};
855
856static inline void
857skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
858{
Tom Herbert61b905d2014-03-24 15:34:47 -0700859 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
Tom Herberta3b18dd2014-07-01 21:33:17 -0700860 skb->sw_hash = 0;
Tom Herbert61b905d2014-03-24 15:34:47 -0700861 skb->hash = hash;
Tom Herbert09323cc2013-12-15 22:16:19 -0800862}
863
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800864void __skb_get_hash(struct sk_buff *skb);
865static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000866{
Tom Herberta3b18dd2014-07-01 21:33:17 -0700867 if (!skb->l4_hash && !skb->sw_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800868 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000869
Tom Herbert61b905d2014-03-24 15:34:47 -0700870 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000871}
872
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800873static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
874{
Tom Herbert61b905d2014-03-24 15:34:47 -0700875 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800876}
877
Tom Herbert7539fad2013-12-15 22:12:18 -0800878static inline void skb_clear_hash(struct sk_buff *skb)
879{
Tom Herbert61b905d2014-03-24 15:34:47 -0700880 skb->hash = 0;
Tom Herberta3b18dd2014-07-01 21:33:17 -0700881 skb->sw_hash = 0;
Tom Herbert61b905d2014-03-24 15:34:47 -0700882 skb->l4_hash = 0;
Tom Herbert7539fad2013-12-15 22:12:18 -0800883}
884
885static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
886{
Tom Herbert61b905d2014-03-24 15:34:47 -0700887 if (!skb->l4_hash)
Tom Herbert7539fad2013-12-15 22:12:18 -0800888 skb_clear_hash(skb);
889}
890
Tom Herbert3df7a742013-12-15 22:16:29 -0800891static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
892{
Tom Herbert61b905d2014-03-24 15:34:47 -0700893 to->hash = from->hash;
Tom Herberta3b18dd2014-07-01 21:33:17 -0700894 to->sw_hash = from->sw_hash;
Tom Herbert61b905d2014-03-24 15:34:47 -0700895 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -0800896};
897
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700898#ifdef NET_SKBUFF_DATA_USES_OFFSET
899static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
900{
901 return skb->head + skb->end;
902}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000903
904static inline unsigned int skb_end_offset(const struct sk_buff *skb)
905{
906 return skb->end;
907}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700908#else
909static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
910{
911 return skb->end;
912}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000913
914static inline unsigned int skb_end_offset(const struct sk_buff *skb)
915{
916 return skb->end - skb->head;
917}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700918#endif
919
Linus Torvalds1da177e2005-04-16 15:20:36 -0700920/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700921#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700922
Patrick Ohlyac45f602009-02-12 05:03:37 +0000923static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
924{
925 return &skb_shinfo(skb)->hwtstamps;
926}
927
Linus Torvalds1da177e2005-04-16 15:20:36 -0700928/**
929 * skb_queue_empty - check if a queue is empty
930 * @list: queue head
931 *
932 * Returns true if the queue is empty, false otherwise.
933 */
934static inline int skb_queue_empty(const struct sk_buff_head *list)
935{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100936 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700937}
938
939/**
David S. Millerfc7ebb22008-09-23 00:34:07 -0700940 * skb_queue_is_last - check if skb is the last entry in the queue
941 * @list: queue head
942 * @skb: buffer
943 *
944 * Returns true if @skb is the last buffer on the list.
945 */
946static inline bool skb_queue_is_last(const struct sk_buff_head *list,
947 const struct sk_buff *skb)
948{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100949 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -0700950}
951
952/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800953 * skb_queue_is_first - check if skb is the first entry in the queue
954 * @list: queue head
955 * @skb: buffer
956 *
957 * Returns true if @skb is the first buffer on the list.
958 */
959static inline bool skb_queue_is_first(const struct sk_buff_head *list,
960 const struct sk_buff *skb)
961{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100962 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800963}
964
965/**
David S. Miller249c8b42008-09-23 00:44:42 -0700966 * skb_queue_next - return the next packet in the queue
967 * @list: queue head
968 * @skb: current buffer
969 *
970 * Return the next packet in @list after @skb. It is only valid to
971 * call this if skb_queue_is_last() evaluates to false.
972 */
973static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
974 const struct sk_buff *skb)
975{
976 /* This BUG_ON may seem severe, but if we just return then we
977 * are going to dereference garbage.
978 */
979 BUG_ON(skb_queue_is_last(list, skb));
980 return skb->next;
981}
982
983/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800984 * skb_queue_prev - return the prev packet in the queue
985 * @list: queue head
986 * @skb: current buffer
987 *
988 * Return the prev packet in @list before @skb. It is only valid to
989 * call this if skb_queue_is_first() evaluates to false.
990 */
991static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
992 const struct sk_buff *skb)
993{
994 /* This BUG_ON may seem severe, but if we just return then we
995 * are going to dereference garbage.
996 */
997 BUG_ON(skb_queue_is_first(list, skb));
998 return skb->prev;
999}
1000
1001/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002 * skb_get - reference buffer
1003 * @skb: buffer to reference
1004 *
1005 * Makes another reference to a socket buffer and returns a pointer
1006 * to the buffer.
1007 */
1008static inline struct sk_buff *skb_get(struct sk_buff *skb)
1009{
1010 atomic_inc(&skb->users);
1011 return skb;
1012}
1013
1014/*
1015 * If users == 1, we are the only owner and are can avoid redundant
1016 * atomic change.
1017 */
1018
1019/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001020 * skb_cloned - is the buffer a clone
1021 * @skb: buffer to check
1022 *
1023 * Returns true if the buffer was generated with skb_clone() and is
1024 * one of multiple shared copies of the buffer. Cloned buffers are
1025 * shared data so must not be written to under normal circumstances.
1026 */
1027static inline int skb_cloned(const struct sk_buff *skb)
1028{
1029 return skb->cloned &&
1030 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1031}
1032
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001033static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1034{
1035 might_sleep_if(pri & __GFP_WAIT);
1036
1037 if (skb_cloned(skb))
1038 return pskb_expand_head(skb, 0, 0, pri);
1039
1040 return 0;
1041}
1042
Linus Torvalds1da177e2005-04-16 15:20:36 -07001043/**
1044 * skb_header_cloned - is the header a clone
1045 * @skb: buffer to check
1046 *
1047 * Returns true if modifying the header part of the buffer requires
1048 * the data to be copied.
1049 */
1050static inline int skb_header_cloned(const struct sk_buff *skb)
1051{
1052 int dataref;
1053
1054 if (!skb->cloned)
1055 return 0;
1056
1057 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1058 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1059 return dataref != 1;
1060}
1061
1062/**
1063 * skb_header_release - release reference to header
1064 * @skb: buffer to operate on
1065 *
1066 * Drop a reference to the header part of the buffer. This is done
1067 * by acquiring a payload reference. You must not read from the header
1068 * part of skb->data after this.
1069 */
1070static inline void skb_header_release(struct sk_buff *skb)
1071{
1072 BUG_ON(skb->nohdr);
1073 skb->nohdr = 1;
1074 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1075}
1076
1077/**
1078 * skb_shared - is the buffer shared
1079 * @skb: buffer to check
1080 *
1081 * Returns true if more than one person has a reference to this
1082 * buffer.
1083 */
1084static inline int skb_shared(const struct sk_buff *skb)
1085{
1086 return atomic_read(&skb->users) != 1;
1087}
1088
1089/**
1090 * skb_share_check - check if buffer is shared and if so clone it
1091 * @skb: buffer to check
1092 * @pri: priority for memory allocation
1093 *
1094 * If the buffer is shared the buffer is cloned and the old copy
1095 * drops a reference. A new clone with a single reference is returned.
1096 * If the buffer is not shared the original buffer is returned. When
1097 * being called from interrupt status or with spinlocks held pri must
1098 * be GFP_ATOMIC.
1099 *
1100 * NULL is returned on a memory allocation failure.
1101 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001102static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001103{
1104 might_sleep_if(pri & __GFP_WAIT);
1105 if (skb_shared(skb)) {
1106 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001107
1108 if (likely(nskb))
1109 consume_skb(skb);
1110 else
1111 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112 skb = nskb;
1113 }
1114 return skb;
1115}
1116
1117/*
1118 * Copy shared buffers into a new sk_buff. We effectively do COW on
1119 * packets to handle cases where we have a local reader and forward
1120 * and a couple of other messy ones. The normal one is tcpdumping
1121 * a packet thats being forwarded.
1122 */
1123
1124/**
1125 * skb_unshare - make a copy of a shared buffer
1126 * @skb: buffer to check
1127 * @pri: priority for memory allocation
1128 *
1129 * If the socket buffer is a clone then this function creates a new
1130 * copy of the data, drops a reference count on the old copy and returns
1131 * the new copy with the reference count at 1. If the buffer is not a clone
1132 * the original buffer is returned. When called with a spinlock held or
1133 * from interrupt state @pri must be %GFP_ATOMIC
1134 *
1135 * %NULL is returned on a memory allocation failure.
1136 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001137static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001138 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139{
1140 might_sleep_if(pri & __GFP_WAIT);
1141 if (skb_cloned(skb)) {
1142 struct sk_buff *nskb = skb_copy(skb, pri);
1143 kfree_skb(skb); /* Free our shared copy */
1144 skb = nskb;
1145 }
1146 return skb;
1147}
1148
1149/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001150 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001151 * @list_: list to peek at
1152 *
1153 * Peek an &sk_buff. Unlike most other operations you _MUST_
1154 * be careful with this one. A peek leaves the buffer on the
1155 * list and someone else may run off with it. You must hold
1156 * the appropriate locks or have a private queue to do this.
1157 *
1158 * Returns %NULL for an empty list or a pointer to the head element.
1159 * The reference count is not incremented and the reference is therefore
1160 * volatile. Use with caution.
1161 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001162static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163{
Eric Dumazet18d07002012-04-30 16:31:46 +00001164 struct sk_buff *skb = list_->next;
1165
1166 if (skb == (struct sk_buff *)list_)
1167 skb = NULL;
1168 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001169}
1170
1171/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001172 * skb_peek_next - peek skb following the given one from a queue
1173 * @skb: skb to start from
1174 * @list_: list to peek at
1175 *
1176 * Returns %NULL when the end of the list is met or a pointer to the
1177 * next element. The reference count is not incremented and the
1178 * reference is therefore volatile. Use with caution.
1179 */
1180static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1181 const struct sk_buff_head *list_)
1182{
1183 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001184
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001185 if (next == (struct sk_buff *)list_)
1186 next = NULL;
1187 return next;
1188}
1189
1190/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001191 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001192 * @list_: list to peek at
1193 *
1194 * Peek an &sk_buff. Unlike most other operations you _MUST_
1195 * be careful with this one. A peek leaves the buffer on the
1196 * list and someone else may run off with it. You must hold
1197 * the appropriate locks or have a private queue to do this.
1198 *
1199 * Returns %NULL for an empty list or a pointer to the tail element.
1200 * The reference count is not incremented and the reference is therefore
1201 * volatile. Use with caution.
1202 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001203static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001204{
Eric Dumazet18d07002012-04-30 16:31:46 +00001205 struct sk_buff *skb = list_->prev;
1206
1207 if (skb == (struct sk_buff *)list_)
1208 skb = NULL;
1209 return skb;
1210
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211}
1212
1213/**
1214 * skb_queue_len - get queue length
1215 * @list_: list to measure
1216 *
1217 * Return the length of an &sk_buff queue.
1218 */
1219static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1220{
1221 return list_->qlen;
1222}
1223
David S. Miller67fed452008-09-21 22:36:24 -07001224/**
1225 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1226 * @list: queue to initialize
1227 *
1228 * This initializes only the list and queue length aspects of
1229 * an sk_buff_head object. This allows to initialize the list
1230 * aspects of an sk_buff_head without reinitializing things like
1231 * the spinlock. It can also be used for on-stack sk_buff_head
1232 * objects where the spinlock is known to not be used.
1233 */
1234static inline void __skb_queue_head_init(struct sk_buff_head *list)
1235{
1236 list->prev = list->next = (struct sk_buff *)list;
1237 list->qlen = 0;
1238}
1239
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001240/*
1241 * This function creates a split out lock class for each invocation;
1242 * this is needed for now since a whole lot of users of the skb-queue
1243 * infrastructure in drivers have different locking usage (in hardirq)
1244 * than the networking core (in softirq only). In the long run either the
1245 * network layer or drivers should need annotation to consolidate the
1246 * main types of usage into 3 classes.
1247 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248static inline void skb_queue_head_init(struct sk_buff_head *list)
1249{
1250 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001251 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001252}
1253
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001254static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1255 struct lock_class_key *class)
1256{
1257 skb_queue_head_init(list);
1258 lockdep_set_class(&list->lock, class);
1259}
1260
Linus Torvalds1da177e2005-04-16 15:20:36 -07001261/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001262 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001263 *
1264 * The "__skb_xxxx()" functions are the non-atomic ones that
1265 * can only be called with interrupts disabled.
1266 */
Joe Perches7965bd42013-09-26 14:48:15 -07001267void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1268 struct sk_buff_head *list);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001269static inline void __skb_insert(struct sk_buff *newsk,
1270 struct sk_buff *prev, struct sk_buff *next,
1271 struct sk_buff_head *list)
1272{
1273 newsk->next = next;
1274 newsk->prev = prev;
1275 next->prev = prev->next = newsk;
1276 list->qlen++;
1277}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001278
David S. Miller67fed452008-09-21 22:36:24 -07001279static inline void __skb_queue_splice(const struct sk_buff_head *list,
1280 struct sk_buff *prev,
1281 struct sk_buff *next)
1282{
1283 struct sk_buff *first = list->next;
1284 struct sk_buff *last = list->prev;
1285
1286 first->prev = prev;
1287 prev->next = first;
1288
1289 last->next = next;
1290 next->prev = last;
1291}
1292
1293/**
1294 * skb_queue_splice - join two skb lists, this is designed for stacks
1295 * @list: the new list to add
1296 * @head: the place to add it in the first list
1297 */
1298static inline void skb_queue_splice(const struct sk_buff_head *list,
1299 struct sk_buff_head *head)
1300{
1301 if (!skb_queue_empty(list)) {
1302 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001303 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001304 }
1305}
1306
1307/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001308 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001309 * @list: the new list to add
1310 * @head: the place to add it in the first list
1311 *
1312 * The list at @list is reinitialised
1313 */
1314static inline void skb_queue_splice_init(struct sk_buff_head *list,
1315 struct sk_buff_head *head)
1316{
1317 if (!skb_queue_empty(list)) {
1318 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001319 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001320 __skb_queue_head_init(list);
1321 }
1322}
1323
1324/**
1325 * skb_queue_splice_tail - join two skb lists, each list being a queue
1326 * @list: the new list to add
1327 * @head: the place to add it in the first list
1328 */
1329static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1330 struct sk_buff_head *head)
1331{
1332 if (!skb_queue_empty(list)) {
1333 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001334 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001335 }
1336}
1337
1338/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001339 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001340 * @list: the new list to add
1341 * @head: the place to add it in the first list
1342 *
1343 * Each of the lists is a queue.
1344 * The list at @list is reinitialised
1345 */
1346static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1347 struct sk_buff_head *head)
1348{
1349 if (!skb_queue_empty(list)) {
1350 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001351 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001352 __skb_queue_head_init(list);
1353 }
1354}
1355
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001357 * __skb_queue_after - queue a buffer at the list head
1358 * @list: list to use
1359 * @prev: place after this buffer
1360 * @newsk: buffer to queue
1361 *
1362 * Queue a buffer int the middle of a list. This function takes no locks
1363 * and you must therefore hold required locks before calling it.
1364 *
1365 * A buffer cannot be placed on two lists at the same time.
1366 */
1367static inline void __skb_queue_after(struct sk_buff_head *list,
1368 struct sk_buff *prev,
1369 struct sk_buff *newsk)
1370{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001371 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001372}
1373
Joe Perches7965bd42013-09-26 14:48:15 -07001374void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1375 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001376
Gerrit Renkerf5572852008-04-14 00:05:28 -07001377static inline void __skb_queue_before(struct sk_buff_head *list,
1378 struct sk_buff *next,
1379 struct sk_buff *newsk)
1380{
1381 __skb_insert(newsk, next->prev, next, list);
1382}
1383
Stephen Hemminger300ce172005-10-30 13:47:34 -08001384/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385 * __skb_queue_head - queue a buffer at the list head
1386 * @list: list to use
1387 * @newsk: buffer to queue
1388 *
1389 * Queue a buffer at the start of a list. This function takes no locks
1390 * and you must therefore hold required locks before calling it.
1391 *
1392 * A buffer cannot be placed on two lists at the same time.
1393 */
Joe Perches7965bd42013-09-26 14:48:15 -07001394void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001395static inline void __skb_queue_head(struct sk_buff_head *list,
1396 struct sk_buff *newsk)
1397{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001398 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399}
1400
1401/**
1402 * __skb_queue_tail - queue a buffer at the list tail
1403 * @list: list to use
1404 * @newsk: buffer to queue
1405 *
1406 * Queue a buffer at the end of a list. This function takes no locks
1407 * and you must therefore hold required locks before calling it.
1408 *
1409 * A buffer cannot be placed on two lists at the same time.
1410 */
Joe Perches7965bd42013-09-26 14:48:15 -07001411void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412static inline void __skb_queue_tail(struct sk_buff_head *list,
1413 struct sk_buff *newsk)
1414{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001415 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001416}
1417
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001419 * remove sk_buff from list. _Must_ be called atomically, and with
1420 * the list known..
1421 */
Joe Perches7965bd42013-09-26 14:48:15 -07001422void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001423static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1424{
1425 struct sk_buff *next, *prev;
1426
1427 list->qlen--;
1428 next = skb->next;
1429 prev = skb->prev;
1430 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431 next->prev = prev;
1432 prev->next = next;
1433}
1434
Gerrit Renkerf525c062008-04-14 00:04:12 -07001435/**
1436 * __skb_dequeue - remove from the head of the queue
1437 * @list: list to dequeue from
1438 *
1439 * Remove the head of the list. This function does not take any locks
1440 * so must be used with appropriate locks held only. The head item is
1441 * returned or %NULL if the list is empty.
1442 */
Joe Perches7965bd42013-09-26 14:48:15 -07001443struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Gerrit Renkerf525c062008-04-14 00:04:12 -07001444static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1445{
1446 struct sk_buff *skb = skb_peek(list);
1447 if (skb)
1448 __skb_unlink(skb, list);
1449 return skb;
1450}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001451
1452/**
1453 * __skb_dequeue_tail - remove from the tail of the queue
1454 * @list: list to dequeue from
1455 *
1456 * Remove the tail of the list. This function does not take any locks
1457 * so must be used with appropriate locks held only. The tail item is
1458 * returned or %NULL if the list is empty.
1459 */
Joe Perches7965bd42013-09-26 14:48:15 -07001460struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001461static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1462{
1463 struct sk_buff *skb = skb_peek_tail(list);
1464 if (skb)
1465 __skb_unlink(skb, list);
1466 return skb;
1467}
1468
1469
David S. Millerbdcc0922012-03-07 20:53:36 -05001470static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001471{
1472 return skb->data_len;
1473}
1474
1475static inline unsigned int skb_headlen(const struct sk_buff *skb)
1476{
1477 return skb->len - skb->data_len;
1478}
1479
1480static inline int skb_pagelen(const struct sk_buff *skb)
1481{
1482 int i, len = 0;
1483
1484 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001485 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001486 return len + skb_headlen(skb);
1487}
1488
Ian Campbell131ea662011-08-19 06:25:00 +00001489/**
1490 * __skb_fill_page_desc - initialise a paged fragment in an skb
1491 * @skb: buffer containing fragment to be initialised
1492 * @i: paged fragment index to initialise
1493 * @page: the page to use for this fragment
1494 * @off: the offset to the data with @page
1495 * @size: the length of the data
1496 *
1497 * Initialises the @i'th fragment of @skb to point to &size bytes at
1498 * offset @off within @page.
1499 *
1500 * Does not take any additional reference on the fragment.
1501 */
1502static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1503 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001504{
1505 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1506
Mel Gormanc48a11c2012-07-31 16:44:23 -07001507 /*
1508 * Propagate page->pfmemalloc to the skb if we can. The problem is
1509 * that not all callers have unique ownership of the page. If
1510 * pfmemalloc is set, we check the mapping as a mapping implies
1511 * page->index is set (index and pfmemalloc share space).
1512 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1513 * do not lose pfmemalloc information as the pages would not be
1514 * allocated using __GFP_MEMALLOC.
1515 */
Ian Campbella8605c62011-10-19 23:01:49 +00001516 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001517 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001518 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001519
1520 page = compound_head(page);
1521 if (page->pfmemalloc && !page->mapping)
1522 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00001523}
1524
1525/**
1526 * skb_fill_page_desc - initialise a paged fragment in an skb
1527 * @skb: buffer containing fragment to be initialised
1528 * @i: paged fragment index to initialise
1529 * @page: the page to use for this fragment
1530 * @off: the offset to the data with @page
1531 * @size: the length of the data
1532 *
1533 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01001534 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00001535 * addition updates @skb such that @i is the last fragment.
1536 *
1537 * Does not take any additional reference on the fragment.
1538 */
1539static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1540 struct page *page, int off, int size)
1541{
1542 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001543 skb_shinfo(skb)->nr_frags = i + 1;
1544}
1545
Joe Perches7965bd42013-09-26 14:48:15 -07001546void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1547 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001548
Jason Wangf8e617e2013-11-01 14:07:47 +08001549void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1550 unsigned int truesize);
1551
Linus Torvalds1da177e2005-04-16 15:20:36 -07001552#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001553#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001554#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1555
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001556#ifdef NET_SKBUFF_DATA_USES_OFFSET
1557static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1558{
1559 return skb->head + skb->tail;
1560}
1561
1562static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1563{
1564 skb->tail = skb->data - skb->head;
1565}
1566
1567static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1568{
1569 skb_reset_tail_pointer(skb);
1570 skb->tail += offset;
1571}
Simon Horman7cc46192013-05-28 20:34:29 +00001572
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001573#else /* NET_SKBUFF_DATA_USES_OFFSET */
1574static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1575{
1576 return skb->tail;
1577}
1578
1579static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1580{
1581 skb->tail = skb->data;
1582}
1583
1584static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1585{
1586 skb->tail = skb->data + offset;
1587}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001588
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001589#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1590
Linus Torvalds1da177e2005-04-16 15:20:36 -07001591/*
1592 * Add data to an sk_buff
1593 */
Mathias Krause0c7ddf32013-11-07 14:18:24 +01001594unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001595unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001596static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1597{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001598 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001599 SKB_LINEAR_ASSERT(skb);
1600 skb->tail += len;
1601 skb->len += len;
1602 return tmp;
1603}
1604
Joe Perches7965bd42013-09-26 14:48:15 -07001605unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001606static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1607{
1608 skb->data -= len;
1609 skb->len += len;
1610 return skb->data;
1611}
1612
Joe Perches7965bd42013-09-26 14:48:15 -07001613unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001614static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1615{
1616 skb->len -= len;
1617 BUG_ON(skb->len < skb->data_len);
1618 return skb->data += len;
1619}
1620
David S. Miller47d29642010-05-02 02:21:44 -07001621static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1622{
1623 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1624}
1625
Joe Perches7965bd42013-09-26 14:48:15 -07001626unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001627
1628static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1629{
1630 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001631 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001632 return NULL;
1633 skb->len -= len;
1634 return skb->data += len;
1635}
1636
1637static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1638{
1639 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1640}
1641
1642static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1643{
1644 if (likely(len <= skb_headlen(skb)))
1645 return 1;
1646 if (unlikely(len > skb->len))
1647 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001648 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001649}
1650
1651/**
1652 * skb_headroom - bytes at buffer head
1653 * @skb: buffer to check
1654 *
1655 * Return the number of bytes of free space at the head of an &sk_buff.
1656 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001657static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001658{
1659 return skb->data - skb->head;
1660}
1661
1662/**
1663 * skb_tailroom - bytes at buffer end
1664 * @skb: buffer to check
1665 *
1666 * Return the number of bytes of free space at the tail of an sk_buff
1667 */
1668static inline int skb_tailroom(const struct sk_buff *skb)
1669{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001670 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001671}
1672
1673/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00001674 * skb_availroom - bytes at buffer end
1675 * @skb: buffer to check
1676 *
1677 * Return the number of bytes of free space at the tail of an sk_buff
1678 * allocated by sk_stream_alloc()
1679 */
1680static inline int skb_availroom(const struct sk_buff *skb)
1681{
Eric Dumazet16fad692013-03-14 05:40:32 +00001682 if (skb_is_nonlinear(skb))
1683 return 0;
1684
1685 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00001686}
1687
1688/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001689 * skb_reserve - adjust headroom
1690 * @skb: buffer to alter
1691 * @len: bytes to move
1692 *
1693 * Increase the headroom of an empty &sk_buff by reducing the tail
1694 * room. This is only allowed for an empty buffer.
1695 */
David S. Miller8243126c2006-01-17 02:54:21 -08001696static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001697{
1698 skb->data += len;
1699 skb->tail += len;
1700}
1701
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001702static inline void skb_reset_inner_headers(struct sk_buff *skb)
1703{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001704 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001705 skb->inner_network_header = skb->network_header;
1706 skb->inner_transport_header = skb->transport_header;
1707}
1708
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00001709static inline void skb_reset_mac_len(struct sk_buff *skb)
1710{
1711 skb->mac_len = skb->network_header - skb->mac_header;
1712}
1713
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001714static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1715 *skb)
1716{
1717 return skb->head + skb->inner_transport_header;
1718}
1719
1720static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1721{
1722 skb->inner_transport_header = skb->data - skb->head;
1723}
1724
1725static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1726 const int offset)
1727{
1728 skb_reset_inner_transport_header(skb);
1729 skb->inner_transport_header += offset;
1730}
1731
1732static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1733{
1734 return skb->head + skb->inner_network_header;
1735}
1736
1737static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1738{
1739 skb->inner_network_header = skb->data - skb->head;
1740}
1741
1742static inline void skb_set_inner_network_header(struct sk_buff *skb,
1743 const int offset)
1744{
1745 skb_reset_inner_network_header(skb);
1746 skb->inner_network_header += offset;
1747}
1748
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001749static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1750{
1751 return skb->head + skb->inner_mac_header;
1752}
1753
1754static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1755{
1756 skb->inner_mac_header = skb->data - skb->head;
1757}
1758
1759static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1760 const int offset)
1761{
1762 skb_reset_inner_mac_header(skb);
1763 skb->inner_mac_header += offset;
1764}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001765static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1766{
Cong Wang35d04612013-05-29 15:16:05 +08001767 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001768}
1769
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001770static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1771{
1772 return skb->head + skb->transport_header;
1773}
1774
1775static inline void skb_reset_transport_header(struct sk_buff *skb)
1776{
1777 skb->transport_header = skb->data - skb->head;
1778}
1779
1780static inline void skb_set_transport_header(struct sk_buff *skb,
1781 const int offset)
1782{
1783 skb_reset_transport_header(skb);
1784 skb->transport_header += offset;
1785}
1786
1787static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1788{
1789 return skb->head + skb->network_header;
1790}
1791
1792static inline void skb_reset_network_header(struct sk_buff *skb)
1793{
1794 skb->network_header = skb->data - skb->head;
1795}
1796
1797static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1798{
1799 skb_reset_network_header(skb);
1800 skb->network_header += offset;
1801}
1802
1803static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1804{
1805 return skb->head + skb->mac_header;
1806}
1807
1808static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1809{
Cong Wang35d04612013-05-29 15:16:05 +08001810 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001811}
1812
1813static inline void skb_reset_mac_header(struct sk_buff *skb)
1814{
1815 skb->mac_header = skb->data - skb->head;
1816}
1817
1818static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1819{
1820 skb_reset_mac_header(skb);
1821 skb->mac_header += offset;
1822}
1823
Timo Teräs0e3da5b2013-12-16 11:02:09 +02001824static inline void skb_pop_mac_header(struct sk_buff *skb)
1825{
1826 skb->mac_header = skb->network_header;
1827}
1828
Ying Xuefbbdb8f2013-03-27 16:46:06 +00001829static inline void skb_probe_transport_header(struct sk_buff *skb,
1830 const int offset_hint)
1831{
1832 struct flow_keys keys;
1833
1834 if (skb_transport_header_was_set(skb))
1835 return;
1836 else if (skb_flow_dissect(skb, &keys))
1837 skb_set_transport_header(skb, keys.thoff);
1838 else
1839 skb_set_transport_header(skb, offset_hint);
1840}
1841
Eric Dumazet03606892012-02-23 10:55:02 +00001842static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1843{
1844 if (skb_mac_header_was_set(skb)) {
1845 const unsigned char *old_mac = skb_mac_header(skb);
1846
1847 skb_set_mac_header(skb, -skb->mac_len);
1848 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1849 }
1850}
1851
Michał Mirosław04fb4512010-12-14 15:24:08 +00001852static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1853{
1854 return skb->csum_start - skb_headroom(skb);
1855}
1856
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001857static inline int skb_transport_offset(const struct sk_buff *skb)
1858{
1859 return skb_transport_header(skb) - skb->data;
1860}
1861
1862static inline u32 skb_network_header_len(const struct sk_buff *skb)
1863{
1864 return skb->transport_header - skb->network_header;
1865}
1866
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001867static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1868{
1869 return skb->inner_transport_header - skb->inner_network_header;
1870}
1871
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001872static inline int skb_network_offset(const struct sk_buff *skb)
1873{
1874 return skb_network_header(skb) - skb->data;
1875}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001876
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001877static inline int skb_inner_network_offset(const struct sk_buff *skb)
1878{
1879 return skb_inner_network_header(skb) - skb->data;
1880}
1881
Changli Gaof9599ce2010-08-04 04:43:44 +00001882static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1883{
1884 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1885}
1886
Linus Torvalds1da177e2005-04-16 15:20:36 -07001887/*
1888 * CPUs often take a performance hit when accessing unaligned memory
1889 * locations. The actual performance hit varies, it can be small if the
1890 * hardware handles it or large if we have to take an exception and fix it
1891 * in software.
1892 *
1893 * Since an ethernet header is 14 bytes network drivers often end up with
1894 * the IP header at an unaligned offset. The IP header can be aligned by
1895 * shifting the start of the packet by 2 bytes. Drivers should do this
1896 * with:
1897 *
Tobias Klauser8660c122009-07-13 22:48:16 +00001898 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001899 *
1900 * The downside to this alignment of the IP header is that the DMA is now
1901 * unaligned. On some architectures the cost of an unaligned DMA is high
1902 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00001903 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001904 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1905 * to be overridden.
1906 */
1907#ifndef NET_IP_ALIGN
1908#define NET_IP_ALIGN 2
1909#endif
1910
Anton Blanchard025be812006-03-31 02:27:06 -08001911/*
1912 * The networking layer reserves some headroom in skb data (via
1913 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1914 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08001915 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08001916 *
1917 * Unfortunately this headroom changes the DMA alignment of the resulting
1918 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1919 * on some architectures. An architecture can override this value,
1920 * perhaps setting it to a cacheline in size (since that will maintain
1921 * cacheline alignment of the DMA). It must be a power of 2.
1922 *
David S. Millerd6301d32009-02-08 19:24:13 -08001923 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08001924 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07001925 *
1926 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1927 * to reduce average number of cache lines per packet.
1928 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07001929 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08001930 */
1931#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07001932#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08001933#endif
1934
Joe Perches7965bd42013-09-26 14:48:15 -07001935int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001936
1937static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1938{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00001939 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07001940 WARN_ON(1);
1941 return;
1942 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001943 skb->len = len;
1944 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001945}
1946
Joe Perches7965bd42013-09-26 14:48:15 -07001947void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001948
1949static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1950{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001951 if (skb->data_len)
1952 return ___pskb_trim(skb, len);
1953 __skb_trim(skb, len);
1954 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001955}
1956
1957static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1958{
1959 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1960}
1961
1962/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07001963 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1964 * @skb: buffer to alter
1965 * @len: new length
1966 *
1967 * This is identical to pskb_trim except that the caller knows that
1968 * the skb is not cloned so we should never get an error due to out-
1969 * of-memory.
1970 */
1971static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1972{
1973 int err = pskb_trim(skb, len);
1974 BUG_ON(err);
1975}
1976
1977/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001978 * skb_orphan - orphan a buffer
1979 * @skb: buffer to orphan
1980 *
1981 * If a buffer currently has an owner then we call the owner's
1982 * destructor function and make the @skb unowned. The buffer continues
1983 * to exist but is no longer charged to its former owner.
1984 */
1985static inline void skb_orphan(struct sk_buff *skb)
1986{
Eric Dumazetc34a7612013-07-30 16:11:15 -07001987 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001988 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07001989 skb->destructor = NULL;
1990 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07001991 } else {
1992 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07001993 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001994}
1995
1996/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00001997 * skb_orphan_frags - orphan the frags contained in a buffer
1998 * @skb: buffer to orphan frags from
1999 * @gfp_mask: allocation mask for replacement pages
2000 *
2001 * For each frag in the SKB which needs a destructor (i.e. has an
2002 * owner) create a copy of that frag and release the original
2003 * page by calling the destructor.
2004 */
2005static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2006{
2007 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2008 return 0;
2009 return skb_copy_ubufs(skb, gfp_mask);
2010}
2011
2012/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002013 * __skb_queue_purge - empty a list
2014 * @list: list to empty
2015 *
2016 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2017 * the list and one reference dropped. This function does not take the
2018 * list lock and the caller must hold the relevant locks to use it.
2019 */
Joe Perches7965bd42013-09-26 14:48:15 -07002020void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021static inline void __skb_queue_purge(struct sk_buff_head *list)
2022{
2023 struct sk_buff *skb;
2024 while ((skb = __skb_dequeue(list)) != NULL)
2025 kfree_skb(skb);
2026}
2027
Alexander Duycke5e67302013-02-08 10:17:15 +00002028#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
2029#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
2030#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
2031
Joe Perches7965bd42013-09-26 14:48:15 -07002032void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002033
Joe Perches7965bd42013-09-26 14:48:15 -07002034struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2035 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07002036
2037/**
2038 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2039 * @dev: network device to receive on
2040 * @length: length to allocate
2041 *
2042 * Allocate a new &sk_buff and assign it a usage count of one. The
2043 * buffer has unspecified headroom built in. Users should allocate
2044 * the headroom they think they need without accounting for the
2045 * built in space. The built in space is used for optimisations.
2046 *
2047 * %NULL is returned if there is no free memory. Although this function
2048 * allocates memory it can be called from an interrupt.
2049 */
2050static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002051 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002052{
2053 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2054}
2055
Eric Dumazet6f532612012-05-18 05:12:12 +00002056/* legacy helper around __netdev_alloc_skb() */
2057static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2058 gfp_t gfp_mask)
2059{
2060 return __netdev_alloc_skb(NULL, length, gfp_mask);
2061}
2062
2063/* legacy helper around netdev_alloc_skb() */
2064static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2065{
2066 return netdev_alloc_skb(NULL, length);
2067}
2068
2069
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002070static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2071 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002072{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002073 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002074
2075 if (NET_IP_ALIGN && skb)
2076 skb_reserve(skb, NET_IP_ALIGN);
2077 return skb;
2078}
2079
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002080static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2081 unsigned int length)
2082{
2083 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2084}
2085
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002086/**
2087 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
Mel Gorman06140022012-07-31 16:44:24 -07002088 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2089 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2090 * @order: size of the allocation
2091 *
2092 * Allocate a new page.
2093 *
2094 * %NULL is returned if there is no free memory.
2095*/
2096static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
2097 struct sk_buff *skb,
2098 unsigned int order)
2099{
2100 struct page *page;
2101
2102 gfp_mask |= __GFP_COLD;
2103
2104 if (!(gfp_mask & __GFP_NOMEMALLOC))
2105 gfp_mask |= __GFP_MEMALLOC;
2106
2107 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2108 if (skb && page && page->pfmemalloc)
2109 skb->pfmemalloc = true;
2110
2111 return page;
2112}
2113
2114/**
2115 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2116 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2117 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2118 *
2119 * Allocate a new page.
2120 *
2121 * %NULL is returned if there is no free memory.
2122 */
2123static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
2124 struct sk_buff *skb)
2125{
2126 return __skb_alloc_pages(gfp_mask, skb, 0);
2127}
2128
2129/**
2130 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2131 * @page: The page that was allocated from skb_alloc_page
2132 * @skb: The skb that may need pfmemalloc set
2133 */
2134static inline void skb_propagate_pfmemalloc(struct page *page,
2135 struct sk_buff *skb)
2136{
2137 if (page && page->pfmemalloc)
2138 skb->pfmemalloc = true;
2139}
2140
Eric Dumazet564824b2010-10-11 19:05:25 +00002141/**
Masanari Iidae2278672014-02-18 22:54:36 +09002142 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002143 * @frag: the paged fragment
2144 *
2145 * Returns the &struct page associated with @frag.
2146 */
2147static inline struct page *skb_frag_page(const skb_frag_t *frag)
2148{
Ian Campbella8605c62011-10-19 23:01:49 +00002149 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00002150}
2151
2152/**
2153 * __skb_frag_ref - take an addition reference on a paged fragment.
2154 * @frag: the paged fragment
2155 *
2156 * Takes an additional reference on the paged fragment @frag.
2157 */
2158static inline void __skb_frag_ref(skb_frag_t *frag)
2159{
2160 get_page(skb_frag_page(frag));
2161}
2162
2163/**
2164 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2165 * @skb: the buffer
2166 * @f: the fragment offset.
2167 *
2168 * Takes an additional reference on the @f'th paged fragment of @skb.
2169 */
2170static inline void skb_frag_ref(struct sk_buff *skb, int f)
2171{
2172 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2173}
2174
2175/**
2176 * __skb_frag_unref - release a reference on a paged fragment.
2177 * @frag: the paged fragment
2178 *
2179 * Releases a reference on the paged fragment @frag.
2180 */
2181static inline void __skb_frag_unref(skb_frag_t *frag)
2182{
2183 put_page(skb_frag_page(frag));
2184}
2185
2186/**
2187 * skb_frag_unref - release a reference on a paged fragment of an skb.
2188 * @skb: the buffer
2189 * @f: the fragment offset
2190 *
2191 * Releases a reference on the @f'th paged fragment of @skb.
2192 */
2193static inline void skb_frag_unref(struct sk_buff *skb, int f)
2194{
2195 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2196}
2197
2198/**
2199 * skb_frag_address - gets the address of the data contained in a paged fragment
2200 * @frag: the paged fragment buffer
2201 *
2202 * Returns the address of the data within @frag. The page must already
2203 * be mapped.
2204 */
2205static inline void *skb_frag_address(const skb_frag_t *frag)
2206{
2207 return page_address(skb_frag_page(frag)) + frag->page_offset;
2208}
2209
2210/**
2211 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2212 * @frag: the paged fragment buffer
2213 *
2214 * Returns the address of the data within @frag. Checks that the page
2215 * is mapped and returns %NULL otherwise.
2216 */
2217static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2218{
2219 void *ptr = page_address(skb_frag_page(frag));
2220 if (unlikely(!ptr))
2221 return NULL;
2222
2223 return ptr + frag->page_offset;
2224}
2225
2226/**
2227 * __skb_frag_set_page - sets the page contained in a paged fragment
2228 * @frag: the paged fragment
2229 * @page: the page to set
2230 *
2231 * Sets the fragment @frag to contain @page.
2232 */
2233static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2234{
Ian Campbella8605c62011-10-19 23:01:49 +00002235 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002236}
2237
2238/**
2239 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2240 * @skb: the buffer
2241 * @f: the fragment offset
2242 * @page: the page to set
2243 *
2244 * Sets the @f'th fragment of @skb to contain @page.
2245 */
2246static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2247 struct page *page)
2248{
2249 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2250}
2251
Eric Dumazet400dfd32013-10-17 16:27:07 -07002252bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2253
Ian Campbell131ea662011-08-19 06:25:00 +00002254/**
2255 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002256 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002257 * @frag: the paged fragment to map
2258 * @offset: the offset within the fragment (starting at the
2259 * fragment's own offset)
2260 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002261 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00002262 *
2263 * Maps the page associated with @frag to @device.
2264 */
2265static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2266 const skb_frag_t *frag,
2267 size_t offset, size_t size,
2268 enum dma_data_direction dir)
2269{
2270 return dma_map_page(dev, skb_frag_page(frag),
2271 frag->page_offset + offset, size, dir);
2272}
2273
Eric Dumazet117632e2011-12-03 21:39:53 +00002274static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2275 gfp_t gfp_mask)
2276{
2277 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2278}
2279
Octavian Purdilabad93e92014-06-12 01:36:26 +03002280
2281static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2282 gfp_t gfp_mask)
2283{
2284 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2285}
2286
2287
Ian Campbell131ea662011-08-19 06:25:00 +00002288/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002289 * skb_clone_writable - is the header of a clone writable
2290 * @skb: buffer to check
2291 * @len: length up to which to write
2292 *
2293 * Returns true if modifying the header part of the cloned buffer
2294 * does not requires the data to be copied.
2295 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002296static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002297{
2298 return !skb_header_cloned(skb) &&
2299 skb_headroom(skb) + len <= skb->hdr_len;
2300}
2301
Herbert Xud9cc2042007-09-16 16:21:16 -07002302static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2303 int cloned)
2304{
2305 int delta = 0;
2306
Herbert Xud9cc2042007-09-16 16:21:16 -07002307 if (headroom > skb_headroom(skb))
2308 delta = headroom - skb_headroom(skb);
2309
2310 if (delta || cloned)
2311 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2312 GFP_ATOMIC);
2313 return 0;
2314}
2315
Patrick McHardy334a8132007-06-25 04:35:20 -07002316/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002317 * skb_cow - copy header of skb when it is required
2318 * @skb: buffer to cow
2319 * @headroom: needed headroom
2320 *
2321 * If the skb passed lacks sufficient headroom or its data part
2322 * is shared, data is reallocated. If reallocation fails, an error
2323 * is returned and original skb is not changed.
2324 *
2325 * The result is skb with writable area skb->head...skb->tail
2326 * and at least @headroom of space at head.
2327 */
2328static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2329{
Herbert Xud9cc2042007-09-16 16:21:16 -07002330 return __skb_cow(skb, headroom, skb_cloned(skb));
2331}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002332
Herbert Xud9cc2042007-09-16 16:21:16 -07002333/**
2334 * skb_cow_head - skb_cow but only making the head writable
2335 * @skb: buffer to cow
2336 * @headroom: needed headroom
2337 *
2338 * This function is identical to skb_cow except that we replace the
2339 * skb_cloned check by skb_header_cloned. It should be used when
2340 * you only need to push on some header and do not need to modify
2341 * the data.
2342 */
2343static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2344{
2345 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002346}
2347
2348/**
2349 * skb_padto - pad an skbuff up to a minimal size
2350 * @skb: buffer to pad
2351 * @len: minimal length
2352 *
2353 * Pads up a buffer to ensure the trailing bytes exist and are
2354 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002355 * is untouched. Otherwise it is extended. Returns zero on
2356 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002357 */
2358
Herbert Xu5b057c62006-06-23 02:06:41 -07002359static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002360{
2361 unsigned int size = skb->len;
2362 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002363 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002364 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002365}
2366
2367static inline int skb_add_data(struct sk_buff *skb,
2368 char __user *from, int copy)
2369{
2370 const int off = skb->len;
2371
2372 if (skb->ip_summed == CHECKSUM_NONE) {
2373 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08002374 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
Linus Torvalds1da177e2005-04-16 15:20:36 -07002375 copy, 0, &err);
2376 if (!err) {
2377 skb->csum = csum_block_add(skb->csum, csum, off);
2378 return 0;
2379 }
2380 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2381 return 0;
2382
2383 __skb_trim(skb, off);
2384 return -EFAULT;
2385}
2386
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002387static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2388 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002389{
2390 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002391 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002392
Ian Campbellea2ab692011-08-22 23:44:58 +00002393 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002394 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002395 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002396 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002397}
2398
Herbert Xu364c6ba2006-06-09 16:10:40 -07002399static inline int __skb_linearize(struct sk_buff *skb)
2400{
2401 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2402}
2403
Linus Torvalds1da177e2005-04-16 15:20:36 -07002404/**
2405 * skb_linearize - convert paged skb to linear one
2406 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002407 *
2408 * If there is no free memory -ENOMEM is returned, otherwise zero
2409 * is returned and the old skb data released.
2410 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002411static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002412{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002413 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2414}
2415
2416/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00002417 * skb_has_shared_frag - can any frag be overwritten
2418 * @skb: buffer to test
2419 *
2420 * Return true if the skb has at least one frag that might be modified
2421 * by an external entity (as in vmsplice()/sendfile())
2422 */
2423static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2424{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00002425 return skb_is_nonlinear(skb) &&
2426 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00002427}
2428
2429/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07002430 * skb_linearize_cow - make sure skb is linear and writable
2431 * @skb: buffer to process
2432 *
2433 * If there is no free memory -ENOMEM is returned, otherwise zero
2434 * is returned and the old skb data released.
2435 */
2436static inline int skb_linearize_cow(struct sk_buff *skb)
2437{
2438 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2439 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002440}
2441
2442/**
2443 * skb_postpull_rcsum - update checksum for received skb after pull
2444 * @skb: buffer to update
2445 * @start: start of data before pull
2446 * @len: length of data pulled
2447 *
2448 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002449 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2450 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002451 */
2452
2453static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002454 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002455{
Patrick McHardy84fa7932006-08-29 16:44:56 -07002456 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002457 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2458}
2459
Herbert Xucbb042f2006-03-20 22:43:56 -08002460unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2461
David S. Miller7ce5a272013-12-02 17:26:05 -05002462/**
2463 * pskb_trim_rcsum - trim received skb and update checksum
2464 * @skb: buffer to trim
2465 * @len: new length
2466 *
2467 * This is exactly the same as pskb_trim except that it ensures the
2468 * checksum of received packets are still valid after the operation.
2469 */
2470
2471static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2472{
2473 if (likely(len >= skb->len))
2474 return 0;
2475 if (skb->ip_summed == CHECKSUM_COMPLETE)
2476 skb->ip_summed = CHECKSUM_NONE;
2477 return __pskb_trim(skb, len);
2478}
2479
Linus Torvalds1da177e2005-04-16 15:20:36 -07002480#define skb_queue_walk(queue, skb) \
2481 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002482 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002483 skb = skb->next)
2484
James Chapman46f89142007-04-30 00:07:31 -07002485#define skb_queue_walk_safe(queue, skb, tmp) \
2486 for (skb = (queue)->next, tmp = skb->next; \
2487 skb != (struct sk_buff *)(queue); \
2488 skb = tmp, tmp = skb->next)
2489
David S. Miller1164f522008-09-23 00:49:44 -07002490#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002491 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002492 skb = skb->next)
2493
2494#define skb_queue_walk_from_safe(queue, skb, tmp) \
2495 for (tmp = skb->next; \
2496 skb != (struct sk_buff *)(queue); \
2497 skb = tmp, tmp = skb->next)
2498
Stephen Hemminger300ce172005-10-30 13:47:34 -08002499#define skb_queue_reverse_walk(queue, skb) \
2500 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002501 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08002502 skb = skb->prev)
2503
David S. Miller686a2952011-01-20 22:47:32 -08002504#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2505 for (skb = (queue)->prev, tmp = skb->prev; \
2506 skb != (struct sk_buff *)(queue); \
2507 skb = tmp, tmp = skb->prev)
2508
2509#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2510 for (tmp = skb->prev; \
2511 skb != (struct sk_buff *)(queue); \
2512 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002513
David S. Miller21dc3302010-08-23 00:13:46 -07002514static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07002515{
2516 return skb_shinfo(skb)->frag_list != NULL;
2517}
2518
2519static inline void skb_frag_list_init(struct sk_buff *skb)
2520{
2521 skb_shinfo(skb)->frag_list = NULL;
2522}
2523
2524static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2525{
2526 frag->next = skb_shinfo(skb)->frag_list;
2527 skb_shinfo(skb)->frag_list = frag;
2528}
2529
2530#define skb_walk_frags(skb, iter) \
2531 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2532
Joe Perches7965bd42013-09-26 14:48:15 -07002533struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2534 int *peeked, int *off, int *err);
2535struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2536 int *err);
2537unsigned int datagram_poll(struct file *file, struct socket *sock,
2538 struct poll_table_struct *wait);
2539int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2540 struct iovec *to, int size);
2541int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2542 struct iovec *iov);
2543int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2544 const struct iovec *from, int from_offset,
2545 int len);
2546int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2547 int offset, size_t count);
2548int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2549 const struct iovec *to, int to_offset,
2550 int size);
2551void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2552void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2553int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07002554int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2555int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2556__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2557 int len, __wsum csum);
2558int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2559 struct pipe_inode_info *pipe, unsigned int len,
2560 unsigned int flags);
2561void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01002562unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00002563int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2564 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07002565void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2566int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2567void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Florian Westphalde960aa2014-01-26 10:58:16 +01002568unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -07002569struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Vlad Yasevich0d5501c2014-08-08 14:42:13 -04002570struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002571
Daniel Borkmann2817a332013-10-30 11:50:51 +01002572struct skb_checksum_ops {
2573 __wsum (*update)(const void *mem, int len, __wsum wsum);
2574 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2575};
2576
2577__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2578 __wsum csum, const struct skb_checksum_ops *ops);
2579__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2580 __wsum csum);
2581
David S. Miller690e36e2014-08-23 12:13:41 -07002582static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset,
2583 int len, void *data, int hlen, void *buffer)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002584{
Patrick McHardy55820ee2005-07-05 14:08:10 -07002585 if (hlen - offset >= len)
David S. Miller690e36e2014-08-23 12:13:41 -07002586 return data + offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002587
David S. Miller690e36e2014-08-23 12:13:41 -07002588 if (!skb ||
2589 skb_copy_bits(skb, offset, buffer, len) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002590 return NULL;
2591
2592 return buffer;
2593}
2594
David S. Miller690e36e2014-08-23 12:13:41 -07002595static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2596 int len, void *buffer)
2597{
2598 return __skb_header_pointer(skb, offset, len, skb->data,
2599 skb_headlen(skb), buffer);
2600}
2601
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01002602/**
2603 * skb_needs_linearize - check if we need to linearize a given skb
2604 * depending on the given device features.
2605 * @skb: socket buffer to check
2606 * @features: net device features
2607 *
2608 * Returns true if either:
2609 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2610 * 2. skb is fragmented and the device does not support SG.
2611 */
2612static inline bool skb_needs_linearize(struct sk_buff *skb,
2613 netdev_features_t features)
2614{
2615 return skb_is_nonlinear(skb) &&
2616 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2617 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2618}
2619
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03002620static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2621 void *to,
2622 const unsigned int len)
2623{
2624 memcpy(to, skb->data, len);
2625}
2626
2627static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2628 const int offset, void *to,
2629 const unsigned int len)
2630{
2631 memcpy(to, skb->data + offset, len);
2632}
2633
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03002634static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2635 const void *from,
2636 const unsigned int len)
2637{
2638 memcpy(skb->data, from, len);
2639}
2640
2641static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2642 const int offset,
2643 const void *from,
2644 const unsigned int len)
2645{
2646 memcpy(skb->data + offset, from, len);
2647}
2648
Joe Perches7965bd42013-09-26 14:48:15 -07002649void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002650
Patrick Ohlyac45f602009-02-12 05:03:37 +00002651static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2652{
2653 return skb->tstamp;
2654}
2655
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002656/**
2657 * skb_get_timestamp - get timestamp from a skb
2658 * @skb: skb to get stamp from
2659 * @stamp: pointer to struct timeval to store stamp in
2660 *
2661 * Timestamps are stored in the skb as offsets to a base timestamp.
2662 * This function converts the offset back to a struct timeval and stores
2663 * it in stamp.
2664 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00002665static inline void skb_get_timestamp(const struct sk_buff *skb,
2666 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002667{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002668 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002669}
2670
Patrick Ohlyac45f602009-02-12 05:03:37 +00002671static inline void skb_get_timestampns(const struct sk_buff *skb,
2672 struct timespec *stamp)
2673{
2674 *stamp = ktime_to_timespec(skb->tstamp);
2675}
2676
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002677static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002678{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002679 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002680}
2681
Stephen Hemminger164891a2007-04-23 22:26:16 -07002682static inline ktime_t net_timedelta(ktime_t t)
2683{
2684 return ktime_sub(ktime_get_real(), t);
2685}
2686
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07002687static inline ktime_t net_invalid_timestamp(void)
2688{
2689 return ktime_set(0, 0);
2690}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002691
Richard Cochranc1f19b52010-07-17 08:49:36 +00002692#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2693
Joe Perches7965bd42013-09-26 14:48:15 -07002694void skb_clone_tx_timestamp(struct sk_buff *skb);
2695bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00002696
2697#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2698
2699static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2700{
2701}
2702
2703static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2704{
2705 return false;
2706}
2707
2708#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2709
2710/**
2711 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2712 *
Richard Cochranda92b192011-10-21 00:49:15 +00002713 * PHY drivers may accept clones of transmitted packets for
2714 * timestamping via their phy_driver.txtstamp method. These drivers
2715 * must call this function to return the skb back to the stack, with
2716 * or without a timestamp.
2717 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00002718 * @skb: clone of the the original outgoing packet
Richard Cochranda92b192011-10-21 00:49:15 +00002719 * @hwtstamps: hardware time stamps, may be NULL if not available
Richard Cochranc1f19b52010-07-17 08:49:36 +00002720 *
2721 */
2722void skb_complete_tx_timestamp(struct sk_buff *skb,
2723 struct skb_shared_hwtstamps *hwtstamps);
2724
Willem de Bruijne7fd2882014-08-04 22:11:48 -04002725void __skb_tstamp_tx(struct sk_buff *orig_skb,
2726 struct skb_shared_hwtstamps *hwtstamps,
2727 struct sock *sk, int tstype);
2728
Patrick Ohlyac45f602009-02-12 05:03:37 +00002729/**
2730 * skb_tstamp_tx - queue clone of skb with send time stamps
2731 * @orig_skb: the original outgoing packet
2732 * @hwtstamps: hardware time stamps, may be NULL if not available
2733 *
2734 * If the skb has a socket associated, then this function clones the
2735 * skb (thus sharing the actual data and optional structures), stores
2736 * the optional hardware time stamping information (if non NULL) or
2737 * generates a software time stamp (otherwise), then queues the clone
2738 * to the error queue of the socket. Errors are silently ignored.
2739 */
Joe Perches7965bd42013-09-26 14:48:15 -07002740void skb_tstamp_tx(struct sk_buff *orig_skb,
2741 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00002742
Richard Cochran4507a712010-07-17 08:48:28 +00002743static inline void sw_tx_timestamp(struct sk_buff *skb)
2744{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002745 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2746 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00002747 skb_tstamp_tx(skb, NULL);
2748}
2749
2750/**
2751 * skb_tx_timestamp() - Driver hook for transmit timestamping
2752 *
2753 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00002754 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00002755 *
David S. Miller73409f32013-12-27 13:04:33 -05002756 * Specifically, one should make absolutely sure that this function is
2757 * called before TX completion of this packet can trigger. Otherwise
2758 * the packet could potentially already be freed.
2759 *
Richard Cochran4507a712010-07-17 08:48:28 +00002760 * @skb: A socket buffer.
2761 */
2762static inline void skb_tx_timestamp(struct sk_buff *skb)
2763{
Richard Cochranc1f19b52010-07-17 08:49:36 +00002764 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002765 sw_tx_timestamp(skb);
2766}
2767
Johannes Berg6e3e9392011-11-09 10:15:42 +01002768/**
2769 * skb_complete_wifi_ack - deliver skb with wifi status
2770 *
2771 * @skb: the original outgoing packet
2772 * @acked: ack status
2773 *
2774 */
2775void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2776
Joe Perches7965bd42013-09-26 14:48:15 -07002777__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2778__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002779
Herbert Xu60476372007-04-09 11:59:39 -07002780static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2781{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002782 return ((skb->ip_summed & CHECKSUM_UNNECESSARY) || skb->csum_valid);
Herbert Xu60476372007-04-09 11:59:39 -07002783}
2784
Herbert Xufb286bb2005-11-10 13:01:24 -08002785/**
2786 * skb_checksum_complete - Calculate checksum of an entire packet
2787 * @skb: packet to process
2788 *
2789 * This function calculates the checksum over the entire packet plus
2790 * the value of skb->csum. The latter can be used to supply the
2791 * checksum of a pseudo header as used by TCP/UDP. It returns the
2792 * checksum.
2793 *
2794 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2795 * this function can be used to verify that checksum on received
2796 * packets. In that case the function should return zero if the
2797 * checksum is correct. In particular, this function will return zero
2798 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2799 * hardware has already verified the correctness of the checksum.
2800 */
Al Viro4381ca32007-07-15 21:00:11 +01002801static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002802{
Herbert Xu60476372007-04-09 11:59:39 -07002803 return skb_csum_unnecessary(skb) ?
2804 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002805}
2806
Tom Herbert77cffe22014-08-27 21:26:46 -07002807static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
2808{
2809 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2810 if (skb->csum_level == 0)
2811 skb->ip_summed = CHECKSUM_NONE;
2812 else
2813 skb->csum_level--;
2814 }
2815}
2816
2817static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
2818{
2819 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2820 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
2821 skb->csum_level++;
2822 } else if (skb->ip_summed == CHECKSUM_NONE) {
2823 skb->ip_summed = CHECKSUM_UNNECESSARY;
2824 skb->csum_level = 0;
2825 }
2826}
2827
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002828/* Check if we need to perform checksum complete validation.
2829 *
2830 * Returns true if checksum complete is needed, false otherwise
2831 * (either checksum is unnecessary or zero checksum is allowed).
2832 */
2833static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
2834 bool zero_okay,
2835 __sum16 check)
2836{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002837 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
2838 skb->csum_valid = 1;
Tom Herbert77cffe22014-08-27 21:26:46 -07002839 __skb_decr_checksum_unnecessary(skb);
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002840 return false;
2841 }
2842
2843 return true;
2844}
2845
2846/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
2847 * in checksum_init.
2848 */
2849#define CHECKSUM_BREAK 76
2850
2851/* Validate (init) checksum based on checksum complete.
2852 *
2853 * Return values:
2854 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
2855 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
2856 * checksum is stored in skb->csum for use in __skb_checksum_complete
2857 * non-zero: value of invalid checksum
2858 *
2859 */
2860static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
2861 bool complete,
2862 __wsum psum)
2863{
2864 if (skb->ip_summed == CHECKSUM_COMPLETE) {
2865 if (!csum_fold(csum_add(psum, skb->csum))) {
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002866 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002867 return 0;
2868 }
2869 }
2870
2871 skb->csum = psum;
2872
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002873 if (complete || skb->len <= CHECKSUM_BREAK) {
2874 __sum16 csum;
2875
2876 csum = __skb_checksum_complete(skb);
2877 skb->csum_valid = !csum;
2878 return csum;
2879 }
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002880
2881 return 0;
2882}
2883
2884static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
2885{
2886 return 0;
2887}
2888
2889/* Perform checksum validate (init). Note that this is a macro since we only
2890 * want to calculate the pseudo header which is an input function if necessary.
2891 * First we try to validate without any computation (checksum unnecessary) and
2892 * then calculate based on checksum complete calling the function to compute
2893 * pseudo header.
2894 *
2895 * Return values:
2896 * 0: checksum is validated or try to in skb_checksum_complete
2897 * non-zero: value of invalid checksum
2898 */
2899#define __skb_checksum_validate(skb, proto, complete, \
2900 zero_okay, check, compute_pseudo) \
2901({ \
2902 __sum16 __ret = 0; \
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002903 skb->csum_valid = 0; \
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002904 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
2905 __ret = __skb_checksum_validate_complete(skb, \
2906 complete, compute_pseudo(skb, proto)); \
2907 __ret; \
2908})
2909
2910#define skb_checksum_init(skb, proto, compute_pseudo) \
2911 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
2912
2913#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
2914 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
2915
2916#define skb_checksum_validate(skb, proto, compute_pseudo) \
2917 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
2918
2919#define skb_checksum_validate_zero_check(skb, proto, check, \
2920 compute_pseudo) \
2921 __skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
2922
2923#define skb_checksum_simple_validate(skb) \
2924 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
2925
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002926#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07002927void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002928static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2929{
2930 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002931 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002932}
2933static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2934{
2935 if (nfct)
2936 atomic_inc(&nfct->use);
2937}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002938#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002939#ifdef CONFIG_BRIDGE_NETFILTER
2940static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2941{
2942 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2943 kfree(nf_bridge);
2944}
2945static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2946{
2947 if (nf_bridge)
2948 atomic_inc(&nf_bridge->use);
2949}
2950#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08002951static inline void nf_reset(struct sk_buff *skb)
2952{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002953#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08002954 nf_conntrack_put(skb->nfct);
2955 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002956#endif
Patrick McHardya193a4a2006-03-20 19:23:05 -08002957#ifdef CONFIG_BRIDGE_NETFILTER
2958 nf_bridge_put(skb->nf_bridge);
2959 skb->nf_bridge = NULL;
2960#endif
2961}
2962
Patrick McHardy124dff02013-04-05 20:42:05 +02002963static inline void nf_reset_trace(struct sk_buff *skb)
2964{
Florian Westphal478b3602014-02-15 23:48:45 +01002965#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00002966 skb->nf_trace = 0;
2967#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002968}
2969
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002970/* Note: This doesn't put any conntrack and bridge info in dst. */
2971static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2972{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002973#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002974 dst->nfct = src->nfct;
2975 nf_conntrack_get(src->nfct);
2976 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002977#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002978#ifdef CONFIG_BRIDGE_NETFILTER
2979 dst->nf_bridge = src->nf_bridge;
2980 nf_bridge_get(src->nf_bridge);
2981#endif
Florian Westphal478b3602014-02-15 23:48:45 +01002982#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
2983 dst->nf_trace = src->nf_trace;
2984#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002985}
2986
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002987static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2988{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002989#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002990 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002991#endif
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002992#ifdef CONFIG_BRIDGE_NETFILTER
2993 nf_bridge_put(dst->nf_bridge);
2994#endif
2995 __nf_copy(dst, src);
2996}
2997
James Morris984bc162006-06-09 00:29:17 -07002998#ifdef CONFIG_NETWORK_SECMARK
2999static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3000{
3001 to->secmark = from->secmark;
3002}
3003
3004static inline void skb_init_secmark(struct sk_buff *skb)
3005{
3006 skb->secmark = 0;
3007}
3008#else
3009static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3010{ }
3011
3012static inline void skb_init_secmark(struct sk_buff *skb)
3013{ }
3014#endif
3015
Eric W. Biederman574f7192014-04-01 12:20:24 -07003016static inline bool skb_irq_freeable(const struct sk_buff *skb)
3017{
3018 return !skb->destructor &&
3019#if IS_ENABLED(CONFIG_XFRM)
3020 !skb->sp &&
3021#endif
3022#if IS_ENABLED(CONFIG_NF_CONNTRACK)
3023 !skb->nfct &&
3024#endif
3025 !skb->_skb_refdst &&
3026 !skb_has_frag_list(skb);
3027}
3028
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003029static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3030{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003031 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003032}
3033
Stephen Hemminger92477442009-03-21 13:39:26 -07003034static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003035{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003036 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003037}
3038
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003039static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3040{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003041 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003042}
3043
David S. Millerd5a9e242009-01-27 16:22:11 -08003044static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3045{
3046 skb->queue_mapping = rx_queue + 1;
3047}
3048
Stephen Hemminger92477442009-03-21 13:39:26 -07003049static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003050{
3051 return skb->queue_mapping - 1;
3052}
3053
Stephen Hemminger92477442009-03-21 13:39:26 -07003054static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003055{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003056 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08003057}
3058
Tom Herbert0e001612014-07-01 21:32:27 -07003059u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
Joe Perches7965bd42013-09-26 14:48:15 -07003060 unsigned int num_tx_queues);
Stephen Hemminger92477442009-03-21 13:39:26 -07003061
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003062static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3063{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003064#ifdef CONFIG_XFRM
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003065 return skb->sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003066#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003067 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003068#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003069}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003070
Pravin B Shelar68c33162013-02-14 14:02:41 +00003071/* Keeps track of mac header offset relative to skb->head.
3072 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3073 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07003074 * tunnel skb it points to outer mac header.
3075 * Keeps track of level of encapsulation of network headers.
3076 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00003077struct skb_gso_cb {
Eric Dumazet3347c962013-10-19 11:42:56 -07003078 int mac_offset;
3079 int encap_level;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003080 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00003081};
3082#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3083
3084static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3085{
3086 return (skb_mac_header(inner_skb) - inner_skb->head) -
3087 SKB_GSO_CB(inner_skb)->mac_offset;
3088}
3089
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00003090static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3091{
3092 int new_headroom, headroom;
3093 int ret;
3094
3095 headroom = skb_headroom(skb);
3096 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3097 if (ret)
3098 return ret;
3099
3100 new_headroom = skb_headroom(skb);
3101 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3102 return 0;
3103}
3104
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003105/* Compute the checksum for a gso segment. First compute the checksum value
3106 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3107 * then add in skb->csum (checksum from csum_start to end of packet).
3108 * skb->csum and csum_start are then updated to reflect the checksum of the
3109 * resultant packet starting from the transport header-- the resultant checksum
3110 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3111 * header.
3112 */
3113static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3114{
3115 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3116 skb_transport_offset(skb);
3117 __u16 csum;
3118
3119 csum = csum_fold(csum_partial(skb_transport_header(skb),
3120 plen, skb->csum));
3121 skb->csum = res;
3122 SKB_GSO_CB(skb)->csum_start -= plen;
3123
3124 return csum;
3125}
3126
David S. Millerbdcc0922012-03-07 20:53:36 -05003127static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07003128{
3129 return skb_shinfo(skb)->gso_size;
3130}
3131
Eric Dumazet36a8f392013-09-29 01:21:32 -07003132/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05003133static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02003134{
3135 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3136}
3137
Joe Perches7965bd42013-09-26 14:48:15 -07003138void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07003139
3140static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3141{
3142 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3143 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003144 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3145
Alexander Duyckb78462e2010-06-02 12:24:37 +00003146 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3147 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07003148 __skb_warn_lro_forwarding(skb);
3149 return true;
3150 }
3151 return false;
3152}
3153
Herbert Xu35fc92a2007-03-26 23:22:20 -07003154static inline void skb_forward_csum(struct sk_buff *skb)
3155{
3156 /* Unfortunately we don't support this one. Any brave souls? */
3157 if (skb->ip_summed == CHECKSUM_COMPLETE)
3158 skb->ip_summed = CHECKSUM_NONE;
3159}
3160
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003161/**
3162 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3163 * @skb: skb to check
3164 *
3165 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3166 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3167 * use this helper, to document places where we make this assertion.
3168 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003169static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003170{
3171#ifdef DEBUG
3172 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3173#endif
3174}
3175
Rusty Russellf35d9d82008-02-04 23:49:54 -05003176bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00003177
Paul Durranted1f50c2014-01-09 10:02:46 +00003178int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3179
Daniel Borkmannf77668d2013-03-19 06:39:30 +00003180u32 __skb_get_poff(const struct sk_buff *skb);
3181
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00003182/**
3183 * skb_head_is_locked - Determine if the skb->head is locked down
3184 * @skb: skb to check
3185 *
3186 * The head on skbs build around a head frag can be removed if they are
3187 * not cloned. This function returns true if the skb head is locked down
3188 * due to either being allocated via kmalloc, or by being a clone with
3189 * multiple references to the head.
3190 */
3191static inline bool skb_head_is_locked(const struct sk_buff *skb)
3192{
3193 return !skb->head_frag || skb_cloned(skb);
3194}
Florian Westphalfe6cc552014-02-13 23:09:12 +01003195
3196/**
3197 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3198 *
3199 * @skb: GSO skb
3200 *
3201 * skb_gso_network_seglen is used to determine the real size of the
3202 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3203 *
3204 * The MAC/L2 header is not accounted for.
3205 */
3206static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3207{
3208 unsigned int hdr_len = skb_transport_header(skb) -
3209 skb_network_header(skb);
3210 return hdr_len + skb_gso_transport_seglen(skb);
3211}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003212#endif /* __KERNEL__ */
3213#endif /* _LINUX_SKBUFF_H */