blob: 50e1e9b3a5a54ff3039c6cc02d070e56fa249d13 [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
50 * for specific protocols e.g. TCP/UDP/SCTP, then, for such packets it will
51 * set CHECKSUM_UNNECESSARY if their checksums are okay. skb->csum is still
52 * undefined in this case though. It is a bad option, but, unfortunately,
53 * nowadays most vendors do this. Apparently with the secret goal to sell
54 * you new devices, when you will add new protocol to your host, f.e. IPv6 8)
55 *
56 * CHECKSUM_COMPLETE:
57 *
58 * This is the most generic way. The device supplied checksum of the _whole_
59 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
60 * hardware doesn't need to parse L3/L4 headers to implement this.
61 *
62 * Note: Even if device supports only some protocols, but is able to produce
63 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
64 *
65 * CHECKSUM_PARTIAL:
66 *
67 * This is identical to the case for output below. This may occur on a packet
68 * received directly from another Linux OS, e.g., a virtualized Linux kernel
69 * on the same host. The packet can be treated in the same way as
70 * CHECKSUM_UNNECESSARY, except that on output (i.e., forwarding) the
71 * checksum must be filled in by the OS or the hardware.
72 *
73 * B. Checksumming on output.
74 *
75 * CHECKSUM_NONE:
76 *
77 * The skb was already checksummed by the protocol, or a checksum is not
78 * required.
79 *
80 * CHECKSUM_PARTIAL:
81 *
82 * The device is required to checksum the packet as seen by hard_start_xmit()
83 * from skb->csum_start up to the end, and to record/write the checksum at
84 * offset skb->csum_start + skb->csum_offset.
85 *
86 * The device must show its capabilities in dev->features, set up at device
87 * setup time, e.g. netdev_features.h:
88 *
89 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
90 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
91 * IPv4. Sigh. Vendors like this way for an unknown reason.
92 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
93 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
94 * NETIF_F_... - Well, you get the picture.
95 *
96 * CHECKSUM_UNNECESSARY:
97 *
98 * Normally, the device will do per protocol specific checksumming. Protocol
99 * implementations that do not want the NIC to perform the checksum
100 * calculation should use this flag in their outgoing skbs.
101 *
102 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
103 * offload. Correspondingly, the FCoE protocol driver
104 * stack should use CHECKSUM_UNNECESSARY.
105 *
106 * Any questions? No questions, good. --ANK
107 */
108
Herbert Xu60476372007-04-09 11:59:39 -0700109/* Don't change this without changing skb_csum_unnecessary! */
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100110#define CHECKSUM_NONE 0
111#define CHECKSUM_UNNECESSARY 1
112#define CHECKSUM_COMPLETE 2
113#define CHECKSUM_PARTIAL 3
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114
Tobias Klauser0bec8c82014-07-22 12:06:23 +0200115#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
David S. Millerfc910a22007-03-25 20:27:59 -0700116#define SKB_WITH_OVERHEAD(X) \
Herbert Xudeea84b2007-10-21 16:27:46 -0700117 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
David S. Millerfc910a22007-03-25 20:27:59 -0700118#define SKB_MAX_ORDER(X, ORDER) \
119 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
121#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
122
Eric Dumazet87fb4b72011-10-13 07:28:54 +0000123/* return minimum truesize of one skb containing X bytes of data */
124#define SKB_TRUESIZE(X) ((X) + \
125 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
126 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
127
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128struct net_device;
David Howells716ea3a2007-04-02 20:19:53 -0700129struct scatterlist;
Jens Axboe9c55e012007-11-06 23:30:13 -0800130struct pipe_inode_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700132#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133struct nf_conntrack {
134 atomic_t use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135};
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700136#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137
138#ifdef CONFIG_BRIDGE_NETFILTER
139struct nf_bridge_info {
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000140 atomic_t use;
141 unsigned int mask;
142 struct net_device *physindev;
143 struct net_device *physoutdev;
144 unsigned long data[32 / sizeof(unsigned long)];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145};
146#endif
147
Linus Torvalds1da177e2005-04-16 15:20:36 -0700148struct sk_buff_head {
149 /* These two members must be first. */
150 struct sk_buff *next;
151 struct sk_buff *prev;
152
153 __u32 qlen;
154 spinlock_t lock;
155};
156
157struct sk_buff;
158
Ian Campbell9d4dde52011-12-22 23:39:14 +0000159/* To allow 64K frame to be packed as single skb without frag_list we
160 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
161 * buffers which do not start on a page boundary.
162 *
163 * Since GRO uses frags we allocate at least 16 regardless of page
164 * size.
Anton Blancharda715dea2011-03-27 14:57:26 +0000165 */
Ian Campbell9d4dde52011-12-22 23:39:14 +0000166#if (65536/PAGE_SIZE + 1) < 16
David S. Millereec00952011-03-29 23:34:08 -0700167#define MAX_SKB_FRAGS 16UL
Anton Blancharda715dea2011-03-27 14:57:26 +0000168#else
Ian Campbell9d4dde52011-12-22 23:39:14 +0000169#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
Anton Blancharda715dea2011-03-27 14:57:26 +0000170#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171
172typedef struct skb_frag_struct skb_frag_t;
173
174struct skb_frag_struct {
Ian Campbella8605c62011-10-19 23:01:49 +0000175 struct {
176 struct page *p;
177 } page;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000178#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
David S. Millera309bb02007-07-30 18:47:03 -0700179 __u32 page_offset;
180 __u32 size;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000181#else
182 __u16 page_offset;
183 __u16 size;
184#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185};
186
Eric Dumazet9e903e02011-10-18 21:00:24 +0000187static inline unsigned int skb_frag_size(const skb_frag_t *frag)
188{
189 return frag->size;
190}
191
192static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
193{
194 frag->size = size;
195}
196
197static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
198{
199 frag->size += delta;
200}
201
202static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
203{
204 frag->size -= delta;
205}
206
Patrick Ohlyac45f602009-02-12 05:03:37 +0000207#define HAVE_HW_TIME_STAMP
208
209/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800210 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000211 * @hwtstamp: hardware time stamp transformed into duration
212 * since arbitrary point in time
Patrick Ohlyac45f602009-02-12 05:03:37 +0000213 *
214 * Software time stamps generated by ktime_get_real() are stored in
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400215 * skb->tstamp.
Patrick Ohlyac45f602009-02-12 05:03:37 +0000216 *
217 * hwtstamps can only be compared against other hwtstamps from
218 * the same device.
219 *
220 * This structure is attached to packets as part of the
221 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
222 */
223struct skb_shared_hwtstamps {
224 ktime_t hwtstamp;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000225};
226
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000227/* Definitions for tx_flags in struct skb_shared_info */
228enum {
229 /* generate hardware time stamp */
230 SKBTX_HW_TSTAMP = 1 << 0,
231
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400232 /* generate software time stamp when queueing packet to NIC */
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000233 SKBTX_SW_TSTAMP = 1 << 1,
234
235 /* device driver is going to provide hardware time stamp */
236 SKBTX_IN_PROGRESS = 1 << 2,
237
Shirley Maa6686f22011-07-06 12:22:12 +0000238 /* device driver supports TX zero-copy buffers */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000239 SKBTX_DEV_ZEROCOPY = 1 << 3,
Johannes Berg6e3e9392011-11-09 10:15:42 +0100240
241 /* generate wifi status information (where possible) */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000242 SKBTX_WIFI_STATUS = 1 << 4,
Pravin B Shelarc9af6db2013-02-11 09:27:41 +0000243
244 /* This indicates at least one fragment might be overwritten
245 * (as in vmsplice(), sendfile() ...)
246 * If we need to compute a TX checksum, we'll need to copy
247 * all frags to avoid possible bad checksum
248 */
249 SKBTX_SHARED_FRAG = 1 << 5,
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400250
251 /* generate software time stamp when entering packet scheduling */
252 SKBTX_SCHED_TSTAMP = 1 << 6,
Shirley Maa6686f22011-07-06 12:22:12 +0000253};
254
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400255#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | SKBTX_SCHED_TSTAMP)
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400256#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
257
Shirley Maa6686f22011-07-06 12:22:12 +0000258/*
259 * The callback notifies userspace to release buffers when skb DMA is done in
260 * lower device, the skb last reference should be 0 when calling this.
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000261 * The zerocopy_success argument is true if zero copy transmit occurred,
262 * false on data copy or out of memory error caused by data copy attempt.
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000263 * The ctx field is used to track device context.
264 * The desc field is used to track userspace buffer index.
Shirley Maa6686f22011-07-06 12:22:12 +0000265 */
266struct ubuf_info {
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000267 void (*callback)(struct ubuf_info *, bool zerocopy_success);
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000268 void *ctx;
Shirley Maa6686f22011-07-06 12:22:12 +0000269 unsigned long desc;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000270};
271
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272/* This data is invariant across clones and lives at
273 * the end of the header data, ie. at skb->end.
274 */
275struct skb_shared_info {
Ian Campbell9f42f122012-01-05 07:13:39 +0000276 unsigned char nr_frags;
277 __u8 tx_flags;
Herbert Xu79671682006-06-22 02:40:14 -0700278 unsigned short gso_size;
279 /* Warning: this field is not always filled in (UFO)! */
280 unsigned short gso_segs;
281 unsigned short gso_type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000283 struct skb_shared_hwtstamps hwtstamps;
Willem de Bruijn09c2d252014-08-04 22:11:47 -0400284 u32 tskey;
Ian Campbell9f42f122012-01-05 07:13:39 +0000285 __be32 ip6_frag_id;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700286
287 /*
288 * Warning : all fields before dataref are cleared in __alloc_skb()
289 */
290 atomic_t dataref;
291
Johann Baudy69e3c752009-05-18 22:11:22 -0700292 /* Intermediate layers must ensure that destructor_arg
293 * remains valid until skb destructor */
294 void * destructor_arg;
Shirley Maa6686f22011-07-06 12:22:12 +0000295
Eric Dumazetfed66382010-07-22 19:09:08 +0000296 /* must be last field, see pskb_expand_head() */
297 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700298};
299
300/* We divide dataref into two halves. The higher 16 bits hold references
301 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700302 * the entire skb->data. A clone of a headerless skb holds the length of
303 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304 *
305 * All users must obey the rule that the skb->data reference count must be
306 * greater than or equal to the payload reference count.
307 *
308 * Holding a reference to the payload part means that the user does not
309 * care about modifications to the header part of skb->data.
310 */
311#define SKB_DATAREF_SHIFT 16
312#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
313
David S. Millerd179cd12005-08-17 14:57:30 -0700314
315enum {
316 SKB_FCLONE_UNAVAILABLE,
317 SKB_FCLONE_ORIG,
318 SKB_FCLONE_CLONE,
319};
320
Herbert Xu79671682006-06-22 02:40:14 -0700321enum {
322 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700323 SKB_GSO_UDP = 1 << 1,
Herbert Xu576a30e2006-06-27 13:22:38 -0700324
325 /* This indicates the skb is from an untrusted source. */
326 SKB_GSO_DODGY = 1 << 2,
Michael Chanb0da85372006-06-29 12:30:00 -0700327
328 /* This indicates the tcp segment has CWR set. */
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700329 SKB_GSO_TCP_ECN = 1 << 3,
330
331 SKB_GSO_TCPV6 = 1 << 4,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800332
333 SKB_GSO_FCOE = 1 << 5,
Pravin B Shelar68c33162013-02-14 14:02:41 +0000334
335 SKB_GSO_GRE = 1 << 6,
Pravin B Shelar73136262013-03-07 13:21:51 +0000336
Tom Herbert4b282522014-06-14 23:23:52 -0700337 SKB_GSO_GRE_CSUM = 1 << 7,
Simon Horman0d89d202013-05-23 21:02:52 +0000338
Tom Herbert4b282522014-06-14 23:23:52 -0700339 SKB_GSO_IPIP = 1 << 8,
Eric Dumazetcb32f512013-10-19 11:42:57 -0700340
Tom Herbert4b282522014-06-14 23:23:52 -0700341 SKB_GSO_SIT = 1 << 9,
Eric Dumazet61c1db72013-10-20 20:47:30 -0700342
Tom Herbert4b282522014-06-14 23:23:52 -0700343 SKB_GSO_UDP_TUNNEL = 1 << 10,
Tom Herbert0f4f4ff2014-06-04 17:20:16 -0700344
345 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
Tom Herbert4749c092014-06-04 17:20:23 -0700346
Tom Herbert4b282522014-06-14 23:23:52 -0700347 SKB_GSO_MPLS = 1 << 12,
348
Herbert Xu79671682006-06-22 02:40:14 -0700349};
350
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700351#if BITS_PER_LONG > 32
352#define NET_SKBUFF_DATA_USES_OFFSET 1
353#endif
354
355#ifdef NET_SKBUFF_DATA_USES_OFFSET
356typedef unsigned int sk_buff_data_t;
357#else
358typedef unsigned char *sk_buff_data_t;
359#endif
360
Eric Dumazet363ec39232014-02-26 14:02:11 -0800361/**
362 * struct skb_mstamp - multi resolution time stamps
363 * @stamp_us: timestamp in us resolution
364 * @stamp_jiffies: timestamp in jiffies
365 */
366struct skb_mstamp {
367 union {
368 u64 v64;
369 struct {
370 u32 stamp_us;
371 u32 stamp_jiffies;
372 };
373 };
374};
375
376/**
377 * skb_mstamp_get - get current timestamp
378 * @cl: place to store timestamps
379 */
380static inline void skb_mstamp_get(struct skb_mstamp *cl)
381{
382 u64 val = local_clock();
383
384 do_div(val, NSEC_PER_USEC);
385 cl->stamp_us = (u32)val;
386 cl->stamp_jiffies = (u32)jiffies;
387}
388
389/**
390 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
391 * @t1: pointer to newest sample
392 * @t0: pointer to oldest sample
393 */
394static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
395 const struct skb_mstamp *t0)
396{
397 s32 delta_us = t1->stamp_us - t0->stamp_us;
398 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
399
400 /* If delta_us is negative, this might be because interval is too big,
401 * or local_clock() drift is too big : fallback using jiffies.
402 */
403 if (delta_us <= 0 ||
404 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
405
406 delta_us = jiffies_to_usecs(delta_jiffies);
407
408 return delta_us;
409}
410
411
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412/**
413 * struct sk_buff - socket buffer
414 * @next: Next buffer in list
415 * @prev: Previous buffer in list
Eric Dumazet363ec39232014-02-26 14:02:11 -0800416 * @tstamp: Time we arrived/left
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700417 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700418 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700419 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000420 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700421 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422 * @len: Length of actual data
423 * @data_len: Data length
424 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700425 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700426 * @csum: Checksum (must include start/offset pair)
427 * @csum_start: Offset from skb->head where checksumming should start
428 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700429 * @priority: Packet queueing priority
WANG Cong60ff7462014-05-04 16:39:18 -0700430 * @ignore_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700431 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700432 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700433 * @nohdr: Payload reference only, must not modify header
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700434 * @nfctinfo: Relationship of this skb to the connection
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700436 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700437 * @ipvs_property: skbuff is owned by ipvs
Randy Dunlap31729362008-02-18 20:52:13 -0800438 * @peeked: this packet has been seen already, so stats have been
439 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700440 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700441 * @protocol: Packet protocol from driver
442 * @destructor: Destruct function
443 * @nfct: Associated connection, if any
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800445 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700446 * @tc_index: Traffic control index
447 * @tc_verd: traffic control verdict
Tom Herbert61b905d2014-03-24 15:34:47 -0700448 * @hash: the packet hash
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700449 * @queue_mapping: Queue mapping for multiqueue devices
Randy Dunlap553a5672008-04-20 10:51:01 -0700450 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700451 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Tom Herbert61b905d2014-03-24 15:34:47 -0700452 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
Changli Gao4ca24622011-08-19 07:26:44 -0700453 * ports.
Tom Herberta3b18dd2014-07-01 21:33:17 -0700454 * @sw_hash: indicates hash was computed in software stack
Johannes Berg6e3e9392011-11-09 10:15:42 +0100455 * @wifi_acked_valid: wifi_acked was set
456 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000457 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Randy Dunlapf4b8ea72006-06-22 16:00:11 -0700458 * @dma_cookie: a cookie to one of several possible DMA operations
459 * done by skb DMA functions
Eliezer Tamir06021292013-06-10 11:39:50 +0300460 * @napi_id: id of the NAPI struct this skb came from
James Morris984bc162006-06-09 00:29:17 -0700461 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700462 * @mark: Generic packet mark
463 * @dropcount: total number of sk_receive_queue overflows
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000464 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b2008-07-14 22:49:06 -0700465 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000466 * @inner_protocol: Protocol (encapsulation)
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000467 * @inner_transport_header: Inner transport layer header (encapsulation)
468 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000469 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700470 * @transport_header: Transport layer header
471 * @network_header: Network layer header
472 * @mac_header: Link layer header
473 * @tail: Tail pointer
474 * @end: End pointer
475 * @head: Head of buffer
476 * @data: Data head pointer
477 * @truesize: Buffer size
478 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479 */
480
481struct sk_buff {
482 /* These two members must be first. */
483 struct sk_buff *next;
484 struct sk_buff *prev;
485
Eric Dumazet363ec39232014-02-26 14:02:11 -0800486 union {
487 ktime_t tstamp;
488 struct skb_mstamp skb_mstamp;
489 };
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000490
491 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 /*
495 * This is the control buffer. It is free to use for every
496 * layer. Please put your private variables there. If you
497 * want to keep them across layers you have to do a skb_clone()
498 * first. This is owned by whoever has the skb queued ATM.
499 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000500 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700501
Eric Dumazet7fee2262010-05-11 23:19:48 +0000502 unsigned long _skb_refdst;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000503#ifdef CONFIG_XFRM
504 struct sec_path *sp;
505#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700507 data_len;
508 __u16 mac_len,
509 hdr_len;
Al Viroff1dcad2006-11-20 18:07:29 -0800510 union {
511 __wsum csum;
Herbert Xu663ead32007-04-09 11:59:07 -0700512 struct {
513 __u16 csum_start;
514 __u16 csum_offset;
515 };
Al Viroff1dcad2006-11-20 18:07:29 -0800516 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700517 __u32 priority;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200518 kmemcheck_bitfield_begin(flags1);
WANG Cong60ff7462014-05-04 16:39:18 -0700519 __u8 ignore_df:1,
Thomas Graf1cbb3382005-07-05 14:13:41 -0700520 cloned:1,
521 ip_summed:2,
Harald Welte6869c4d2005-08-09 19:24:19 -0700522 nohdr:1,
523 nfctinfo:3;
David S. Millerd179cd12005-08-17 14:57:30 -0700524 __u8 pkt_type:3,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800525 fclone:2,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700526 ipvs_property:1,
Herbert Xua59322b2007-12-05 01:53:40 -0800527 peeked:1,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700528 nf_trace:1;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200529 kmemcheck_bitfield_end(flags1);
Eric Dumazet4ab408d2010-03-01 03:09:26 +0000530 __be16 protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531
532 void (*destructor)(struct sk_buff *skb);
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800533#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700534 struct nf_conntrack *nfct;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100535#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536#ifdef CONFIG_BRIDGE_NETFILTER
537 struct nf_bridge_info *nf_bridge;
538#endif
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700539
Eric Dumazet8964be42009-11-20 15:35:04 -0800540 int skb_iif;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000541
Tom Herbert61b905d2014-03-24 15:34:47 -0700542 __u32 hash;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000543
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000544 __be16 vlan_proto;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000545 __u16 vlan_tci;
546
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700548 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700550 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200553
Eric Dumazet0a148422011-04-20 09:27:32 +0000554 __u16 queue_mapping;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200555 kmemcheck_bitfield_begin(flags2);
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400556#ifdef CONFIG_IPV6_NDISC_NODETYPE
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000557 __u8 ndisc_nodetype:2;
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400558#endif
Mel Gormanc93bdd02012-07-31 16:44:19 -0700559 __u8 pfmemalloc:1;
Tom Herbert3853b582010-11-21 13:17:29 +0000560 __u8 ooo_okay:1;
Tom Herbert61b905d2014-03-24 15:34:47 -0700561 __u8 l4_hash:1;
Tom Herberta3b18dd2014-07-01 21:33:17 -0700562 __u8 sw_hash:1;
Johannes Berg6e3e9392011-11-09 10:15:42 +0100563 __u8 wifi_acked_valid:1;
564 __u8 wifi_acked:1;
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000565 __u8 no_fcs:1;
Eric Dumazetd3836f22012-04-27 00:33:38 +0000566 __u8 head_frag:1;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000567 /* Encapsulation protocol and NIC drivers should use
568 * this flag to indicate to each other if the skb contains
569 * encapsulated packet or not and maybe use the inner packet
570 * headers if needed
571 */
572 __u8 encapsulation:1;
Tom Herbert7e2b10c2014-06-04 17:20:02 -0700573 __u8 encap_hdr_csum:1;
Tom Herbert5d0c2b92014-06-10 18:54:13 -0700574 __u8 csum_valid:1;
Tom Herbert7e3cead2014-06-10 18:54:19 -0700575 __u8 csum_complete_sw:1;
Tom Herberta3b18dd2014-07-01 21:33:17 -0700576 /* 2/4 bit hole (depending on ndisc_nodetype presence) */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200577 kmemcheck_bitfield_end(flags2);
578
Cong Wange0d10952013-08-01 11:10:25 +0800579#if defined CONFIG_NET_DMA || defined CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir06021292013-06-10 11:39:50 +0300580 union {
581 unsigned int napi_id;
582 dma_cookie_t dma_cookie;
583 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700584#endif
James Morris984bc162006-06-09 00:29:17 -0700585#ifdef CONFIG_NETWORK_SECMARK
586 __u32 secmark;
587#endif
Neil Horman3b885782009-10-12 13:26:31 -0700588 union {
589 __u32 mark;
590 __u32 dropcount;
Eric Dumazet16fad692013-03-14 05:40:32 +0000591 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700592 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700593
Simon Horman0d89d202013-05-23 21:02:52 +0000594 __be16 inner_protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000595 __u16 inner_transport_header;
596 __u16 inner_network_header;
597 __u16 inner_mac_header;
598 __u16 transport_header;
599 __u16 network_header;
600 __u16 mac_header;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700602 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700603 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700605 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700606 unsigned int truesize;
607 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700608};
609
610#ifdef __KERNEL__
611/*
612 * Handling routines are only of interest to the kernel
613 */
614#include <linux/slab.h>
615
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616
Mel Gormanc93bdd02012-07-31 16:44:19 -0700617#define SKB_ALLOC_FCLONE 0x01
618#define SKB_ALLOC_RX 0x02
619
620/* Returns true if the skb was allocated from PFMEMALLOC reserves */
621static inline bool skb_pfmemalloc(const struct sk_buff *skb)
622{
623 return unlikely(skb->pfmemalloc);
624}
625
Eric Dumazet7fee2262010-05-11 23:19:48 +0000626/*
627 * skb might have a dst pointer attached, refcounted or not.
628 * _skb_refdst low order bit is set if refcount was _not_ taken
629 */
630#define SKB_DST_NOREF 1UL
631#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
632
633/**
634 * skb_dst - returns skb dst_entry
635 * @skb: buffer
636 *
637 * Returns skb dst_entry, regardless of reference taken or not.
638 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000639static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
640{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000641 /* If refdst was not refcounted, check we still are in a
642 * rcu_read_lock section
643 */
644 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
645 !rcu_read_lock_held() &&
646 !rcu_read_lock_bh_held());
647 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000648}
649
Eric Dumazet7fee2262010-05-11 23:19:48 +0000650/**
651 * skb_dst_set - sets skb dst
652 * @skb: buffer
653 * @dst: dst entry
654 *
655 * Sets skb dst, assuming a reference was taken on dst and should
656 * be released by skb_dst_drop()
657 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000658static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
659{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000660 skb->_skb_refdst = (unsigned long)dst;
661}
662
Joe Perches7965bd42013-09-26 14:48:15 -0700663void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
664 bool force);
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200665
666/**
667 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
668 * @skb: buffer
669 * @dst: dst entry
670 *
671 * Sets skb dst, assuming a reference was not taken on dst.
672 * If dst entry is cached, we do not take reference and dst_release
673 * will be avoided by refdst_drop. If dst entry is not cached, we take
674 * reference, so that last dst_release can destroy the dst immediately.
675 */
676static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
677{
678 __skb_dst_set_noref(skb, dst, false);
679}
680
681/**
682 * skb_dst_set_noref_force - sets skb dst, without taking reference
683 * @skb: buffer
684 * @dst: dst entry
685 *
686 * Sets skb dst, assuming a reference was not taken on dst.
687 * No reference is taken and no dst_release will be called. While for
688 * cached dsts deferred reclaim is a basic feature, for entries that are
689 * not cached it is caller's job to guarantee that last dst_release for
690 * provided dst happens when nobody uses it, eg. after a RCU grace period.
691 */
692static inline void skb_dst_set_noref_force(struct sk_buff *skb,
693 struct dst_entry *dst)
694{
695 __skb_dst_set_noref(skb, dst, true);
696}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000697
698/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300699 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000700 * @skb: buffer
701 */
702static inline bool skb_dst_is_noref(const struct sk_buff *skb)
703{
704 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000705}
706
Eric Dumazet511c3f92009-06-02 05:14:27 +0000707static inline struct rtable *skb_rtable(const struct sk_buff *skb)
708{
Eric Dumazetadf30902009-06-02 05:19:30 +0000709 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000710}
711
Joe Perches7965bd42013-09-26 14:48:15 -0700712void kfree_skb(struct sk_buff *skb);
713void kfree_skb_list(struct sk_buff *segs);
714void skb_tx_error(struct sk_buff *skb);
715void consume_skb(struct sk_buff *skb);
716void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000717extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000718
Joe Perches7965bd42013-09-26 14:48:15 -0700719void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
720bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
721 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000722
Joe Perches7965bd42013-09-26 14:48:15 -0700723struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
724 int node);
725struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700726static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100727 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700728{
Eric Dumazet564824b2010-10-11 19:05:25 +0000729 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700730}
731
732static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100733 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700734{
Mel Gormanc93bdd02012-07-31 16:44:19 -0700735 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700736}
737
Joe Perches7965bd42013-09-26 14:48:15 -0700738struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
Patrick McHardy0ebd0ac2013-04-17 06:46:58 +0000739static inline struct sk_buff *alloc_skb_head(gfp_t priority)
740{
741 return __alloc_skb_head(priority, -1);
742}
743
Joe Perches7965bd42013-09-26 14:48:15 -0700744struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
745int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
746struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
747struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
Octavian Purdilabad93e92014-06-12 01:36:26 +0300748struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
749 gfp_t gfp_mask, bool fclone);
750static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
751 gfp_t gfp_mask)
752{
753 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
754}
Eric Dumazet117632e2011-12-03 21:39:53 +0000755
Joe Perches7965bd42013-09-26 14:48:15 -0700756int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
757struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
758 unsigned int headroom);
759struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
760 int newtailroom, gfp_t priority);
Fan Du25a91d82014-01-18 09:54:23 +0800761int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
762 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -0700763int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
764 int len);
765int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
766int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000767#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768
Joe Perches7965bd42013-09-26 14:48:15 -0700769int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
770 int getfrag(void *from, char *to, int offset,
771 int len, int odd, struct sk_buff *skb),
772 void *from, int length);
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700773
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800774struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700775 __u32 lower_offset;
776 __u32 upper_offset;
777 __u32 frag_idx;
778 __u32 stepped_offset;
779 struct sk_buff *root_skb;
780 struct sk_buff *cur_skb;
781 __u8 *frag_data;
782};
783
Joe Perches7965bd42013-09-26 14:48:15 -0700784void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
785 unsigned int to, struct skb_seq_state *st);
786unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
787 struct skb_seq_state *st);
788void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -0700789
Joe Perches7965bd42013-09-26 14:48:15 -0700790unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
791 unsigned int to, struct ts_config *config,
792 struct ts_state *state);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700793
Tom Herbert09323cc2013-12-15 22:16:19 -0800794/*
795 * Packet hash types specify the type of hash in skb_set_hash.
796 *
797 * Hash types refer to the protocol layer addresses which are used to
798 * construct a packet's hash. The hashes are used to differentiate or identify
799 * flows of the protocol layer for the hash type. Hash types are either
800 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
801 *
802 * Properties of hashes:
803 *
804 * 1) Two packets in different flows have different hash values
805 * 2) Two packets in the same flow should have the same hash value
806 *
807 * A hash at a higher layer is considered to be more specific. A driver should
808 * set the most specific hash possible.
809 *
810 * A driver cannot indicate a more specific hash than the layer at which a hash
811 * was computed. For instance an L3 hash cannot be set as an L4 hash.
812 *
813 * A driver may indicate a hash level which is less specific than the
814 * actual layer the hash was computed on. For instance, a hash computed
815 * at L4 may be considered an L3 hash. This should only be done if the
816 * driver can't unambiguously determine that the HW computed the hash at
817 * the higher layer. Note that the "should" in the second property above
818 * permits this.
819 */
820enum pkt_hash_types {
821 PKT_HASH_TYPE_NONE, /* Undefined type */
822 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
823 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
824 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
825};
826
827static inline void
828skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
829{
Tom Herbert61b905d2014-03-24 15:34:47 -0700830 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
Tom Herberta3b18dd2014-07-01 21:33:17 -0700831 skb->sw_hash = 0;
Tom Herbert61b905d2014-03-24 15:34:47 -0700832 skb->hash = hash;
Tom Herbert09323cc2013-12-15 22:16:19 -0800833}
834
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800835void __skb_get_hash(struct sk_buff *skb);
836static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000837{
Tom Herberta3b18dd2014-07-01 21:33:17 -0700838 if (!skb->l4_hash && !skb->sw_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800839 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000840
Tom Herbert61b905d2014-03-24 15:34:47 -0700841 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000842}
843
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800844static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
845{
Tom Herbert61b905d2014-03-24 15:34:47 -0700846 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800847}
848
Tom Herbert7539fad2013-12-15 22:12:18 -0800849static inline void skb_clear_hash(struct sk_buff *skb)
850{
Tom Herbert61b905d2014-03-24 15:34:47 -0700851 skb->hash = 0;
Tom Herberta3b18dd2014-07-01 21:33:17 -0700852 skb->sw_hash = 0;
Tom Herbert61b905d2014-03-24 15:34:47 -0700853 skb->l4_hash = 0;
Tom Herbert7539fad2013-12-15 22:12:18 -0800854}
855
856static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
857{
Tom Herbert61b905d2014-03-24 15:34:47 -0700858 if (!skb->l4_hash)
Tom Herbert7539fad2013-12-15 22:12:18 -0800859 skb_clear_hash(skb);
860}
861
Tom Herbert3df7a742013-12-15 22:16:29 -0800862static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
863{
Tom Herbert61b905d2014-03-24 15:34:47 -0700864 to->hash = from->hash;
Tom Herberta3b18dd2014-07-01 21:33:17 -0700865 to->sw_hash = from->sw_hash;
Tom Herbert61b905d2014-03-24 15:34:47 -0700866 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -0800867};
868
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700869#ifdef NET_SKBUFF_DATA_USES_OFFSET
870static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
871{
872 return skb->head + skb->end;
873}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000874
875static inline unsigned int skb_end_offset(const struct sk_buff *skb)
876{
877 return skb->end;
878}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700879#else
880static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
881{
882 return skb->end;
883}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000884
885static inline unsigned int skb_end_offset(const struct sk_buff *skb)
886{
887 return skb->end - skb->head;
888}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700889#endif
890
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700892#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893
Patrick Ohlyac45f602009-02-12 05:03:37 +0000894static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
895{
896 return &skb_shinfo(skb)->hwtstamps;
897}
898
Linus Torvalds1da177e2005-04-16 15:20:36 -0700899/**
900 * skb_queue_empty - check if a queue is empty
901 * @list: queue head
902 *
903 * Returns true if the queue is empty, false otherwise.
904 */
905static inline int skb_queue_empty(const struct sk_buff_head *list)
906{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100907 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700908}
909
910/**
David S. Millerfc7ebb22008-09-23 00:34:07 -0700911 * skb_queue_is_last - check if skb is the last entry in the queue
912 * @list: queue head
913 * @skb: buffer
914 *
915 * Returns true if @skb is the last buffer on the list.
916 */
917static inline bool skb_queue_is_last(const struct sk_buff_head *list,
918 const struct sk_buff *skb)
919{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100920 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -0700921}
922
923/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800924 * skb_queue_is_first - check if skb is the first entry in the queue
925 * @list: queue head
926 * @skb: buffer
927 *
928 * Returns true if @skb is the first buffer on the list.
929 */
930static inline bool skb_queue_is_first(const struct sk_buff_head *list,
931 const struct sk_buff *skb)
932{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100933 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800934}
935
936/**
David S. Miller249c8b42008-09-23 00:44:42 -0700937 * skb_queue_next - return the next packet in the queue
938 * @list: queue head
939 * @skb: current buffer
940 *
941 * Return the next packet in @list after @skb. It is only valid to
942 * call this if skb_queue_is_last() evaluates to false.
943 */
944static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
945 const struct sk_buff *skb)
946{
947 /* This BUG_ON may seem severe, but if we just return then we
948 * are going to dereference garbage.
949 */
950 BUG_ON(skb_queue_is_last(list, skb));
951 return skb->next;
952}
953
954/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800955 * skb_queue_prev - return the prev packet in the queue
956 * @list: queue head
957 * @skb: current buffer
958 *
959 * Return the prev packet in @list before @skb. It is only valid to
960 * call this if skb_queue_is_first() evaluates to false.
961 */
962static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
963 const struct sk_buff *skb)
964{
965 /* This BUG_ON may seem severe, but if we just return then we
966 * are going to dereference garbage.
967 */
968 BUG_ON(skb_queue_is_first(list, skb));
969 return skb->prev;
970}
971
972/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700973 * skb_get - reference buffer
974 * @skb: buffer to reference
975 *
976 * Makes another reference to a socket buffer and returns a pointer
977 * to the buffer.
978 */
979static inline struct sk_buff *skb_get(struct sk_buff *skb)
980{
981 atomic_inc(&skb->users);
982 return skb;
983}
984
985/*
986 * If users == 1, we are the only owner and are can avoid redundant
987 * atomic change.
988 */
989
990/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700991 * skb_cloned - is the buffer a clone
992 * @skb: buffer to check
993 *
994 * Returns true if the buffer was generated with skb_clone() and is
995 * one of multiple shared copies of the buffer. Cloned buffers are
996 * shared data so must not be written to under normal circumstances.
997 */
998static inline int skb_cloned(const struct sk_buff *skb)
999{
1000 return skb->cloned &&
1001 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1002}
1003
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001004static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1005{
1006 might_sleep_if(pri & __GFP_WAIT);
1007
1008 if (skb_cloned(skb))
1009 return pskb_expand_head(skb, 0, 0, pri);
1010
1011 return 0;
1012}
1013
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014/**
1015 * skb_header_cloned - is the header a clone
1016 * @skb: buffer to check
1017 *
1018 * Returns true if modifying the header part of the buffer requires
1019 * the data to be copied.
1020 */
1021static inline int skb_header_cloned(const struct sk_buff *skb)
1022{
1023 int dataref;
1024
1025 if (!skb->cloned)
1026 return 0;
1027
1028 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1029 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1030 return dataref != 1;
1031}
1032
1033/**
1034 * skb_header_release - release reference to header
1035 * @skb: buffer to operate on
1036 *
1037 * Drop a reference to the header part of the buffer. This is done
1038 * by acquiring a payload reference. You must not read from the header
1039 * part of skb->data after this.
1040 */
1041static inline void skb_header_release(struct sk_buff *skb)
1042{
1043 BUG_ON(skb->nohdr);
1044 skb->nohdr = 1;
1045 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1046}
1047
1048/**
1049 * skb_shared - is the buffer shared
1050 * @skb: buffer to check
1051 *
1052 * Returns true if more than one person has a reference to this
1053 * buffer.
1054 */
1055static inline int skb_shared(const struct sk_buff *skb)
1056{
1057 return atomic_read(&skb->users) != 1;
1058}
1059
1060/**
1061 * skb_share_check - check if buffer is shared and if so clone it
1062 * @skb: buffer to check
1063 * @pri: priority for memory allocation
1064 *
1065 * If the buffer is shared the buffer is cloned and the old copy
1066 * drops a reference. A new clone with a single reference is returned.
1067 * If the buffer is not shared the original buffer is returned. When
1068 * being called from interrupt status or with spinlocks held pri must
1069 * be GFP_ATOMIC.
1070 *
1071 * NULL is returned on a memory allocation failure.
1072 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001073static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001074{
1075 might_sleep_if(pri & __GFP_WAIT);
1076 if (skb_shared(skb)) {
1077 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001078
1079 if (likely(nskb))
1080 consume_skb(skb);
1081 else
1082 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001083 skb = nskb;
1084 }
1085 return skb;
1086}
1087
1088/*
1089 * Copy shared buffers into a new sk_buff. We effectively do COW on
1090 * packets to handle cases where we have a local reader and forward
1091 * and a couple of other messy ones. The normal one is tcpdumping
1092 * a packet thats being forwarded.
1093 */
1094
1095/**
1096 * skb_unshare - make a copy of a shared buffer
1097 * @skb: buffer to check
1098 * @pri: priority for memory allocation
1099 *
1100 * If the socket buffer is a clone then this function creates a new
1101 * copy of the data, drops a reference count on the old copy and returns
1102 * the new copy with the reference count at 1. If the buffer is not a clone
1103 * the original buffer is returned. When called with a spinlock held or
1104 * from interrupt state @pri must be %GFP_ATOMIC
1105 *
1106 * %NULL is returned on a memory allocation failure.
1107 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001108static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001109 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001110{
1111 might_sleep_if(pri & __GFP_WAIT);
1112 if (skb_cloned(skb)) {
1113 struct sk_buff *nskb = skb_copy(skb, pri);
1114 kfree_skb(skb); /* Free our shared copy */
1115 skb = nskb;
1116 }
1117 return skb;
1118}
1119
1120/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001121 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001122 * @list_: list to peek at
1123 *
1124 * Peek an &sk_buff. Unlike most other operations you _MUST_
1125 * be careful with this one. A peek leaves the buffer on the
1126 * list and someone else may run off with it. You must hold
1127 * the appropriate locks or have a private queue to do this.
1128 *
1129 * Returns %NULL for an empty list or a pointer to the head element.
1130 * The reference count is not incremented and the reference is therefore
1131 * volatile. Use with caution.
1132 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001133static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001134{
Eric Dumazet18d07002012-04-30 16:31:46 +00001135 struct sk_buff *skb = list_->next;
1136
1137 if (skb == (struct sk_buff *)list_)
1138 skb = NULL;
1139 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001140}
1141
1142/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001143 * skb_peek_next - peek skb following the given one from a queue
1144 * @skb: skb to start from
1145 * @list_: list to peek at
1146 *
1147 * Returns %NULL when the end of the list is met or a pointer to the
1148 * next element. The reference count is not incremented and the
1149 * reference is therefore volatile. Use with caution.
1150 */
1151static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1152 const struct sk_buff_head *list_)
1153{
1154 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001155
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001156 if (next == (struct sk_buff *)list_)
1157 next = NULL;
1158 return next;
1159}
1160
1161/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001162 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001163 * @list_: list to peek at
1164 *
1165 * Peek an &sk_buff. Unlike most other operations you _MUST_
1166 * be careful with this one. A peek leaves the buffer on the
1167 * list and someone else may run off with it. You must hold
1168 * the appropriate locks or have a private queue to do this.
1169 *
1170 * Returns %NULL for an empty list or a pointer to the tail element.
1171 * The reference count is not incremented and the reference is therefore
1172 * volatile. Use with caution.
1173 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001174static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175{
Eric Dumazet18d07002012-04-30 16:31:46 +00001176 struct sk_buff *skb = list_->prev;
1177
1178 if (skb == (struct sk_buff *)list_)
1179 skb = NULL;
1180 return skb;
1181
Linus Torvalds1da177e2005-04-16 15:20:36 -07001182}
1183
1184/**
1185 * skb_queue_len - get queue length
1186 * @list_: list to measure
1187 *
1188 * Return the length of an &sk_buff queue.
1189 */
1190static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1191{
1192 return list_->qlen;
1193}
1194
David S. Miller67fed452008-09-21 22:36:24 -07001195/**
1196 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1197 * @list: queue to initialize
1198 *
1199 * This initializes only the list and queue length aspects of
1200 * an sk_buff_head object. This allows to initialize the list
1201 * aspects of an sk_buff_head without reinitializing things like
1202 * the spinlock. It can also be used for on-stack sk_buff_head
1203 * objects where the spinlock is known to not be used.
1204 */
1205static inline void __skb_queue_head_init(struct sk_buff_head *list)
1206{
1207 list->prev = list->next = (struct sk_buff *)list;
1208 list->qlen = 0;
1209}
1210
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001211/*
1212 * This function creates a split out lock class for each invocation;
1213 * this is needed for now since a whole lot of users of the skb-queue
1214 * infrastructure in drivers have different locking usage (in hardirq)
1215 * than the networking core (in softirq only). In the long run either the
1216 * network layer or drivers should need annotation to consolidate the
1217 * main types of usage into 3 classes.
1218 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001219static inline void skb_queue_head_init(struct sk_buff_head *list)
1220{
1221 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001222 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001223}
1224
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001225static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1226 struct lock_class_key *class)
1227{
1228 skb_queue_head_init(list);
1229 lockdep_set_class(&list->lock, class);
1230}
1231
Linus Torvalds1da177e2005-04-16 15:20:36 -07001232/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001233 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001234 *
1235 * The "__skb_xxxx()" functions are the non-atomic ones that
1236 * can only be called with interrupts disabled.
1237 */
Joe Perches7965bd42013-09-26 14:48:15 -07001238void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1239 struct sk_buff_head *list);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001240static inline void __skb_insert(struct sk_buff *newsk,
1241 struct sk_buff *prev, struct sk_buff *next,
1242 struct sk_buff_head *list)
1243{
1244 newsk->next = next;
1245 newsk->prev = prev;
1246 next->prev = prev->next = newsk;
1247 list->qlen++;
1248}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001249
David S. Miller67fed452008-09-21 22:36:24 -07001250static inline void __skb_queue_splice(const struct sk_buff_head *list,
1251 struct sk_buff *prev,
1252 struct sk_buff *next)
1253{
1254 struct sk_buff *first = list->next;
1255 struct sk_buff *last = list->prev;
1256
1257 first->prev = prev;
1258 prev->next = first;
1259
1260 last->next = next;
1261 next->prev = last;
1262}
1263
1264/**
1265 * skb_queue_splice - join two skb lists, this is designed for stacks
1266 * @list: the new list to add
1267 * @head: the place to add it in the first list
1268 */
1269static inline void skb_queue_splice(const struct sk_buff_head *list,
1270 struct sk_buff_head *head)
1271{
1272 if (!skb_queue_empty(list)) {
1273 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001274 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001275 }
1276}
1277
1278/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001279 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001280 * @list: the new list to add
1281 * @head: the place to add it in the first list
1282 *
1283 * The list at @list is reinitialised
1284 */
1285static inline void skb_queue_splice_init(struct sk_buff_head *list,
1286 struct sk_buff_head *head)
1287{
1288 if (!skb_queue_empty(list)) {
1289 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001290 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001291 __skb_queue_head_init(list);
1292 }
1293}
1294
1295/**
1296 * skb_queue_splice_tail - join two skb lists, each list being a queue
1297 * @list: the new list to add
1298 * @head: the place to add it in the first list
1299 */
1300static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1301 struct sk_buff_head *head)
1302{
1303 if (!skb_queue_empty(list)) {
1304 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001305 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001306 }
1307}
1308
1309/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001310 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001311 * @list: the new list to add
1312 * @head: the place to add it in the first list
1313 *
1314 * Each of the lists is a queue.
1315 * The list at @list is reinitialised
1316 */
1317static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1318 struct sk_buff_head *head)
1319{
1320 if (!skb_queue_empty(list)) {
1321 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001322 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001323 __skb_queue_head_init(list);
1324 }
1325}
1326
Linus Torvalds1da177e2005-04-16 15:20:36 -07001327/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001328 * __skb_queue_after - queue a buffer at the list head
1329 * @list: list to use
1330 * @prev: place after this buffer
1331 * @newsk: buffer to queue
1332 *
1333 * Queue a buffer int the middle of a list. This function takes no locks
1334 * and you must therefore hold required locks before calling it.
1335 *
1336 * A buffer cannot be placed on two lists at the same time.
1337 */
1338static inline void __skb_queue_after(struct sk_buff_head *list,
1339 struct sk_buff *prev,
1340 struct sk_buff *newsk)
1341{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001342 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001343}
1344
Joe Perches7965bd42013-09-26 14:48:15 -07001345void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1346 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001347
Gerrit Renkerf5572852008-04-14 00:05:28 -07001348static inline void __skb_queue_before(struct sk_buff_head *list,
1349 struct sk_buff *next,
1350 struct sk_buff *newsk)
1351{
1352 __skb_insert(newsk, next->prev, next, list);
1353}
1354
Stephen Hemminger300ce172005-10-30 13:47:34 -08001355/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356 * __skb_queue_head - queue a buffer at the list head
1357 * @list: list to use
1358 * @newsk: buffer to queue
1359 *
1360 * Queue a buffer at the start of a list. This function takes no locks
1361 * and you must therefore hold required locks before calling it.
1362 *
1363 * A buffer cannot be placed on two lists at the same time.
1364 */
Joe Perches7965bd42013-09-26 14:48:15 -07001365void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001366static inline void __skb_queue_head(struct sk_buff_head *list,
1367 struct sk_buff *newsk)
1368{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001369 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001370}
1371
1372/**
1373 * __skb_queue_tail - queue a buffer at the list tail
1374 * @list: list to use
1375 * @newsk: buffer to queue
1376 *
1377 * Queue a buffer at the end of a list. This function takes no locks
1378 * and you must therefore hold required locks before calling it.
1379 *
1380 * A buffer cannot be placed on two lists at the same time.
1381 */
Joe Perches7965bd42013-09-26 14:48:15 -07001382void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383static inline void __skb_queue_tail(struct sk_buff_head *list,
1384 struct sk_buff *newsk)
1385{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001386 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001387}
1388
Linus Torvalds1da177e2005-04-16 15:20:36 -07001389/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001390 * remove sk_buff from list. _Must_ be called atomically, and with
1391 * the list known..
1392 */
Joe Perches7965bd42013-09-26 14:48:15 -07001393void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1395{
1396 struct sk_buff *next, *prev;
1397
1398 list->qlen--;
1399 next = skb->next;
1400 prev = skb->prev;
1401 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001402 next->prev = prev;
1403 prev->next = next;
1404}
1405
Gerrit Renkerf525c062008-04-14 00:04:12 -07001406/**
1407 * __skb_dequeue - remove from the head of the queue
1408 * @list: list to dequeue from
1409 *
1410 * Remove the head of the list. This function does not take any locks
1411 * so must be used with appropriate locks held only. The head item is
1412 * returned or %NULL if the list is empty.
1413 */
Joe Perches7965bd42013-09-26 14:48:15 -07001414struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Gerrit Renkerf525c062008-04-14 00:04:12 -07001415static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1416{
1417 struct sk_buff *skb = skb_peek(list);
1418 if (skb)
1419 __skb_unlink(skb, list);
1420 return skb;
1421}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001422
1423/**
1424 * __skb_dequeue_tail - remove from the tail of the queue
1425 * @list: list to dequeue from
1426 *
1427 * Remove the tail of the list. This function does not take any locks
1428 * so must be used with appropriate locks held only. The tail item is
1429 * returned or %NULL if the list is empty.
1430 */
Joe Perches7965bd42013-09-26 14:48:15 -07001431struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001432static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1433{
1434 struct sk_buff *skb = skb_peek_tail(list);
1435 if (skb)
1436 __skb_unlink(skb, list);
1437 return skb;
1438}
1439
1440
David S. Millerbdcc0922012-03-07 20:53:36 -05001441static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001442{
1443 return skb->data_len;
1444}
1445
1446static inline unsigned int skb_headlen(const struct sk_buff *skb)
1447{
1448 return skb->len - skb->data_len;
1449}
1450
1451static inline int skb_pagelen(const struct sk_buff *skb)
1452{
1453 int i, len = 0;
1454
1455 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001456 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001457 return len + skb_headlen(skb);
1458}
1459
Ian Campbell131ea662011-08-19 06:25:00 +00001460/**
1461 * __skb_fill_page_desc - initialise a paged fragment in an skb
1462 * @skb: buffer containing fragment to be initialised
1463 * @i: paged fragment index to initialise
1464 * @page: the page to use for this fragment
1465 * @off: the offset to the data with @page
1466 * @size: the length of the data
1467 *
1468 * Initialises the @i'th fragment of @skb to point to &size bytes at
1469 * offset @off within @page.
1470 *
1471 * Does not take any additional reference on the fragment.
1472 */
1473static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1474 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001475{
1476 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1477
Mel Gormanc48a11c2012-07-31 16:44:23 -07001478 /*
1479 * Propagate page->pfmemalloc to the skb if we can. The problem is
1480 * that not all callers have unique ownership of the page. If
1481 * pfmemalloc is set, we check the mapping as a mapping implies
1482 * page->index is set (index and pfmemalloc share space).
1483 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1484 * do not lose pfmemalloc information as the pages would not be
1485 * allocated using __GFP_MEMALLOC.
1486 */
Ian Campbella8605c62011-10-19 23:01:49 +00001487 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001488 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001489 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001490
1491 page = compound_head(page);
1492 if (page->pfmemalloc && !page->mapping)
1493 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00001494}
1495
1496/**
1497 * skb_fill_page_desc - initialise a paged fragment in an skb
1498 * @skb: buffer containing fragment to be initialised
1499 * @i: paged fragment index to initialise
1500 * @page: the page to use for this fragment
1501 * @off: the offset to the data with @page
1502 * @size: the length of the data
1503 *
1504 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01001505 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00001506 * addition updates @skb such that @i is the last fragment.
1507 *
1508 * Does not take any additional reference on the fragment.
1509 */
1510static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1511 struct page *page, int off, int size)
1512{
1513 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001514 skb_shinfo(skb)->nr_frags = i + 1;
1515}
1516
Joe Perches7965bd42013-09-26 14:48:15 -07001517void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1518 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001519
Jason Wangf8e617e2013-11-01 14:07:47 +08001520void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1521 unsigned int truesize);
1522
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001524#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001525#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1526
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001527#ifdef NET_SKBUFF_DATA_USES_OFFSET
1528static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1529{
1530 return skb->head + skb->tail;
1531}
1532
1533static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1534{
1535 skb->tail = skb->data - skb->head;
1536}
1537
1538static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1539{
1540 skb_reset_tail_pointer(skb);
1541 skb->tail += offset;
1542}
Simon Horman7cc46192013-05-28 20:34:29 +00001543
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001544#else /* NET_SKBUFF_DATA_USES_OFFSET */
1545static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1546{
1547 return skb->tail;
1548}
1549
1550static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1551{
1552 skb->tail = skb->data;
1553}
1554
1555static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1556{
1557 skb->tail = skb->data + offset;
1558}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001559
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001560#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1561
Linus Torvalds1da177e2005-04-16 15:20:36 -07001562/*
1563 * Add data to an sk_buff
1564 */
Mathias Krause0c7ddf32013-11-07 14:18:24 +01001565unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001566unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1568{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001569 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001570 SKB_LINEAR_ASSERT(skb);
1571 skb->tail += len;
1572 skb->len += len;
1573 return tmp;
1574}
1575
Joe Perches7965bd42013-09-26 14:48:15 -07001576unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001577static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1578{
1579 skb->data -= len;
1580 skb->len += len;
1581 return skb->data;
1582}
1583
Joe Perches7965bd42013-09-26 14:48:15 -07001584unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001585static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1586{
1587 skb->len -= len;
1588 BUG_ON(skb->len < skb->data_len);
1589 return skb->data += len;
1590}
1591
David S. Miller47d29642010-05-02 02:21:44 -07001592static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1593{
1594 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1595}
1596
Joe Perches7965bd42013-09-26 14:48:15 -07001597unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001598
1599static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1600{
1601 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001602 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001603 return NULL;
1604 skb->len -= len;
1605 return skb->data += len;
1606}
1607
1608static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1609{
1610 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1611}
1612
1613static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1614{
1615 if (likely(len <= skb_headlen(skb)))
1616 return 1;
1617 if (unlikely(len > skb->len))
1618 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001619 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001620}
1621
1622/**
1623 * skb_headroom - bytes at buffer head
1624 * @skb: buffer to check
1625 *
1626 * Return the number of bytes of free space at the head of an &sk_buff.
1627 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001628static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001629{
1630 return skb->data - skb->head;
1631}
1632
1633/**
1634 * skb_tailroom - bytes at buffer end
1635 * @skb: buffer to check
1636 *
1637 * Return the number of bytes of free space at the tail of an sk_buff
1638 */
1639static inline int skb_tailroom(const struct sk_buff *skb)
1640{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001641 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001642}
1643
1644/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00001645 * skb_availroom - bytes at buffer end
1646 * @skb: buffer to check
1647 *
1648 * Return the number of bytes of free space at the tail of an sk_buff
1649 * allocated by sk_stream_alloc()
1650 */
1651static inline int skb_availroom(const struct sk_buff *skb)
1652{
Eric Dumazet16fad692013-03-14 05:40:32 +00001653 if (skb_is_nonlinear(skb))
1654 return 0;
1655
1656 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00001657}
1658
1659/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001660 * skb_reserve - adjust headroom
1661 * @skb: buffer to alter
1662 * @len: bytes to move
1663 *
1664 * Increase the headroom of an empty &sk_buff by reducing the tail
1665 * room. This is only allowed for an empty buffer.
1666 */
David S. Miller8243126c2006-01-17 02:54:21 -08001667static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001668{
1669 skb->data += len;
1670 skb->tail += len;
1671}
1672
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001673static inline void skb_reset_inner_headers(struct sk_buff *skb)
1674{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001675 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001676 skb->inner_network_header = skb->network_header;
1677 skb->inner_transport_header = skb->transport_header;
1678}
1679
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00001680static inline void skb_reset_mac_len(struct sk_buff *skb)
1681{
1682 skb->mac_len = skb->network_header - skb->mac_header;
1683}
1684
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001685static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1686 *skb)
1687{
1688 return skb->head + skb->inner_transport_header;
1689}
1690
1691static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1692{
1693 skb->inner_transport_header = skb->data - skb->head;
1694}
1695
1696static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1697 const int offset)
1698{
1699 skb_reset_inner_transport_header(skb);
1700 skb->inner_transport_header += offset;
1701}
1702
1703static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1704{
1705 return skb->head + skb->inner_network_header;
1706}
1707
1708static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1709{
1710 skb->inner_network_header = skb->data - skb->head;
1711}
1712
1713static inline void skb_set_inner_network_header(struct sk_buff *skb,
1714 const int offset)
1715{
1716 skb_reset_inner_network_header(skb);
1717 skb->inner_network_header += offset;
1718}
1719
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001720static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1721{
1722 return skb->head + skb->inner_mac_header;
1723}
1724
1725static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1726{
1727 skb->inner_mac_header = skb->data - skb->head;
1728}
1729
1730static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1731 const int offset)
1732{
1733 skb_reset_inner_mac_header(skb);
1734 skb->inner_mac_header += offset;
1735}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001736static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1737{
Cong Wang35d04612013-05-29 15:16:05 +08001738 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001739}
1740
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001741static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1742{
1743 return skb->head + skb->transport_header;
1744}
1745
1746static inline void skb_reset_transport_header(struct sk_buff *skb)
1747{
1748 skb->transport_header = skb->data - skb->head;
1749}
1750
1751static inline void skb_set_transport_header(struct sk_buff *skb,
1752 const int offset)
1753{
1754 skb_reset_transport_header(skb);
1755 skb->transport_header += offset;
1756}
1757
1758static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1759{
1760 return skb->head + skb->network_header;
1761}
1762
1763static inline void skb_reset_network_header(struct sk_buff *skb)
1764{
1765 skb->network_header = skb->data - skb->head;
1766}
1767
1768static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1769{
1770 skb_reset_network_header(skb);
1771 skb->network_header += offset;
1772}
1773
1774static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1775{
1776 return skb->head + skb->mac_header;
1777}
1778
1779static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1780{
Cong Wang35d04612013-05-29 15:16:05 +08001781 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001782}
1783
1784static inline void skb_reset_mac_header(struct sk_buff *skb)
1785{
1786 skb->mac_header = skb->data - skb->head;
1787}
1788
1789static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1790{
1791 skb_reset_mac_header(skb);
1792 skb->mac_header += offset;
1793}
1794
Timo Teräs0e3da5b2013-12-16 11:02:09 +02001795static inline void skb_pop_mac_header(struct sk_buff *skb)
1796{
1797 skb->mac_header = skb->network_header;
1798}
1799
Ying Xuefbbdb8f2013-03-27 16:46:06 +00001800static inline void skb_probe_transport_header(struct sk_buff *skb,
1801 const int offset_hint)
1802{
1803 struct flow_keys keys;
1804
1805 if (skb_transport_header_was_set(skb))
1806 return;
1807 else if (skb_flow_dissect(skb, &keys))
1808 skb_set_transport_header(skb, keys.thoff);
1809 else
1810 skb_set_transport_header(skb, offset_hint);
1811}
1812
Eric Dumazet03606892012-02-23 10:55:02 +00001813static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1814{
1815 if (skb_mac_header_was_set(skb)) {
1816 const unsigned char *old_mac = skb_mac_header(skb);
1817
1818 skb_set_mac_header(skb, -skb->mac_len);
1819 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1820 }
1821}
1822
Michał Mirosław04fb4512010-12-14 15:24:08 +00001823static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1824{
1825 return skb->csum_start - skb_headroom(skb);
1826}
1827
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001828static inline int skb_transport_offset(const struct sk_buff *skb)
1829{
1830 return skb_transport_header(skb) - skb->data;
1831}
1832
1833static inline u32 skb_network_header_len(const struct sk_buff *skb)
1834{
1835 return skb->transport_header - skb->network_header;
1836}
1837
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001838static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1839{
1840 return skb->inner_transport_header - skb->inner_network_header;
1841}
1842
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001843static inline int skb_network_offset(const struct sk_buff *skb)
1844{
1845 return skb_network_header(skb) - skb->data;
1846}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001847
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001848static inline int skb_inner_network_offset(const struct sk_buff *skb)
1849{
1850 return skb_inner_network_header(skb) - skb->data;
1851}
1852
Changli Gaof9599ce2010-08-04 04:43:44 +00001853static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1854{
1855 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1856}
1857
Tom Herberte5eb4e32014-06-14 23:24:28 -07001858static inline void skb_pop_rcv_encapsulation(struct sk_buff *skb)
1859{
1860 /* Only continue with checksum unnecessary if device indicated
1861 * it is valid across encapsulation (skb->encapsulation was set).
1862 */
1863 if (skb->ip_summed == CHECKSUM_UNNECESSARY && !skb->encapsulation)
1864 skb->ip_summed = CHECKSUM_NONE;
1865
1866 skb->encapsulation = 0;
1867 skb->csum_valid = 0;
1868}
1869
Linus Torvalds1da177e2005-04-16 15:20:36 -07001870/*
1871 * CPUs often take a performance hit when accessing unaligned memory
1872 * locations. The actual performance hit varies, it can be small if the
1873 * hardware handles it or large if we have to take an exception and fix it
1874 * in software.
1875 *
1876 * Since an ethernet header is 14 bytes network drivers often end up with
1877 * the IP header at an unaligned offset. The IP header can be aligned by
1878 * shifting the start of the packet by 2 bytes. Drivers should do this
1879 * with:
1880 *
Tobias Klauser8660c122009-07-13 22:48:16 +00001881 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001882 *
1883 * The downside to this alignment of the IP header is that the DMA is now
1884 * unaligned. On some architectures the cost of an unaligned DMA is high
1885 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00001886 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001887 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1888 * to be overridden.
1889 */
1890#ifndef NET_IP_ALIGN
1891#define NET_IP_ALIGN 2
1892#endif
1893
Anton Blanchard025be812006-03-31 02:27:06 -08001894/*
1895 * The networking layer reserves some headroom in skb data (via
1896 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1897 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08001898 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08001899 *
1900 * Unfortunately this headroom changes the DMA alignment of the resulting
1901 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1902 * on some architectures. An architecture can override this value,
1903 * perhaps setting it to a cacheline in size (since that will maintain
1904 * cacheline alignment of the DMA). It must be a power of 2.
1905 *
David S. Millerd6301d32009-02-08 19:24:13 -08001906 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08001907 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07001908 *
1909 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1910 * to reduce average number of cache lines per packet.
1911 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07001912 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08001913 */
1914#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07001915#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08001916#endif
1917
Joe Perches7965bd42013-09-26 14:48:15 -07001918int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001919
1920static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1921{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00001922 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07001923 WARN_ON(1);
1924 return;
1925 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001926 skb->len = len;
1927 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001928}
1929
Joe Perches7965bd42013-09-26 14:48:15 -07001930void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001931
1932static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1933{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001934 if (skb->data_len)
1935 return ___pskb_trim(skb, len);
1936 __skb_trim(skb, len);
1937 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001938}
1939
1940static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1941{
1942 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1943}
1944
1945/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07001946 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1947 * @skb: buffer to alter
1948 * @len: new length
1949 *
1950 * This is identical to pskb_trim except that the caller knows that
1951 * the skb is not cloned so we should never get an error due to out-
1952 * of-memory.
1953 */
1954static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1955{
1956 int err = pskb_trim(skb, len);
1957 BUG_ON(err);
1958}
1959
1960/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001961 * skb_orphan - orphan a buffer
1962 * @skb: buffer to orphan
1963 *
1964 * If a buffer currently has an owner then we call the owner's
1965 * destructor function and make the @skb unowned. The buffer continues
1966 * to exist but is no longer charged to its former owner.
1967 */
1968static inline void skb_orphan(struct sk_buff *skb)
1969{
Eric Dumazetc34a7612013-07-30 16:11:15 -07001970 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001971 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07001972 skb->destructor = NULL;
1973 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07001974 } else {
1975 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07001976 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001977}
1978
1979/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00001980 * skb_orphan_frags - orphan the frags contained in a buffer
1981 * @skb: buffer to orphan frags from
1982 * @gfp_mask: allocation mask for replacement pages
1983 *
1984 * For each frag in the SKB which needs a destructor (i.e. has an
1985 * owner) create a copy of that frag and release the original
1986 * page by calling the destructor.
1987 */
1988static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1989{
1990 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1991 return 0;
1992 return skb_copy_ubufs(skb, gfp_mask);
1993}
1994
1995/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001996 * __skb_queue_purge - empty a list
1997 * @list: list to empty
1998 *
1999 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2000 * the list and one reference dropped. This function does not take the
2001 * list lock and the caller must hold the relevant locks to use it.
2002 */
Joe Perches7965bd42013-09-26 14:48:15 -07002003void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002004static inline void __skb_queue_purge(struct sk_buff_head *list)
2005{
2006 struct sk_buff *skb;
2007 while ((skb = __skb_dequeue(list)) != NULL)
2008 kfree_skb(skb);
2009}
2010
Alexander Duycke5e67302013-02-08 10:17:15 +00002011#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
2012#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
2013#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
2014
Joe Perches7965bd42013-09-26 14:48:15 -07002015void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002016
Joe Perches7965bd42013-09-26 14:48:15 -07002017struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2018 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07002019
2020/**
2021 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2022 * @dev: network device to receive on
2023 * @length: length to allocate
2024 *
2025 * Allocate a new &sk_buff and assign it a usage count of one. The
2026 * buffer has unspecified headroom built in. Users should allocate
2027 * the headroom they think they need without accounting for the
2028 * built in space. The built in space is used for optimisations.
2029 *
2030 * %NULL is returned if there is no free memory. Although this function
2031 * allocates memory it can be called from an interrupt.
2032 */
2033static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002034 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002035{
2036 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2037}
2038
Eric Dumazet6f532612012-05-18 05:12:12 +00002039/* legacy helper around __netdev_alloc_skb() */
2040static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2041 gfp_t gfp_mask)
2042{
2043 return __netdev_alloc_skb(NULL, length, gfp_mask);
2044}
2045
2046/* legacy helper around netdev_alloc_skb() */
2047static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2048{
2049 return netdev_alloc_skb(NULL, length);
2050}
2051
2052
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002053static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2054 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002055{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002056 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002057
2058 if (NET_IP_ALIGN && skb)
2059 skb_reserve(skb, NET_IP_ALIGN);
2060 return skb;
2061}
2062
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002063static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2064 unsigned int length)
2065{
2066 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2067}
2068
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002069/**
2070 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
Mel Gorman06140022012-07-31 16:44:24 -07002071 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2072 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2073 * @order: size of the allocation
2074 *
2075 * Allocate a new page.
2076 *
2077 * %NULL is returned if there is no free memory.
2078*/
2079static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
2080 struct sk_buff *skb,
2081 unsigned int order)
2082{
2083 struct page *page;
2084
2085 gfp_mask |= __GFP_COLD;
2086
2087 if (!(gfp_mask & __GFP_NOMEMALLOC))
2088 gfp_mask |= __GFP_MEMALLOC;
2089
2090 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2091 if (skb && page && page->pfmemalloc)
2092 skb->pfmemalloc = true;
2093
2094 return page;
2095}
2096
2097/**
2098 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2099 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2100 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2101 *
2102 * Allocate a new page.
2103 *
2104 * %NULL is returned if there is no free memory.
2105 */
2106static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
2107 struct sk_buff *skb)
2108{
2109 return __skb_alloc_pages(gfp_mask, skb, 0);
2110}
2111
2112/**
2113 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2114 * @page: The page that was allocated from skb_alloc_page
2115 * @skb: The skb that may need pfmemalloc set
2116 */
2117static inline void skb_propagate_pfmemalloc(struct page *page,
2118 struct sk_buff *skb)
2119{
2120 if (page && page->pfmemalloc)
2121 skb->pfmemalloc = true;
2122}
2123
Eric Dumazet564824b2010-10-11 19:05:25 +00002124/**
Masanari Iidae2278672014-02-18 22:54:36 +09002125 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002126 * @frag: the paged fragment
2127 *
2128 * Returns the &struct page associated with @frag.
2129 */
2130static inline struct page *skb_frag_page(const skb_frag_t *frag)
2131{
Ian Campbella8605c62011-10-19 23:01:49 +00002132 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00002133}
2134
2135/**
2136 * __skb_frag_ref - take an addition reference on a paged fragment.
2137 * @frag: the paged fragment
2138 *
2139 * Takes an additional reference on the paged fragment @frag.
2140 */
2141static inline void __skb_frag_ref(skb_frag_t *frag)
2142{
2143 get_page(skb_frag_page(frag));
2144}
2145
2146/**
2147 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2148 * @skb: the buffer
2149 * @f: the fragment offset.
2150 *
2151 * Takes an additional reference on the @f'th paged fragment of @skb.
2152 */
2153static inline void skb_frag_ref(struct sk_buff *skb, int f)
2154{
2155 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2156}
2157
2158/**
2159 * __skb_frag_unref - release a reference on a paged fragment.
2160 * @frag: the paged fragment
2161 *
2162 * Releases a reference on the paged fragment @frag.
2163 */
2164static inline void __skb_frag_unref(skb_frag_t *frag)
2165{
2166 put_page(skb_frag_page(frag));
2167}
2168
2169/**
2170 * skb_frag_unref - release a reference on a paged fragment of an skb.
2171 * @skb: the buffer
2172 * @f: the fragment offset
2173 *
2174 * Releases a reference on the @f'th paged fragment of @skb.
2175 */
2176static inline void skb_frag_unref(struct sk_buff *skb, int f)
2177{
2178 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2179}
2180
2181/**
2182 * skb_frag_address - gets the address of the data contained in a paged fragment
2183 * @frag: the paged fragment buffer
2184 *
2185 * Returns the address of the data within @frag. The page must already
2186 * be mapped.
2187 */
2188static inline void *skb_frag_address(const skb_frag_t *frag)
2189{
2190 return page_address(skb_frag_page(frag)) + frag->page_offset;
2191}
2192
2193/**
2194 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2195 * @frag: the paged fragment buffer
2196 *
2197 * Returns the address of the data within @frag. Checks that the page
2198 * is mapped and returns %NULL otherwise.
2199 */
2200static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2201{
2202 void *ptr = page_address(skb_frag_page(frag));
2203 if (unlikely(!ptr))
2204 return NULL;
2205
2206 return ptr + frag->page_offset;
2207}
2208
2209/**
2210 * __skb_frag_set_page - sets the page contained in a paged fragment
2211 * @frag: the paged fragment
2212 * @page: the page to set
2213 *
2214 * Sets the fragment @frag to contain @page.
2215 */
2216static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2217{
Ian Campbella8605c62011-10-19 23:01:49 +00002218 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002219}
2220
2221/**
2222 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2223 * @skb: the buffer
2224 * @f: the fragment offset
2225 * @page: the page to set
2226 *
2227 * Sets the @f'th fragment of @skb to contain @page.
2228 */
2229static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2230 struct page *page)
2231{
2232 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2233}
2234
Eric Dumazet400dfd32013-10-17 16:27:07 -07002235bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2236
Ian Campbell131ea662011-08-19 06:25:00 +00002237/**
2238 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002239 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002240 * @frag: the paged fragment to map
2241 * @offset: the offset within the fragment (starting at the
2242 * fragment's own offset)
2243 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002244 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00002245 *
2246 * Maps the page associated with @frag to @device.
2247 */
2248static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2249 const skb_frag_t *frag,
2250 size_t offset, size_t size,
2251 enum dma_data_direction dir)
2252{
2253 return dma_map_page(dev, skb_frag_page(frag),
2254 frag->page_offset + offset, size, dir);
2255}
2256
Eric Dumazet117632e2011-12-03 21:39:53 +00002257static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2258 gfp_t gfp_mask)
2259{
2260 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2261}
2262
Octavian Purdilabad93e92014-06-12 01:36:26 +03002263
2264static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2265 gfp_t gfp_mask)
2266{
2267 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2268}
2269
2270
Ian Campbell131ea662011-08-19 06:25:00 +00002271/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002272 * skb_clone_writable - is the header of a clone writable
2273 * @skb: buffer to check
2274 * @len: length up to which to write
2275 *
2276 * Returns true if modifying the header part of the cloned buffer
2277 * does not requires the data to be copied.
2278 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002279static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002280{
2281 return !skb_header_cloned(skb) &&
2282 skb_headroom(skb) + len <= skb->hdr_len;
2283}
2284
Herbert Xud9cc2042007-09-16 16:21:16 -07002285static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2286 int cloned)
2287{
2288 int delta = 0;
2289
Herbert Xud9cc2042007-09-16 16:21:16 -07002290 if (headroom > skb_headroom(skb))
2291 delta = headroom - skb_headroom(skb);
2292
2293 if (delta || cloned)
2294 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2295 GFP_ATOMIC);
2296 return 0;
2297}
2298
Patrick McHardy334a8132007-06-25 04:35:20 -07002299/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002300 * skb_cow - copy header of skb when it is required
2301 * @skb: buffer to cow
2302 * @headroom: needed headroom
2303 *
2304 * If the skb passed lacks sufficient headroom or its data part
2305 * is shared, data is reallocated. If reallocation fails, an error
2306 * is returned and original skb is not changed.
2307 *
2308 * The result is skb with writable area skb->head...skb->tail
2309 * and at least @headroom of space at head.
2310 */
2311static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2312{
Herbert Xud9cc2042007-09-16 16:21:16 -07002313 return __skb_cow(skb, headroom, skb_cloned(skb));
2314}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002315
Herbert Xud9cc2042007-09-16 16:21:16 -07002316/**
2317 * skb_cow_head - skb_cow but only making the head writable
2318 * @skb: buffer to cow
2319 * @headroom: needed headroom
2320 *
2321 * This function is identical to skb_cow except that we replace the
2322 * skb_cloned check by skb_header_cloned. It should be used when
2323 * you only need to push on some header and do not need to modify
2324 * the data.
2325 */
2326static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2327{
2328 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002329}
2330
2331/**
2332 * skb_padto - pad an skbuff up to a minimal size
2333 * @skb: buffer to pad
2334 * @len: minimal length
2335 *
2336 * Pads up a buffer to ensure the trailing bytes exist and are
2337 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002338 * is untouched. Otherwise it is extended. Returns zero on
2339 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002340 */
2341
Herbert Xu5b057c62006-06-23 02:06:41 -07002342static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002343{
2344 unsigned int size = skb->len;
2345 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002346 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002347 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002348}
2349
2350static inline int skb_add_data(struct sk_buff *skb,
2351 char __user *from, int copy)
2352{
2353 const int off = skb->len;
2354
2355 if (skb->ip_summed == CHECKSUM_NONE) {
2356 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08002357 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
Linus Torvalds1da177e2005-04-16 15:20:36 -07002358 copy, 0, &err);
2359 if (!err) {
2360 skb->csum = csum_block_add(skb->csum, csum, off);
2361 return 0;
2362 }
2363 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2364 return 0;
2365
2366 __skb_trim(skb, off);
2367 return -EFAULT;
2368}
2369
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002370static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2371 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002372{
2373 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002374 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002375
Ian Campbellea2ab692011-08-22 23:44:58 +00002376 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002377 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002378 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002379 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002380}
2381
Herbert Xu364c6ba2006-06-09 16:10:40 -07002382static inline int __skb_linearize(struct sk_buff *skb)
2383{
2384 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2385}
2386
Linus Torvalds1da177e2005-04-16 15:20:36 -07002387/**
2388 * skb_linearize - convert paged skb to linear one
2389 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002390 *
2391 * If there is no free memory -ENOMEM is returned, otherwise zero
2392 * is returned and the old skb data released.
2393 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002394static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002395{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002396 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2397}
2398
2399/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00002400 * skb_has_shared_frag - can any frag be overwritten
2401 * @skb: buffer to test
2402 *
2403 * Return true if the skb has at least one frag that might be modified
2404 * by an external entity (as in vmsplice()/sendfile())
2405 */
2406static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2407{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00002408 return skb_is_nonlinear(skb) &&
2409 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00002410}
2411
2412/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07002413 * skb_linearize_cow - make sure skb is linear and writable
2414 * @skb: buffer to process
2415 *
2416 * If there is no free memory -ENOMEM is returned, otherwise zero
2417 * is returned and the old skb data released.
2418 */
2419static inline int skb_linearize_cow(struct sk_buff *skb)
2420{
2421 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2422 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002423}
2424
2425/**
2426 * skb_postpull_rcsum - update checksum for received skb after pull
2427 * @skb: buffer to update
2428 * @start: start of data before pull
2429 * @len: length of data pulled
2430 *
2431 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002432 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2433 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002434 */
2435
2436static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f92006-03-20 22:43:56 -08002437 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002438{
Patrick McHardy84fa7932006-08-29 16:44:56 -07002439 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002440 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2441}
2442
Herbert Xucbb042f92006-03-20 22:43:56 -08002443unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2444
David S. Miller7ce5a272013-12-02 17:26:05 -05002445/**
2446 * pskb_trim_rcsum - trim received skb and update checksum
2447 * @skb: buffer to trim
2448 * @len: new length
2449 *
2450 * This is exactly the same as pskb_trim except that it ensures the
2451 * checksum of received packets are still valid after the operation.
2452 */
2453
2454static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2455{
2456 if (likely(len >= skb->len))
2457 return 0;
2458 if (skb->ip_summed == CHECKSUM_COMPLETE)
2459 skb->ip_summed = CHECKSUM_NONE;
2460 return __pskb_trim(skb, len);
2461}
2462
Linus Torvalds1da177e2005-04-16 15:20:36 -07002463#define skb_queue_walk(queue, skb) \
2464 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002465 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002466 skb = skb->next)
2467
James Chapman46f89142007-04-30 00:07:31 -07002468#define skb_queue_walk_safe(queue, skb, tmp) \
2469 for (skb = (queue)->next, tmp = skb->next; \
2470 skb != (struct sk_buff *)(queue); \
2471 skb = tmp, tmp = skb->next)
2472
David S. Miller1164f522008-09-23 00:49:44 -07002473#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002474 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002475 skb = skb->next)
2476
2477#define skb_queue_walk_from_safe(queue, skb, tmp) \
2478 for (tmp = skb->next; \
2479 skb != (struct sk_buff *)(queue); \
2480 skb = tmp, tmp = skb->next)
2481
Stephen Hemminger300ce172005-10-30 13:47:34 -08002482#define skb_queue_reverse_walk(queue, skb) \
2483 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002484 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08002485 skb = skb->prev)
2486
David S. Miller686a2952011-01-20 22:47:32 -08002487#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2488 for (skb = (queue)->prev, tmp = skb->prev; \
2489 skb != (struct sk_buff *)(queue); \
2490 skb = tmp, tmp = skb->prev)
2491
2492#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2493 for (tmp = skb->prev; \
2494 skb != (struct sk_buff *)(queue); \
2495 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002496
David S. Miller21dc3302010-08-23 00:13:46 -07002497static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07002498{
2499 return skb_shinfo(skb)->frag_list != NULL;
2500}
2501
2502static inline void skb_frag_list_init(struct sk_buff *skb)
2503{
2504 skb_shinfo(skb)->frag_list = NULL;
2505}
2506
2507static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2508{
2509 frag->next = skb_shinfo(skb)->frag_list;
2510 skb_shinfo(skb)->frag_list = frag;
2511}
2512
2513#define skb_walk_frags(skb, iter) \
2514 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2515
Joe Perches7965bd42013-09-26 14:48:15 -07002516struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2517 int *peeked, int *off, int *err);
2518struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2519 int *err);
2520unsigned int datagram_poll(struct file *file, struct socket *sock,
2521 struct poll_table_struct *wait);
2522int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2523 struct iovec *to, int size);
2524int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2525 struct iovec *iov);
2526int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2527 const struct iovec *from, int from_offset,
2528 int len);
2529int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2530 int offset, size_t count);
2531int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2532 const struct iovec *to, int to_offset,
2533 int size);
2534void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2535void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2536int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07002537int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2538int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2539__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2540 int len, __wsum csum);
2541int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2542 struct pipe_inode_info *pipe, unsigned int len,
2543 unsigned int flags);
2544void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01002545unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00002546int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2547 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07002548void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2549int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2550void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Florian Westphalde960aa2014-01-26 10:58:16 +01002551unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -07002552struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002553
Daniel Borkmann2817a332013-10-30 11:50:51 +01002554struct skb_checksum_ops {
2555 __wsum (*update)(const void *mem, int len, __wsum wsum);
2556 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2557};
2558
2559__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2560 __wsum csum, const struct skb_checksum_ops *ops);
2561__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2562 __wsum csum);
2563
Linus Torvalds1da177e2005-04-16 15:20:36 -07002564static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2565 int len, void *buffer)
2566{
2567 int hlen = skb_headlen(skb);
2568
Patrick McHardy55820ee2005-07-05 14:08:10 -07002569 if (hlen - offset >= len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002570 return skb->data + offset;
2571
2572 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2573 return NULL;
2574
2575 return buffer;
2576}
2577
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01002578/**
2579 * skb_needs_linearize - check if we need to linearize a given skb
2580 * depending on the given device features.
2581 * @skb: socket buffer to check
2582 * @features: net device features
2583 *
2584 * Returns true if either:
2585 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2586 * 2. skb is fragmented and the device does not support SG.
2587 */
2588static inline bool skb_needs_linearize(struct sk_buff *skb,
2589 netdev_features_t features)
2590{
2591 return skb_is_nonlinear(skb) &&
2592 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2593 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2594}
2595
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03002596static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2597 void *to,
2598 const unsigned int len)
2599{
2600 memcpy(to, skb->data, len);
2601}
2602
2603static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2604 const int offset, void *to,
2605 const unsigned int len)
2606{
2607 memcpy(to, skb->data + offset, len);
2608}
2609
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03002610static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2611 const void *from,
2612 const unsigned int len)
2613{
2614 memcpy(skb->data, from, len);
2615}
2616
2617static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2618 const int offset,
2619 const void *from,
2620 const unsigned int len)
2621{
2622 memcpy(skb->data + offset, from, len);
2623}
2624
Joe Perches7965bd42013-09-26 14:48:15 -07002625void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002626
Patrick Ohlyac45f602009-02-12 05:03:37 +00002627static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2628{
2629 return skb->tstamp;
2630}
2631
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002632/**
2633 * skb_get_timestamp - get timestamp from a skb
2634 * @skb: skb to get stamp from
2635 * @stamp: pointer to struct timeval to store stamp in
2636 *
2637 * Timestamps are stored in the skb as offsets to a base timestamp.
2638 * This function converts the offset back to a struct timeval and stores
2639 * it in stamp.
2640 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00002641static inline void skb_get_timestamp(const struct sk_buff *skb,
2642 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002643{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002644 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002645}
2646
Patrick Ohlyac45f602009-02-12 05:03:37 +00002647static inline void skb_get_timestampns(const struct sk_buff *skb,
2648 struct timespec *stamp)
2649{
2650 *stamp = ktime_to_timespec(skb->tstamp);
2651}
2652
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002653static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002654{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002655 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002656}
2657
Stephen Hemminger164891a2007-04-23 22:26:16 -07002658static inline ktime_t net_timedelta(ktime_t t)
2659{
2660 return ktime_sub(ktime_get_real(), t);
2661}
2662
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07002663static inline ktime_t net_invalid_timestamp(void)
2664{
2665 return ktime_set(0, 0);
2666}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002667
Richard Cochranc1f19b52010-07-17 08:49:36 +00002668#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2669
Joe Perches7965bd42013-09-26 14:48:15 -07002670void skb_clone_tx_timestamp(struct sk_buff *skb);
2671bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00002672
2673#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2674
2675static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2676{
2677}
2678
2679static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2680{
2681 return false;
2682}
2683
2684#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2685
2686/**
2687 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2688 *
Richard Cochranda92b192011-10-21 00:49:15 +00002689 * PHY drivers may accept clones of transmitted packets for
2690 * timestamping via their phy_driver.txtstamp method. These drivers
2691 * must call this function to return the skb back to the stack, with
2692 * or without a timestamp.
2693 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00002694 * @skb: clone of the the original outgoing packet
Richard Cochranda92b192011-10-21 00:49:15 +00002695 * @hwtstamps: hardware time stamps, may be NULL if not available
Richard Cochranc1f19b52010-07-17 08:49:36 +00002696 *
2697 */
2698void skb_complete_tx_timestamp(struct sk_buff *skb,
2699 struct skb_shared_hwtstamps *hwtstamps);
2700
Willem de Bruijne7fd2882014-08-04 22:11:48 -04002701void __skb_tstamp_tx(struct sk_buff *orig_skb,
2702 struct skb_shared_hwtstamps *hwtstamps,
2703 struct sock *sk, int tstype);
2704
Patrick Ohlyac45f602009-02-12 05:03:37 +00002705/**
2706 * skb_tstamp_tx - queue clone of skb with send time stamps
2707 * @orig_skb: the original outgoing packet
2708 * @hwtstamps: hardware time stamps, may be NULL if not available
2709 *
2710 * If the skb has a socket associated, then this function clones the
2711 * skb (thus sharing the actual data and optional structures), stores
2712 * the optional hardware time stamping information (if non NULL) or
2713 * generates a software time stamp (otherwise), then queues the clone
2714 * to the error queue of the socket. Errors are silently ignored.
2715 */
Joe Perches7965bd42013-09-26 14:48:15 -07002716void skb_tstamp_tx(struct sk_buff *orig_skb,
2717 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00002718
Richard Cochran4507a712010-07-17 08:48:28 +00002719static inline void sw_tx_timestamp(struct sk_buff *skb)
2720{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002721 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2722 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00002723 skb_tstamp_tx(skb, NULL);
2724}
2725
2726/**
2727 * skb_tx_timestamp() - Driver hook for transmit timestamping
2728 *
2729 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00002730 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00002731 *
David S. Miller73409f32013-12-27 13:04:33 -05002732 * Specifically, one should make absolutely sure that this function is
2733 * called before TX completion of this packet can trigger. Otherwise
2734 * the packet could potentially already be freed.
2735 *
Richard Cochran4507a712010-07-17 08:48:28 +00002736 * @skb: A socket buffer.
2737 */
2738static inline void skb_tx_timestamp(struct sk_buff *skb)
2739{
Richard Cochranc1f19b52010-07-17 08:49:36 +00002740 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002741 sw_tx_timestamp(skb);
2742}
2743
Johannes Berg6e3e9392011-11-09 10:15:42 +01002744/**
2745 * skb_complete_wifi_ack - deliver skb with wifi status
2746 *
2747 * @skb: the original outgoing packet
2748 * @acked: ack status
2749 *
2750 */
2751void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2752
Joe Perches7965bd42013-09-26 14:48:15 -07002753__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2754__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002755
Herbert Xu60476372007-04-09 11:59:39 -07002756static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2757{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002758 return ((skb->ip_summed & CHECKSUM_UNNECESSARY) || skb->csum_valid);
Herbert Xu60476372007-04-09 11:59:39 -07002759}
2760
Herbert Xufb286bb2005-11-10 13:01:24 -08002761/**
2762 * skb_checksum_complete - Calculate checksum of an entire packet
2763 * @skb: packet to process
2764 *
2765 * This function calculates the checksum over the entire packet plus
2766 * the value of skb->csum. The latter can be used to supply the
2767 * checksum of a pseudo header as used by TCP/UDP. It returns the
2768 * checksum.
2769 *
2770 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2771 * this function can be used to verify that checksum on received
2772 * packets. In that case the function should return zero if the
2773 * checksum is correct. In particular, this function will return zero
2774 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2775 * hardware has already verified the correctness of the checksum.
2776 */
Al Viro4381ca32007-07-15 21:00:11 +01002777static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002778{
Herbert Xu60476372007-04-09 11:59:39 -07002779 return skb_csum_unnecessary(skb) ?
2780 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002781}
2782
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002783/* Check if we need to perform checksum complete validation.
2784 *
2785 * Returns true if checksum complete is needed, false otherwise
2786 * (either checksum is unnecessary or zero checksum is allowed).
2787 */
2788static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
2789 bool zero_okay,
2790 __sum16 check)
2791{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002792 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
2793 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002794 return false;
2795 }
2796
2797 return true;
2798}
2799
2800/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
2801 * in checksum_init.
2802 */
2803#define CHECKSUM_BREAK 76
2804
2805/* Validate (init) checksum based on checksum complete.
2806 *
2807 * Return values:
2808 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
2809 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
2810 * checksum is stored in skb->csum for use in __skb_checksum_complete
2811 * non-zero: value of invalid checksum
2812 *
2813 */
2814static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
2815 bool complete,
2816 __wsum psum)
2817{
2818 if (skb->ip_summed == CHECKSUM_COMPLETE) {
2819 if (!csum_fold(csum_add(psum, skb->csum))) {
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002820 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002821 return 0;
2822 }
2823 }
2824
2825 skb->csum = psum;
2826
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002827 if (complete || skb->len <= CHECKSUM_BREAK) {
2828 __sum16 csum;
2829
2830 csum = __skb_checksum_complete(skb);
2831 skb->csum_valid = !csum;
2832 return csum;
2833 }
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002834
2835 return 0;
2836}
2837
2838static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
2839{
2840 return 0;
2841}
2842
2843/* Perform checksum validate (init). Note that this is a macro since we only
2844 * want to calculate the pseudo header which is an input function if necessary.
2845 * First we try to validate without any computation (checksum unnecessary) and
2846 * then calculate based on checksum complete calling the function to compute
2847 * pseudo header.
2848 *
2849 * Return values:
2850 * 0: checksum is validated or try to in skb_checksum_complete
2851 * non-zero: value of invalid checksum
2852 */
2853#define __skb_checksum_validate(skb, proto, complete, \
2854 zero_okay, check, compute_pseudo) \
2855({ \
2856 __sum16 __ret = 0; \
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002857 skb->csum_valid = 0; \
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002858 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
2859 __ret = __skb_checksum_validate_complete(skb, \
2860 complete, compute_pseudo(skb, proto)); \
2861 __ret; \
2862})
2863
2864#define skb_checksum_init(skb, proto, compute_pseudo) \
2865 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
2866
2867#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
2868 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
2869
2870#define skb_checksum_validate(skb, proto, compute_pseudo) \
2871 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
2872
2873#define skb_checksum_validate_zero_check(skb, proto, check, \
2874 compute_pseudo) \
2875 __skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
2876
2877#define skb_checksum_simple_validate(skb) \
2878 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
2879
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002880#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07002881void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002882static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2883{
2884 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002885 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002886}
2887static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2888{
2889 if (nfct)
2890 atomic_inc(&nfct->use);
2891}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002892#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002893#ifdef CONFIG_BRIDGE_NETFILTER
2894static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2895{
2896 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2897 kfree(nf_bridge);
2898}
2899static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2900{
2901 if (nf_bridge)
2902 atomic_inc(&nf_bridge->use);
2903}
2904#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08002905static inline void nf_reset(struct sk_buff *skb)
2906{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002907#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08002908 nf_conntrack_put(skb->nfct);
2909 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002910#endif
Patrick McHardya193a4a2006-03-20 19:23:05 -08002911#ifdef CONFIG_BRIDGE_NETFILTER
2912 nf_bridge_put(skb->nf_bridge);
2913 skb->nf_bridge = NULL;
2914#endif
2915}
2916
Patrick McHardy124dff02013-04-05 20:42:05 +02002917static inline void nf_reset_trace(struct sk_buff *skb)
2918{
Florian Westphal478b3602014-02-15 23:48:45 +01002919#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00002920 skb->nf_trace = 0;
2921#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002922}
2923
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002924/* Note: This doesn't put any conntrack and bridge info in dst. */
2925static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2926{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002927#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002928 dst->nfct = src->nfct;
2929 nf_conntrack_get(src->nfct);
2930 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002931#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002932#ifdef CONFIG_BRIDGE_NETFILTER
2933 dst->nf_bridge = src->nf_bridge;
2934 nf_bridge_get(src->nf_bridge);
2935#endif
Florian Westphal478b3602014-02-15 23:48:45 +01002936#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
2937 dst->nf_trace = src->nf_trace;
2938#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002939}
2940
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002941static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2942{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002943#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002944 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002945#endif
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002946#ifdef CONFIG_BRIDGE_NETFILTER
2947 nf_bridge_put(dst->nf_bridge);
2948#endif
2949 __nf_copy(dst, src);
2950}
2951
James Morris984bc162006-06-09 00:29:17 -07002952#ifdef CONFIG_NETWORK_SECMARK
2953static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2954{
2955 to->secmark = from->secmark;
2956}
2957
2958static inline void skb_init_secmark(struct sk_buff *skb)
2959{
2960 skb->secmark = 0;
2961}
2962#else
2963static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2964{ }
2965
2966static inline void skb_init_secmark(struct sk_buff *skb)
2967{ }
2968#endif
2969
Eric W. Biederman574f7192014-04-01 12:20:24 -07002970static inline bool skb_irq_freeable(const struct sk_buff *skb)
2971{
2972 return !skb->destructor &&
2973#if IS_ENABLED(CONFIG_XFRM)
2974 !skb->sp &&
2975#endif
2976#if IS_ENABLED(CONFIG_NF_CONNTRACK)
2977 !skb->nfct &&
2978#endif
2979 !skb->_skb_refdst &&
2980 !skb_has_frag_list(skb);
2981}
2982
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002983static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2984{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002985 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002986}
2987
Stephen Hemminger92477442009-03-21 13:39:26 -07002988static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002989{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002990 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002991}
2992
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002993static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2994{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002995 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002996}
2997
David S. Millerd5a9e242009-01-27 16:22:11 -08002998static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2999{
3000 skb->queue_mapping = rx_queue + 1;
3001}
3002
Stephen Hemminger92477442009-03-21 13:39:26 -07003003static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003004{
3005 return skb->queue_mapping - 1;
3006}
3007
Stephen Hemminger92477442009-03-21 13:39:26 -07003008static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003009{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003010 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08003011}
3012
Tom Herbert0e001612014-07-01 21:32:27 -07003013u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
Joe Perches7965bd42013-09-26 14:48:15 -07003014 unsigned int num_tx_queues);
Stephen Hemminger92477442009-03-21 13:39:26 -07003015
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003016static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3017{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003018#ifdef CONFIG_XFRM
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003019 return skb->sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003020#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003021 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003022#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003023}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003024
Pravin B Shelar68c33162013-02-14 14:02:41 +00003025/* Keeps track of mac header offset relative to skb->head.
3026 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3027 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07003028 * tunnel skb it points to outer mac header.
3029 * Keeps track of level of encapsulation of network headers.
3030 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00003031struct skb_gso_cb {
Eric Dumazet3347c962013-10-19 11:42:56 -07003032 int mac_offset;
3033 int encap_level;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003034 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00003035};
3036#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3037
3038static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3039{
3040 return (skb_mac_header(inner_skb) - inner_skb->head) -
3041 SKB_GSO_CB(inner_skb)->mac_offset;
3042}
3043
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00003044static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3045{
3046 int new_headroom, headroom;
3047 int ret;
3048
3049 headroom = skb_headroom(skb);
3050 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3051 if (ret)
3052 return ret;
3053
3054 new_headroom = skb_headroom(skb);
3055 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3056 return 0;
3057}
3058
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003059/* Compute the checksum for a gso segment. First compute the checksum value
3060 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3061 * then add in skb->csum (checksum from csum_start to end of packet).
3062 * skb->csum and csum_start are then updated to reflect the checksum of the
3063 * resultant packet starting from the transport header-- the resultant checksum
3064 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3065 * header.
3066 */
3067static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3068{
3069 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3070 skb_transport_offset(skb);
3071 __u16 csum;
3072
3073 csum = csum_fold(csum_partial(skb_transport_header(skb),
3074 plen, skb->csum));
3075 skb->csum = res;
3076 SKB_GSO_CB(skb)->csum_start -= plen;
3077
3078 return csum;
3079}
3080
David S. Millerbdcc0922012-03-07 20:53:36 -05003081static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07003082{
3083 return skb_shinfo(skb)->gso_size;
3084}
3085
Eric Dumazet36a8f392013-09-29 01:21:32 -07003086/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05003087static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02003088{
3089 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3090}
3091
Joe Perches7965bd42013-09-26 14:48:15 -07003092void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07003093
3094static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3095{
3096 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3097 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003098 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3099
Alexander Duyckb78462e2010-06-02 12:24:37 +00003100 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3101 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07003102 __skb_warn_lro_forwarding(skb);
3103 return true;
3104 }
3105 return false;
3106}
3107
Herbert Xu35fc92a2007-03-26 23:22:20 -07003108static inline void skb_forward_csum(struct sk_buff *skb)
3109{
3110 /* Unfortunately we don't support this one. Any brave souls? */
3111 if (skb->ip_summed == CHECKSUM_COMPLETE)
3112 skb->ip_summed = CHECKSUM_NONE;
3113}
3114
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003115/**
3116 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3117 * @skb: skb to check
3118 *
3119 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3120 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3121 * use this helper, to document places where we make this assertion.
3122 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003123static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003124{
3125#ifdef DEBUG
3126 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3127#endif
3128}
3129
Rusty Russellf35d9d82008-02-04 23:49:54 -05003130bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00003131
Paul Durranted1f50c2014-01-09 10:02:46 +00003132int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3133
Daniel Borkmannf77668d2013-03-19 06:39:30 +00003134u32 __skb_get_poff(const struct sk_buff *skb);
3135
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00003136/**
3137 * skb_head_is_locked - Determine if the skb->head is locked down
3138 * @skb: skb to check
3139 *
3140 * The head on skbs build around a head frag can be removed if they are
3141 * not cloned. This function returns true if the skb head is locked down
3142 * due to either being allocated via kmalloc, or by being a clone with
3143 * multiple references to the head.
3144 */
3145static inline bool skb_head_is_locked(const struct sk_buff *skb)
3146{
3147 return !skb->head_frag || skb_cloned(skb);
3148}
Florian Westphalfe6cc552014-02-13 23:09:12 +01003149
3150/**
3151 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3152 *
3153 * @skb: GSO skb
3154 *
3155 * skb_gso_network_seglen is used to determine the real size of the
3156 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3157 *
3158 * The MAC/L2 header is not accounted for.
3159 */
3160static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3161{
3162 unsigned int hdr_len = skb_transport_header(skb) -
3163 skb_network_header(skb);
3164 return hdr_len + skb_gso_transport_seglen(skb);
3165}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003166#endif /* __KERNEL__ */
3167#endif /* _LINUX_SKBUFF_H */