blob: 18ddf9684a277501b0e5df12e62955380ef29cbd [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
Randy Dunlap553a5672008-04-20 10:51:01 -0700455 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700456 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Tom Herbert61b905d2014-03-24 15:34:47 -0700457 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
Changli Gao4ca24622011-08-19 07:26:44 -0700458 * ports.
Tom Herberta3b18dd2014-07-01 21:33:17 -0700459 * @sw_hash: indicates hash was computed in software stack
Johannes Berg6e3e9392011-11-09 10:15:42 +0100460 * @wifi_acked_valid: wifi_acked was set
461 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000462 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Randy Dunlapf4b8ea72006-06-22 16:00:11 -0700463 * @dma_cookie: a cookie to one of several possible DMA operations
464 * done by skb DMA functions
Eliezer Tamir06021292013-06-10 11:39:50 +0300465 * @napi_id: id of the NAPI struct this skb came from
James Morris984bc162006-06-09 00:29:17 -0700466 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700467 * @mark: Generic packet mark
468 * @dropcount: total number of sk_receive_queue overflows
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000469 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b02008-07-14 22:49:06 -0700470 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000471 * @inner_protocol: Protocol (encapsulation)
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000472 * @inner_transport_header: Inner transport layer header (encapsulation)
473 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000474 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700475 * @transport_header: Transport layer header
476 * @network_header: Network layer header
477 * @mac_header: Link layer header
478 * @tail: Tail pointer
479 * @end: End pointer
480 * @head: Head of buffer
481 * @data: Data head pointer
482 * @truesize: Buffer size
483 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484 */
485
486struct sk_buff {
487 /* These two members must be first. */
488 struct sk_buff *next;
489 struct sk_buff *prev;
490
Eric Dumazet363ec39232014-02-26 14:02:11 -0800491 union {
492 ktime_t tstamp;
493 struct skb_mstamp skb_mstamp;
494 };
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000495
496 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499 /*
500 * This is the control buffer. It is free to use for every
501 * layer. Please put your private variables there. If you
502 * want to keep them across layers you have to do a skb_clone()
503 * first. This is owned by whoever has the skb queued ATM.
504 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000505 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700506
Eric Dumazet7fee2262010-05-11 23:19:48 +0000507 unsigned long _skb_refdst;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000508#ifdef CONFIG_XFRM
509 struct sec_path *sp;
510#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700512 data_len;
513 __u16 mac_len,
514 hdr_len;
Al Viroff1dcad2006-11-20 18:07:29 -0800515 union {
516 __wsum csum;
Herbert Xu663ead32007-04-09 11:59:07 -0700517 struct {
518 __u16 csum_start;
519 __u16 csum_offset;
520 };
Al Viroff1dcad2006-11-20 18:07:29 -0800521 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522 __u32 priority;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200523 kmemcheck_bitfield_begin(flags1);
WANG Cong60ff7462014-05-04 16:39:18 -0700524 __u8 ignore_df:1,
Thomas Graf1cbb3382005-07-05 14:13:41 -0700525 cloned:1,
526 ip_summed:2,
Harald Welte6869c4d2005-08-09 19:24:19 -0700527 nohdr:1,
528 nfctinfo:3;
David S. Millerd179cd12005-08-17 14:57:30 -0700529 __u8 pkt_type:3,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800530 fclone:2,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700531 ipvs_property:1,
Herbert Xua59322b2007-12-05 01:53:40 -0800532 peeked:1,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700533 nf_trace:1;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200534 kmemcheck_bitfield_end(flags1);
Eric Dumazet4ab408d2010-03-01 03:09:26 +0000535 __be16 protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536
537 void (*destructor)(struct sk_buff *skb);
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800538#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700539 struct nf_conntrack *nfct;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100540#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700541#ifdef CONFIG_BRIDGE_NETFILTER
542 struct nf_bridge_info *nf_bridge;
543#endif
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700544
Eric Dumazet8964be42009-11-20 15:35:04 -0800545 int skb_iif;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000546
Tom Herbert61b905d2014-03-24 15:34:47 -0700547 __u32 hash;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000548
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000549 __be16 vlan_proto;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000550 __u16 vlan_tci;
551
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700553 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700555 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200558
Eric Dumazet0a148422011-04-20 09:27:32 +0000559 __u16 queue_mapping;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200560 kmemcheck_bitfield_begin(flags2);
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400561#ifdef CONFIG_IPV6_NDISC_NODETYPE
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000562 __u8 ndisc_nodetype:2;
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400563#endif
Mel Gormanc93bdd02012-07-31 16:44:19 -0700564 __u8 pfmemalloc:1;
Tom Herbert3853b582010-11-21 13:17:29 +0000565 __u8 ooo_okay:1;
Tom Herbert61b905d2014-03-24 15:34:47 -0700566 __u8 l4_hash:1;
Tom Herberta3b18dd2014-07-01 21:33:17 -0700567 __u8 sw_hash:1;
Johannes Berg6e3e9392011-11-09 10:15:42 +0100568 __u8 wifi_acked_valid:1;
569 __u8 wifi_acked:1;
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000570 __u8 no_fcs:1;
Eric Dumazetd3836f22012-04-27 00:33:38 +0000571 __u8 head_frag:1;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000572 /* Encapsulation protocol and NIC drivers should use
573 * this flag to indicate to each other if the skb contains
574 * encapsulated packet or not and maybe use the inner packet
575 * headers if needed
576 */
577 __u8 encapsulation:1;
Tom Herbert7e2b10c2014-06-04 17:20:02 -0700578 __u8 encap_hdr_csum:1;
Tom Herbert5d0c2b92014-06-10 18:54:13 -0700579 __u8 csum_valid:1;
Tom Herbert7e3cead2014-06-10 18:54:19 -0700580 __u8 csum_complete_sw:1;
Tom Herberta3b18dd2014-07-01 21:33:17 -0700581 /* 2/4 bit hole (depending on ndisc_nodetype presence) */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200582 kmemcheck_bitfield_end(flags2);
583
Cong Wange0d10952013-08-01 11:10:25 +0800584#if defined CONFIG_NET_DMA || defined CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir06021292013-06-10 11:39:50 +0300585 union {
586 unsigned int napi_id;
587 dma_cookie_t dma_cookie;
588 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700589#endif
James Morris984bc162006-06-09 00:29:17 -0700590#ifdef CONFIG_NETWORK_SECMARK
591 __u32 secmark;
592#endif
Neil Horman3b885782009-10-12 13:26:31 -0700593 union {
594 __u32 mark;
595 __u32 dropcount;
Eric Dumazet16fad692013-03-14 05:40:32 +0000596 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700597 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700598
Simon Horman0d89d202013-05-23 21:02:52 +0000599 __be16 inner_protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000600 __u16 inner_transport_header;
601 __u16 inner_network_header;
602 __u16 inner_mac_header;
603 __u16 transport_header;
604 __u16 network_header;
605 __u16 mac_header;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700607 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700608 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700610 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700611 unsigned int truesize;
612 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700613};
614
615#ifdef __KERNEL__
616/*
617 * Handling routines are only of interest to the kernel
618 */
619#include <linux/slab.h>
620
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621
Mel Gormanc93bdd02012-07-31 16:44:19 -0700622#define SKB_ALLOC_FCLONE 0x01
623#define SKB_ALLOC_RX 0x02
624
625/* Returns true if the skb was allocated from PFMEMALLOC reserves */
626static inline bool skb_pfmemalloc(const struct sk_buff *skb)
627{
628 return unlikely(skb->pfmemalloc);
629}
630
Eric Dumazet7fee2262010-05-11 23:19:48 +0000631/*
632 * skb might have a dst pointer attached, refcounted or not.
633 * _skb_refdst low order bit is set if refcount was _not_ taken
634 */
635#define SKB_DST_NOREF 1UL
636#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
637
638/**
639 * skb_dst - returns skb dst_entry
640 * @skb: buffer
641 *
642 * Returns skb dst_entry, regardless of reference taken or not.
643 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000644static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
645{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000646 /* If refdst was not refcounted, check we still are in a
647 * rcu_read_lock section
648 */
649 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
650 !rcu_read_lock_held() &&
651 !rcu_read_lock_bh_held());
652 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000653}
654
Eric Dumazet7fee2262010-05-11 23:19:48 +0000655/**
656 * skb_dst_set - sets skb dst
657 * @skb: buffer
658 * @dst: dst entry
659 *
660 * Sets skb dst, assuming a reference was taken on dst and should
661 * be released by skb_dst_drop()
662 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000663static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
664{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000665 skb->_skb_refdst = (unsigned long)dst;
666}
667
Joe Perches7965bd42013-09-26 14:48:15 -0700668void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
669 bool force);
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200670
671/**
672 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
673 * @skb: buffer
674 * @dst: dst entry
675 *
676 * Sets skb dst, assuming a reference was not taken on dst.
677 * If dst entry is cached, we do not take reference and dst_release
678 * will be avoided by refdst_drop. If dst entry is not cached, we take
679 * reference, so that last dst_release can destroy the dst immediately.
680 */
681static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
682{
683 __skb_dst_set_noref(skb, dst, false);
684}
685
686/**
687 * skb_dst_set_noref_force - sets skb dst, without taking reference
688 * @skb: buffer
689 * @dst: dst entry
690 *
691 * Sets skb dst, assuming a reference was not taken on dst.
692 * No reference is taken and no dst_release will be called. While for
693 * cached dsts deferred reclaim is a basic feature, for entries that are
694 * not cached it is caller's job to guarantee that last dst_release for
695 * provided dst happens when nobody uses it, eg. after a RCU grace period.
696 */
697static inline void skb_dst_set_noref_force(struct sk_buff *skb,
698 struct dst_entry *dst)
699{
700 __skb_dst_set_noref(skb, dst, true);
701}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000702
703/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300704 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000705 * @skb: buffer
706 */
707static inline bool skb_dst_is_noref(const struct sk_buff *skb)
708{
709 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000710}
711
Eric Dumazet511c3f92009-06-02 05:14:27 +0000712static inline struct rtable *skb_rtable(const struct sk_buff *skb)
713{
Eric Dumazetadf30902009-06-02 05:19:30 +0000714 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000715}
716
Joe Perches7965bd42013-09-26 14:48:15 -0700717void kfree_skb(struct sk_buff *skb);
718void kfree_skb_list(struct sk_buff *segs);
719void skb_tx_error(struct sk_buff *skb);
720void consume_skb(struct sk_buff *skb);
721void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000722extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000723
Joe Perches7965bd42013-09-26 14:48:15 -0700724void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
725bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
726 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000727
Joe Perches7965bd42013-09-26 14:48:15 -0700728struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
729 int node);
730struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700731static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100732 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700733{
Eric Dumazet564824b2010-10-11 19:05:25 +0000734 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700735}
736
737static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100738 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700739{
Mel Gormanc93bdd02012-07-31 16:44:19 -0700740 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700741}
742
Joe Perches7965bd42013-09-26 14:48:15 -0700743struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
Patrick McHardy0ebd0ac2013-04-17 06:46:58 +0000744static inline struct sk_buff *alloc_skb_head(gfp_t priority)
745{
746 return __alloc_skb_head(priority, -1);
747}
748
Joe Perches7965bd42013-09-26 14:48:15 -0700749struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
750int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
751struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
752struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
Octavian Purdilabad93e92014-06-12 01:36:26 +0300753struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
754 gfp_t gfp_mask, bool fclone);
755static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
756 gfp_t gfp_mask)
757{
758 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
759}
Eric Dumazet117632e2011-12-03 21:39:53 +0000760
Joe Perches7965bd42013-09-26 14:48:15 -0700761int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
762struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
763 unsigned int headroom);
764struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
765 int newtailroom, gfp_t priority);
Fan Du25a91d82014-01-18 09:54:23 +0800766int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
767 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -0700768int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
769 int len);
770int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
771int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000772#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773
Joe Perches7965bd42013-09-26 14:48:15 -0700774int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
775 int getfrag(void *from, char *to, int offset,
776 int len, int odd, struct sk_buff *skb),
777 void *from, int length);
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700778
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800779struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700780 __u32 lower_offset;
781 __u32 upper_offset;
782 __u32 frag_idx;
783 __u32 stepped_offset;
784 struct sk_buff *root_skb;
785 struct sk_buff *cur_skb;
786 __u8 *frag_data;
787};
788
Joe Perches7965bd42013-09-26 14:48:15 -0700789void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
790 unsigned int to, struct skb_seq_state *st);
791unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
792 struct skb_seq_state *st);
793void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -0700794
Joe Perches7965bd42013-09-26 14:48:15 -0700795unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
796 unsigned int to, struct ts_config *config,
797 struct ts_state *state);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700798
Tom Herbert09323cc2013-12-15 22:16:19 -0800799/*
800 * Packet hash types specify the type of hash in skb_set_hash.
801 *
802 * Hash types refer to the protocol layer addresses which are used to
803 * construct a packet's hash. The hashes are used to differentiate or identify
804 * flows of the protocol layer for the hash type. Hash types are either
805 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
806 *
807 * Properties of hashes:
808 *
809 * 1) Two packets in different flows have different hash values
810 * 2) Two packets in the same flow should have the same hash value
811 *
812 * A hash at a higher layer is considered to be more specific. A driver should
813 * set the most specific hash possible.
814 *
815 * A driver cannot indicate a more specific hash than the layer at which a hash
816 * was computed. For instance an L3 hash cannot be set as an L4 hash.
817 *
818 * A driver may indicate a hash level which is less specific than the
819 * actual layer the hash was computed on. For instance, a hash computed
820 * at L4 may be considered an L3 hash. This should only be done if the
821 * driver can't unambiguously determine that the HW computed the hash at
822 * the higher layer. Note that the "should" in the second property above
823 * permits this.
824 */
825enum pkt_hash_types {
826 PKT_HASH_TYPE_NONE, /* Undefined type */
827 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
828 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
829 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
830};
831
832static inline void
833skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
834{
Tom Herbert61b905d2014-03-24 15:34:47 -0700835 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
Tom Herberta3b18dd2014-07-01 21:33:17 -0700836 skb->sw_hash = 0;
Tom Herbert61b905d2014-03-24 15:34:47 -0700837 skb->hash = hash;
Tom Herbert09323cc2013-12-15 22:16:19 -0800838}
839
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800840void __skb_get_hash(struct sk_buff *skb);
841static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000842{
Tom Herberta3b18dd2014-07-01 21:33:17 -0700843 if (!skb->l4_hash && !skb->sw_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800844 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000845
Tom Herbert61b905d2014-03-24 15:34:47 -0700846 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000847}
848
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800849static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
850{
Tom Herbert61b905d2014-03-24 15:34:47 -0700851 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800852}
853
Tom Herbert7539fad2013-12-15 22:12:18 -0800854static inline void skb_clear_hash(struct sk_buff *skb)
855{
Tom Herbert61b905d2014-03-24 15:34:47 -0700856 skb->hash = 0;
Tom Herberta3b18dd2014-07-01 21:33:17 -0700857 skb->sw_hash = 0;
Tom Herbert61b905d2014-03-24 15:34:47 -0700858 skb->l4_hash = 0;
Tom Herbert7539fad2013-12-15 22:12:18 -0800859}
860
861static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
862{
Tom Herbert61b905d2014-03-24 15:34:47 -0700863 if (!skb->l4_hash)
Tom Herbert7539fad2013-12-15 22:12:18 -0800864 skb_clear_hash(skb);
865}
866
Tom Herbert3df7a742013-12-15 22:16:29 -0800867static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
868{
Tom Herbert61b905d2014-03-24 15:34:47 -0700869 to->hash = from->hash;
Tom Herberta3b18dd2014-07-01 21:33:17 -0700870 to->sw_hash = from->sw_hash;
Tom Herbert61b905d2014-03-24 15:34:47 -0700871 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -0800872};
873
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700874#ifdef NET_SKBUFF_DATA_USES_OFFSET
875static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
876{
877 return skb->head + skb->end;
878}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000879
880static inline unsigned int skb_end_offset(const struct sk_buff *skb)
881{
882 return skb->end;
883}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700884#else
885static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
886{
887 return skb->end;
888}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000889
890static inline unsigned int skb_end_offset(const struct sk_buff *skb)
891{
892 return skb->end - skb->head;
893}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700894#endif
895
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700897#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898
Patrick Ohlyac45f602009-02-12 05:03:37 +0000899static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
900{
901 return &skb_shinfo(skb)->hwtstamps;
902}
903
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904/**
905 * skb_queue_empty - check if a queue is empty
906 * @list: queue head
907 *
908 * Returns true if the queue is empty, false otherwise.
909 */
910static inline int skb_queue_empty(const struct sk_buff_head *list)
911{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100912 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700913}
914
915/**
David S. Millerfc7ebb22008-09-23 00:34:07 -0700916 * skb_queue_is_last - check if skb is the last entry in the queue
917 * @list: queue head
918 * @skb: buffer
919 *
920 * Returns true if @skb is the last buffer on the list.
921 */
922static inline bool skb_queue_is_last(const struct sk_buff_head *list,
923 const struct sk_buff *skb)
924{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100925 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -0700926}
927
928/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800929 * skb_queue_is_first - check if skb is the first entry in the queue
930 * @list: queue head
931 * @skb: buffer
932 *
933 * Returns true if @skb is the first buffer on the list.
934 */
935static inline bool skb_queue_is_first(const struct sk_buff_head *list,
936 const struct sk_buff *skb)
937{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100938 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800939}
940
941/**
David S. Miller249c8b42008-09-23 00:44:42 -0700942 * skb_queue_next - return the next packet in the queue
943 * @list: queue head
944 * @skb: current buffer
945 *
946 * Return the next packet in @list after @skb. It is only valid to
947 * call this if skb_queue_is_last() evaluates to false.
948 */
949static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
950 const struct sk_buff *skb)
951{
952 /* This BUG_ON may seem severe, but if we just return then we
953 * are going to dereference garbage.
954 */
955 BUG_ON(skb_queue_is_last(list, skb));
956 return skb->next;
957}
958
959/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800960 * skb_queue_prev - return the prev packet in the queue
961 * @list: queue head
962 * @skb: current buffer
963 *
964 * Return the prev packet in @list before @skb. It is only valid to
965 * call this if skb_queue_is_first() evaluates to false.
966 */
967static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
968 const struct sk_buff *skb)
969{
970 /* This BUG_ON may seem severe, but if we just return then we
971 * are going to dereference garbage.
972 */
973 BUG_ON(skb_queue_is_first(list, skb));
974 return skb->prev;
975}
976
977/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700978 * skb_get - reference buffer
979 * @skb: buffer to reference
980 *
981 * Makes another reference to a socket buffer and returns a pointer
982 * to the buffer.
983 */
984static inline struct sk_buff *skb_get(struct sk_buff *skb)
985{
986 atomic_inc(&skb->users);
987 return skb;
988}
989
990/*
991 * If users == 1, we are the only owner and are can avoid redundant
992 * atomic change.
993 */
994
995/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996 * skb_cloned - is the buffer a clone
997 * @skb: buffer to check
998 *
999 * Returns true if the buffer was generated with skb_clone() and is
1000 * one of multiple shared copies of the buffer. Cloned buffers are
1001 * shared data so must not be written to under normal circumstances.
1002 */
1003static inline int skb_cloned(const struct sk_buff *skb)
1004{
1005 return skb->cloned &&
1006 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1007}
1008
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001009static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1010{
1011 might_sleep_if(pri & __GFP_WAIT);
1012
1013 if (skb_cloned(skb))
1014 return pskb_expand_head(skb, 0, 0, pri);
1015
1016 return 0;
1017}
1018
Linus Torvalds1da177e2005-04-16 15:20:36 -07001019/**
1020 * skb_header_cloned - is the header a clone
1021 * @skb: buffer to check
1022 *
1023 * Returns true if modifying the header part of the buffer requires
1024 * the data to be copied.
1025 */
1026static inline int skb_header_cloned(const struct sk_buff *skb)
1027{
1028 int dataref;
1029
1030 if (!skb->cloned)
1031 return 0;
1032
1033 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1034 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1035 return dataref != 1;
1036}
1037
1038/**
1039 * skb_header_release - release reference to header
1040 * @skb: buffer to operate on
1041 *
1042 * Drop a reference to the header part of the buffer. This is done
1043 * by acquiring a payload reference. You must not read from the header
1044 * part of skb->data after this.
1045 */
1046static inline void skb_header_release(struct sk_buff *skb)
1047{
1048 BUG_ON(skb->nohdr);
1049 skb->nohdr = 1;
1050 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1051}
1052
1053/**
1054 * skb_shared - is the buffer shared
1055 * @skb: buffer to check
1056 *
1057 * Returns true if more than one person has a reference to this
1058 * buffer.
1059 */
1060static inline int skb_shared(const struct sk_buff *skb)
1061{
1062 return atomic_read(&skb->users) != 1;
1063}
1064
1065/**
1066 * skb_share_check - check if buffer is shared and if so clone it
1067 * @skb: buffer to check
1068 * @pri: priority for memory allocation
1069 *
1070 * If the buffer is shared the buffer is cloned and the old copy
1071 * drops a reference. A new clone with a single reference is returned.
1072 * If the buffer is not shared the original buffer is returned. When
1073 * being called from interrupt status or with spinlocks held pri must
1074 * be GFP_ATOMIC.
1075 *
1076 * NULL is returned on a memory allocation failure.
1077 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001078static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079{
1080 might_sleep_if(pri & __GFP_WAIT);
1081 if (skb_shared(skb)) {
1082 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001083
1084 if (likely(nskb))
1085 consume_skb(skb);
1086 else
1087 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001088 skb = nskb;
1089 }
1090 return skb;
1091}
1092
1093/*
1094 * Copy shared buffers into a new sk_buff. We effectively do COW on
1095 * packets to handle cases where we have a local reader and forward
1096 * and a couple of other messy ones. The normal one is tcpdumping
1097 * a packet thats being forwarded.
1098 */
1099
1100/**
1101 * skb_unshare - make a copy of a shared buffer
1102 * @skb: buffer to check
1103 * @pri: priority for memory allocation
1104 *
1105 * If the socket buffer is a clone then this function creates a new
1106 * copy of the data, drops a reference count on the old copy and returns
1107 * the new copy with the reference count at 1. If the buffer is not a clone
1108 * the original buffer is returned. When called with a spinlock held or
1109 * from interrupt state @pri must be %GFP_ATOMIC
1110 *
1111 * %NULL is returned on a memory allocation failure.
1112 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001113static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001114 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001115{
1116 might_sleep_if(pri & __GFP_WAIT);
1117 if (skb_cloned(skb)) {
1118 struct sk_buff *nskb = skb_copy(skb, pri);
1119 kfree_skb(skb); /* Free our shared copy */
1120 skb = nskb;
1121 }
1122 return skb;
1123}
1124
1125/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001126 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001127 * @list_: list to peek at
1128 *
1129 * Peek an &sk_buff. Unlike most other operations you _MUST_
1130 * be careful with this one. A peek leaves the buffer on the
1131 * list and someone else may run off with it. You must hold
1132 * the appropriate locks or have a private queue to do this.
1133 *
1134 * Returns %NULL for an empty list or a pointer to the head element.
1135 * The reference count is not incremented and the reference is therefore
1136 * volatile. Use with caution.
1137 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001138static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139{
Eric Dumazet18d07002012-04-30 16:31:46 +00001140 struct sk_buff *skb = list_->next;
1141
1142 if (skb == (struct sk_buff *)list_)
1143 skb = NULL;
1144 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001145}
1146
1147/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001148 * skb_peek_next - peek skb following the given one from a queue
1149 * @skb: skb to start from
1150 * @list_: list to peek at
1151 *
1152 * Returns %NULL when the end of the list is met or a pointer to the
1153 * next element. The reference count is not incremented and the
1154 * reference is therefore volatile. Use with caution.
1155 */
1156static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1157 const struct sk_buff_head *list_)
1158{
1159 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001160
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001161 if (next == (struct sk_buff *)list_)
1162 next = NULL;
1163 return next;
1164}
1165
1166/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001167 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168 * @list_: list to peek at
1169 *
1170 * Peek an &sk_buff. Unlike most other operations you _MUST_
1171 * be careful with this one. A peek leaves the buffer on the
1172 * list and someone else may run off with it. You must hold
1173 * the appropriate locks or have a private queue to do this.
1174 *
1175 * Returns %NULL for an empty list or a pointer to the tail element.
1176 * The reference count is not incremented and the reference is therefore
1177 * volatile. Use with caution.
1178 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001179static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180{
Eric Dumazet18d07002012-04-30 16:31:46 +00001181 struct sk_buff *skb = list_->prev;
1182
1183 if (skb == (struct sk_buff *)list_)
1184 skb = NULL;
1185 return skb;
1186
Linus Torvalds1da177e2005-04-16 15:20:36 -07001187}
1188
1189/**
1190 * skb_queue_len - get queue length
1191 * @list_: list to measure
1192 *
1193 * Return the length of an &sk_buff queue.
1194 */
1195static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1196{
1197 return list_->qlen;
1198}
1199
David S. Miller67fed452008-09-21 22:36:24 -07001200/**
1201 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1202 * @list: queue to initialize
1203 *
1204 * This initializes only the list and queue length aspects of
1205 * an sk_buff_head object. This allows to initialize the list
1206 * aspects of an sk_buff_head without reinitializing things like
1207 * the spinlock. It can also be used for on-stack sk_buff_head
1208 * objects where the spinlock is known to not be used.
1209 */
1210static inline void __skb_queue_head_init(struct sk_buff_head *list)
1211{
1212 list->prev = list->next = (struct sk_buff *)list;
1213 list->qlen = 0;
1214}
1215
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001216/*
1217 * This function creates a split out lock class for each invocation;
1218 * this is needed for now since a whole lot of users of the skb-queue
1219 * infrastructure in drivers have different locking usage (in hardirq)
1220 * than the networking core (in softirq only). In the long run either the
1221 * network layer or drivers should need annotation to consolidate the
1222 * main types of usage into 3 classes.
1223 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001224static inline void skb_queue_head_init(struct sk_buff_head *list)
1225{
1226 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001227 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001228}
1229
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001230static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1231 struct lock_class_key *class)
1232{
1233 skb_queue_head_init(list);
1234 lockdep_set_class(&list->lock, class);
1235}
1236
Linus Torvalds1da177e2005-04-16 15:20:36 -07001237/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001238 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001239 *
1240 * The "__skb_xxxx()" functions are the non-atomic ones that
1241 * can only be called with interrupts disabled.
1242 */
Joe Perches7965bd42013-09-26 14:48:15 -07001243void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1244 struct sk_buff_head *list);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001245static inline void __skb_insert(struct sk_buff *newsk,
1246 struct sk_buff *prev, struct sk_buff *next,
1247 struct sk_buff_head *list)
1248{
1249 newsk->next = next;
1250 newsk->prev = prev;
1251 next->prev = prev->next = newsk;
1252 list->qlen++;
1253}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001254
David S. Miller67fed452008-09-21 22:36:24 -07001255static inline void __skb_queue_splice(const struct sk_buff_head *list,
1256 struct sk_buff *prev,
1257 struct sk_buff *next)
1258{
1259 struct sk_buff *first = list->next;
1260 struct sk_buff *last = list->prev;
1261
1262 first->prev = prev;
1263 prev->next = first;
1264
1265 last->next = next;
1266 next->prev = last;
1267}
1268
1269/**
1270 * skb_queue_splice - join two skb lists, this is designed for stacks
1271 * @list: the new list to add
1272 * @head: the place to add it in the first list
1273 */
1274static inline void skb_queue_splice(const struct sk_buff_head *list,
1275 struct sk_buff_head *head)
1276{
1277 if (!skb_queue_empty(list)) {
1278 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001279 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001280 }
1281}
1282
1283/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001284 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001285 * @list: the new list to add
1286 * @head: the place to add it in the first list
1287 *
1288 * The list at @list is reinitialised
1289 */
1290static inline void skb_queue_splice_init(struct sk_buff_head *list,
1291 struct sk_buff_head *head)
1292{
1293 if (!skb_queue_empty(list)) {
1294 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001295 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001296 __skb_queue_head_init(list);
1297 }
1298}
1299
1300/**
1301 * skb_queue_splice_tail - join two skb lists, each list being a queue
1302 * @list: the new list to add
1303 * @head: the place to add it in the first list
1304 */
1305static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1306 struct sk_buff_head *head)
1307{
1308 if (!skb_queue_empty(list)) {
1309 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001310 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001311 }
1312}
1313
1314/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001315 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001316 * @list: the new list to add
1317 * @head: the place to add it in the first list
1318 *
1319 * Each of the lists is a queue.
1320 * The list at @list is reinitialised
1321 */
1322static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1323 struct sk_buff_head *head)
1324{
1325 if (!skb_queue_empty(list)) {
1326 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001327 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001328 __skb_queue_head_init(list);
1329 }
1330}
1331
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001333 * __skb_queue_after - queue a buffer at the list head
1334 * @list: list to use
1335 * @prev: place after this buffer
1336 * @newsk: buffer to queue
1337 *
1338 * Queue a buffer int the middle of a list. This function takes no locks
1339 * and you must therefore hold required locks before calling it.
1340 *
1341 * A buffer cannot be placed on two lists at the same time.
1342 */
1343static inline void __skb_queue_after(struct sk_buff_head *list,
1344 struct sk_buff *prev,
1345 struct sk_buff *newsk)
1346{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001347 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001348}
1349
Joe Perches7965bd42013-09-26 14:48:15 -07001350void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1351 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001352
Gerrit Renkerf5572852008-04-14 00:05:28 -07001353static inline void __skb_queue_before(struct sk_buff_head *list,
1354 struct sk_buff *next,
1355 struct sk_buff *newsk)
1356{
1357 __skb_insert(newsk, next->prev, next, list);
1358}
1359
Stephen Hemminger300ce172005-10-30 13:47:34 -08001360/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001361 * __skb_queue_head - queue a buffer at the list head
1362 * @list: list to use
1363 * @newsk: buffer to queue
1364 *
1365 * Queue a buffer at the start of a list. This function takes no locks
1366 * and you must therefore hold required locks before calling it.
1367 *
1368 * A buffer cannot be placed on two lists at the same time.
1369 */
Joe Perches7965bd42013-09-26 14:48:15 -07001370void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001371static inline void __skb_queue_head(struct sk_buff_head *list,
1372 struct sk_buff *newsk)
1373{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001374 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375}
1376
1377/**
1378 * __skb_queue_tail - queue a buffer at the list tail
1379 * @list: list to use
1380 * @newsk: buffer to queue
1381 *
1382 * Queue a buffer at the end of a list. This function takes no locks
1383 * and you must therefore hold required locks before calling it.
1384 *
1385 * A buffer cannot be placed on two lists at the same time.
1386 */
Joe Perches7965bd42013-09-26 14:48:15 -07001387void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001388static inline void __skb_queue_tail(struct sk_buff_head *list,
1389 struct sk_buff *newsk)
1390{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001391 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392}
1393
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001395 * remove sk_buff from list. _Must_ be called atomically, and with
1396 * the list known..
1397 */
Joe Perches7965bd42013-09-26 14:48:15 -07001398void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1400{
1401 struct sk_buff *next, *prev;
1402
1403 list->qlen--;
1404 next = skb->next;
1405 prev = skb->prev;
1406 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407 next->prev = prev;
1408 prev->next = next;
1409}
1410
Gerrit Renkerf525c062008-04-14 00:04:12 -07001411/**
1412 * __skb_dequeue - remove from the head of the queue
1413 * @list: list to dequeue from
1414 *
1415 * Remove the head of the list. This function does not take any locks
1416 * so must be used with appropriate locks held only. The head item is
1417 * returned or %NULL if the list is empty.
1418 */
Joe Perches7965bd42013-09-26 14:48:15 -07001419struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Gerrit Renkerf525c062008-04-14 00:04:12 -07001420static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1421{
1422 struct sk_buff *skb = skb_peek(list);
1423 if (skb)
1424 __skb_unlink(skb, list);
1425 return skb;
1426}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001427
1428/**
1429 * __skb_dequeue_tail - remove from the tail of the queue
1430 * @list: list to dequeue from
1431 *
1432 * Remove the tail of the list. This function does not take any locks
1433 * so must be used with appropriate locks held only. The tail item is
1434 * returned or %NULL if the list is empty.
1435 */
Joe Perches7965bd42013-09-26 14:48:15 -07001436struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1438{
1439 struct sk_buff *skb = skb_peek_tail(list);
1440 if (skb)
1441 __skb_unlink(skb, list);
1442 return skb;
1443}
1444
1445
David S. Millerbdcc0922012-03-07 20:53:36 -05001446static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001447{
1448 return skb->data_len;
1449}
1450
1451static inline unsigned int skb_headlen(const struct sk_buff *skb)
1452{
1453 return skb->len - skb->data_len;
1454}
1455
1456static inline int skb_pagelen(const struct sk_buff *skb)
1457{
1458 int i, len = 0;
1459
1460 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001461 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001462 return len + skb_headlen(skb);
1463}
1464
Ian Campbell131ea662011-08-19 06:25:00 +00001465/**
1466 * __skb_fill_page_desc - initialise a paged fragment in an skb
1467 * @skb: buffer containing fragment to be initialised
1468 * @i: paged fragment index to initialise
1469 * @page: the page to use for this fragment
1470 * @off: the offset to the data with @page
1471 * @size: the length of the data
1472 *
1473 * Initialises the @i'th fragment of @skb to point to &size bytes at
1474 * offset @off within @page.
1475 *
1476 * Does not take any additional reference on the fragment.
1477 */
1478static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1479 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001480{
1481 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1482
Mel Gormanc48a11c2012-07-31 16:44:23 -07001483 /*
1484 * Propagate page->pfmemalloc to the skb if we can. The problem is
1485 * that not all callers have unique ownership of the page. If
1486 * pfmemalloc is set, we check the mapping as a mapping implies
1487 * page->index is set (index and pfmemalloc share space).
1488 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1489 * do not lose pfmemalloc information as the pages would not be
1490 * allocated using __GFP_MEMALLOC.
1491 */
Ian Campbella8605c62011-10-19 23:01:49 +00001492 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001494 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001495
1496 page = compound_head(page);
1497 if (page->pfmemalloc && !page->mapping)
1498 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00001499}
1500
1501/**
1502 * skb_fill_page_desc - initialise a paged fragment in an skb
1503 * @skb: buffer containing fragment to be initialised
1504 * @i: paged fragment index to initialise
1505 * @page: the page to use for this fragment
1506 * @off: the offset to the data with @page
1507 * @size: the length of the data
1508 *
1509 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01001510 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00001511 * addition updates @skb such that @i is the last fragment.
1512 *
1513 * Does not take any additional reference on the fragment.
1514 */
1515static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1516 struct page *page, int off, int size)
1517{
1518 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519 skb_shinfo(skb)->nr_frags = i + 1;
1520}
1521
Joe Perches7965bd42013-09-26 14:48:15 -07001522void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1523 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001524
Jason Wangf8e617e2013-11-01 14:07:47 +08001525void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1526 unsigned int truesize);
1527
Linus Torvalds1da177e2005-04-16 15:20:36 -07001528#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001529#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001530#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1531
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001532#ifdef NET_SKBUFF_DATA_USES_OFFSET
1533static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1534{
1535 return skb->head + skb->tail;
1536}
1537
1538static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1539{
1540 skb->tail = skb->data - skb->head;
1541}
1542
1543static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1544{
1545 skb_reset_tail_pointer(skb);
1546 skb->tail += offset;
1547}
Simon Horman7cc46192013-05-28 20:34:29 +00001548
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001549#else /* NET_SKBUFF_DATA_USES_OFFSET */
1550static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1551{
1552 return skb->tail;
1553}
1554
1555static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1556{
1557 skb->tail = skb->data;
1558}
1559
1560static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1561{
1562 skb->tail = skb->data + offset;
1563}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001564
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001565#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1566
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567/*
1568 * Add data to an sk_buff
1569 */
Mathias Krause0c7ddf32013-11-07 14:18:24 +01001570unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001571unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001572static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1573{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001574 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001575 SKB_LINEAR_ASSERT(skb);
1576 skb->tail += len;
1577 skb->len += len;
1578 return tmp;
1579}
1580
Joe Perches7965bd42013-09-26 14:48:15 -07001581unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1583{
1584 skb->data -= len;
1585 skb->len += len;
1586 return skb->data;
1587}
1588
Joe Perches7965bd42013-09-26 14:48:15 -07001589unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001590static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1591{
1592 skb->len -= len;
1593 BUG_ON(skb->len < skb->data_len);
1594 return skb->data += len;
1595}
1596
David S. Miller47d29642010-05-02 02:21:44 -07001597static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1598{
1599 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1600}
1601
Joe Perches7965bd42013-09-26 14:48:15 -07001602unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001603
1604static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1605{
1606 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001607 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001608 return NULL;
1609 skb->len -= len;
1610 return skb->data += len;
1611}
1612
1613static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1614{
1615 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1616}
1617
1618static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1619{
1620 if (likely(len <= skb_headlen(skb)))
1621 return 1;
1622 if (unlikely(len > skb->len))
1623 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001624 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001625}
1626
1627/**
1628 * skb_headroom - bytes at buffer head
1629 * @skb: buffer to check
1630 *
1631 * Return the number of bytes of free space at the head of an &sk_buff.
1632 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001633static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001634{
1635 return skb->data - skb->head;
1636}
1637
1638/**
1639 * skb_tailroom - bytes at buffer end
1640 * @skb: buffer to check
1641 *
1642 * Return the number of bytes of free space at the tail of an sk_buff
1643 */
1644static inline int skb_tailroom(const struct sk_buff *skb)
1645{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001646 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647}
1648
1649/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00001650 * skb_availroom - bytes at buffer end
1651 * @skb: buffer to check
1652 *
1653 * Return the number of bytes of free space at the tail of an sk_buff
1654 * allocated by sk_stream_alloc()
1655 */
1656static inline int skb_availroom(const struct sk_buff *skb)
1657{
Eric Dumazet16fad692013-03-14 05:40:32 +00001658 if (skb_is_nonlinear(skb))
1659 return 0;
1660
1661 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00001662}
1663
1664/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001665 * skb_reserve - adjust headroom
1666 * @skb: buffer to alter
1667 * @len: bytes to move
1668 *
1669 * Increase the headroom of an empty &sk_buff by reducing the tail
1670 * room. This is only allowed for an empty buffer.
1671 */
David S. Miller8243126c2006-01-17 02:54:21 -08001672static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001673{
1674 skb->data += len;
1675 skb->tail += len;
1676}
1677
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001678static inline void skb_reset_inner_headers(struct sk_buff *skb)
1679{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001680 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001681 skb->inner_network_header = skb->network_header;
1682 skb->inner_transport_header = skb->transport_header;
1683}
1684
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00001685static inline void skb_reset_mac_len(struct sk_buff *skb)
1686{
1687 skb->mac_len = skb->network_header - skb->mac_header;
1688}
1689
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001690static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1691 *skb)
1692{
1693 return skb->head + skb->inner_transport_header;
1694}
1695
1696static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1697{
1698 skb->inner_transport_header = skb->data - skb->head;
1699}
1700
1701static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1702 const int offset)
1703{
1704 skb_reset_inner_transport_header(skb);
1705 skb->inner_transport_header += offset;
1706}
1707
1708static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1709{
1710 return skb->head + skb->inner_network_header;
1711}
1712
1713static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1714{
1715 skb->inner_network_header = skb->data - skb->head;
1716}
1717
1718static inline void skb_set_inner_network_header(struct sk_buff *skb,
1719 const int offset)
1720{
1721 skb_reset_inner_network_header(skb);
1722 skb->inner_network_header += offset;
1723}
1724
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001725static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1726{
1727 return skb->head + skb->inner_mac_header;
1728}
1729
1730static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1731{
1732 skb->inner_mac_header = skb->data - skb->head;
1733}
1734
1735static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1736 const int offset)
1737{
1738 skb_reset_inner_mac_header(skb);
1739 skb->inner_mac_header += offset;
1740}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001741static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1742{
Cong Wang35d04612013-05-29 15:16:05 +08001743 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001744}
1745
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001746static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1747{
1748 return skb->head + skb->transport_header;
1749}
1750
1751static inline void skb_reset_transport_header(struct sk_buff *skb)
1752{
1753 skb->transport_header = skb->data - skb->head;
1754}
1755
1756static inline void skb_set_transport_header(struct sk_buff *skb,
1757 const int offset)
1758{
1759 skb_reset_transport_header(skb);
1760 skb->transport_header += offset;
1761}
1762
1763static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1764{
1765 return skb->head + skb->network_header;
1766}
1767
1768static inline void skb_reset_network_header(struct sk_buff *skb)
1769{
1770 skb->network_header = skb->data - skb->head;
1771}
1772
1773static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1774{
1775 skb_reset_network_header(skb);
1776 skb->network_header += offset;
1777}
1778
1779static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1780{
1781 return skb->head + skb->mac_header;
1782}
1783
1784static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1785{
Cong Wang35d04612013-05-29 15:16:05 +08001786 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001787}
1788
1789static inline void skb_reset_mac_header(struct sk_buff *skb)
1790{
1791 skb->mac_header = skb->data - skb->head;
1792}
1793
1794static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1795{
1796 skb_reset_mac_header(skb);
1797 skb->mac_header += offset;
1798}
1799
Timo Teräs0e3da5b2013-12-16 11:02:09 +02001800static inline void skb_pop_mac_header(struct sk_buff *skb)
1801{
1802 skb->mac_header = skb->network_header;
1803}
1804
Ying Xuefbbdb8f2013-03-27 16:46:06 +00001805static inline void skb_probe_transport_header(struct sk_buff *skb,
1806 const int offset_hint)
1807{
1808 struct flow_keys keys;
1809
1810 if (skb_transport_header_was_set(skb))
1811 return;
1812 else if (skb_flow_dissect(skb, &keys))
1813 skb_set_transport_header(skb, keys.thoff);
1814 else
1815 skb_set_transport_header(skb, offset_hint);
1816}
1817
Eric Dumazet03606892012-02-23 10:55:02 +00001818static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1819{
1820 if (skb_mac_header_was_set(skb)) {
1821 const unsigned char *old_mac = skb_mac_header(skb);
1822
1823 skb_set_mac_header(skb, -skb->mac_len);
1824 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1825 }
1826}
1827
Michał Mirosław04fb4512010-12-14 15:24:08 +00001828static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1829{
1830 return skb->csum_start - skb_headroom(skb);
1831}
1832
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001833static inline int skb_transport_offset(const struct sk_buff *skb)
1834{
1835 return skb_transport_header(skb) - skb->data;
1836}
1837
1838static inline u32 skb_network_header_len(const struct sk_buff *skb)
1839{
1840 return skb->transport_header - skb->network_header;
1841}
1842
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001843static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1844{
1845 return skb->inner_transport_header - skb->inner_network_header;
1846}
1847
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001848static inline int skb_network_offset(const struct sk_buff *skb)
1849{
1850 return skb_network_header(skb) - skb->data;
1851}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001852
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001853static inline int skb_inner_network_offset(const struct sk_buff *skb)
1854{
1855 return skb_inner_network_header(skb) - skb->data;
1856}
1857
Changli Gaof9599ce2010-08-04 04:43:44 +00001858static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1859{
1860 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1861}
1862
Tom Herberte5eb4e32014-06-14 23:24:28 -07001863static inline void skb_pop_rcv_encapsulation(struct sk_buff *skb)
1864{
1865 /* Only continue with checksum unnecessary if device indicated
1866 * it is valid across encapsulation (skb->encapsulation was set).
1867 */
1868 if (skb->ip_summed == CHECKSUM_UNNECESSARY && !skb->encapsulation)
1869 skb->ip_summed = CHECKSUM_NONE;
1870
1871 skb->encapsulation = 0;
1872 skb->csum_valid = 0;
1873}
1874
Linus Torvalds1da177e2005-04-16 15:20:36 -07001875/*
1876 * CPUs often take a performance hit when accessing unaligned memory
1877 * locations. The actual performance hit varies, it can be small if the
1878 * hardware handles it or large if we have to take an exception and fix it
1879 * in software.
1880 *
1881 * Since an ethernet header is 14 bytes network drivers often end up with
1882 * the IP header at an unaligned offset. The IP header can be aligned by
1883 * shifting the start of the packet by 2 bytes. Drivers should do this
1884 * with:
1885 *
Tobias Klauser8660c122009-07-13 22:48:16 +00001886 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001887 *
1888 * The downside to this alignment of the IP header is that the DMA is now
1889 * unaligned. On some architectures the cost of an unaligned DMA is high
1890 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00001891 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001892 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1893 * to be overridden.
1894 */
1895#ifndef NET_IP_ALIGN
1896#define NET_IP_ALIGN 2
1897#endif
1898
Anton Blanchard025be812006-03-31 02:27:06 -08001899/*
1900 * The networking layer reserves some headroom in skb data (via
1901 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1902 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08001903 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08001904 *
1905 * Unfortunately this headroom changes the DMA alignment of the resulting
1906 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1907 * on some architectures. An architecture can override this value,
1908 * perhaps setting it to a cacheline in size (since that will maintain
1909 * cacheline alignment of the DMA). It must be a power of 2.
1910 *
David S. Millerd6301d32009-02-08 19:24:13 -08001911 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08001912 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07001913 *
1914 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1915 * to reduce average number of cache lines per packet.
1916 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07001917 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08001918 */
1919#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07001920#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08001921#endif
1922
Joe Perches7965bd42013-09-26 14:48:15 -07001923int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001924
1925static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1926{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00001927 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07001928 WARN_ON(1);
1929 return;
1930 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001931 skb->len = len;
1932 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001933}
1934
Joe Perches7965bd42013-09-26 14:48:15 -07001935void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001936
1937static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1938{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001939 if (skb->data_len)
1940 return ___pskb_trim(skb, len);
1941 __skb_trim(skb, len);
1942 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001943}
1944
1945static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1946{
1947 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1948}
1949
1950/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07001951 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1952 * @skb: buffer to alter
1953 * @len: new length
1954 *
1955 * This is identical to pskb_trim except that the caller knows that
1956 * the skb is not cloned so we should never get an error due to out-
1957 * of-memory.
1958 */
1959static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1960{
1961 int err = pskb_trim(skb, len);
1962 BUG_ON(err);
1963}
1964
1965/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001966 * skb_orphan - orphan a buffer
1967 * @skb: buffer to orphan
1968 *
1969 * If a buffer currently has an owner then we call the owner's
1970 * destructor function and make the @skb unowned. The buffer continues
1971 * to exist but is no longer charged to its former owner.
1972 */
1973static inline void skb_orphan(struct sk_buff *skb)
1974{
Eric Dumazetc34a7612013-07-30 16:11:15 -07001975 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001976 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07001977 skb->destructor = NULL;
1978 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07001979 } else {
1980 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07001981 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001982}
1983
1984/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00001985 * skb_orphan_frags - orphan the frags contained in a buffer
1986 * @skb: buffer to orphan frags from
1987 * @gfp_mask: allocation mask for replacement pages
1988 *
1989 * For each frag in the SKB which needs a destructor (i.e. has an
1990 * owner) create a copy of that frag and release the original
1991 * page by calling the destructor.
1992 */
1993static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1994{
1995 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1996 return 0;
1997 return skb_copy_ubufs(skb, gfp_mask);
1998}
1999
2000/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002001 * __skb_queue_purge - empty a list
2002 * @list: list to empty
2003 *
2004 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2005 * the list and one reference dropped. This function does not take the
2006 * list lock and the caller must hold the relevant locks to use it.
2007 */
Joe Perches7965bd42013-09-26 14:48:15 -07002008void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009static inline void __skb_queue_purge(struct sk_buff_head *list)
2010{
2011 struct sk_buff *skb;
2012 while ((skb = __skb_dequeue(list)) != NULL)
2013 kfree_skb(skb);
2014}
2015
Alexander Duycke5e67302013-02-08 10:17:15 +00002016#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
2017#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
2018#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
2019
Joe Perches7965bd42013-09-26 14:48:15 -07002020void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021
Joe Perches7965bd42013-09-26 14:48:15 -07002022struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2023 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07002024
2025/**
2026 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2027 * @dev: network device to receive on
2028 * @length: length to allocate
2029 *
2030 * Allocate a new &sk_buff and assign it a usage count of one. The
2031 * buffer has unspecified headroom built in. Users should allocate
2032 * the headroom they think they need without accounting for the
2033 * built in space. The built in space is used for optimisations.
2034 *
2035 * %NULL is returned if there is no free memory. Although this function
2036 * allocates memory it can be called from an interrupt.
2037 */
2038static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002039 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002040{
2041 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2042}
2043
Eric Dumazet6f532612012-05-18 05:12:12 +00002044/* legacy helper around __netdev_alloc_skb() */
2045static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2046 gfp_t gfp_mask)
2047{
2048 return __netdev_alloc_skb(NULL, length, gfp_mask);
2049}
2050
2051/* legacy helper around netdev_alloc_skb() */
2052static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2053{
2054 return netdev_alloc_skb(NULL, length);
2055}
2056
2057
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002058static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2059 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002060{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002061 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002062
2063 if (NET_IP_ALIGN && skb)
2064 skb_reserve(skb, NET_IP_ALIGN);
2065 return skb;
2066}
2067
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002068static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2069 unsigned int length)
2070{
2071 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2072}
2073
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002074/**
2075 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
Mel Gorman06140022012-07-31 16:44:24 -07002076 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2077 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2078 * @order: size of the allocation
2079 *
2080 * Allocate a new page.
2081 *
2082 * %NULL is returned if there is no free memory.
2083*/
2084static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
2085 struct sk_buff *skb,
2086 unsigned int order)
2087{
2088 struct page *page;
2089
2090 gfp_mask |= __GFP_COLD;
2091
2092 if (!(gfp_mask & __GFP_NOMEMALLOC))
2093 gfp_mask |= __GFP_MEMALLOC;
2094
2095 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2096 if (skb && page && page->pfmemalloc)
2097 skb->pfmemalloc = true;
2098
2099 return page;
2100}
2101
2102/**
2103 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2104 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2105 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2106 *
2107 * Allocate a new page.
2108 *
2109 * %NULL is returned if there is no free memory.
2110 */
2111static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
2112 struct sk_buff *skb)
2113{
2114 return __skb_alloc_pages(gfp_mask, skb, 0);
2115}
2116
2117/**
2118 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2119 * @page: The page that was allocated from skb_alloc_page
2120 * @skb: The skb that may need pfmemalloc set
2121 */
2122static inline void skb_propagate_pfmemalloc(struct page *page,
2123 struct sk_buff *skb)
2124{
2125 if (page && page->pfmemalloc)
2126 skb->pfmemalloc = true;
2127}
2128
Eric Dumazet564824b2010-10-11 19:05:25 +00002129/**
Masanari Iidae2278672014-02-18 22:54:36 +09002130 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002131 * @frag: the paged fragment
2132 *
2133 * Returns the &struct page associated with @frag.
2134 */
2135static inline struct page *skb_frag_page(const skb_frag_t *frag)
2136{
Ian Campbella8605c62011-10-19 23:01:49 +00002137 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00002138}
2139
2140/**
2141 * __skb_frag_ref - take an addition reference on a paged fragment.
2142 * @frag: the paged fragment
2143 *
2144 * Takes an additional reference on the paged fragment @frag.
2145 */
2146static inline void __skb_frag_ref(skb_frag_t *frag)
2147{
2148 get_page(skb_frag_page(frag));
2149}
2150
2151/**
2152 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2153 * @skb: the buffer
2154 * @f: the fragment offset.
2155 *
2156 * Takes an additional reference on the @f'th paged fragment of @skb.
2157 */
2158static inline void skb_frag_ref(struct sk_buff *skb, int f)
2159{
2160 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2161}
2162
2163/**
2164 * __skb_frag_unref - release a reference on a paged fragment.
2165 * @frag: the paged fragment
2166 *
2167 * Releases a reference on the paged fragment @frag.
2168 */
2169static inline void __skb_frag_unref(skb_frag_t *frag)
2170{
2171 put_page(skb_frag_page(frag));
2172}
2173
2174/**
2175 * skb_frag_unref - release a reference on a paged fragment of an skb.
2176 * @skb: the buffer
2177 * @f: the fragment offset
2178 *
2179 * Releases a reference on the @f'th paged fragment of @skb.
2180 */
2181static inline void skb_frag_unref(struct sk_buff *skb, int f)
2182{
2183 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2184}
2185
2186/**
2187 * skb_frag_address - gets the address of the data contained in a paged fragment
2188 * @frag: the paged fragment buffer
2189 *
2190 * Returns the address of the data within @frag. The page must already
2191 * be mapped.
2192 */
2193static inline void *skb_frag_address(const skb_frag_t *frag)
2194{
2195 return page_address(skb_frag_page(frag)) + frag->page_offset;
2196}
2197
2198/**
2199 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2200 * @frag: the paged fragment buffer
2201 *
2202 * Returns the address of the data within @frag. Checks that the page
2203 * is mapped and returns %NULL otherwise.
2204 */
2205static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2206{
2207 void *ptr = page_address(skb_frag_page(frag));
2208 if (unlikely(!ptr))
2209 return NULL;
2210
2211 return ptr + frag->page_offset;
2212}
2213
2214/**
2215 * __skb_frag_set_page - sets the page contained in a paged fragment
2216 * @frag: the paged fragment
2217 * @page: the page to set
2218 *
2219 * Sets the fragment @frag to contain @page.
2220 */
2221static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2222{
Ian Campbella8605c62011-10-19 23:01:49 +00002223 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002224}
2225
2226/**
2227 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2228 * @skb: the buffer
2229 * @f: the fragment offset
2230 * @page: the page to set
2231 *
2232 * Sets the @f'th fragment of @skb to contain @page.
2233 */
2234static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2235 struct page *page)
2236{
2237 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2238}
2239
Eric Dumazet400dfd32013-10-17 16:27:07 -07002240bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2241
Ian Campbell131ea662011-08-19 06:25:00 +00002242/**
2243 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002244 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002245 * @frag: the paged fragment to map
2246 * @offset: the offset within the fragment (starting at the
2247 * fragment's own offset)
2248 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002249 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00002250 *
2251 * Maps the page associated with @frag to @device.
2252 */
2253static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2254 const skb_frag_t *frag,
2255 size_t offset, size_t size,
2256 enum dma_data_direction dir)
2257{
2258 return dma_map_page(dev, skb_frag_page(frag),
2259 frag->page_offset + offset, size, dir);
2260}
2261
Eric Dumazet117632e2011-12-03 21:39:53 +00002262static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2263 gfp_t gfp_mask)
2264{
2265 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2266}
2267
Octavian Purdilabad93e92014-06-12 01:36:26 +03002268
2269static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2270 gfp_t gfp_mask)
2271{
2272 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2273}
2274
2275
Ian Campbell131ea662011-08-19 06:25:00 +00002276/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002277 * skb_clone_writable - is the header of a clone writable
2278 * @skb: buffer to check
2279 * @len: length up to which to write
2280 *
2281 * Returns true if modifying the header part of the cloned buffer
2282 * does not requires the data to be copied.
2283 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002284static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002285{
2286 return !skb_header_cloned(skb) &&
2287 skb_headroom(skb) + len <= skb->hdr_len;
2288}
2289
Herbert Xud9cc2042007-09-16 16:21:16 -07002290static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2291 int cloned)
2292{
2293 int delta = 0;
2294
Herbert Xud9cc2042007-09-16 16:21:16 -07002295 if (headroom > skb_headroom(skb))
2296 delta = headroom - skb_headroom(skb);
2297
2298 if (delta || cloned)
2299 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2300 GFP_ATOMIC);
2301 return 0;
2302}
2303
Patrick McHardy334a8132007-06-25 04:35:20 -07002304/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002305 * skb_cow - copy header of skb when it is required
2306 * @skb: buffer to cow
2307 * @headroom: needed headroom
2308 *
2309 * If the skb passed lacks sufficient headroom or its data part
2310 * is shared, data is reallocated. If reallocation fails, an error
2311 * is returned and original skb is not changed.
2312 *
2313 * The result is skb with writable area skb->head...skb->tail
2314 * and at least @headroom of space at head.
2315 */
2316static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2317{
Herbert Xud9cc2042007-09-16 16:21:16 -07002318 return __skb_cow(skb, headroom, skb_cloned(skb));
2319}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002320
Herbert Xud9cc2042007-09-16 16:21:16 -07002321/**
2322 * skb_cow_head - skb_cow but only making the head writable
2323 * @skb: buffer to cow
2324 * @headroom: needed headroom
2325 *
2326 * This function is identical to skb_cow except that we replace the
2327 * skb_cloned check by skb_header_cloned. It should be used when
2328 * you only need to push on some header and do not need to modify
2329 * the data.
2330 */
2331static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2332{
2333 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002334}
2335
2336/**
2337 * skb_padto - pad an skbuff up to a minimal size
2338 * @skb: buffer to pad
2339 * @len: minimal length
2340 *
2341 * Pads up a buffer to ensure the trailing bytes exist and are
2342 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002343 * is untouched. Otherwise it is extended. Returns zero on
2344 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002345 */
2346
Herbert Xu5b057c62006-06-23 02:06:41 -07002347static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002348{
2349 unsigned int size = skb->len;
2350 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002351 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002352 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002353}
2354
2355static inline int skb_add_data(struct sk_buff *skb,
2356 char __user *from, int copy)
2357{
2358 const int off = skb->len;
2359
2360 if (skb->ip_summed == CHECKSUM_NONE) {
2361 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08002362 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
Linus Torvalds1da177e2005-04-16 15:20:36 -07002363 copy, 0, &err);
2364 if (!err) {
2365 skb->csum = csum_block_add(skb->csum, csum, off);
2366 return 0;
2367 }
2368 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2369 return 0;
2370
2371 __skb_trim(skb, off);
2372 return -EFAULT;
2373}
2374
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002375static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2376 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002377{
2378 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002379 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002380
Ian Campbellea2ab692011-08-22 23:44:58 +00002381 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002382 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002383 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002384 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002385}
2386
Herbert Xu364c6ba2006-06-09 16:10:40 -07002387static inline int __skb_linearize(struct sk_buff *skb)
2388{
2389 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2390}
2391
Linus Torvalds1da177e2005-04-16 15:20:36 -07002392/**
2393 * skb_linearize - convert paged skb to linear one
2394 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002395 *
2396 * If there is no free memory -ENOMEM is returned, otherwise zero
2397 * is returned and the old skb data released.
2398 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002399static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002400{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002401 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2402}
2403
2404/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00002405 * skb_has_shared_frag - can any frag be overwritten
2406 * @skb: buffer to test
2407 *
2408 * Return true if the skb has at least one frag that might be modified
2409 * by an external entity (as in vmsplice()/sendfile())
2410 */
2411static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2412{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00002413 return skb_is_nonlinear(skb) &&
2414 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00002415}
2416
2417/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07002418 * skb_linearize_cow - make sure skb is linear and writable
2419 * @skb: buffer to process
2420 *
2421 * If there is no free memory -ENOMEM is returned, otherwise zero
2422 * is returned and the old skb data released.
2423 */
2424static inline int skb_linearize_cow(struct sk_buff *skb)
2425{
2426 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2427 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002428}
2429
2430/**
2431 * skb_postpull_rcsum - update checksum for received skb after pull
2432 * @skb: buffer to update
2433 * @start: start of data before pull
2434 * @len: length of data pulled
2435 *
2436 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002437 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2438 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002439 */
2440
2441static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002442 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002443{
Patrick McHardy84fa7932006-08-29 16:44:56 -07002444 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002445 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2446}
2447
Herbert Xucbb042f2006-03-20 22:43:56 -08002448unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2449
David S. Miller7ce5a272013-12-02 17:26:05 -05002450/**
2451 * pskb_trim_rcsum - trim received skb and update checksum
2452 * @skb: buffer to trim
2453 * @len: new length
2454 *
2455 * This is exactly the same as pskb_trim except that it ensures the
2456 * checksum of received packets are still valid after the operation.
2457 */
2458
2459static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2460{
2461 if (likely(len >= skb->len))
2462 return 0;
2463 if (skb->ip_summed == CHECKSUM_COMPLETE)
2464 skb->ip_summed = CHECKSUM_NONE;
2465 return __pskb_trim(skb, len);
2466}
2467
Linus Torvalds1da177e2005-04-16 15:20:36 -07002468#define skb_queue_walk(queue, skb) \
2469 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002470 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002471 skb = skb->next)
2472
James Chapman46f89142007-04-30 00:07:31 -07002473#define skb_queue_walk_safe(queue, skb, tmp) \
2474 for (skb = (queue)->next, tmp = skb->next; \
2475 skb != (struct sk_buff *)(queue); \
2476 skb = tmp, tmp = skb->next)
2477
David S. Miller1164f522008-09-23 00:49:44 -07002478#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002479 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002480 skb = skb->next)
2481
2482#define skb_queue_walk_from_safe(queue, skb, tmp) \
2483 for (tmp = skb->next; \
2484 skb != (struct sk_buff *)(queue); \
2485 skb = tmp, tmp = skb->next)
2486
Stephen Hemminger300ce172005-10-30 13:47:34 -08002487#define skb_queue_reverse_walk(queue, skb) \
2488 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002489 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08002490 skb = skb->prev)
2491
David S. Miller686a2952011-01-20 22:47:32 -08002492#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2493 for (skb = (queue)->prev, tmp = skb->prev; \
2494 skb != (struct sk_buff *)(queue); \
2495 skb = tmp, tmp = skb->prev)
2496
2497#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2498 for (tmp = skb->prev; \
2499 skb != (struct sk_buff *)(queue); \
2500 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002501
David S. Miller21dc3302010-08-23 00:13:46 -07002502static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07002503{
2504 return skb_shinfo(skb)->frag_list != NULL;
2505}
2506
2507static inline void skb_frag_list_init(struct sk_buff *skb)
2508{
2509 skb_shinfo(skb)->frag_list = NULL;
2510}
2511
2512static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2513{
2514 frag->next = skb_shinfo(skb)->frag_list;
2515 skb_shinfo(skb)->frag_list = frag;
2516}
2517
2518#define skb_walk_frags(skb, iter) \
2519 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2520
Joe Perches7965bd42013-09-26 14:48:15 -07002521struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2522 int *peeked, int *off, int *err);
2523struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2524 int *err);
2525unsigned int datagram_poll(struct file *file, struct socket *sock,
2526 struct poll_table_struct *wait);
2527int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2528 struct iovec *to, int size);
2529int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2530 struct iovec *iov);
2531int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2532 const struct iovec *from, int from_offset,
2533 int len);
2534int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2535 int offset, size_t count);
2536int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2537 const struct iovec *to, int to_offset,
2538 int size);
2539void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2540void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2541int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07002542int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2543int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2544__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2545 int len, __wsum csum);
2546int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2547 struct pipe_inode_info *pipe, unsigned int len,
2548 unsigned int flags);
2549void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01002550unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00002551int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2552 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07002553void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2554int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2555void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Florian Westphalde960aa2014-01-26 10:58:16 +01002556unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -07002557struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Vlad Yasevich0d5501c2014-08-08 14:42:13 -04002558struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002559
Daniel Borkmann2817a332013-10-30 11:50:51 +01002560struct skb_checksum_ops {
2561 __wsum (*update)(const void *mem, int len, __wsum wsum);
2562 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2563};
2564
2565__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2566 __wsum csum, const struct skb_checksum_ops *ops);
2567__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2568 __wsum csum);
2569
David S. Miller690e36e2014-08-23 12:13:41 -07002570static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset,
2571 int len, void *data, int hlen, void *buffer)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002572{
Patrick McHardy55820ee2005-07-05 14:08:10 -07002573 if (hlen - offset >= len)
David S. Miller690e36e2014-08-23 12:13:41 -07002574 return data + offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002575
David S. Miller690e36e2014-08-23 12:13:41 -07002576 if (!skb ||
2577 skb_copy_bits(skb, offset, buffer, len) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002578 return NULL;
2579
2580 return buffer;
2581}
2582
David S. Miller690e36e2014-08-23 12:13:41 -07002583static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2584 int len, void *buffer)
2585{
2586 return __skb_header_pointer(skb, offset, len, skb->data,
2587 skb_headlen(skb), buffer);
2588}
2589
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01002590/**
2591 * skb_needs_linearize - check if we need to linearize a given skb
2592 * depending on the given device features.
2593 * @skb: socket buffer to check
2594 * @features: net device features
2595 *
2596 * Returns true if either:
2597 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2598 * 2. skb is fragmented and the device does not support SG.
2599 */
2600static inline bool skb_needs_linearize(struct sk_buff *skb,
2601 netdev_features_t features)
2602{
2603 return skb_is_nonlinear(skb) &&
2604 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2605 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2606}
2607
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03002608static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2609 void *to,
2610 const unsigned int len)
2611{
2612 memcpy(to, skb->data, len);
2613}
2614
2615static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2616 const int offset, void *to,
2617 const unsigned int len)
2618{
2619 memcpy(to, skb->data + offset, len);
2620}
2621
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03002622static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2623 const void *from,
2624 const unsigned int len)
2625{
2626 memcpy(skb->data, from, len);
2627}
2628
2629static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2630 const int offset,
2631 const void *from,
2632 const unsigned int len)
2633{
2634 memcpy(skb->data + offset, from, len);
2635}
2636
Joe Perches7965bd42013-09-26 14:48:15 -07002637void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002638
Patrick Ohlyac45f602009-02-12 05:03:37 +00002639static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2640{
2641 return skb->tstamp;
2642}
2643
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002644/**
2645 * skb_get_timestamp - get timestamp from a skb
2646 * @skb: skb to get stamp from
2647 * @stamp: pointer to struct timeval to store stamp in
2648 *
2649 * Timestamps are stored in the skb as offsets to a base timestamp.
2650 * This function converts the offset back to a struct timeval and stores
2651 * it in stamp.
2652 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00002653static inline void skb_get_timestamp(const struct sk_buff *skb,
2654 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002655{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002656 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002657}
2658
Patrick Ohlyac45f602009-02-12 05:03:37 +00002659static inline void skb_get_timestampns(const struct sk_buff *skb,
2660 struct timespec *stamp)
2661{
2662 *stamp = ktime_to_timespec(skb->tstamp);
2663}
2664
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002665static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002666{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002667 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002668}
2669
Stephen Hemminger164891a2007-04-23 22:26:16 -07002670static inline ktime_t net_timedelta(ktime_t t)
2671{
2672 return ktime_sub(ktime_get_real(), t);
2673}
2674
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07002675static inline ktime_t net_invalid_timestamp(void)
2676{
2677 return ktime_set(0, 0);
2678}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002679
Richard Cochranc1f19b52010-07-17 08:49:36 +00002680#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2681
Joe Perches7965bd42013-09-26 14:48:15 -07002682void skb_clone_tx_timestamp(struct sk_buff *skb);
2683bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00002684
2685#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2686
2687static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2688{
2689}
2690
2691static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2692{
2693 return false;
2694}
2695
2696#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2697
2698/**
2699 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2700 *
Richard Cochranda92b192011-10-21 00:49:15 +00002701 * PHY drivers may accept clones of transmitted packets for
2702 * timestamping via their phy_driver.txtstamp method. These drivers
2703 * must call this function to return the skb back to the stack, with
2704 * or without a timestamp.
2705 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00002706 * @skb: clone of the the original outgoing packet
Richard Cochranda92b192011-10-21 00:49:15 +00002707 * @hwtstamps: hardware time stamps, may be NULL if not available
Richard Cochranc1f19b52010-07-17 08:49:36 +00002708 *
2709 */
2710void skb_complete_tx_timestamp(struct sk_buff *skb,
2711 struct skb_shared_hwtstamps *hwtstamps);
2712
Willem de Bruijne7fd2882014-08-04 22:11:48 -04002713void __skb_tstamp_tx(struct sk_buff *orig_skb,
2714 struct skb_shared_hwtstamps *hwtstamps,
2715 struct sock *sk, int tstype);
2716
Patrick Ohlyac45f602009-02-12 05:03:37 +00002717/**
2718 * skb_tstamp_tx - queue clone of skb with send time stamps
2719 * @orig_skb: the original outgoing packet
2720 * @hwtstamps: hardware time stamps, may be NULL if not available
2721 *
2722 * If the skb has a socket associated, then this function clones the
2723 * skb (thus sharing the actual data and optional structures), stores
2724 * the optional hardware time stamping information (if non NULL) or
2725 * generates a software time stamp (otherwise), then queues the clone
2726 * to the error queue of the socket. Errors are silently ignored.
2727 */
Joe Perches7965bd42013-09-26 14:48:15 -07002728void skb_tstamp_tx(struct sk_buff *orig_skb,
2729 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00002730
Richard Cochran4507a712010-07-17 08:48:28 +00002731static inline void sw_tx_timestamp(struct sk_buff *skb)
2732{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002733 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2734 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00002735 skb_tstamp_tx(skb, NULL);
2736}
2737
2738/**
2739 * skb_tx_timestamp() - Driver hook for transmit timestamping
2740 *
2741 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00002742 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00002743 *
David S. Miller73409f32013-12-27 13:04:33 -05002744 * Specifically, one should make absolutely sure that this function is
2745 * called before TX completion of this packet can trigger. Otherwise
2746 * the packet could potentially already be freed.
2747 *
Richard Cochran4507a712010-07-17 08:48:28 +00002748 * @skb: A socket buffer.
2749 */
2750static inline void skb_tx_timestamp(struct sk_buff *skb)
2751{
Richard Cochranc1f19b52010-07-17 08:49:36 +00002752 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002753 sw_tx_timestamp(skb);
2754}
2755
Johannes Berg6e3e9392011-11-09 10:15:42 +01002756/**
2757 * skb_complete_wifi_ack - deliver skb with wifi status
2758 *
2759 * @skb: the original outgoing packet
2760 * @acked: ack status
2761 *
2762 */
2763void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2764
Joe Perches7965bd42013-09-26 14:48:15 -07002765__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2766__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002767
Herbert Xu60476372007-04-09 11:59:39 -07002768static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2769{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002770 return ((skb->ip_summed & CHECKSUM_UNNECESSARY) || skb->csum_valid);
Herbert Xu60476372007-04-09 11:59:39 -07002771}
2772
Herbert Xufb286bb2005-11-10 13:01:24 -08002773/**
2774 * skb_checksum_complete - Calculate checksum of an entire packet
2775 * @skb: packet to process
2776 *
2777 * This function calculates the checksum over the entire packet plus
2778 * the value of skb->csum. The latter can be used to supply the
2779 * checksum of a pseudo header as used by TCP/UDP. It returns the
2780 * checksum.
2781 *
2782 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2783 * this function can be used to verify that checksum on received
2784 * packets. In that case the function should return zero if the
2785 * checksum is correct. In particular, this function will return zero
2786 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2787 * hardware has already verified the correctness of the checksum.
2788 */
Al Viro4381ca32007-07-15 21:00:11 +01002789static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002790{
Herbert Xu60476372007-04-09 11:59:39 -07002791 return skb_csum_unnecessary(skb) ?
2792 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002793}
2794
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002795/* Check if we need to perform checksum complete validation.
2796 *
2797 * Returns true if checksum complete is needed, false otherwise
2798 * (either checksum is unnecessary or zero checksum is allowed).
2799 */
2800static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
2801 bool zero_okay,
2802 __sum16 check)
2803{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002804 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
2805 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002806 return false;
2807 }
2808
2809 return true;
2810}
2811
2812/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
2813 * in checksum_init.
2814 */
2815#define CHECKSUM_BREAK 76
2816
2817/* Validate (init) checksum based on checksum complete.
2818 *
2819 * Return values:
2820 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
2821 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
2822 * checksum is stored in skb->csum for use in __skb_checksum_complete
2823 * non-zero: value of invalid checksum
2824 *
2825 */
2826static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
2827 bool complete,
2828 __wsum psum)
2829{
2830 if (skb->ip_summed == CHECKSUM_COMPLETE) {
2831 if (!csum_fold(csum_add(psum, skb->csum))) {
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002832 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002833 return 0;
2834 }
2835 }
2836
2837 skb->csum = psum;
2838
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002839 if (complete || skb->len <= CHECKSUM_BREAK) {
2840 __sum16 csum;
2841
2842 csum = __skb_checksum_complete(skb);
2843 skb->csum_valid = !csum;
2844 return csum;
2845 }
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002846
2847 return 0;
2848}
2849
2850static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
2851{
2852 return 0;
2853}
2854
2855/* Perform checksum validate (init). Note that this is a macro since we only
2856 * want to calculate the pseudo header which is an input function if necessary.
2857 * First we try to validate without any computation (checksum unnecessary) and
2858 * then calculate based on checksum complete calling the function to compute
2859 * pseudo header.
2860 *
2861 * Return values:
2862 * 0: checksum is validated or try to in skb_checksum_complete
2863 * non-zero: value of invalid checksum
2864 */
2865#define __skb_checksum_validate(skb, proto, complete, \
2866 zero_okay, check, compute_pseudo) \
2867({ \
2868 __sum16 __ret = 0; \
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002869 skb->csum_valid = 0; \
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002870 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
2871 __ret = __skb_checksum_validate_complete(skb, \
2872 complete, compute_pseudo(skb, proto)); \
2873 __ret; \
2874})
2875
2876#define skb_checksum_init(skb, proto, compute_pseudo) \
2877 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
2878
2879#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
2880 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
2881
2882#define skb_checksum_validate(skb, proto, compute_pseudo) \
2883 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
2884
2885#define skb_checksum_validate_zero_check(skb, proto, check, \
2886 compute_pseudo) \
2887 __skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
2888
2889#define skb_checksum_simple_validate(skb) \
2890 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
2891
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002892#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07002893void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002894static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2895{
2896 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002897 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002898}
2899static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2900{
2901 if (nfct)
2902 atomic_inc(&nfct->use);
2903}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002904#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002905#ifdef CONFIG_BRIDGE_NETFILTER
2906static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2907{
2908 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2909 kfree(nf_bridge);
2910}
2911static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2912{
2913 if (nf_bridge)
2914 atomic_inc(&nf_bridge->use);
2915}
2916#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08002917static inline void nf_reset(struct sk_buff *skb)
2918{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002919#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08002920 nf_conntrack_put(skb->nfct);
2921 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002922#endif
Patrick McHardya193a4a2006-03-20 19:23:05 -08002923#ifdef CONFIG_BRIDGE_NETFILTER
2924 nf_bridge_put(skb->nf_bridge);
2925 skb->nf_bridge = NULL;
2926#endif
2927}
2928
Patrick McHardy124dff02013-04-05 20:42:05 +02002929static inline void nf_reset_trace(struct sk_buff *skb)
2930{
Florian Westphal478b3602014-02-15 23:48:45 +01002931#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00002932 skb->nf_trace = 0;
2933#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002934}
2935
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002936/* Note: This doesn't put any conntrack and bridge info in dst. */
2937static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2938{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002939#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002940 dst->nfct = src->nfct;
2941 nf_conntrack_get(src->nfct);
2942 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002943#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002944#ifdef CONFIG_BRIDGE_NETFILTER
2945 dst->nf_bridge = src->nf_bridge;
2946 nf_bridge_get(src->nf_bridge);
2947#endif
Florian Westphal478b3602014-02-15 23:48:45 +01002948#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
2949 dst->nf_trace = src->nf_trace;
2950#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002951}
2952
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002953static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2954{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002955#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002956 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002957#endif
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002958#ifdef CONFIG_BRIDGE_NETFILTER
2959 nf_bridge_put(dst->nf_bridge);
2960#endif
2961 __nf_copy(dst, src);
2962}
2963
James Morris984bc162006-06-09 00:29:17 -07002964#ifdef CONFIG_NETWORK_SECMARK
2965static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2966{
2967 to->secmark = from->secmark;
2968}
2969
2970static inline void skb_init_secmark(struct sk_buff *skb)
2971{
2972 skb->secmark = 0;
2973}
2974#else
2975static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2976{ }
2977
2978static inline void skb_init_secmark(struct sk_buff *skb)
2979{ }
2980#endif
2981
Eric W. Biederman574f7192014-04-01 12:20:24 -07002982static inline bool skb_irq_freeable(const struct sk_buff *skb)
2983{
2984 return !skb->destructor &&
2985#if IS_ENABLED(CONFIG_XFRM)
2986 !skb->sp &&
2987#endif
2988#if IS_ENABLED(CONFIG_NF_CONNTRACK)
2989 !skb->nfct &&
2990#endif
2991 !skb->_skb_refdst &&
2992 !skb_has_frag_list(skb);
2993}
2994
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002995static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2996{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002997 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002998}
2999
Stephen Hemminger92477442009-03-21 13:39:26 -07003000static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003001{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003002 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003003}
3004
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003005static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3006{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003007 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003008}
3009
David S. Millerd5a9e242009-01-27 16:22:11 -08003010static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3011{
3012 skb->queue_mapping = rx_queue + 1;
3013}
3014
Stephen Hemminger92477442009-03-21 13:39:26 -07003015static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003016{
3017 return skb->queue_mapping - 1;
3018}
3019
Stephen Hemminger92477442009-03-21 13:39:26 -07003020static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003021{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003022 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08003023}
3024
Tom Herbert0e001612014-07-01 21:32:27 -07003025u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
Joe Perches7965bd42013-09-26 14:48:15 -07003026 unsigned int num_tx_queues);
Stephen Hemminger92477442009-03-21 13:39:26 -07003027
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003028static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3029{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003030#ifdef CONFIG_XFRM
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003031 return skb->sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003032#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003033 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003034#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003035}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003036
Pravin B Shelar68c33162013-02-14 14:02:41 +00003037/* Keeps track of mac header offset relative to skb->head.
3038 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3039 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07003040 * tunnel skb it points to outer mac header.
3041 * Keeps track of level of encapsulation of network headers.
3042 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00003043struct skb_gso_cb {
Eric Dumazet3347c962013-10-19 11:42:56 -07003044 int mac_offset;
3045 int encap_level;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003046 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00003047};
3048#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3049
3050static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3051{
3052 return (skb_mac_header(inner_skb) - inner_skb->head) -
3053 SKB_GSO_CB(inner_skb)->mac_offset;
3054}
3055
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00003056static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3057{
3058 int new_headroom, headroom;
3059 int ret;
3060
3061 headroom = skb_headroom(skb);
3062 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3063 if (ret)
3064 return ret;
3065
3066 new_headroom = skb_headroom(skb);
3067 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3068 return 0;
3069}
3070
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003071/* Compute the checksum for a gso segment. First compute the checksum value
3072 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3073 * then add in skb->csum (checksum from csum_start to end of packet).
3074 * skb->csum and csum_start are then updated to reflect the checksum of the
3075 * resultant packet starting from the transport header-- the resultant checksum
3076 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3077 * header.
3078 */
3079static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3080{
3081 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3082 skb_transport_offset(skb);
3083 __u16 csum;
3084
3085 csum = csum_fold(csum_partial(skb_transport_header(skb),
3086 plen, skb->csum));
3087 skb->csum = res;
3088 SKB_GSO_CB(skb)->csum_start -= plen;
3089
3090 return csum;
3091}
3092
David S. Millerbdcc0922012-03-07 20:53:36 -05003093static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07003094{
3095 return skb_shinfo(skb)->gso_size;
3096}
3097
Eric Dumazet36a8f392013-09-29 01:21:32 -07003098/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05003099static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02003100{
3101 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3102}
3103
Joe Perches7965bd42013-09-26 14:48:15 -07003104void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07003105
3106static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3107{
3108 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3109 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003110 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3111
Alexander Duyckb78462e2010-06-02 12:24:37 +00003112 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3113 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07003114 __skb_warn_lro_forwarding(skb);
3115 return true;
3116 }
3117 return false;
3118}
3119
Herbert Xu35fc92a2007-03-26 23:22:20 -07003120static inline void skb_forward_csum(struct sk_buff *skb)
3121{
3122 /* Unfortunately we don't support this one. Any brave souls? */
3123 if (skb->ip_summed == CHECKSUM_COMPLETE)
3124 skb->ip_summed = CHECKSUM_NONE;
3125}
3126
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003127/**
3128 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3129 * @skb: skb to check
3130 *
3131 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3132 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3133 * use this helper, to document places where we make this assertion.
3134 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003135static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003136{
3137#ifdef DEBUG
3138 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3139#endif
3140}
3141
Rusty Russellf35d9d82008-02-04 23:49:54 -05003142bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00003143
Paul Durranted1f50c2014-01-09 10:02:46 +00003144int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3145
Daniel Borkmannf77668d2013-03-19 06:39:30 +00003146u32 __skb_get_poff(const struct sk_buff *skb);
3147
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00003148/**
3149 * skb_head_is_locked - Determine if the skb->head is locked down
3150 * @skb: skb to check
3151 *
3152 * The head on skbs build around a head frag can be removed if they are
3153 * not cloned. This function returns true if the skb head is locked down
3154 * due to either being allocated via kmalloc, or by being a clone with
3155 * multiple references to the head.
3156 */
3157static inline bool skb_head_is_locked(const struct sk_buff *skb)
3158{
3159 return !skb->head_frag || skb_cloned(skb);
3160}
Florian Westphalfe6cc552014-02-13 23:09:12 +01003161
3162/**
3163 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3164 *
3165 * @skb: GSO skb
3166 *
3167 * skb_gso_network_seglen is used to determine the real size of the
3168 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3169 *
3170 * The MAC/L2 header is not accounted for.
3171 */
3172static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3173{
3174 unsigned int hdr_len = skb_transport_header(skb) -
3175 skb_network_header(skb);
3176 return hdr_len + skb_gso_transport_seglen(skb);
3177}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003178#endif /* __KERNEL__ */
3179#endif /* _LINUX_SKBUFF_H */