blob: 3c9574c80933ac526e929aa1ada52598d165ed09 [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,
Willem de Bruijne1c8a602014-08-04 22:11:50 -0400253
254 /* generate software timestamp on peer data acknowledgment */
255 SKBTX_ACK_TSTAMP = 1 << 7,
Shirley Maa6686f22011-07-06 12:22:12 +0000256};
257
Willem de Bruijne1c8a602014-08-04 22:11:50 -0400258#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
259 SKBTX_SCHED_TSTAMP | \
260 SKBTX_ACK_TSTAMP)
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400261#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
262
Shirley Maa6686f22011-07-06 12:22:12 +0000263/*
264 * The callback notifies userspace to release buffers when skb DMA is done in
265 * lower device, the skb last reference should be 0 when calling this.
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000266 * The zerocopy_success argument is true if zero copy transmit occurred,
267 * false on data copy or out of memory error caused by data copy attempt.
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000268 * The ctx field is used to track device context.
269 * The desc field is used to track userspace buffer index.
Shirley Maa6686f22011-07-06 12:22:12 +0000270 */
271struct ubuf_info {
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000272 void (*callback)(struct ubuf_info *, bool zerocopy_success);
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000273 void *ctx;
Shirley Maa6686f22011-07-06 12:22:12 +0000274 unsigned long desc;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000275};
276
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277/* This data is invariant across clones and lives at
278 * the end of the header data, ie. at skb->end.
279 */
280struct skb_shared_info {
Ian Campbell9f42f122012-01-05 07:13:39 +0000281 unsigned char nr_frags;
282 __u8 tx_flags;
Herbert Xu79671682006-06-22 02:40:14 -0700283 unsigned short gso_size;
284 /* Warning: this field is not always filled in (UFO)! */
285 unsigned short gso_segs;
286 unsigned short gso_type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000288 struct skb_shared_hwtstamps hwtstamps;
Willem de Bruijn09c2d252014-08-04 22:11:47 -0400289 u32 tskey;
Ian Campbell9f42f122012-01-05 07:13:39 +0000290 __be32 ip6_frag_id;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700291
292 /*
293 * Warning : all fields before dataref are cleared in __alloc_skb()
294 */
295 atomic_t dataref;
296
Johann Baudy69e3c752009-05-18 22:11:22 -0700297 /* Intermediate layers must ensure that destructor_arg
298 * remains valid until skb destructor */
299 void * destructor_arg;
Shirley Maa6686f22011-07-06 12:22:12 +0000300
Eric Dumazetfed66382010-07-22 19:09:08 +0000301 /* must be last field, see pskb_expand_head() */
302 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700303};
304
305/* We divide dataref into two halves. The higher 16 bits hold references
306 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700307 * the entire skb->data. A clone of a headerless skb holds the length of
308 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700309 *
310 * All users must obey the rule that the skb->data reference count must be
311 * greater than or equal to the payload reference count.
312 *
313 * Holding a reference to the payload part means that the user does not
314 * care about modifications to the header part of skb->data.
315 */
316#define SKB_DATAREF_SHIFT 16
317#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
318
David S. Millerd179cd12005-08-17 14:57:30 -0700319
320enum {
321 SKB_FCLONE_UNAVAILABLE,
322 SKB_FCLONE_ORIG,
323 SKB_FCLONE_CLONE,
324};
325
Herbert Xu79671682006-06-22 02:40:14 -0700326enum {
327 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700328 SKB_GSO_UDP = 1 << 1,
Herbert Xu576a30e2006-06-27 13:22:38 -0700329
330 /* This indicates the skb is from an untrusted source. */
331 SKB_GSO_DODGY = 1 << 2,
Michael Chanb0da85372006-06-29 12:30:00 -0700332
333 /* This indicates the tcp segment has CWR set. */
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700334 SKB_GSO_TCP_ECN = 1 << 3,
335
336 SKB_GSO_TCPV6 = 1 << 4,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800337
338 SKB_GSO_FCOE = 1 << 5,
Pravin B Shelar68c33162013-02-14 14:02:41 +0000339
340 SKB_GSO_GRE = 1 << 6,
Pravin B Shelar73136262013-03-07 13:21:51 +0000341
Tom Herbert4b282522014-06-14 23:23:52 -0700342 SKB_GSO_GRE_CSUM = 1 << 7,
Simon Horman0d89d202013-05-23 21:02:52 +0000343
Tom Herbert4b282522014-06-14 23:23:52 -0700344 SKB_GSO_IPIP = 1 << 8,
Eric Dumazetcb32f512013-10-19 11:42:57 -0700345
Tom Herbert4b282522014-06-14 23:23:52 -0700346 SKB_GSO_SIT = 1 << 9,
Eric Dumazet61c1db72013-10-20 20:47:30 -0700347
Tom Herbert4b282522014-06-14 23:23:52 -0700348 SKB_GSO_UDP_TUNNEL = 1 << 10,
Tom Herbert0f4f4ff2014-06-04 17:20:16 -0700349
350 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
Tom Herbert4749c092014-06-04 17:20:23 -0700351
Tom Herbert4b282522014-06-14 23:23:52 -0700352 SKB_GSO_MPLS = 1 << 12,
353
Herbert Xu79671682006-06-22 02:40:14 -0700354};
355
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700356#if BITS_PER_LONG > 32
357#define NET_SKBUFF_DATA_USES_OFFSET 1
358#endif
359
360#ifdef NET_SKBUFF_DATA_USES_OFFSET
361typedef unsigned int sk_buff_data_t;
362#else
363typedef unsigned char *sk_buff_data_t;
364#endif
365
Eric Dumazet363ec39232014-02-26 14:02:11 -0800366/**
367 * struct skb_mstamp - multi resolution time stamps
368 * @stamp_us: timestamp in us resolution
369 * @stamp_jiffies: timestamp in jiffies
370 */
371struct skb_mstamp {
372 union {
373 u64 v64;
374 struct {
375 u32 stamp_us;
376 u32 stamp_jiffies;
377 };
378 };
379};
380
381/**
382 * skb_mstamp_get - get current timestamp
383 * @cl: place to store timestamps
384 */
385static inline void skb_mstamp_get(struct skb_mstamp *cl)
386{
387 u64 val = local_clock();
388
389 do_div(val, NSEC_PER_USEC);
390 cl->stamp_us = (u32)val;
391 cl->stamp_jiffies = (u32)jiffies;
392}
393
394/**
395 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
396 * @t1: pointer to newest sample
397 * @t0: pointer to oldest sample
398 */
399static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
400 const struct skb_mstamp *t0)
401{
402 s32 delta_us = t1->stamp_us - t0->stamp_us;
403 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
404
405 /* If delta_us is negative, this might be because interval is too big,
406 * or local_clock() drift is too big : fallback using jiffies.
407 */
408 if (delta_us <= 0 ||
409 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
410
411 delta_us = jiffies_to_usecs(delta_jiffies);
412
413 return delta_us;
414}
415
416
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417/**
418 * struct sk_buff - socket buffer
419 * @next: Next buffer in list
420 * @prev: Previous buffer in list
Eric Dumazet363ec39232014-02-26 14:02:11 -0800421 * @tstamp: Time we arrived/left
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700422 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700423 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700424 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000425 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700426 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427 * @len: Length of actual data
428 * @data_len: Data length
429 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700430 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700431 * @csum: Checksum (must include start/offset pair)
432 * @csum_start: Offset from skb->head where checksumming should start
433 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700434 * @priority: Packet queueing priority
WANG Cong60ff7462014-05-04 16:39:18 -0700435 * @ignore_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700437 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438 * @nohdr: Payload reference only, must not modify header
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700439 * @nfctinfo: Relationship of this skb to the connection
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700441 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700442 * @ipvs_property: skbuff is owned by ipvs
Randy Dunlap31729362008-02-18 20:52:13 -0800443 * @peeked: this packet has been seen already, so stats have been
444 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700445 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700446 * @protocol: Packet protocol from driver
447 * @destructor: Destruct function
448 * @nfct: Associated connection, if any
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800450 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700451 * @tc_index: Traffic control index
452 * @tc_verd: traffic control verdict
Tom Herbert61b905d2014-03-24 15:34:47 -0700453 * @hash: the packet hash
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700454 * @queue_mapping: Queue mapping for multiqueue devices
David S. Miller0b725a22014-08-25 15:51:53 -0700455 * @xmit_more: More SKBs are pending for this queue
Randy Dunlap553a5672008-04-20 10:51:01 -0700456 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700457 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Tom Herbert61b905d2014-03-24 15:34:47 -0700458 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
Changli Gao4ca24622011-08-19 07:26:44 -0700459 * ports.
Tom Herberta3b18dd2014-07-01 21:33:17 -0700460 * @sw_hash: indicates hash was computed in software stack
Johannes Berg6e3e9392011-11-09 10:15:42 +0100461 * @wifi_acked_valid: wifi_acked was set
462 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000463 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Randy Dunlapf4b8ea72006-06-22 16:00:11 -0700464 * @dma_cookie: a cookie to one of several possible DMA operations
465 * done by skb DMA functions
Eliezer Tamir06021292013-06-10 11:39:50 +0300466 * @napi_id: id of the NAPI struct this skb came from
James Morris984bc162006-06-09 00:29:17 -0700467 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700468 * @mark: Generic packet mark
469 * @dropcount: total number of sk_receive_queue overflows
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000470 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b02008-07-14 22:49:06 -0700471 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000472 * @inner_protocol: Protocol (encapsulation)
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000473 * @inner_transport_header: Inner transport layer header (encapsulation)
474 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000475 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700476 * @transport_header: Transport layer header
477 * @network_header: Network layer header
478 * @mac_header: Link layer header
479 * @tail: Tail pointer
480 * @end: End pointer
481 * @head: Head of buffer
482 * @data: Data head pointer
483 * @truesize: Buffer size
484 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485 */
486
487struct sk_buff {
488 /* These two members must be first. */
489 struct sk_buff *next;
490 struct sk_buff *prev;
491
Eric Dumazet363ec39232014-02-26 14:02:11 -0800492 union {
493 ktime_t tstamp;
494 struct skb_mstamp skb_mstamp;
495 };
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000496
497 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499
Linus Torvalds1da177e2005-04-16 15:20:36 -0700500 /*
501 * This is the control buffer. It is free to use for every
502 * layer. Please put your private variables there. If you
503 * want to keep them across layers you have to do a skb_clone()
504 * first. This is owned by whoever has the skb queued ATM.
505 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000506 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507
Eric Dumazet7fee2262010-05-11 23:19:48 +0000508 unsigned long _skb_refdst;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000509#ifdef CONFIG_XFRM
510 struct sec_path *sp;
511#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700512 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700513 data_len;
514 __u16 mac_len,
515 hdr_len;
Al Viroff1dcad2006-11-20 18:07:29 -0800516 union {
517 __wsum csum;
Herbert Xu663ead32007-04-09 11:59:07 -0700518 struct {
519 __u16 csum_start;
520 __u16 csum_offset;
521 };
Al Viroff1dcad2006-11-20 18:07:29 -0800522 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700523 __u32 priority;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200524 kmemcheck_bitfield_begin(flags1);
WANG Cong60ff7462014-05-04 16:39:18 -0700525 __u8 ignore_df:1,
Thomas Graf1cbb3382005-07-05 14:13:41 -0700526 cloned:1,
527 ip_summed:2,
Harald Welte6869c4d2005-08-09 19:24:19 -0700528 nohdr:1,
529 nfctinfo:3;
David S. Millerd179cd12005-08-17 14:57:30 -0700530 __u8 pkt_type:3,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800531 fclone:2,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700532 ipvs_property:1,
Herbert Xua59322b2007-12-05 01:53:40 -0800533 peeked:1,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700534 nf_trace:1;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200535 kmemcheck_bitfield_end(flags1);
Eric Dumazet4ab408d2010-03-01 03:09:26 +0000536 __be16 protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537
538 void (*destructor)(struct sk_buff *skb);
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800539#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700540 struct nf_conntrack *nfct;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100541#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542#ifdef CONFIG_BRIDGE_NETFILTER
543 struct nf_bridge_info *nf_bridge;
544#endif
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700545
Eric Dumazet8964be42009-11-20 15:35:04 -0800546 int skb_iif;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000547
Tom Herbert61b905d2014-03-24 15:34:47 -0700548 __u32 hash;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000549
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000550 __be16 vlan_proto;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000551 __u16 vlan_tci;
552
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700554 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700556 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700558#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200559
Eric Dumazet0a148422011-04-20 09:27:32 +0000560 __u16 queue_mapping;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200561 kmemcheck_bitfield_begin(flags2);
David S. Miller0b725a22014-08-25 15:51:53 -0700562 __u8 xmit_more:1;
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400563#ifdef CONFIG_IPV6_NDISC_NODETYPE
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000564 __u8 ndisc_nodetype:2;
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400565#endif
Mel Gormanc93bdd02012-07-31 16:44:19 -0700566 __u8 pfmemalloc:1;
Tom Herbert3853b582010-11-21 13:17:29 +0000567 __u8 ooo_okay:1;
Tom Herbert61b905d2014-03-24 15:34:47 -0700568 __u8 l4_hash:1;
Tom Herberta3b18dd2014-07-01 21:33:17 -0700569 __u8 sw_hash:1;
Johannes Berg6e3e9392011-11-09 10:15:42 +0100570 __u8 wifi_acked_valid:1;
571 __u8 wifi_acked:1;
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000572 __u8 no_fcs:1;
Eric Dumazetd3836f22012-04-27 00:33:38 +0000573 __u8 head_frag:1;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000574 /* Encapsulation protocol and NIC drivers should use
575 * this flag to indicate to each other if the skb contains
576 * encapsulated packet or not and maybe use the inner packet
577 * headers if needed
578 */
579 __u8 encapsulation:1;
Tom Herbert7e2b10c2014-06-04 17:20:02 -0700580 __u8 encap_hdr_csum:1;
Tom Herbert5d0c2b92014-06-10 18:54:13 -0700581 __u8 csum_valid:1;
Tom Herbert7e3cead2014-06-10 18:54:19 -0700582 __u8 csum_complete_sw:1;
David S. Miller5c214032014-08-27 14:39:04 -0700583 /* 1/3 bit hole (depending on ndisc_nodetype presence) */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200584 kmemcheck_bitfield_end(flags2);
585
Cong Wange0d10952013-08-01 11:10:25 +0800586#if defined CONFIG_NET_DMA || defined CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir06021292013-06-10 11:39:50 +0300587 union {
588 unsigned int napi_id;
589 dma_cookie_t dma_cookie;
590 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700591#endif
James Morris984bc162006-06-09 00:29:17 -0700592#ifdef CONFIG_NETWORK_SECMARK
593 __u32 secmark;
594#endif
Neil Horman3b885782009-10-12 13:26:31 -0700595 union {
596 __u32 mark;
597 __u32 dropcount;
Eric Dumazet16fad692013-03-14 05:40:32 +0000598 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700599 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600
Tom Herbertde20fe82014-08-27 21:26:35 -0700601 kmemcheck_bitfield_begin(flags3);
602 /* 16 bit hole */
603 kmemcheck_bitfield_end(flags3);
604
Simon Horman0d89d202013-05-23 21:02:52 +0000605 __be16 inner_protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000606 __u16 inner_transport_header;
607 __u16 inner_network_header;
608 __u16 inner_mac_header;
609 __u16 transport_header;
610 __u16 network_header;
611 __u16 mac_header;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700612 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700613 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700614 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700616 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700617 unsigned int truesize;
618 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700619};
620
621#ifdef __KERNEL__
622/*
623 * Handling routines are only of interest to the kernel
624 */
625#include <linux/slab.h>
626
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627
Mel Gormanc93bdd02012-07-31 16:44:19 -0700628#define SKB_ALLOC_FCLONE 0x01
629#define SKB_ALLOC_RX 0x02
630
631/* Returns true if the skb was allocated from PFMEMALLOC reserves */
632static inline bool skb_pfmemalloc(const struct sk_buff *skb)
633{
634 return unlikely(skb->pfmemalloc);
635}
636
Eric Dumazet7fee2262010-05-11 23:19:48 +0000637/*
638 * skb might have a dst pointer attached, refcounted or not.
639 * _skb_refdst low order bit is set if refcount was _not_ taken
640 */
641#define SKB_DST_NOREF 1UL
642#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
643
644/**
645 * skb_dst - returns skb dst_entry
646 * @skb: buffer
647 *
648 * Returns skb dst_entry, regardless of reference taken or not.
649 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000650static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
651{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000652 /* If refdst was not refcounted, check we still are in a
653 * rcu_read_lock section
654 */
655 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
656 !rcu_read_lock_held() &&
657 !rcu_read_lock_bh_held());
658 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000659}
660
Eric Dumazet7fee2262010-05-11 23:19:48 +0000661/**
662 * skb_dst_set - sets skb dst
663 * @skb: buffer
664 * @dst: dst entry
665 *
666 * Sets skb dst, assuming a reference was taken on dst and should
667 * be released by skb_dst_drop()
668 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000669static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
670{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000671 skb->_skb_refdst = (unsigned long)dst;
672}
673
Joe Perches7965bd42013-09-26 14:48:15 -0700674void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
675 bool force);
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200676
677/**
678 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
679 * @skb: buffer
680 * @dst: dst entry
681 *
682 * Sets skb dst, assuming a reference was not taken on dst.
683 * If dst entry is cached, we do not take reference and dst_release
684 * will be avoided by refdst_drop. If dst entry is not cached, we take
685 * reference, so that last dst_release can destroy the dst immediately.
686 */
687static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
688{
689 __skb_dst_set_noref(skb, dst, false);
690}
691
692/**
693 * skb_dst_set_noref_force - sets skb dst, without taking reference
694 * @skb: buffer
695 * @dst: dst entry
696 *
697 * Sets skb dst, assuming a reference was not taken on dst.
698 * No reference is taken and no dst_release will be called. While for
699 * cached dsts deferred reclaim is a basic feature, for entries that are
700 * not cached it is caller's job to guarantee that last dst_release for
701 * provided dst happens when nobody uses it, eg. after a RCU grace period.
702 */
703static inline void skb_dst_set_noref_force(struct sk_buff *skb,
704 struct dst_entry *dst)
705{
706 __skb_dst_set_noref(skb, dst, true);
707}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000708
709/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300710 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000711 * @skb: buffer
712 */
713static inline bool skb_dst_is_noref(const struct sk_buff *skb)
714{
715 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000716}
717
Eric Dumazet511c3f92009-06-02 05:14:27 +0000718static inline struct rtable *skb_rtable(const struct sk_buff *skb)
719{
Eric Dumazetadf30902009-06-02 05:19:30 +0000720 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000721}
722
Joe Perches7965bd42013-09-26 14:48:15 -0700723void kfree_skb(struct sk_buff *skb);
724void kfree_skb_list(struct sk_buff *segs);
725void skb_tx_error(struct sk_buff *skb);
726void consume_skb(struct sk_buff *skb);
727void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000728extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000729
Joe Perches7965bd42013-09-26 14:48:15 -0700730void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
731bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
732 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000733
Joe Perches7965bd42013-09-26 14:48:15 -0700734struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
735 int node);
736struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700737static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100738 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700739{
Eric Dumazet564824b2010-10-11 19:05:25 +0000740 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700741}
742
743static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100744 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700745{
Mel Gormanc93bdd02012-07-31 16:44:19 -0700746 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700747}
748
Joe Perches7965bd42013-09-26 14:48:15 -0700749struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
Patrick McHardy0ebd0ac2013-04-17 06:46:58 +0000750static inline struct sk_buff *alloc_skb_head(gfp_t priority)
751{
752 return __alloc_skb_head(priority, -1);
753}
754
Joe Perches7965bd42013-09-26 14:48:15 -0700755struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
756int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
757struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
758struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
Octavian Purdilabad93e92014-06-12 01:36:26 +0300759struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
760 gfp_t gfp_mask, bool fclone);
761static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
762 gfp_t gfp_mask)
763{
764 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
765}
Eric Dumazet117632e2011-12-03 21:39:53 +0000766
Joe Perches7965bd42013-09-26 14:48:15 -0700767int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
768struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
769 unsigned int headroom);
770struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
771 int newtailroom, gfp_t priority);
Fan Du25a91d82014-01-18 09:54:23 +0800772int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
773 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -0700774int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
775 int len);
776int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
777int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000778#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700779
Joe Perches7965bd42013-09-26 14:48:15 -0700780int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
781 int getfrag(void *from, char *to, int offset,
782 int len, int odd, struct sk_buff *skb),
783 void *from, int length);
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700784
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800785struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700786 __u32 lower_offset;
787 __u32 upper_offset;
788 __u32 frag_idx;
789 __u32 stepped_offset;
790 struct sk_buff *root_skb;
791 struct sk_buff *cur_skb;
792 __u8 *frag_data;
793};
794
Joe Perches7965bd42013-09-26 14:48:15 -0700795void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
796 unsigned int to, struct skb_seq_state *st);
797unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
798 struct skb_seq_state *st);
799void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -0700800
Joe Perches7965bd42013-09-26 14:48:15 -0700801unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
802 unsigned int to, struct ts_config *config,
803 struct ts_state *state);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700804
Tom Herbert09323cc2013-12-15 22:16:19 -0800805/*
806 * Packet hash types specify the type of hash in skb_set_hash.
807 *
808 * Hash types refer to the protocol layer addresses which are used to
809 * construct a packet's hash. The hashes are used to differentiate or identify
810 * flows of the protocol layer for the hash type. Hash types are either
811 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
812 *
813 * Properties of hashes:
814 *
815 * 1) Two packets in different flows have different hash values
816 * 2) Two packets in the same flow should have the same hash value
817 *
818 * A hash at a higher layer is considered to be more specific. A driver should
819 * set the most specific hash possible.
820 *
821 * A driver cannot indicate a more specific hash than the layer at which a hash
822 * was computed. For instance an L3 hash cannot be set as an L4 hash.
823 *
824 * A driver may indicate a hash level which is less specific than the
825 * actual layer the hash was computed on. For instance, a hash computed
826 * at L4 may be considered an L3 hash. This should only be done if the
827 * driver can't unambiguously determine that the HW computed the hash at
828 * the higher layer. Note that the "should" in the second property above
829 * permits this.
830 */
831enum pkt_hash_types {
832 PKT_HASH_TYPE_NONE, /* Undefined type */
833 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
834 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
835 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
836};
837
838static inline void
839skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
840{
Tom Herbert61b905d2014-03-24 15:34:47 -0700841 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
Tom Herberta3b18dd2014-07-01 21:33:17 -0700842 skb->sw_hash = 0;
Tom Herbert61b905d2014-03-24 15:34:47 -0700843 skb->hash = hash;
Tom Herbert09323cc2013-12-15 22:16:19 -0800844}
845
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800846void __skb_get_hash(struct sk_buff *skb);
847static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000848{
Tom Herberta3b18dd2014-07-01 21:33:17 -0700849 if (!skb->l4_hash && !skb->sw_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800850 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000851
Tom Herbert61b905d2014-03-24 15:34:47 -0700852 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000853}
854
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800855static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
856{
Tom Herbert61b905d2014-03-24 15:34:47 -0700857 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800858}
859
Tom Herbert7539fad2013-12-15 22:12:18 -0800860static inline void skb_clear_hash(struct sk_buff *skb)
861{
Tom Herbert61b905d2014-03-24 15:34:47 -0700862 skb->hash = 0;
Tom Herberta3b18dd2014-07-01 21:33:17 -0700863 skb->sw_hash = 0;
Tom Herbert61b905d2014-03-24 15:34:47 -0700864 skb->l4_hash = 0;
Tom Herbert7539fad2013-12-15 22:12:18 -0800865}
866
867static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
868{
Tom Herbert61b905d2014-03-24 15:34:47 -0700869 if (!skb->l4_hash)
Tom Herbert7539fad2013-12-15 22:12:18 -0800870 skb_clear_hash(skb);
871}
872
Tom Herbert3df7a742013-12-15 22:16:29 -0800873static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
874{
Tom Herbert61b905d2014-03-24 15:34:47 -0700875 to->hash = from->hash;
Tom Herberta3b18dd2014-07-01 21:33:17 -0700876 to->sw_hash = from->sw_hash;
Tom Herbert61b905d2014-03-24 15:34:47 -0700877 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -0800878};
879
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700880#ifdef NET_SKBUFF_DATA_USES_OFFSET
881static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
882{
883 return skb->head + skb->end;
884}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000885
886static inline unsigned int skb_end_offset(const struct sk_buff *skb)
887{
888 return skb->end;
889}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700890#else
891static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
892{
893 return skb->end;
894}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000895
896static inline unsigned int skb_end_offset(const struct sk_buff *skb)
897{
898 return skb->end - skb->head;
899}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700900#endif
901
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700903#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904
Patrick Ohlyac45f602009-02-12 05:03:37 +0000905static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
906{
907 return &skb_shinfo(skb)->hwtstamps;
908}
909
Linus Torvalds1da177e2005-04-16 15:20:36 -0700910/**
911 * skb_queue_empty - check if a queue is empty
912 * @list: queue head
913 *
914 * Returns true if the queue is empty, false otherwise.
915 */
916static inline int skb_queue_empty(const struct sk_buff_head *list)
917{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100918 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700919}
920
921/**
David S. Millerfc7ebb22008-09-23 00:34:07 -0700922 * skb_queue_is_last - check if skb is the last entry in the queue
923 * @list: queue head
924 * @skb: buffer
925 *
926 * Returns true if @skb is the last buffer on the list.
927 */
928static inline bool skb_queue_is_last(const struct sk_buff_head *list,
929 const struct sk_buff *skb)
930{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100931 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -0700932}
933
934/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800935 * skb_queue_is_first - check if skb is the first entry in the queue
936 * @list: queue head
937 * @skb: buffer
938 *
939 * Returns true if @skb is the first buffer on the list.
940 */
941static inline bool skb_queue_is_first(const struct sk_buff_head *list,
942 const struct sk_buff *skb)
943{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100944 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800945}
946
947/**
David S. Miller249c8b42008-09-23 00:44:42 -0700948 * skb_queue_next - return the next packet in the queue
949 * @list: queue head
950 * @skb: current buffer
951 *
952 * Return the next packet in @list after @skb. It is only valid to
953 * call this if skb_queue_is_last() evaluates to false.
954 */
955static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
956 const struct sk_buff *skb)
957{
958 /* This BUG_ON may seem severe, but if we just return then we
959 * are going to dereference garbage.
960 */
961 BUG_ON(skb_queue_is_last(list, skb));
962 return skb->next;
963}
964
965/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800966 * skb_queue_prev - return the prev packet in the queue
967 * @list: queue head
968 * @skb: current buffer
969 *
970 * Return the prev packet in @list before @skb. It is only valid to
971 * call this if skb_queue_is_first() evaluates to false.
972 */
973static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
974 const struct sk_buff *skb)
975{
976 /* This BUG_ON may seem severe, but if we just return then we
977 * are going to dereference garbage.
978 */
979 BUG_ON(skb_queue_is_first(list, skb));
980 return skb->prev;
981}
982
983/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700984 * skb_get - reference buffer
985 * @skb: buffer to reference
986 *
987 * Makes another reference to a socket buffer and returns a pointer
988 * to the buffer.
989 */
990static inline struct sk_buff *skb_get(struct sk_buff *skb)
991{
992 atomic_inc(&skb->users);
993 return skb;
994}
995
996/*
997 * If users == 1, we are the only owner and are can avoid redundant
998 * atomic change.
999 */
1000
1001/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001002 * skb_cloned - is the buffer a clone
1003 * @skb: buffer to check
1004 *
1005 * Returns true if the buffer was generated with skb_clone() and is
1006 * one of multiple shared copies of the buffer. Cloned buffers are
1007 * shared data so must not be written to under normal circumstances.
1008 */
1009static inline int skb_cloned(const struct sk_buff *skb)
1010{
1011 return skb->cloned &&
1012 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1013}
1014
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001015static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1016{
1017 might_sleep_if(pri & __GFP_WAIT);
1018
1019 if (skb_cloned(skb))
1020 return pskb_expand_head(skb, 0, 0, pri);
1021
1022 return 0;
1023}
1024
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025/**
1026 * skb_header_cloned - is the header a clone
1027 * @skb: buffer to check
1028 *
1029 * Returns true if modifying the header part of the buffer requires
1030 * the data to be copied.
1031 */
1032static inline int skb_header_cloned(const struct sk_buff *skb)
1033{
1034 int dataref;
1035
1036 if (!skb->cloned)
1037 return 0;
1038
1039 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1040 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1041 return dataref != 1;
1042}
1043
1044/**
1045 * skb_header_release - release reference to header
1046 * @skb: buffer to operate on
1047 *
1048 * Drop a reference to the header part of the buffer. This is done
1049 * by acquiring a payload reference. You must not read from the header
1050 * part of skb->data after this.
1051 */
1052static inline void skb_header_release(struct sk_buff *skb)
1053{
1054 BUG_ON(skb->nohdr);
1055 skb->nohdr = 1;
1056 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1057}
1058
1059/**
1060 * skb_shared - is the buffer shared
1061 * @skb: buffer to check
1062 *
1063 * Returns true if more than one person has a reference to this
1064 * buffer.
1065 */
1066static inline int skb_shared(const struct sk_buff *skb)
1067{
1068 return atomic_read(&skb->users) != 1;
1069}
1070
1071/**
1072 * skb_share_check - check if buffer is shared and if so clone it
1073 * @skb: buffer to check
1074 * @pri: priority for memory allocation
1075 *
1076 * If the buffer is shared the buffer is cloned and the old copy
1077 * drops a reference. A new clone with a single reference is returned.
1078 * If the buffer is not shared the original buffer is returned. When
1079 * being called from interrupt status or with spinlocks held pri must
1080 * be GFP_ATOMIC.
1081 *
1082 * NULL is returned on a memory allocation failure.
1083 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001084static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001085{
1086 might_sleep_if(pri & __GFP_WAIT);
1087 if (skb_shared(skb)) {
1088 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001089
1090 if (likely(nskb))
1091 consume_skb(skb);
1092 else
1093 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001094 skb = nskb;
1095 }
1096 return skb;
1097}
1098
1099/*
1100 * Copy shared buffers into a new sk_buff. We effectively do COW on
1101 * packets to handle cases where we have a local reader and forward
1102 * and a couple of other messy ones. The normal one is tcpdumping
1103 * a packet thats being forwarded.
1104 */
1105
1106/**
1107 * skb_unshare - make a copy of a shared buffer
1108 * @skb: buffer to check
1109 * @pri: priority for memory allocation
1110 *
1111 * If the socket buffer is a clone then this function creates a new
1112 * copy of the data, drops a reference count on the old copy and returns
1113 * the new copy with the reference count at 1. If the buffer is not a clone
1114 * the original buffer is returned. When called with a spinlock held or
1115 * from interrupt state @pri must be %GFP_ATOMIC
1116 *
1117 * %NULL is returned on a memory allocation failure.
1118 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001119static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001120 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121{
1122 might_sleep_if(pri & __GFP_WAIT);
1123 if (skb_cloned(skb)) {
1124 struct sk_buff *nskb = skb_copy(skb, pri);
1125 kfree_skb(skb); /* Free our shared copy */
1126 skb = nskb;
1127 }
1128 return skb;
1129}
1130
1131/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001132 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133 * @list_: list to peek at
1134 *
1135 * Peek an &sk_buff. Unlike most other operations you _MUST_
1136 * be careful with this one. A peek leaves the buffer on the
1137 * list and someone else may run off with it. You must hold
1138 * the appropriate locks or have a private queue to do this.
1139 *
1140 * Returns %NULL for an empty list or a pointer to the head element.
1141 * The reference count is not incremented and the reference is therefore
1142 * volatile. Use with caution.
1143 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001144static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001145{
Eric Dumazet18d07002012-04-30 16:31:46 +00001146 struct sk_buff *skb = list_->next;
1147
1148 if (skb == (struct sk_buff *)list_)
1149 skb = NULL;
1150 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001151}
1152
1153/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001154 * skb_peek_next - peek skb following the given one from a queue
1155 * @skb: skb to start from
1156 * @list_: list to peek at
1157 *
1158 * Returns %NULL when the end of the list is met or a pointer to the
1159 * next element. The reference count is not incremented and the
1160 * reference is therefore volatile. Use with caution.
1161 */
1162static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1163 const struct sk_buff_head *list_)
1164{
1165 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001166
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001167 if (next == (struct sk_buff *)list_)
1168 next = NULL;
1169 return next;
1170}
1171
1172/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001173 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001174 * @list_: list to peek at
1175 *
1176 * Peek an &sk_buff. Unlike most other operations you _MUST_
1177 * be careful with this one. A peek leaves the buffer on the
1178 * list and someone else may run off with it. You must hold
1179 * the appropriate locks or have a private queue to do this.
1180 *
1181 * Returns %NULL for an empty list or a pointer to the tail element.
1182 * The reference count is not incremented and the reference is therefore
1183 * volatile. Use with caution.
1184 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001185static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001186{
Eric Dumazet18d07002012-04-30 16:31:46 +00001187 struct sk_buff *skb = list_->prev;
1188
1189 if (skb == (struct sk_buff *)list_)
1190 skb = NULL;
1191 return skb;
1192
Linus Torvalds1da177e2005-04-16 15:20:36 -07001193}
1194
1195/**
1196 * skb_queue_len - get queue length
1197 * @list_: list to measure
1198 *
1199 * Return the length of an &sk_buff queue.
1200 */
1201static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1202{
1203 return list_->qlen;
1204}
1205
David S. Miller67fed452008-09-21 22:36:24 -07001206/**
1207 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1208 * @list: queue to initialize
1209 *
1210 * This initializes only the list and queue length aspects of
1211 * an sk_buff_head object. This allows to initialize the list
1212 * aspects of an sk_buff_head without reinitializing things like
1213 * the spinlock. It can also be used for on-stack sk_buff_head
1214 * objects where the spinlock is known to not be used.
1215 */
1216static inline void __skb_queue_head_init(struct sk_buff_head *list)
1217{
1218 list->prev = list->next = (struct sk_buff *)list;
1219 list->qlen = 0;
1220}
1221
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001222/*
1223 * This function creates a split out lock class for each invocation;
1224 * this is needed for now since a whole lot of users of the skb-queue
1225 * infrastructure in drivers have different locking usage (in hardirq)
1226 * than the networking core (in softirq only). In the long run either the
1227 * network layer or drivers should need annotation to consolidate the
1228 * main types of usage into 3 classes.
1229 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230static inline void skb_queue_head_init(struct sk_buff_head *list)
1231{
1232 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001233 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001234}
1235
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001236static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1237 struct lock_class_key *class)
1238{
1239 skb_queue_head_init(list);
1240 lockdep_set_class(&list->lock, class);
1241}
1242
Linus Torvalds1da177e2005-04-16 15:20:36 -07001243/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001244 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001245 *
1246 * The "__skb_xxxx()" functions are the non-atomic ones that
1247 * can only be called with interrupts disabled.
1248 */
Joe Perches7965bd42013-09-26 14:48:15 -07001249void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1250 struct sk_buff_head *list);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001251static inline void __skb_insert(struct sk_buff *newsk,
1252 struct sk_buff *prev, struct sk_buff *next,
1253 struct sk_buff_head *list)
1254{
1255 newsk->next = next;
1256 newsk->prev = prev;
1257 next->prev = prev->next = newsk;
1258 list->qlen++;
1259}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001260
David S. Miller67fed452008-09-21 22:36:24 -07001261static inline void __skb_queue_splice(const struct sk_buff_head *list,
1262 struct sk_buff *prev,
1263 struct sk_buff *next)
1264{
1265 struct sk_buff *first = list->next;
1266 struct sk_buff *last = list->prev;
1267
1268 first->prev = prev;
1269 prev->next = first;
1270
1271 last->next = next;
1272 next->prev = last;
1273}
1274
1275/**
1276 * skb_queue_splice - join two skb lists, this is designed for stacks
1277 * @list: the new list to add
1278 * @head: the place to add it in the first list
1279 */
1280static inline void skb_queue_splice(const struct sk_buff_head *list,
1281 struct sk_buff_head *head)
1282{
1283 if (!skb_queue_empty(list)) {
1284 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001285 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001286 }
1287}
1288
1289/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001290 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001291 * @list: the new list to add
1292 * @head: the place to add it in the first list
1293 *
1294 * The list at @list is reinitialised
1295 */
1296static inline void skb_queue_splice_init(struct sk_buff_head *list,
1297 struct sk_buff_head *head)
1298{
1299 if (!skb_queue_empty(list)) {
1300 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001301 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001302 __skb_queue_head_init(list);
1303 }
1304}
1305
1306/**
1307 * skb_queue_splice_tail - join two skb lists, each list being a queue
1308 * @list: the new list to add
1309 * @head: the place to add it in the first list
1310 */
1311static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1312 struct sk_buff_head *head)
1313{
1314 if (!skb_queue_empty(list)) {
1315 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001316 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001317 }
1318}
1319
1320/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001321 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001322 * @list: the new list to add
1323 * @head: the place to add it in the first list
1324 *
1325 * Each of the lists is a queue.
1326 * The list at @list is reinitialised
1327 */
1328static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1329 struct sk_buff_head *head)
1330{
1331 if (!skb_queue_empty(list)) {
1332 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001333 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001334 __skb_queue_head_init(list);
1335 }
1336}
1337
Linus Torvalds1da177e2005-04-16 15:20:36 -07001338/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001339 * __skb_queue_after - queue a buffer at the list head
1340 * @list: list to use
1341 * @prev: place after this buffer
1342 * @newsk: buffer to queue
1343 *
1344 * Queue a buffer int the middle of a list. This function takes no locks
1345 * and you must therefore hold required locks before calling it.
1346 *
1347 * A buffer cannot be placed on two lists at the same time.
1348 */
1349static inline void __skb_queue_after(struct sk_buff_head *list,
1350 struct sk_buff *prev,
1351 struct sk_buff *newsk)
1352{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001353 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001354}
1355
Joe Perches7965bd42013-09-26 14:48:15 -07001356void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1357 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001358
Gerrit Renkerf5572852008-04-14 00:05:28 -07001359static inline void __skb_queue_before(struct sk_buff_head *list,
1360 struct sk_buff *next,
1361 struct sk_buff *newsk)
1362{
1363 __skb_insert(newsk, next->prev, next, list);
1364}
1365
Stephen Hemminger300ce172005-10-30 13:47:34 -08001366/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001367 * __skb_queue_head - queue a buffer at the list head
1368 * @list: list to use
1369 * @newsk: buffer to queue
1370 *
1371 * Queue a buffer at the start of a list. This function takes no locks
1372 * and you must therefore hold required locks before calling it.
1373 *
1374 * A buffer cannot be placed on two lists at the same time.
1375 */
Joe Perches7965bd42013-09-26 14:48:15 -07001376void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001377static inline void __skb_queue_head(struct sk_buff_head *list,
1378 struct sk_buff *newsk)
1379{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001380 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001381}
1382
1383/**
1384 * __skb_queue_tail - queue a buffer at the list tail
1385 * @list: list to use
1386 * @newsk: buffer to queue
1387 *
1388 * Queue a buffer at the end of a list. This function takes no locks
1389 * and you must therefore hold required locks before calling it.
1390 *
1391 * A buffer cannot be placed on two lists at the same time.
1392 */
Joe Perches7965bd42013-09-26 14:48:15 -07001393void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394static inline void __skb_queue_tail(struct sk_buff_head *list,
1395 struct sk_buff *newsk)
1396{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001397 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398}
1399
Linus Torvalds1da177e2005-04-16 15:20:36 -07001400/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001401 * remove sk_buff from list. _Must_ be called atomically, and with
1402 * the list known..
1403 */
Joe Perches7965bd42013-09-26 14:48:15 -07001404void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001405static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1406{
1407 struct sk_buff *next, *prev;
1408
1409 list->qlen--;
1410 next = skb->next;
1411 prev = skb->prev;
1412 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001413 next->prev = prev;
1414 prev->next = next;
1415}
1416
Gerrit Renkerf525c062008-04-14 00:04:12 -07001417/**
1418 * __skb_dequeue - remove from the head of the queue
1419 * @list: list to dequeue from
1420 *
1421 * Remove the head of the list. This function does not take any locks
1422 * so must be used with appropriate locks held only. The head item is
1423 * returned or %NULL if the list is empty.
1424 */
Joe Perches7965bd42013-09-26 14:48:15 -07001425struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Gerrit Renkerf525c062008-04-14 00:04:12 -07001426static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1427{
1428 struct sk_buff *skb = skb_peek(list);
1429 if (skb)
1430 __skb_unlink(skb, list);
1431 return skb;
1432}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001433
1434/**
1435 * __skb_dequeue_tail - remove from the tail of the queue
1436 * @list: list to dequeue from
1437 *
1438 * Remove the tail of the list. This function does not take any locks
1439 * so must be used with appropriate locks held only. The tail item is
1440 * returned or %NULL if the list is empty.
1441 */
Joe Perches7965bd42013-09-26 14:48:15 -07001442struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001443static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1444{
1445 struct sk_buff *skb = skb_peek_tail(list);
1446 if (skb)
1447 __skb_unlink(skb, list);
1448 return skb;
1449}
1450
1451
David S. Millerbdcc0922012-03-07 20:53:36 -05001452static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001453{
1454 return skb->data_len;
1455}
1456
1457static inline unsigned int skb_headlen(const struct sk_buff *skb)
1458{
1459 return skb->len - skb->data_len;
1460}
1461
1462static inline int skb_pagelen(const struct sk_buff *skb)
1463{
1464 int i, len = 0;
1465
1466 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001467 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001468 return len + skb_headlen(skb);
1469}
1470
Ian Campbell131ea662011-08-19 06:25:00 +00001471/**
1472 * __skb_fill_page_desc - initialise a paged fragment in an skb
1473 * @skb: buffer containing fragment to be initialised
1474 * @i: paged fragment index to initialise
1475 * @page: the page to use for this fragment
1476 * @off: the offset to the data with @page
1477 * @size: the length of the data
1478 *
1479 * Initialises the @i'th fragment of @skb to point to &size bytes at
1480 * offset @off within @page.
1481 *
1482 * Does not take any additional reference on the fragment.
1483 */
1484static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1485 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001486{
1487 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1488
Mel Gormanc48a11c2012-07-31 16:44:23 -07001489 /*
1490 * Propagate page->pfmemalloc to the skb if we can. The problem is
1491 * that not all callers have unique ownership of the page. If
1492 * pfmemalloc is set, we check the mapping as a mapping implies
1493 * page->index is set (index and pfmemalloc share space).
1494 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1495 * do not lose pfmemalloc information as the pages would not be
1496 * allocated using __GFP_MEMALLOC.
1497 */
Ian Campbella8605c62011-10-19 23:01:49 +00001498 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001500 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001501
1502 page = compound_head(page);
1503 if (page->pfmemalloc && !page->mapping)
1504 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00001505}
1506
1507/**
1508 * skb_fill_page_desc - initialise a paged fragment in an skb
1509 * @skb: buffer containing fragment to be initialised
1510 * @i: paged fragment index to initialise
1511 * @page: the page to use for this fragment
1512 * @off: the offset to the data with @page
1513 * @size: the length of the data
1514 *
1515 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01001516 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00001517 * addition updates @skb such that @i is the last fragment.
1518 *
1519 * Does not take any additional reference on the fragment.
1520 */
1521static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1522 struct page *page, int off, int size)
1523{
1524 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001525 skb_shinfo(skb)->nr_frags = i + 1;
1526}
1527
Joe Perches7965bd42013-09-26 14:48:15 -07001528void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1529 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001530
Jason Wangf8e617e2013-11-01 14:07:47 +08001531void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1532 unsigned int truesize);
1533
Linus Torvalds1da177e2005-04-16 15:20:36 -07001534#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001535#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001536#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1537
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001538#ifdef NET_SKBUFF_DATA_USES_OFFSET
1539static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1540{
1541 return skb->head + skb->tail;
1542}
1543
1544static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1545{
1546 skb->tail = skb->data - skb->head;
1547}
1548
1549static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1550{
1551 skb_reset_tail_pointer(skb);
1552 skb->tail += offset;
1553}
Simon Horman7cc46192013-05-28 20:34:29 +00001554
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001555#else /* NET_SKBUFF_DATA_USES_OFFSET */
1556static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1557{
1558 return skb->tail;
1559}
1560
1561static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1562{
1563 skb->tail = skb->data;
1564}
1565
1566static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1567{
1568 skb->tail = skb->data + offset;
1569}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001570
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001571#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1572
Linus Torvalds1da177e2005-04-16 15:20:36 -07001573/*
1574 * Add data to an sk_buff
1575 */
Mathias Krause0c7ddf32013-11-07 14:18:24 +01001576unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001577unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001578static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1579{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001580 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001581 SKB_LINEAR_ASSERT(skb);
1582 skb->tail += len;
1583 skb->len += len;
1584 return tmp;
1585}
1586
Joe Perches7965bd42013-09-26 14:48:15 -07001587unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001588static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1589{
1590 skb->data -= len;
1591 skb->len += len;
1592 return skb->data;
1593}
1594
Joe Perches7965bd42013-09-26 14:48:15 -07001595unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001596static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1597{
1598 skb->len -= len;
1599 BUG_ON(skb->len < skb->data_len);
1600 return skb->data += len;
1601}
1602
David S. Miller47d29642010-05-02 02:21:44 -07001603static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1604{
1605 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1606}
1607
Joe Perches7965bd42013-09-26 14:48:15 -07001608unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001609
1610static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1611{
1612 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001613 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001614 return NULL;
1615 skb->len -= len;
1616 return skb->data += len;
1617}
1618
1619static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1620{
1621 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1622}
1623
1624static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1625{
1626 if (likely(len <= skb_headlen(skb)))
1627 return 1;
1628 if (unlikely(len > skb->len))
1629 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001630 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001631}
1632
1633/**
1634 * skb_headroom - bytes at buffer head
1635 * @skb: buffer to check
1636 *
1637 * Return the number of bytes of free space at the head of an &sk_buff.
1638 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001639static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001640{
1641 return skb->data - skb->head;
1642}
1643
1644/**
1645 * skb_tailroom - 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 */
1650static inline int skb_tailroom(const struct sk_buff *skb)
1651{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001652 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653}
1654
1655/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00001656 * skb_availroom - bytes at buffer end
1657 * @skb: buffer to check
1658 *
1659 * Return the number of bytes of free space at the tail of an sk_buff
1660 * allocated by sk_stream_alloc()
1661 */
1662static inline int skb_availroom(const struct sk_buff *skb)
1663{
Eric Dumazet16fad692013-03-14 05:40:32 +00001664 if (skb_is_nonlinear(skb))
1665 return 0;
1666
1667 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00001668}
1669
1670/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001671 * skb_reserve - adjust headroom
1672 * @skb: buffer to alter
1673 * @len: bytes to move
1674 *
1675 * Increase the headroom of an empty &sk_buff by reducing the tail
1676 * room. This is only allowed for an empty buffer.
1677 */
David S. Miller8243126c2006-01-17 02:54:21 -08001678static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001679{
1680 skb->data += len;
1681 skb->tail += len;
1682}
1683
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001684static inline void skb_reset_inner_headers(struct sk_buff *skb)
1685{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001686 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001687 skb->inner_network_header = skb->network_header;
1688 skb->inner_transport_header = skb->transport_header;
1689}
1690
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00001691static inline void skb_reset_mac_len(struct sk_buff *skb)
1692{
1693 skb->mac_len = skb->network_header - skb->mac_header;
1694}
1695
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001696static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1697 *skb)
1698{
1699 return skb->head + skb->inner_transport_header;
1700}
1701
1702static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1703{
1704 skb->inner_transport_header = skb->data - skb->head;
1705}
1706
1707static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1708 const int offset)
1709{
1710 skb_reset_inner_transport_header(skb);
1711 skb->inner_transport_header += offset;
1712}
1713
1714static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1715{
1716 return skb->head + skb->inner_network_header;
1717}
1718
1719static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1720{
1721 skb->inner_network_header = skb->data - skb->head;
1722}
1723
1724static inline void skb_set_inner_network_header(struct sk_buff *skb,
1725 const int offset)
1726{
1727 skb_reset_inner_network_header(skb);
1728 skb->inner_network_header += offset;
1729}
1730
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001731static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1732{
1733 return skb->head + skb->inner_mac_header;
1734}
1735
1736static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1737{
1738 skb->inner_mac_header = skb->data - skb->head;
1739}
1740
1741static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1742 const int offset)
1743{
1744 skb_reset_inner_mac_header(skb);
1745 skb->inner_mac_header += offset;
1746}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001747static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1748{
Cong Wang35d04612013-05-29 15:16:05 +08001749 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001750}
1751
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001752static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1753{
1754 return skb->head + skb->transport_header;
1755}
1756
1757static inline void skb_reset_transport_header(struct sk_buff *skb)
1758{
1759 skb->transport_header = skb->data - skb->head;
1760}
1761
1762static inline void skb_set_transport_header(struct sk_buff *skb,
1763 const int offset)
1764{
1765 skb_reset_transport_header(skb);
1766 skb->transport_header += offset;
1767}
1768
1769static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1770{
1771 return skb->head + skb->network_header;
1772}
1773
1774static inline void skb_reset_network_header(struct sk_buff *skb)
1775{
1776 skb->network_header = skb->data - skb->head;
1777}
1778
1779static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1780{
1781 skb_reset_network_header(skb);
1782 skb->network_header += offset;
1783}
1784
1785static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1786{
1787 return skb->head + skb->mac_header;
1788}
1789
1790static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1791{
Cong Wang35d04612013-05-29 15:16:05 +08001792 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001793}
1794
1795static inline void skb_reset_mac_header(struct sk_buff *skb)
1796{
1797 skb->mac_header = skb->data - skb->head;
1798}
1799
1800static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1801{
1802 skb_reset_mac_header(skb);
1803 skb->mac_header += offset;
1804}
1805
Timo Teräs0e3da5b2013-12-16 11:02:09 +02001806static inline void skb_pop_mac_header(struct sk_buff *skb)
1807{
1808 skb->mac_header = skb->network_header;
1809}
1810
Ying Xuefbbdb8f2013-03-27 16:46:06 +00001811static inline void skb_probe_transport_header(struct sk_buff *skb,
1812 const int offset_hint)
1813{
1814 struct flow_keys keys;
1815
1816 if (skb_transport_header_was_set(skb))
1817 return;
1818 else if (skb_flow_dissect(skb, &keys))
1819 skb_set_transport_header(skb, keys.thoff);
1820 else
1821 skb_set_transport_header(skb, offset_hint);
1822}
1823
Eric Dumazet03606892012-02-23 10:55:02 +00001824static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1825{
1826 if (skb_mac_header_was_set(skb)) {
1827 const unsigned char *old_mac = skb_mac_header(skb);
1828
1829 skb_set_mac_header(skb, -skb->mac_len);
1830 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1831 }
1832}
1833
Michał Mirosław04fb4512010-12-14 15:24:08 +00001834static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1835{
1836 return skb->csum_start - skb_headroom(skb);
1837}
1838
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001839static inline int skb_transport_offset(const struct sk_buff *skb)
1840{
1841 return skb_transport_header(skb) - skb->data;
1842}
1843
1844static inline u32 skb_network_header_len(const struct sk_buff *skb)
1845{
1846 return skb->transport_header - skb->network_header;
1847}
1848
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001849static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1850{
1851 return skb->inner_transport_header - skb->inner_network_header;
1852}
1853
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001854static inline int skb_network_offset(const struct sk_buff *skb)
1855{
1856 return skb_network_header(skb) - skb->data;
1857}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001858
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001859static inline int skb_inner_network_offset(const struct sk_buff *skb)
1860{
1861 return skb_inner_network_header(skb) - skb->data;
1862}
1863
Changli Gaof9599ce2010-08-04 04:43:44 +00001864static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1865{
1866 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1867}
1868
Tom Herberte5eb4e32014-06-14 23:24:28 -07001869static inline void skb_pop_rcv_encapsulation(struct sk_buff *skb)
1870{
1871 /* Only continue with checksum unnecessary if device indicated
1872 * it is valid across encapsulation (skb->encapsulation was set).
1873 */
1874 if (skb->ip_summed == CHECKSUM_UNNECESSARY && !skb->encapsulation)
1875 skb->ip_summed = CHECKSUM_NONE;
1876
1877 skb->encapsulation = 0;
1878 skb->csum_valid = 0;
1879}
1880
Linus Torvalds1da177e2005-04-16 15:20:36 -07001881/*
1882 * CPUs often take a performance hit when accessing unaligned memory
1883 * locations. The actual performance hit varies, it can be small if the
1884 * hardware handles it or large if we have to take an exception and fix it
1885 * in software.
1886 *
1887 * Since an ethernet header is 14 bytes network drivers often end up with
1888 * the IP header at an unaligned offset. The IP header can be aligned by
1889 * shifting the start of the packet by 2 bytes. Drivers should do this
1890 * with:
1891 *
Tobias Klauser8660c122009-07-13 22:48:16 +00001892 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001893 *
1894 * The downside to this alignment of the IP header is that the DMA is now
1895 * unaligned. On some architectures the cost of an unaligned DMA is high
1896 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00001897 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001898 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1899 * to be overridden.
1900 */
1901#ifndef NET_IP_ALIGN
1902#define NET_IP_ALIGN 2
1903#endif
1904
Anton Blanchard025be812006-03-31 02:27:06 -08001905/*
1906 * The networking layer reserves some headroom in skb data (via
1907 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1908 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08001909 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08001910 *
1911 * Unfortunately this headroom changes the DMA alignment of the resulting
1912 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1913 * on some architectures. An architecture can override this value,
1914 * perhaps setting it to a cacheline in size (since that will maintain
1915 * cacheline alignment of the DMA). It must be a power of 2.
1916 *
David S. Millerd6301d32009-02-08 19:24:13 -08001917 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08001918 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07001919 *
1920 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1921 * to reduce average number of cache lines per packet.
1922 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07001923 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08001924 */
1925#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07001926#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08001927#endif
1928
Joe Perches7965bd42013-09-26 14:48:15 -07001929int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001930
1931static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1932{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00001933 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07001934 WARN_ON(1);
1935 return;
1936 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001937 skb->len = len;
1938 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001939}
1940
Joe Perches7965bd42013-09-26 14:48:15 -07001941void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001942
1943static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1944{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001945 if (skb->data_len)
1946 return ___pskb_trim(skb, len);
1947 __skb_trim(skb, len);
1948 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001949}
1950
1951static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1952{
1953 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1954}
1955
1956/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07001957 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1958 * @skb: buffer to alter
1959 * @len: new length
1960 *
1961 * This is identical to pskb_trim except that the caller knows that
1962 * the skb is not cloned so we should never get an error due to out-
1963 * of-memory.
1964 */
1965static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1966{
1967 int err = pskb_trim(skb, len);
1968 BUG_ON(err);
1969}
1970
1971/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001972 * skb_orphan - orphan a buffer
1973 * @skb: buffer to orphan
1974 *
1975 * If a buffer currently has an owner then we call the owner's
1976 * destructor function and make the @skb unowned. The buffer continues
1977 * to exist but is no longer charged to its former owner.
1978 */
1979static inline void skb_orphan(struct sk_buff *skb)
1980{
Eric Dumazetc34a7612013-07-30 16:11:15 -07001981 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001982 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07001983 skb->destructor = NULL;
1984 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07001985 } else {
1986 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07001987 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001988}
1989
1990/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00001991 * skb_orphan_frags - orphan the frags contained in a buffer
1992 * @skb: buffer to orphan frags from
1993 * @gfp_mask: allocation mask for replacement pages
1994 *
1995 * For each frag in the SKB which needs a destructor (i.e. has an
1996 * owner) create a copy of that frag and release the original
1997 * page by calling the destructor.
1998 */
1999static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2000{
2001 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2002 return 0;
2003 return skb_copy_ubufs(skb, gfp_mask);
2004}
2005
2006/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002007 * __skb_queue_purge - empty a list
2008 * @list: list to empty
2009 *
2010 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2011 * the list and one reference dropped. This function does not take the
2012 * list lock and the caller must hold the relevant locks to use it.
2013 */
Joe Perches7965bd42013-09-26 14:48:15 -07002014void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002015static inline void __skb_queue_purge(struct sk_buff_head *list)
2016{
2017 struct sk_buff *skb;
2018 while ((skb = __skb_dequeue(list)) != NULL)
2019 kfree_skb(skb);
2020}
2021
Alexander Duycke5e67302013-02-08 10:17:15 +00002022#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
2023#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
2024#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
2025
Joe Perches7965bd42013-09-26 14:48:15 -07002026void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002027
Joe Perches7965bd42013-09-26 14:48:15 -07002028struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2029 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07002030
2031/**
2032 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2033 * @dev: network device to receive on
2034 * @length: length to allocate
2035 *
2036 * Allocate a new &sk_buff and assign it a usage count of one. The
2037 * buffer has unspecified headroom built in. Users should allocate
2038 * the headroom they think they need without accounting for the
2039 * built in space. The built in space is used for optimisations.
2040 *
2041 * %NULL is returned if there is no free memory. Although this function
2042 * allocates memory it can be called from an interrupt.
2043 */
2044static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002045 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002046{
2047 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2048}
2049
Eric Dumazet6f532612012-05-18 05:12:12 +00002050/* legacy helper around __netdev_alloc_skb() */
2051static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2052 gfp_t gfp_mask)
2053{
2054 return __netdev_alloc_skb(NULL, length, gfp_mask);
2055}
2056
2057/* legacy helper around netdev_alloc_skb() */
2058static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2059{
2060 return netdev_alloc_skb(NULL, length);
2061}
2062
2063
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002064static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2065 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002066{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002067 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002068
2069 if (NET_IP_ALIGN && skb)
2070 skb_reserve(skb, NET_IP_ALIGN);
2071 return skb;
2072}
2073
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002074static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2075 unsigned int length)
2076{
2077 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2078}
2079
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002080/**
2081 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
Mel Gorman06140022012-07-31 16:44:24 -07002082 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2083 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2084 * @order: size of the allocation
2085 *
2086 * Allocate a new page.
2087 *
2088 * %NULL is returned if there is no free memory.
2089*/
2090static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
2091 struct sk_buff *skb,
2092 unsigned int order)
2093{
2094 struct page *page;
2095
2096 gfp_mask |= __GFP_COLD;
2097
2098 if (!(gfp_mask & __GFP_NOMEMALLOC))
2099 gfp_mask |= __GFP_MEMALLOC;
2100
2101 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2102 if (skb && page && page->pfmemalloc)
2103 skb->pfmemalloc = true;
2104
2105 return page;
2106}
2107
2108/**
2109 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2110 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2111 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2112 *
2113 * Allocate a new page.
2114 *
2115 * %NULL is returned if there is no free memory.
2116 */
2117static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
2118 struct sk_buff *skb)
2119{
2120 return __skb_alloc_pages(gfp_mask, skb, 0);
2121}
2122
2123/**
2124 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2125 * @page: The page that was allocated from skb_alloc_page
2126 * @skb: The skb that may need pfmemalloc set
2127 */
2128static inline void skb_propagate_pfmemalloc(struct page *page,
2129 struct sk_buff *skb)
2130{
2131 if (page && page->pfmemalloc)
2132 skb->pfmemalloc = true;
2133}
2134
Eric Dumazet564824b2010-10-11 19:05:25 +00002135/**
Masanari Iidae2278672014-02-18 22:54:36 +09002136 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002137 * @frag: the paged fragment
2138 *
2139 * Returns the &struct page associated with @frag.
2140 */
2141static inline struct page *skb_frag_page(const skb_frag_t *frag)
2142{
Ian Campbella8605c62011-10-19 23:01:49 +00002143 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00002144}
2145
2146/**
2147 * __skb_frag_ref - take an addition reference on a paged fragment.
2148 * @frag: the paged fragment
2149 *
2150 * Takes an additional reference on the paged fragment @frag.
2151 */
2152static inline void __skb_frag_ref(skb_frag_t *frag)
2153{
2154 get_page(skb_frag_page(frag));
2155}
2156
2157/**
2158 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2159 * @skb: the buffer
2160 * @f: the fragment offset.
2161 *
2162 * Takes an additional reference on the @f'th paged fragment of @skb.
2163 */
2164static inline void skb_frag_ref(struct sk_buff *skb, int f)
2165{
2166 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2167}
2168
2169/**
2170 * __skb_frag_unref - release a reference on a paged fragment.
2171 * @frag: the paged fragment
2172 *
2173 * Releases a reference on the paged fragment @frag.
2174 */
2175static inline void __skb_frag_unref(skb_frag_t *frag)
2176{
2177 put_page(skb_frag_page(frag));
2178}
2179
2180/**
2181 * skb_frag_unref - release a reference on a paged fragment of an skb.
2182 * @skb: the buffer
2183 * @f: the fragment offset
2184 *
2185 * Releases a reference on the @f'th paged fragment of @skb.
2186 */
2187static inline void skb_frag_unref(struct sk_buff *skb, int f)
2188{
2189 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2190}
2191
2192/**
2193 * skb_frag_address - gets the address of the data contained in a paged fragment
2194 * @frag: the paged fragment buffer
2195 *
2196 * Returns the address of the data within @frag. The page must already
2197 * be mapped.
2198 */
2199static inline void *skb_frag_address(const skb_frag_t *frag)
2200{
2201 return page_address(skb_frag_page(frag)) + frag->page_offset;
2202}
2203
2204/**
2205 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2206 * @frag: the paged fragment buffer
2207 *
2208 * Returns the address of the data within @frag. Checks that the page
2209 * is mapped and returns %NULL otherwise.
2210 */
2211static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2212{
2213 void *ptr = page_address(skb_frag_page(frag));
2214 if (unlikely(!ptr))
2215 return NULL;
2216
2217 return ptr + frag->page_offset;
2218}
2219
2220/**
2221 * __skb_frag_set_page - sets the page contained in a paged fragment
2222 * @frag: the paged fragment
2223 * @page: the page to set
2224 *
2225 * Sets the fragment @frag to contain @page.
2226 */
2227static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2228{
Ian Campbella8605c62011-10-19 23:01:49 +00002229 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002230}
2231
2232/**
2233 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2234 * @skb: the buffer
2235 * @f: the fragment offset
2236 * @page: the page to set
2237 *
2238 * Sets the @f'th fragment of @skb to contain @page.
2239 */
2240static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2241 struct page *page)
2242{
2243 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2244}
2245
Eric Dumazet400dfd32013-10-17 16:27:07 -07002246bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2247
Ian Campbell131ea662011-08-19 06:25:00 +00002248/**
2249 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002250 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002251 * @frag: the paged fragment to map
2252 * @offset: the offset within the fragment (starting at the
2253 * fragment's own offset)
2254 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002255 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00002256 *
2257 * Maps the page associated with @frag to @device.
2258 */
2259static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2260 const skb_frag_t *frag,
2261 size_t offset, size_t size,
2262 enum dma_data_direction dir)
2263{
2264 return dma_map_page(dev, skb_frag_page(frag),
2265 frag->page_offset + offset, size, dir);
2266}
2267
Eric Dumazet117632e2011-12-03 21:39:53 +00002268static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2269 gfp_t gfp_mask)
2270{
2271 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2272}
2273
Octavian Purdilabad93e92014-06-12 01:36:26 +03002274
2275static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2276 gfp_t gfp_mask)
2277{
2278 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2279}
2280
2281
Ian Campbell131ea662011-08-19 06:25:00 +00002282/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002283 * skb_clone_writable - is the header of a clone writable
2284 * @skb: buffer to check
2285 * @len: length up to which to write
2286 *
2287 * Returns true if modifying the header part of the cloned buffer
2288 * does not requires the data to be copied.
2289 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002290static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002291{
2292 return !skb_header_cloned(skb) &&
2293 skb_headroom(skb) + len <= skb->hdr_len;
2294}
2295
Herbert Xud9cc2042007-09-16 16:21:16 -07002296static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2297 int cloned)
2298{
2299 int delta = 0;
2300
Herbert Xud9cc2042007-09-16 16:21:16 -07002301 if (headroom > skb_headroom(skb))
2302 delta = headroom - skb_headroom(skb);
2303
2304 if (delta || cloned)
2305 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2306 GFP_ATOMIC);
2307 return 0;
2308}
2309
Patrick McHardy334a8132007-06-25 04:35:20 -07002310/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002311 * skb_cow - copy header of skb when it is required
2312 * @skb: buffer to cow
2313 * @headroom: needed headroom
2314 *
2315 * If the skb passed lacks sufficient headroom or its data part
2316 * is shared, data is reallocated. If reallocation fails, an error
2317 * is returned and original skb is not changed.
2318 *
2319 * The result is skb with writable area skb->head...skb->tail
2320 * and at least @headroom of space at head.
2321 */
2322static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2323{
Herbert Xud9cc2042007-09-16 16:21:16 -07002324 return __skb_cow(skb, headroom, skb_cloned(skb));
2325}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002326
Herbert Xud9cc2042007-09-16 16:21:16 -07002327/**
2328 * skb_cow_head - skb_cow but only making the head writable
2329 * @skb: buffer to cow
2330 * @headroom: needed headroom
2331 *
2332 * This function is identical to skb_cow except that we replace the
2333 * skb_cloned check by skb_header_cloned. It should be used when
2334 * you only need to push on some header and do not need to modify
2335 * the data.
2336 */
2337static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2338{
2339 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002340}
2341
2342/**
2343 * skb_padto - pad an skbuff up to a minimal size
2344 * @skb: buffer to pad
2345 * @len: minimal length
2346 *
2347 * Pads up a buffer to ensure the trailing bytes exist and are
2348 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002349 * is untouched. Otherwise it is extended. Returns zero on
2350 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002351 */
2352
Herbert Xu5b057c62006-06-23 02:06:41 -07002353static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002354{
2355 unsigned int size = skb->len;
2356 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002357 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002358 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002359}
2360
2361static inline int skb_add_data(struct sk_buff *skb,
2362 char __user *from, int copy)
2363{
2364 const int off = skb->len;
2365
2366 if (skb->ip_summed == CHECKSUM_NONE) {
2367 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08002368 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
Linus Torvalds1da177e2005-04-16 15:20:36 -07002369 copy, 0, &err);
2370 if (!err) {
2371 skb->csum = csum_block_add(skb->csum, csum, off);
2372 return 0;
2373 }
2374 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2375 return 0;
2376
2377 __skb_trim(skb, off);
2378 return -EFAULT;
2379}
2380
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002381static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2382 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002383{
2384 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002385 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002386
Ian Campbellea2ab692011-08-22 23:44:58 +00002387 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002388 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002389 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002390 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002391}
2392
Herbert Xu364c6ba2006-06-09 16:10:40 -07002393static inline int __skb_linearize(struct sk_buff *skb)
2394{
2395 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2396}
2397
Linus Torvalds1da177e2005-04-16 15:20:36 -07002398/**
2399 * skb_linearize - convert paged skb to linear one
2400 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002401 *
2402 * If there is no free memory -ENOMEM is returned, otherwise zero
2403 * is returned and the old skb data released.
2404 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002405static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002406{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002407 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2408}
2409
2410/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00002411 * skb_has_shared_frag - can any frag be overwritten
2412 * @skb: buffer to test
2413 *
2414 * Return true if the skb has at least one frag that might be modified
2415 * by an external entity (as in vmsplice()/sendfile())
2416 */
2417static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2418{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00002419 return skb_is_nonlinear(skb) &&
2420 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00002421}
2422
2423/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07002424 * skb_linearize_cow - make sure skb is linear and writable
2425 * @skb: buffer to process
2426 *
2427 * If there is no free memory -ENOMEM is returned, otherwise zero
2428 * is returned and the old skb data released.
2429 */
2430static inline int skb_linearize_cow(struct sk_buff *skb)
2431{
2432 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2433 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002434}
2435
2436/**
2437 * skb_postpull_rcsum - update checksum for received skb after pull
2438 * @skb: buffer to update
2439 * @start: start of data before pull
2440 * @len: length of data pulled
2441 *
2442 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002443 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2444 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002445 */
2446
2447static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002448 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002449{
Patrick McHardy84fa7932006-08-29 16:44:56 -07002450 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002451 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2452}
2453
Herbert Xucbb042f2006-03-20 22:43:56 -08002454unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2455
David S. Miller7ce5a272013-12-02 17:26:05 -05002456/**
2457 * pskb_trim_rcsum - trim received skb and update checksum
2458 * @skb: buffer to trim
2459 * @len: new length
2460 *
2461 * This is exactly the same as pskb_trim except that it ensures the
2462 * checksum of received packets are still valid after the operation.
2463 */
2464
2465static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2466{
2467 if (likely(len >= skb->len))
2468 return 0;
2469 if (skb->ip_summed == CHECKSUM_COMPLETE)
2470 skb->ip_summed = CHECKSUM_NONE;
2471 return __pskb_trim(skb, len);
2472}
2473
Linus Torvalds1da177e2005-04-16 15:20:36 -07002474#define skb_queue_walk(queue, skb) \
2475 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002476 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002477 skb = skb->next)
2478
James Chapman46f89142007-04-30 00:07:31 -07002479#define skb_queue_walk_safe(queue, skb, tmp) \
2480 for (skb = (queue)->next, tmp = skb->next; \
2481 skb != (struct sk_buff *)(queue); \
2482 skb = tmp, tmp = skb->next)
2483
David S. Miller1164f522008-09-23 00:49:44 -07002484#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002485 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002486 skb = skb->next)
2487
2488#define skb_queue_walk_from_safe(queue, skb, tmp) \
2489 for (tmp = skb->next; \
2490 skb != (struct sk_buff *)(queue); \
2491 skb = tmp, tmp = skb->next)
2492
Stephen Hemminger300ce172005-10-30 13:47:34 -08002493#define skb_queue_reverse_walk(queue, skb) \
2494 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002495 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08002496 skb = skb->prev)
2497
David S. Miller686a2952011-01-20 22:47:32 -08002498#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2499 for (skb = (queue)->prev, tmp = skb->prev; \
2500 skb != (struct sk_buff *)(queue); \
2501 skb = tmp, tmp = skb->prev)
2502
2503#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2504 for (tmp = skb->prev; \
2505 skb != (struct sk_buff *)(queue); \
2506 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002507
David S. Miller21dc3302010-08-23 00:13:46 -07002508static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07002509{
2510 return skb_shinfo(skb)->frag_list != NULL;
2511}
2512
2513static inline void skb_frag_list_init(struct sk_buff *skb)
2514{
2515 skb_shinfo(skb)->frag_list = NULL;
2516}
2517
2518static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2519{
2520 frag->next = skb_shinfo(skb)->frag_list;
2521 skb_shinfo(skb)->frag_list = frag;
2522}
2523
2524#define skb_walk_frags(skb, iter) \
2525 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2526
Joe Perches7965bd42013-09-26 14:48:15 -07002527struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2528 int *peeked, int *off, int *err);
2529struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2530 int *err);
2531unsigned int datagram_poll(struct file *file, struct socket *sock,
2532 struct poll_table_struct *wait);
2533int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2534 struct iovec *to, int size);
2535int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2536 struct iovec *iov);
2537int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2538 const struct iovec *from, int from_offset,
2539 int len);
2540int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2541 int offset, size_t count);
2542int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2543 const struct iovec *to, int to_offset,
2544 int size);
2545void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2546void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2547int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07002548int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2549int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2550__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2551 int len, __wsum csum);
2552int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2553 struct pipe_inode_info *pipe, unsigned int len,
2554 unsigned int flags);
2555void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01002556unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00002557int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2558 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07002559void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2560int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2561void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Florian Westphalde960aa2014-01-26 10:58:16 +01002562unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -07002563struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Vlad Yasevich0d5501c2014-08-08 14:42:13 -04002564struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002565
Daniel Borkmann2817a332013-10-30 11:50:51 +01002566struct skb_checksum_ops {
2567 __wsum (*update)(const void *mem, int len, __wsum wsum);
2568 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2569};
2570
2571__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2572 __wsum csum, const struct skb_checksum_ops *ops);
2573__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2574 __wsum csum);
2575
David S. Miller690e36e2014-08-23 12:13:41 -07002576static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset,
2577 int len, void *data, int hlen, void *buffer)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002578{
Patrick McHardy55820ee2005-07-05 14:08:10 -07002579 if (hlen - offset >= len)
David S. Miller690e36e2014-08-23 12:13:41 -07002580 return data + offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002581
David S. Miller690e36e2014-08-23 12:13:41 -07002582 if (!skb ||
2583 skb_copy_bits(skb, offset, buffer, len) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002584 return NULL;
2585
2586 return buffer;
2587}
2588
David S. Miller690e36e2014-08-23 12:13:41 -07002589static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2590 int len, void *buffer)
2591{
2592 return __skb_header_pointer(skb, offset, len, skb->data,
2593 skb_headlen(skb), buffer);
2594}
2595
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01002596/**
2597 * skb_needs_linearize - check if we need to linearize a given skb
2598 * depending on the given device features.
2599 * @skb: socket buffer to check
2600 * @features: net device features
2601 *
2602 * Returns true if either:
2603 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2604 * 2. skb is fragmented and the device does not support SG.
2605 */
2606static inline bool skb_needs_linearize(struct sk_buff *skb,
2607 netdev_features_t features)
2608{
2609 return skb_is_nonlinear(skb) &&
2610 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2611 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2612}
2613
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03002614static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2615 void *to,
2616 const unsigned int len)
2617{
2618 memcpy(to, skb->data, len);
2619}
2620
2621static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2622 const int offset, void *to,
2623 const unsigned int len)
2624{
2625 memcpy(to, skb->data + offset, len);
2626}
2627
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03002628static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2629 const void *from,
2630 const unsigned int len)
2631{
2632 memcpy(skb->data, from, len);
2633}
2634
2635static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2636 const int offset,
2637 const void *from,
2638 const unsigned int len)
2639{
2640 memcpy(skb->data + offset, from, len);
2641}
2642
Joe Perches7965bd42013-09-26 14:48:15 -07002643void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002644
Patrick Ohlyac45f602009-02-12 05:03:37 +00002645static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2646{
2647 return skb->tstamp;
2648}
2649
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002650/**
2651 * skb_get_timestamp - get timestamp from a skb
2652 * @skb: skb to get stamp from
2653 * @stamp: pointer to struct timeval to store stamp in
2654 *
2655 * Timestamps are stored in the skb as offsets to a base timestamp.
2656 * This function converts the offset back to a struct timeval and stores
2657 * it in stamp.
2658 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00002659static inline void skb_get_timestamp(const struct sk_buff *skb,
2660 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002661{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002662 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002663}
2664
Patrick Ohlyac45f602009-02-12 05:03:37 +00002665static inline void skb_get_timestampns(const struct sk_buff *skb,
2666 struct timespec *stamp)
2667{
2668 *stamp = ktime_to_timespec(skb->tstamp);
2669}
2670
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002671static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002672{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002673 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002674}
2675
Stephen Hemminger164891a2007-04-23 22:26:16 -07002676static inline ktime_t net_timedelta(ktime_t t)
2677{
2678 return ktime_sub(ktime_get_real(), t);
2679}
2680
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07002681static inline ktime_t net_invalid_timestamp(void)
2682{
2683 return ktime_set(0, 0);
2684}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002685
Richard Cochranc1f19b52010-07-17 08:49:36 +00002686#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2687
Joe Perches7965bd42013-09-26 14:48:15 -07002688void skb_clone_tx_timestamp(struct sk_buff *skb);
2689bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00002690
2691#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2692
2693static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2694{
2695}
2696
2697static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2698{
2699 return false;
2700}
2701
2702#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2703
2704/**
2705 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2706 *
Richard Cochranda92b192011-10-21 00:49:15 +00002707 * PHY drivers may accept clones of transmitted packets for
2708 * timestamping via their phy_driver.txtstamp method. These drivers
2709 * must call this function to return the skb back to the stack, with
2710 * or without a timestamp.
2711 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00002712 * @skb: clone of the the original outgoing packet
Richard Cochranda92b192011-10-21 00:49:15 +00002713 * @hwtstamps: hardware time stamps, may be NULL if not available
Richard Cochranc1f19b52010-07-17 08:49:36 +00002714 *
2715 */
2716void skb_complete_tx_timestamp(struct sk_buff *skb,
2717 struct skb_shared_hwtstamps *hwtstamps);
2718
Willem de Bruijne7fd2882014-08-04 22:11:48 -04002719void __skb_tstamp_tx(struct sk_buff *orig_skb,
2720 struct skb_shared_hwtstamps *hwtstamps,
2721 struct sock *sk, int tstype);
2722
Patrick Ohlyac45f602009-02-12 05:03:37 +00002723/**
2724 * skb_tstamp_tx - queue clone of skb with send time stamps
2725 * @orig_skb: the original outgoing packet
2726 * @hwtstamps: hardware time stamps, may be NULL if not available
2727 *
2728 * If the skb has a socket associated, then this function clones the
2729 * skb (thus sharing the actual data and optional structures), stores
2730 * the optional hardware time stamping information (if non NULL) or
2731 * generates a software time stamp (otherwise), then queues the clone
2732 * to the error queue of the socket. Errors are silently ignored.
2733 */
Joe Perches7965bd42013-09-26 14:48:15 -07002734void skb_tstamp_tx(struct sk_buff *orig_skb,
2735 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00002736
Richard Cochran4507a712010-07-17 08:48:28 +00002737static inline void sw_tx_timestamp(struct sk_buff *skb)
2738{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002739 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2740 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00002741 skb_tstamp_tx(skb, NULL);
2742}
2743
2744/**
2745 * skb_tx_timestamp() - Driver hook for transmit timestamping
2746 *
2747 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00002748 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00002749 *
David S. Miller73409f32013-12-27 13:04:33 -05002750 * Specifically, one should make absolutely sure that this function is
2751 * called before TX completion of this packet can trigger. Otherwise
2752 * the packet could potentially already be freed.
2753 *
Richard Cochran4507a712010-07-17 08:48:28 +00002754 * @skb: A socket buffer.
2755 */
2756static inline void skb_tx_timestamp(struct sk_buff *skb)
2757{
Richard Cochranc1f19b52010-07-17 08:49:36 +00002758 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002759 sw_tx_timestamp(skb);
2760}
2761
Johannes Berg6e3e9392011-11-09 10:15:42 +01002762/**
2763 * skb_complete_wifi_ack - deliver skb with wifi status
2764 *
2765 * @skb: the original outgoing packet
2766 * @acked: ack status
2767 *
2768 */
2769void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2770
Joe Perches7965bd42013-09-26 14:48:15 -07002771__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2772__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002773
Herbert Xu60476372007-04-09 11:59:39 -07002774static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2775{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002776 return ((skb->ip_summed & CHECKSUM_UNNECESSARY) || skb->csum_valid);
Herbert Xu60476372007-04-09 11:59:39 -07002777}
2778
Herbert Xufb286bb2005-11-10 13:01:24 -08002779/**
2780 * skb_checksum_complete - Calculate checksum of an entire packet
2781 * @skb: packet to process
2782 *
2783 * This function calculates the checksum over the entire packet plus
2784 * the value of skb->csum. The latter can be used to supply the
2785 * checksum of a pseudo header as used by TCP/UDP. It returns the
2786 * checksum.
2787 *
2788 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2789 * this function can be used to verify that checksum on received
2790 * packets. In that case the function should return zero if the
2791 * checksum is correct. In particular, this function will return zero
2792 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2793 * hardware has already verified the correctness of the checksum.
2794 */
Al Viro4381ca32007-07-15 21:00:11 +01002795static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002796{
Herbert Xu60476372007-04-09 11:59:39 -07002797 return skb_csum_unnecessary(skb) ?
2798 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002799}
2800
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002801/* Check if we need to perform checksum complete validation.
2802 *
2803 * Returns true if checksum complete is needed, false otherwise
2804 * (either checksum is unnecessary or zero checksum is allowed).
2805 */
2806static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
2807 bool zero_okay,
2808 __sum16 check)
2809{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002810 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
2811 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002812 return false;
2813 }
2814
2815 return true;
2816}
2817
2818/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
2819 * in checksum_init.
2820 */
2821#define CHECKSUM_BREAK 76
2822
2823/* Validate (init) checksum based on checksum complete.
2824 *
2825 * Return values:
2826 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
2827 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
2828 * checksum is stored in skb->csum for use in __skb_checksum_complete
2829 * non-zero: value of invalid checksum
2830 *
2831 */
2832static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
2833 bool complete,
2834 __wsum psum)
2835{
2836 if (skb->ip_summed == CHECKSUM_COMPLETE) {
2837 if (!csum_fold(csum_add(psum, skb->csum))) {
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002838 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002839 return 0;
2840 }
2841 }
2842
2843 skb->csum = psum;
2844
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002845 if (complete || skb->len <= CHECKSUM_BREAK) {
2846 __sum16 csum;
2847
2848 csum = __skb_checksum_complete(skb);
2849 skb->csum_valid = !csum;
2850 return csum;
2851 }
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002852
2853 return 0;
2854}
2855
2856static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
2857{
2858 return 0;
2859}
2860
2861/* Perform checksum validate (init). Note that this is a macro since we only
2862 * want to calculate the pseudo header which is an input function if necessary.
2863 * First we try to validate without any computation (checksum unnecessary) and
2864 * then calculate based on checksum complete calling the function to compute
2865 * pseudo header.
2866 *
2867 * Return values:
2868 * 0: checksum is validated or try to in skb_checksum_complete
2869 * non-zero: value of invalid checksum
2870 */
2871#define __skb_checksum_validate(skb, proto, complete, \
2872 zero_okay, check, compute_pseudo) \
2873({ \
2874 __sum16 __ret = 0; \
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002875 skb->csum_valid = 0; \
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002876 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
2877 __ret = __skb_checksum_validate_complete(skb, \
2878 complete, compute_pseudo(skb, proto)); \
2879 __ret; \
2880})
2881
2882#define skb_checksum_init(skb, proto, compute_pseudo) \
2883 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
2884
2885#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
2886 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
2887
2888#define skb_checksum_validate(skb, proto, compute_pseudo) \
2889 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
2890
2891#define skb_checksum_validate_zero_check(skb, proto, check, \
2892 compute_pseudo) \
2893 __skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
2894
2895#define skb_checksum_simple_validate(skb) \
2896 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
2897
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002898#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07002899void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002900static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2901{
2902 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002903 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002904}
2905static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2906{
2907 if (nfct)
2908 atomic_inc(&nfct->use);
2909}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002910#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002911#ifdef CONFIG_BRIDGE_NETFILTER
2912static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2913{
2914 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2915 kfree(nf_bridge);
2916}
2917static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2918{
2919 if (nf_bridge)
2920 atomic_inc(&nf_bridge->use);
2921}
2922#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08002923static inline void nf_reset(struct sk_buff *skb)
2924{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002925#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08002926 nf_conntrack_put(skb->nfct);
2927 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002928#endif
Patrick McHardya193a4a2006-03-20 19:23:05 -08002929#ifdef CONFIG_BRIDGE_NETFILTER
2930 nf_bridge_put(skb->nf_bridge);
2931 skb->nf_bridge = NULL;
2932#endif
2933}
2934
Patrick McHardy124dff02013-04-05 20:42:05 +02002935static inline void nf_reset_trace(struct sk_buff *skb)
2936{
Florian Westphal478b3602014-02-15 23:48:45 +01002937#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00002938 skb->nf_trace = 0;
2939#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002940}
2941
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002942/* Note: This doesn't put any conntrack and bridge info in dst. */
2943static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2944{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002945#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002946 dst->nfct = src->nfct;
2947 nf_conntrack_get(src->nfct);
2948 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002949#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002950#ifdef CONFIG_BRIDGE_NETFILTER
2951 dst->nf_bridge = src->nf_bridge;
2952 nf_bridge_get(src->nf_bridge);
2953#endif
Florian Westphal478b3602014-02-15 23:48:45 +01002954#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
2955 dst->nf_trace = src->nf_trace;
2956#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002957}
2958
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002959static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2960{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002961#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002962 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002963#endif
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002964#ifdef CONFIG_BRIDGE_NETFILTER
2965 nf_bridge_put(dst->nf_bridge);
2966#endif
2967 __nf_copy(dst, src);
2968}
2969
James Morris984bc162006-06-09 00:29:17 -07002970#ifdef CONFIG_NETWORK_SECMARK
2971static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2972{
2973 to->secmark = from->secmark;
2974}
2975
2976static inline void skb_init_secmark(struct sk_buff *skb)
2977{
2978 skb->secmark = 0;
2979}
2980#else
2981static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2982{ }
2983
2984static inline void skb_init_secmark(struct sk_buff *skb)
2985{ }
2986#endif
2987
Eric W. Biederman574f7192014-04-01 12:20:24 -07002988static inline bool skb_irq_freeable(const struct sk_buff *skb)
2989{
2990 return !skb->destructor &&
2991#if IS_ENABLED(CONFIG_XFRM)
2992 !skb->sp &&
2993#endif
2994#if IS_ENABLED(CONFIG_NF_CONNTRACK)
2995 !skb->nfct &&
2996#endif
2997 !skb->_skb_refdst &&
2998 !skb_has_frag_list(skb);
2999}
3000
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003001static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3002{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003003 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003004}
3005
Stephen Hemminger92477442009-03-21 13:39:26 -07003006static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003007{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003008 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003009}
3010
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003011static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3012{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003013 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003014}
3015
David S. Millerd5a9e242009-01-27 16:22:11 -08003016static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3017{
3018 skb->queue_mapping = rx_queue + 1;
3019}
3020
Stephen Hemminger92477442009-03-21 13:39:26 -07003021static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003022{
3023 return skb->queue_mapping - 1;
3024}
3025
Stephen Hemminger92477442009-03-21 13:39:26 -07003026static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003027{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003028 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08003029}
3030
Tom Herbert0e001612014-07-01 21:32:27 -07003031u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
Joe Perches7965bd42013-09-26 14:48:15 -07003032 unsigned int num_tx_queues);
Stephen Hemminger92477442009-03-21 13:39:26 -07003033
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003034static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3035{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003036#ifdef CONFIG_XFRM
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003037 return skb->sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003038#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003039 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003040#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003041}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003042
Pravin B Shelar68c33162013-02-14 14:02:41 +00003043/* Keeps track of mac header offset relative to skb->head.
3044 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3045 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07003046 * tunnel skb it points to outer mac header.
3047 * Keeps track of level of encapsulation of network headers.
3048 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00003049struct skb_gso_cb {
Eric Dumazet3347c962013-10-19 11:42:56 -07003050 int mac_offset;
3051 int encap_level;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003052 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00003053};
3054#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3055
3056static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3057{
3058 return (skb_mac_header(inner_skb) - inner_skb->head) -
3059 SKB_GSO_CB(inner_skb)->mac_offset;
3060}
3061
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00003062static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3063{
3064 int new_headroom, headroom;
3065 int ret;
3066
3067 headroom = skb_headroom(skb);
3068 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3069 if (ret)
3070 return ret;
3071
3072 new_headroom = skb_headroom(skb);
3073 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3074 return 0;
3075}
3076
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003077/* Compute the checksum for a gso segment. First compute the checksum value
3078 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3079 * then add in skb->csum (checksum from csum_start to end of packet).
3080 * skb->csum and csum_start are then updated to reflect the checksum of the
3081 * resultant packet starting from the transport header-- the resultant checksum
3082 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3083 * header.
3084 */
3085static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3086{
3087 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3088 skb_transport_offset(skb);
3089 __u16 csum;
3090
3091 csum = csum_fold(csum_partial(skb_transport_header(skb),
3092 plen, skb->csum));
3093 skb->csum = res;
3094 SKB_GSO_CB(skb)->csum_start -= plen;
3095
3096 return csum;
3097}
3098
David S. Millerbdcc0922012-03-07 20:53:36 -05003099static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07003100{
3101 return skb_shinfo(skb)->gso_size;
3102}
3103
Eric Dumazet36a8f392013-09-29 01:21:32 -07003104/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05003105static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02003106{
3107 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3108}
3109
Joe Perches7965bd42013-09-26 14:48:15 -07003110void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07003111
3112static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3113{
3114 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3115 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003116 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3117
Alexander Duyckb78462e2010-06-02 12:24:37 +00003118 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3119 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07003120 __skb_warn_lro_forwarding(skb);
3121 return true;
3122 }
3123 return false;
3124}
3125
Herbert Xu35fc92a2007-03-26 23:22:20 -07003126static inline void skb_forward_csum(struct sk_buff *skb)
3127{
3128 /* Unfortunately we don't support this one. Any brave souls? */
3129 if (skb->ip_summed == CHECKSUM_COMPLETE)
3130 skb->ip_summed = CHECKSUM_NONE;
3131}
3132
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003133/**
3134 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3135 * @skb: skb to check
3136 *
3137 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3138 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3139 * use this helper, to document places where we make this assertion.
3140 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003141static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003142{
3143#ifdef DEBUG
3144 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3145#endif
3146}
3147
Rusty Russellf35d9d82008-02-04 23:49:54 -05003148bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00003149
Paul Durranted1f50c2014-01-09 10:02:46 +00003150int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3151
Daniel Borkmannf77668d2013-03-19 06:39:30 +00003152u32 __skb_get_poff(const struct sk_buff *skb);
3153
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00003154/**
3155 * skb_head_is_locked - Determine if the skb->head is locked down
3156 * @skb: skb to check
3157 *
3158 * The head on skbs build around a head frag can be removed if they are
3159 * not cloned. This function returns true if the skb head is locked down
3160 * due to either being allocated via kmalloc, or by being a clone with
3161 * multiple references to the head.
3162 */
3163static inline bool skb_head_is_locked(const struct sk_buff *skb)
3164{
3165 return !skb->head_frag || skb_cloned(skb);
3166}
Florian Westphalfe6cc552014-02-13 23:09:12 +01003167
3168/**
3169 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3170 *
3171 * @skb: GSO skb
3172 *
3173 * skb_gso_network_seglen is used to determine the real size of the
3174 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3175 *
3176 * The MAC/L2 header is not accounted for.
3177 */
3178static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3179{
3180 unsigned int hdr_len = skb_transport_header(skb) -
3181 skb_network_header(skb);
3182 return hdr_len + skb_gso_transport_seglen(skb);
3183}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003184#endif /* __KERNEL__ */
3185#endif /* _LINUX_SKBUFF_H */