blob: 0e35b3af7317389b0093346f7f5334c0c4e86731 [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
232 /* generate software time stamp */
233 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,
Shirley Maa6686f22011-07-06 12:22:12 +0000250};
251
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400252#define SKBTX_ANY_SW_TSTAMP SKBTX_SW_TSTAMP
253#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
254
Shirley Maa6686f22011-07-06 12:22:12 +0000255/*
256 * The callback notifies userspace to release buffers when skb DMA is done in
257 * lower device, the skb last reference should be 0 when calling this.
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000258 * The zerocopy_success argument is true if zero copy transmit occurred,
259 * false on data copy or out of memory error caused by data copy attempt.
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000260 * The ctx field is used to track device context.
261 * The desc field is used to track userspace buffer index.
Shirley Maa6686f22011-07-06 12:22:12 +0000262 */
263struct ubuf_info {
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000264 void (*callback)(struct ubuf_info *, bool zerocopy_success);
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000265 void *ctx;
Shirley Maa6686f22011-07-06 12:22:12 +0000266 unsigned long desc;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000267};
268
Linus Torvalds1da177e2005-04-16 15:20:36 -0700269/* This data is invariant across clones and lives at
270 * the end of the header data, ie. at skb->end.
271 */
272struct skb_shared_info {
Ian Campbell9f42f122012-01-05 07:13:39 +0000273 unsigned char nr_frags;
274 __u8 tx_flags;
Herbert Xu79671682006-06-22 02:40:14 -0700275 unsigned short gso_size;
276 /* Warning: this field is not always filled in (UFO)! */
277 unsigned short gso_segs;
278 unsigned short gso_type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000280 struct skb_shared_hwtstamps hwtstamps;
Willem de Bruijn09c2d252014-08-04 22:11:47 -0400281 u32 tskey;
Ian Campbell9f42f122012-01-05 07:13:39 +0000282 __be32 ip6_frag_id;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700283
284 /*
285 * Warning : all fields before dataref are cleared in __alloc_skb()
286 */
287 atomic_t dataref;
288
Johann Baudy69e3c752009-05-18 22:11:22 -0700289 /* Intermediate layers must ensure that destructor_arg
290 * remains valid until skb destructor */
291 void * destructor_arg;
Shirley Maa6686f22011-07-06 12:22:12 +0000292
Eric Dumazetfed66382010-07-22 19:09:08 +0000293 /* must be last field, see pskb_expand_head() */
294 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700295};
296
297/* We divide dataref into two halves. The higher 16 bits hold references
298 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700299 * the entire skb->data. A clone of a headerless skb holds the length of
300 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700301 *
302 * All users must obey the rule that the skb->data reference count must be
303 * greater than or equal to the payload reference count.
304 *
305 * Holding a reference to the payload part means that the user does not
306 * care about modifications to the header part of skb->data.
307 */
308#define SKB_DATAREF_SHIFT 16
309#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
310
David S. Millerd179cd12005-08-17 14:57:30 -0700311
312enum {
313 SKB_FCLONE_UNAVAILABLE,
314 SKB_FCLONE_ORIG,
315 SKB_FCLONE_CLONE,
316};
317
Herbert Xu79671682006-06-22 02:40:14 -0700318enum {
319 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700320 SKB_GSO_UDP = 1 << 1,
Herbert Xu576a30e2006-06-27 13:22:38 -0700321
322 /* This indicates the skb is from an untrusted source. */
323 SKB_GSO_DODGY = 1 << 2,
Michael Chanb0da85372006-06-29 12:30:00 -0700324
325 /* This indicates the tcp segment has CWR set. */
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700326 SKB_GSO_TCP_ECN = 1 << 3,
327
328 SKB_GSO_TCPV6 = 1 << 4,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800329
330 SKB_GSO_FCOE = 1 << 5,
Pravin B Shelar68c33162013-02-14 14:02:41 +0000331
332 SKB_GSO_GRE = 1 << 6,
Pravin B Shelar73136262013-03-07 13:21:51 +0000333
Tom Herbert4b282522014-06-14 23:23:52 -0700334 SKB_GSO_GRE_CSUM = 1 << 7,
Simon Horman0d89d202013-05-23 21:02:52 +0000335
Tom Herbert4b282522014-06-14 23:23:52 -0700336 SKB_GSO_IPIP = 1 << 8,
Eric Dumazetcb32f512013-10-19 11:42:57 -0700337
Tom Herbert4b282522014-06-14 23:23:52 -0700338 SKB_GSO_SIT = 1 << 9,
Eric Dumazet61c1db72013-10-20 20:47:30 -0700339
Tom Herbert4b282522014-06-14 23:23:52 -0700340 SKB_GSO_UDP_TUNNEL = 1 << 10,
Tom Herbert0f4f4ff2014-06-04 17:20:16 -0700341
342 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
Tom Herbert4749c092014-06-04 17:20:23 -0700343
Tom Herbert4b282522014-06-14 23:23:52 -0700344 SKB_GSO_MPLS = 1 << 12,
345
Herbert Xu79671682006-06-22 02:40:14 -0700346};
347
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700348#if BITS_PER_LONG > 32
349#define NET_SKBUFF_DATA_USES_OFFSET 1
350#endif
351
352#ifdef NET_SKBUFF_DATA_USES_OFFSET
353typedef unsigned int sk_buff_data_t;
354#else
355typedef unsigned char *sk_buff_data_t;
356#endif
357
Eric Dumazet363ec39232014-02-26 14:02:11 -0800358/**
359 * struct skb_mstamp - multi resolution time stamps
360 * @stamp_us: timestamp in us resolution
361 * @stamp_jiffies: timestamp in jiffies
362 */
363struct skb_mstamp {
364 union {
365 u64 v64;
366 struct {
367 u32 stamp_us;
368 u32 stamp_jiffies;
369 };
370 };
371};
372
373/**
374 * skb_mstamp_get - get current timestamp
375 * @cl: place to store timestamps
376 */
377static inline void skb_mstamp_get(struct skb_mstamp *cl)
378{
379 u64 val = local_clock();
380
381 do_div(val, NSEC_PER_USEC);
382 cl->stamp_us = (u32)val;
383 cl->stamp_jiffies = (u32)jiffies;
384}
385
386/**
387 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
388 * @t1: pointer to newest sample
389 * @t0: pointer to oldest sample
390 */
391static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
392 const struct skb_mstamp *t0)
393{
394 s32 delta_us = t1->stamp_us - t0->stamp_us;
395 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
396
397 /* If delta_us is negative, this might be because interval is too big,
398 * or local_clock() drift is too big : fallback using jiffies.
399 */
400 if (delta_us <= 0 ||
401 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
402
403 delta_us = jiffies_to_usecs(delta_jiffies);
404
405 return delta_us;
406}
407
408
Linus Torvalds1da177e2005-04-16 15:20:36 -0700409/**
410 * struct sk_buff - socket buffer
411 * @next: Next buffer in list
412 * @prev: Previous buffer in list
Eric Dumazet363ec39232014-02-26 14:02:11 -0800413 * @tstamp: Time we arrived/left
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700414 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700416 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000417 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700418 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700419 * @len: Length of actual data
420 * @data_len: Data length
421 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700422 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700423 * @csum: Checksum (must include start/offset pair)
424 * @csum_start: Offset from skb->head where checksumming should start
425 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700426 * @priority: Packet queueing priority
WANG Cong60ff7462014-05-04 16:39:18 -0700427 * @ignore_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700428 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700429 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430 * @nohdr: Payload reference only, must not modify header
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700431 * @nfctinfo: Relationship of this skb to the connection
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700433 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700434 * @ipvs_property: skbuff is owned by ipvs
Randy Dunlap31729362008-02-18 20:52:13 -0800435 * @peeked: this packet has been seen already, so stats have been
436 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700437 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700438 * @protocol: Packet protocol from driver
439 * @destructor: Destruct function
440 * @nfct: Associated connection, if any
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800442 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 * @tc_index: Traffic control index
444 * @tc_verd: traffic control verdict
Tom Herbert61b905d2014-03-24 15:34:47 -0700445 * @hash: the packet hash
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700446 * @queue_mapping: Queue mapping for multiqueue devices
Randy Dunlap553a5672008-04-20 10:51:01 -0700447 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700448 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Tom Herbert61b905d2014-03-24 15:34:47 -0700449 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
Changli Gao4ca24622011-08-19 07:26:44 -0700450 * ports.
Tom Herberta3b18dd2014-07-01 21:33:17 -0700451 * @sw_hash: indicates hash was computed in software stack
Johannes Berg6e3e9392011-11-09 10:15:42 +0100452 * @wifi_acked_valid: wifi_acked was set
453 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000454 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Randy Dunlapf4b8ea72006-06-22 16:00:11 -0700455 * @dma_cookie: a cookie to one of several possible DMA operations
456 * done by skb DMA functions
Eliezer Tamir06021292013-06-10 11:39:50 +0300457 * @napi_id: id of the NAPI struct this skb came from
James Morris984bc162006-06-09 00:29:17 -0700458 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700459 * @mark: Generic packet mark
460 * @dropcount: total number of sk_receive_queue overflows
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000461 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b02008-07-14 22:49:06 -0700462 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000463 * @inner_protocol: Protocol (encapsulation)
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000464 * @inner_transport_header: Inner transport layer header (encapsulation)
465 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000466 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700467 * @transport_header: Transport layer header
468 * @network_header: Network layer header
469 * @mac_header: Link layer header
470 * @tail: Tail pointer
471 * @end: End pointer
472 * @head: Head of buffer
473 * @data: Data head pointer
474 * @truesize: Buffer size
475 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476 */
477
478struct sk_buff {
479 /* These two members must be first. */
480 struct sk_buff *next;
481 struct sk_buff *prev;
482
Eric Dumazet363ec39232014-02-26 14:02:11 -0800483 union {
484 ktime_t tstamp;
485 struct skb_mstamp skb_mstamp;
486 };
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000487
488 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700489 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491 /*
492 * This is the control buffer. It is free to use for every
493 * layer. Please put your private variables there. If you
494 * want to keep them across layers you have to do a skb_clone()
495 * first. This is owned by whoever has the skb queued ATM.
496 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000497 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700498
Eric Dumazet7fee2262010-05-11 23:19:48 +0000499 unsigned long _skb_refdst;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000500#ifdef CONFIG_XFRM
501 struct sec_path *sp;
502#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700504 data_len;
505 __u16 mac_len,
506 hdr_len;
Al Viroff1dcad2006-11-20 18:07:29 -0800507 union {
508 __wsum csum;
Herbert Xu663ead32007-04-09 11:59:07 -0700509 struct {
510 __u16 csum_start;
511 __u16 csum_offset;
512 };
Al Viroff1dcad2006-11-20 18:07:29 -0800513 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700514 __u32 priority;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200515 kmemcheck_bitfield_begin(flags1);
WANG Cong60ff7462014-05-04 16:39:18 -0700516 __u8 ignore_df:1,
Thomas Graf1cbb3382005-07-05 14:13:41 -0700517 cloned:1,
518 ip_summed:2,
Harald Welte6869c4d2005-08-09 19:24:19 -0700519 nohdr:1,
520 nfctinfo:3;
David S. Millerd179cd12005-08-17 14:57:30 -0700521 __u8 pkt_type:3,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800522 fclone:2,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700523 ipvs_property:1,
Herbert Xua59322b2007-12-05 01:53:40 -0800524 peeked:1,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700525 nf_trace:1;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200526 kmemcheck_bitfield_end(flags1);
Eric Dumazet4ab408d2010-03-01 03:09:26 +0000527 __be16 protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700528
529 void (*destructor)(struct sk_buff *skb);
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800530#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700531 struct nf_conntrack *nfct;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100532#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533#ifdef CONFIG_BRIDGE_NETFILTER
534 struct nf_bridge_info *nf_bridge;
535#endif
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700536
Eric Dumazet8964be42009-11-20 15:35:04 -0800537 int skb_iif;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000538
Tom Herbert61b905d2014-03-24 15:34:47 -0700539 __u32 hash;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000540
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000541 __be16 vlan_proto;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000542 __u16 vlan_tci;
543
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700545 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700547 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700548#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200550
Eric Dumazet0a148422011-04-20 09:27:32 +0000551 __u16 queue_mapping;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200552 kmemcheck_bitfield_begin(flags2);
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400553#ifdef CONFIG_IPV6_NDISC_NODETYPE
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000554 __u8 ndisc_nodetype:2;
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400555#endif
Mel Gormanc93bdd02012-07-31 16:44:19 -0700556 __u8 pfmemalloc:1;
Tom Herbert3853b582010-11-21 13:17:29 +0000557 __u8 ooo_okay:1;
Tom Herbert61b905d2014-03-24 15:34:47 -0700558 __u8 l4_hash:1;
Tom Herberta3b18dd2014-07-01 21:33:17 -0700559 __u8 sw_hash:1;
Johannes Berg6e3e9392011-11-09 10:15:42 +0100560 __u8 wifi_acked_valid:1;
561 __u8 wifi_acked:1;
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000562 __u8 no_fcs:1;
Eric Dumazetd3836f22012-04-27 00:33:38 +0000563 __u8 head_frag:1;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000564 /* Encapsulation protocol and NIC drivers should use
565 * this flag to indicate to each other if the skb contains
566 * encapsulated packet or not and maybe use the inner packet
567 * headers if needed
568 */
569 __u8 encapsulation:1;
Tom Herbert7e2b10c2014-06-04 17:20:02 -0700570 __u8 encap_hdr_csum:1;
Tom Herbert5d0c2b92014-06-10 18:54:13 -0700571 __u8 csum_valid:1;
Tom Herbert7e3cead2014-06-10 18:54:19 -0700572 __u8 csum_complete_sw:1;
Tom Herberta3b18dd2014-07-01 21:33:17 -0700573 /* 2/4 bit hole (depending on ndisc_nodetype presence) */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200574 kmemcheck_bitfield_end(flags2);
575
Cong Wange0d10952013-08-01 11:10:25 +0800576#if defined CONFIG_NET_DMA || defined CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir06021292013-06-10 11:39:50 +0300577 union {
578 unsigned int napi_id;
579 dma_cookie_t dma_cookie;
580 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700581#endif
James Morris984bc162006-06-09 00:29:17 -0700582#ifdef CONFIG_NETWORK_SECMARK
583 __u32 secmark;
584#endif
Neil Horman3b885782009-10-12 13:26:31 -0700585 union {
586 __u32 mark;
587 __u32 dropcount;
Eric Dumazet16fad692013-03-14 05:40:32 +0000588 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700589 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700590
Simon Horman0d89d202013-05-23 21:02:52 +0000591 __be16 inner_protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000592 __u16 inner_transport_header;
593 __u16 inner_network_header;
594 __u16 inner_mac_header;
595 __u16 transport_header;
596 __u16 network_header;
597 __u16 mac_header;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700598 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700599 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700600 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700602 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700603 unsigned int truesize;
604 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700605};
606
607#ifdef __KERNEL__
608/*
609 * Handling routines are only of interest to the kernel
610 */
611#include <linux/slab.h>
612
Linus Torvalds1da177e2005-04-16 15:20:36 -0700613
Mel Gormanc93bdd02012-07-31 16:44:19 -0700614#define SKB_ALLOC_FCLONE 0x01
615#define SKB_ALLOC_RX 0x02
616
617/* Returns true if the skb was allocated from PFMEMALLOC reserves */
618static inline bool skb_pfmemalloc(const struct sk_buff *skb)
619{
620 return unlikely(skb->pfmemalloc);
621}
622
Eric Dumazet7fee2262010-05-11 23:19:48 +0000623/*
624 * skb might have a dst pointer attached, refcounted or not.
625 * _skb_refdst low order bit is set if refcount was _not_ taken
626 */
627#define SKB_DST_NOREF 1UL
628#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
629
630/**
631 * skb_dst - returns skb dst_entry
632 * @skb: buffer
633 *
634 * Returns skb dst_entry, regardless of reference taken or not.
635 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000636static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
637{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000638 /* If refdst was not refcounted, check we still are in a
639 * rcu_read_lock section
640 */
641 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
642 !rcu_read_lock_held() &&
643 !rcu_read_lock_bh_held());
644 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000645}
646
Eric Dumazet7fee2262010-05-11 23:19:48 +0000647/**
648 * skb_dst_set - sets skb dst
649 * @skb: buffer
650 * @dst: dst entry
651 *
652 * Sets skb dst, assuming a reference was taken on dst and should
653 * be released by skb_dst_drop()
654 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000655static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
656{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000657 skb->_skb_refdst = (unsigned long)dst;
658}
659
Joe Perches7965bd42013-09-26 14:48:15 -0700660void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
661 bool force);
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200662
663/**
664 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
665 * @skb: buffer
666 * @dst: dst entry
667 *
668 * Sets skb dst, assuming a reference was not taken on dst.
669 * If dst entry is cached, we do not take reference and dst_release
670 * will be avoided by refdst_drop. If dst entry is not cached, we take
671 * reference, so that last dst_release can destroy the dst immediately.
672 */
673static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
674{
675 __skb_dst_set_noref(skb, dst, false);
676}
677
678/**
679 * skb_dst_set_noref_force - sets skb dst, without taking reference
680 * @skb: buffer
681 * @dst: dst entry
682 *
683 * Sets skb dst, assuming a reference was not taken on dst.
684 * No reference is taken and no dst_release will be called. While for
685 * cached dsts deferred reclaim is a basic feature, for entries that are
686 * not cached it is caller's job to guarantee that last dst_release for
687 * provided dst happens when nobody uses it, eg. after a RCU grace period.
688 */
689static inline void skb_dst_set_noref_force(struct sk_buff *skb,
690 struct dst_entry *dst)
691{
692 __skb_dst_set_noref(skb, dst, true);
693}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000694
695/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300696 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000697 * @skb: buffer
698 */
699static inline bool skb_dst_is_noref(const struct sk_buff *skb)
700{
701 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000702}
703
Eric Dumazet511c3f92009-06-02 05:14:27 +0000704static inline struct rtable *skb_rtable(const struct sk_buff *skb)
705{
Eric Dumazetadf30902009-06-02 05:19:30 +0000706 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000707}
708
Joe Perches7965bd42013-09-26 14:48:15 -0700709void kfree_skb(struct sk_buff *skb);
710void kfree_skb_list(struct sk_buff *segs);
711void skb_tx_error(struct sk_buff *skb);
712void consume_skb(struct sk_buff *skb);
713void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000714extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000715
Joe Perches7965bd42013-09-26 14:48:15 -0700716void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
717bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
718 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000719
Joe Perches7965bd42013-09-26 14:48:15 -0700720struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
721 int node);
722struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700723static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100724 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700725{
Eric Dumazet564824b2010-10-11 19:05:25 +0000726 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700727}
728
729static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100730 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700731{
Mel Gormanc93bdd02012-07-31 16:44:19 -0700732 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700733}
734
Joe Perches7965bd42013-09-26 14:48:15 -0700735struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
Patrick McHardy0ebd0ac2013-04-17 06:46:58 +0000736static inline struct sk_buff *alloc_skb_head(gfp_t priority)
737{
738 return __alloc_skb_head(priority, -1);
739}
740
Joe Perches7965bd42013-09-26 14:48:15 -0700741struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
742int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
743struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
744struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
Octavian Purdilabad93e92014-06-12 01:36:26 +0300745struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
746 gfp_t gfp_mask, bool fclone);
747static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
748 gfp_t gfp_mask)
749{
750 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
751}
Eric Dumazet117632e2011-12-03 21:39:53 +0000752
Joe Perches7965bd42013-09-26 14:48:15 -0700753int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
754struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
755 unsigned int headroom);
756struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
757 int newtailroom, gfp_t priority);
Fan Du25a91d82014-01-18 09:54:23 +0800758int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
759 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -0700760int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
761 int len);
762int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
763int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000764#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765
Joe Perches7965bd42013-09-26 14:48:15 -0700766int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
767 int getfrag(void *from, char *to, int offset,
768 int len, int odd, struct sk_buff *skb),
769 void *from, int length);
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700770
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800771struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700772 __u32 lower_offset;
773 __u32 upper_offset;
774 __u32 frag_idx;
775 __u32 stepped_offset;
776 struct sk_buff *root_skb;
777 struct sk_buff *cur_skb;
778 __u8 *frag_data;
779};
780
Joe Perches7965bd42013-09-26 14:48:15 -0700781void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
782 unsigned int to, struct skb_seq_state *st);
783unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
784 struct skb_seq_state *st);
785void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -0700786
Joe Perches7965bd42013-09-26 14:48:15 -0700787unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
788 unsigned int to, struct ts_config *config,
789 struct ts_state *state);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700790
Tom Herbert09323cc2013-12-15 22:16:19 -0800791/*
792 * Packet hash types specify the type of hash in skb_set_hash.
793 *
794 * Hash types refer to the protocol layer addresses which are used to
795 * construct a packet's hash. The hashes are used to differentiate or identify
796 * flows of the protocol layer for the hash type. Hash types are either
797 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
798 *
799 * Properties of hashes:
800 *
801 * 1) Two packets in different flows have different hash values
802 * 2) Two packets in the same flow should have the same hash value
803 *
804 * A hash at a higher layer is considered to be more specific. A driver should
805 * set the most specific hash possible.
806 *
807 * A driver cannot indicate a more specific hash than the layer at which a hash
808 * was computed. For instance an L3 hash cannot be set as an L4 hash.
809 *
810 * A driver may indicate a hash level which is less specific than the
811 * actual layer the hash was computed on. For instance, a hash computed
812 * at L4 may be considered an L3 hash. This should only be done if the
813 * driver can't unambiguously determine that the HW computed the hash at
814 * the higher layer. Note that the "should" in the second property above
815 * permits this.
816 */
817enum pkt_hash_types {
818 PKT_HASH_TYPE_NONE, /* Undefined type */
819 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
820 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
821 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
822};
823
824static inline void
825skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
826{
Tom Herbert61b905d2014-03-24 15:34:47 -0700827 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
Tom Herberta3b18dd2014-07-01 21:33:17 -0700828 skb->sw_hash = 0;
Tom Herbert61b905d2014-03-24 15:34:47 -0700829 skb->hash = hash;
Tom Herbert09323cc2013-12-15 22:16:19 -0800830}
831
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800832void __skb_get_hash(struct sk_buff *skb);
833static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000834{
Tom Herberta3b18dd2014-07-01 21:33:17 -0700835 if (!skb->l4_hash && !skb->sw_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800836 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000837
Tom Herbert61b905d2014-03-24 15:34:47 -0700838 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000839}
840
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800841static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
842{
Tom Herbert61b905d2014-03-24 15:34:47 -0700843 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800844}
845
Tom Herbert7539fad2013-12-15 22:12:18 -0800846static inline void skb_clear_hash(struct sk_buff *skb)
847{
Tom Herbert61b905d2014-03-24 15:34:47 -0700848 skb->hash = 0;
Tom Herberta3b18dd2014-07-01 21:33:17 -0700849 skb->sw_hash = 0;
Tom Herbert61b905d2014-03-24 15:34:47 -0700850 skb->l4_hash = 0;
Tom Herbert7539fad2013-12-15 22:12:18 -0800851}
852
853static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
854{
Tom Herbert61b905d2014-03-24 15:34:47 -0700855 if (!skb->l4_hash)
Tom Herbert7539fad2013-12-15 22:12:18 -0800856 skb_clear_hash(skb);
857}
858
Tom Herbert3df7a742013-12-15 22:16:29 -0800859static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
860{
Tom Herbert61b905d2014-03-24 15:34:47 -0700861 to->hash = from->hash;
Tom Herberta3b18dd2014-07-01 21:33:17 -0700862 to->sw_hash = from->sw_hash;
Tom Herbert61b905d2014-03-24 15:34:47 -0700863 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -0800864};
865
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700866#ifdef NET_SKBUFF_DATA_USES_OFFSET
867static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
868{
869 return skb->head + skb->end;
870}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000871
872static inline unsigned int skb_end_offset(const struct sk_buff *skb)
873{
874 return skb->end;
875}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700876#else
877static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
878{
879 return skb->end;
880}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000881
882static inline unsigned int skb_end_offset(const struct sk_buff *skb)
883{
884 return skb->end - skb->head;
885}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700886#endif
887
Linus Torvalds1da177e2005-04-16 15:20:36 -0700888/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700889#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890
Patrick Ohlyac45f602009-02-12 05:03:37 +0000891static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
892{
893 return &skb_shinfo(skb)->hwtstamps;
894}
895
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896/**
897 * skb_queue_empty - check if a queue is empty
898 * @list: queue head
899 *
900 * Returns true if the queue is empty, false otherwise.
901 */
902static inline int skb_queue_empty(const struct sk_buff_head *list)
903{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100904 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700905}
906
907/**
David S. Millerfc7ebb22008-09-23 00:34:07 -0700908 * skb_queue_is_last - check if skb is the last entry in the queue
909 * @list: queue head
910 * @skb: buffer
911 *
912 * Returns true if @skb is the last buffer on the list.
913 */
914static inline bool skb_queue_is_last(const struct sk_buff_head *list,
915 const struct sk_buff *skb)
916{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100917 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -0700918}
919
920/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800921 * skb_queue_is_first - check if skb is the first entry in the queue
922 * @list: queue head
923 * @skb: buffer
924 *
925 * Returns true if @skb is the first buffer on the list.
926 */
927static inline bool skb_queue_is_first(const struct sk_buff_head *list,
928 const struct sk_buff *skb)
929{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100930 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800931}
932
933/**
David S. Miller249c8b42008-09-23 00:44:42 -0700934 * skb_queue_next - return the next packet in the queue
935 * @list: queue head
936 * @skb: current buffer
937 *
938 * Return the next packet in @list after @skb. It is only valid to
939 * call this if skb_queue_is_last() evaluates to false.
940 */
941static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
942 const struct sk_buff *skb)
943{
944 /* This BUG_ON may seem severe, but if we just return then we
945 * are going to dereference garbage.
946 */
947 BUG_ON(skb_queue_is_last(list, skb));
948 return skb->next;
949}
950
951/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800952 * skb_queue_prev - return the prev packet in the queue
953 * @list: queue head
954 * @skb: current buffer
955 *
956 * Return the prev packet in @list before @skb. It is only valid to
957 * call this if skb_queue_is_first() evaluates to false.
958 */
959static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
960 const struct sk_buff *skb)
961{
962 /* This BUG_ON may seem severe, but if we just return then we
963 * are going to dereference garbage.
964 */
965 BUG_ON(skb_queue_is_first(list, skb));
966 return skb->prev;
967}
968
969/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700970 * skb_get - reference buffer
971 * @skb: buffer to reference
972 *
973 * Makes another reference to a socket buffer and returns a pointer
974 * to the buffer.
975 */
976static inline struct sk_buff *skb_get(struct sk_buff *skb)
977{
978 atomic_inc(&skb->users);
979 return skb;
980}
981
982/*
983 * If users == 1, we are the only owner and are can avoid redundant
984 * atomic change.
985 */
986
987/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700988 * skb_cloned - is the buffer a clone
989 * @skb: buffer to check
990 *
991 * Returns true if the buffer was generated with skb_clone() and is
992 * one of multiple shared copies of the buffer. Cloned buffers are
993 * shared data so must not be written to under normal circumstances.
994 */
995static inline int skb_cloned(const struct sk_buff *skb)
996{
997 return skb->cloned &&
998 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
999}
1000
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001001static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1002{
1003 might_sleep_if(pri & __GFP_WAIT);
1004
1005 if (skb_cloned(skb))
1006 return pskb_expand_head(skb, 0, 0, pri);
1007
1008 return 0;
1009}
1010
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011/**
1012 * skb_header_cloned - is the header a clone
1013 * @skb: buffer to check
1014 *
1015 * Returns true if modifying the header part of the buffer requires
1016 * the data to be copied.
1017 */
1018static inline int skb_header_cloned(const struct sk_buff *skb)
1019{
1020 int dataref;
1021
1022 if (!skb->cloned)
1023 return 0;
1024
1025 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1026 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1027 return dataref != 1;
1028}
1029
1030/**
1031 * skb_header_release - release reference to header
1032 * @skb: buffer to operate on
1033 *
1034 * Drop a reference to the header part of the buffer. This is done
1035 * by acquiring a payload reference. You must not read from the header
1036 * part of skb->data after this.
1037 */
1038static inline void skb_header_release(struct sk_buff *skb)
1039{
1040 BUG_ON(skb->nohdr);
1041 skb->nohdr = 1;
1042 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1043}
1044
1045/**
1046 * skb_shared - is the buffer shared
1047 * @skb: buffer to check
1048 *
1049 * Returns true if more than one person has a reference to this
1050 * buffer.
1051 */
1052static inline int skb_shared(const struct sk_buff *skb)
1053{
1054 return atomic_read(&skb->users) != 1;
1055}
1056
1057/**
1058 * skb_share_check - check if buffer is shared and if so clone it
1059 * @skb: buffer to check
1060 * @pri: priority for memory allocation
1061 *
1062 * If the buffer is shared the buffer is cloned and the old copy
1063 * drops a reference. A new clone with a single reference is returned.
1064 * If the buffer is not shared the original buffer is returned. When
1065 * being called from interrupt status or with spinlocks held pri must
1066 * be GFP_ATOMIC.
1067 *
1068 * NULL is returned on a memory allocation failure.
1069 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001070static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001071{
1072 might_sleep_if(pri & __GFP_WAIT);
1073 if (skb_shared(skb)) {
1074 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001075
1076 if (likely(nskb))
1077 consume_skb(skb);
1078 else
1079 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001080 skb = nskb;
1081 }
1082 return skb;
1083}
1084
1085/*
1086 * Copy shared buffers into a new sk_buff. We effectively do COW on
1087 * packets to handle cases where we have a local reader and forward
1088 * and a couple of other messy ones. The normal one is tcpdumping
1089 * a packet thats being forwarded.
1090 */
1091
1092/**
1093 * skb_unshare - make a copy of a shared buffer
1094 * @skb: buffer to check
1095 * @pri: priority for memory allocation
1096 *
1097 * If the socket buffer is a clone then this function creates a new
1098 * copy of the data, drops a reference count on the old copy and returns
1099 * the new copy with the reference count at 1. If the buffer is not a clone
1100 * the original buffer is returned. When called with a spinlock held or
1101 * from interrupt state @pri must be %GFP_ATOMIC
1102 *
1103 * %NULL is returned on a memory allocation failure.
1104 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001105static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001106 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107{
1108 might_sleep_if(pri & __GFP_WAIT);
1109 if (skb_cloned(skb)) {
1110 struct sk_buff *nskb = skb_copy(skb, pri);
1111 kfree_skb(skb); /* Free our shared copy */
1112 skb = nskb;
1113 }
1114 return skb;
1115}
1116
1117/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001118 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119 * @list_: list to peek at
1120 *
1121 * Peek an &sk_buff. Unlike most other operations you _MUST_
1122 * be careful with this one. A peek leaves the buffer on the
1123 * list and someone else may run off with it. You must hold
1124 * the appropriate locks or have a private queue to do this.
1125 *
1126 * Returns %NULL for an empty list or a pointer to the head element.
1127 * The reference count is not incremented and the reference is therefore
1128 * volatile. Use with caution.
1129 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001130static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001131{
Eric Dumazet18d07002012-04-30 16:31:46 +00001132 struct sk_buff *skb = list_->next;
1133
1134 if (skb == (struct sk_buff *)list_)
1135 skb = NULL;
1136 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001137}
1138
1139/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001140 * skb_peek_next - peek skb following the given one from a queue
1141 * @skb: skb to start from
1142 * @list_: list to peek at
1143 *
1144 * Returns %NULL when the end of the list is met or a pointer to the
1145 * next element. The reference count is not incremented and the
1146 * reference is therefore volatile. Use with caution.
1147 */
1148static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1149 const struct sk_buff_head *list_)
1150{
1151 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001152
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001153 if (next == (struct sk_buff *)list_)
1154 next = NULL;
1155 return next;
1156}
1157
1158/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001159 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001160 * @list_: list to peek at
1161 *
1162 * Peek an &sk_buff. Unlike most other operations you _MUST_
1163 * be careful with this one. A peek leaves the buffer on the
1164 * list and someone else may run off with it. You must hold
1165 * the appropriate locks or have a private queue to do this.
1166 *
1167 * Returns %NULL for an empty list or a pointer to the tail element.
1168 * The reference count is not incremented and the reference is therefore
1169 * volatile. Use with caution.
1170 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001171static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172{
Eric Dumazet18d07002012-04-30 16:31:46 +00001173 struct sk_buff *skb = list_->prev;
1174
1175 if (skb == (struct sk_buff *)list_)
1176 skb = NULL;
1177 return skb;
1178
Linus Torvalds1da177e2005-04-16 15:20:36 -07001179}
1180
1181/**
1182 * skb_queue_len - get queue length
1183 * @list_: list to measure
1184 *
1185 * Return the length of an &sk_buff queue.
1186 */
1187static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1188{
1189 return list_->qlen;
1190}
1191
David S. Miller67fed452008-09-21 22:36:24 -07001192/**
1193 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1194 * @list: queue to initialize
1195 *
1196 * This initializes only the list and queue length aspects of
1197 * an sk_buff_head object. This allows to initialize the list
1198 * aspects of an sk_buff_head without reinitializing things like
1199 * the spinlock. It can also be used for on-stack sk_buff_head
1200 * objects where the spinlock is known to not be used.
1201 */
1202static inline void __skb_queue_head_init(struct sk_buff_head *list)
1203{
1204 list->prev = list->next = (struct sk_buff *)list;
1205 list->qlen = 0;
1206}
1207
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001208/*
1209 * This function creates a split out lock class for each invocation;
1210 * this is needed for now since a whole lot of users of the skb-queue
1211 * infrastructure in drivers have different locking usage (in hardirq)
1212 * than the networking core (in softirq only). In the long run either the
1213 * network layer or drivers should need annotation to consolidate the
1214 * main types of usage into 3 classes.
1215 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001216static inline void skb_queue_head_init(struct sk_buff_head *list)
1217{
1218 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001219 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001220}
1221
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001222static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1223 struct lock_class_key *class)
1224{
1225 skb_queue_head_init(list);
1226 lockdep_set_class(&list->lock, class);
1227}
1228
Linus Torvalds1da177e2005-04-16 15:20:36 -07001229/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001230 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001231 *
1232 * The "__skb_xxxx()" functions are the non-atomic ones that
1233 * can only be called with interrupts disabled.
1234 */
Joe Perches7965bd42013-09-26 14:48:15 -07001235void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1236 struct sk_buff_head *list);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001237static inline void __skb_insert(struct sk_buff *newsk,
1238 struct sk_buff *prev, struct sk_buff *next,
1239 struct sk_buff_head *list)
1240{
1241 newsk->next = next;
1242 newsk->prev = prev;
1243 next->prev = prev->next = newsk;
1244 list->qlen++;
1245}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001246
David S. Miller67fed452008-09-21 22:36:24 -07001247static inline void __skb_queue_splice(const struct sk_buff_head *list,
1248 struct sk_buff *prev,
1249 struct sk_buff *next)
1250{
1251 struct sk_buff *first = list->next;
1252 struct sk_buff *last = list->prev;
1253
1254 first->prev = prev;
1255 prev->next = first;
1256
1257 last->next = next;
1258 next->prev = last;
1259}
1260
1261/**
1262 * skb_queue_splice - join two skb lists, this is designed for stacks
1263 * @list: the new list to add
1264 * @head: the place to add it in the first list
1265 */
1266static inline void skb_queue_splice(const struct sk_buff_head *list,
1267 struct sk_buff_head *head)
1268{
1269 if (!skb_queue_empty(list)) {
1270 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001271 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001272 }
1273}
1274
1275/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001276 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001277 * @list: the new list to add
1278 * @head: the place to add it in the first list
1279 *
1280 * The list at @list is reinitialised
1281 */
1282static inline void skb_queue_splice_init(struct sk_buff_head *list,
1283 struct sk_buff_head *head)
1284{
1285 if (!skb_queue_empty(list)) {
1286 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001287 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001288 __skb_queue_head_init(list);
1289 }
1290}
1291
1292/**
1293 * skb_queue_splice_tail - join two skb lists, each list being a queue
1294 * @list: the new list to add
1295 * @head: the place to add it in the first list
1296 */
1297static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1298 struct sk_buff_head *head)
1299{
1300 if (!skb_queue_empty(list)) {
1301 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001302 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001303 }
1304}
1305
1306/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001307 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001308 * @list: the new list to add
1309 * @head: the place to add it in the first list
1310 *
1311 * Each of the lists is a queue.
1312 * The list at @list is reinitialised
1313 */
1314static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1315 struct sk_buff_head *head)
1316{
1317 if (!skb_queue_empty(list)) {
1318 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001319 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001320 __skb_queue_head_init(list);
1321 }
1322}
1323
Linus Torvalds1da177e2005-04-16 15:20:36 -07001324/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001325 * __skb_queue_after - queue a buffer at the list head
1326 * @list: list to use
1327 * @prev: place after this buffer
1328 * @newsk: buffer to queue
1329 *
1330 * Queue a buffer int the middle of a list. This function takes no locks
1331 * and you must therefore hold required locks before calling it.
1332 *
1333 * A buffer cannot be placed on two lists at the same time.
1334 */
1335static inline void __skb_queue_after(struct sk_buff_head *list,
1336 struct sk_buff *prev,
1337 struct sk_buff *newsk)
1338{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001339 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001340}
1341
Joe Perches7965bd42013-09-26 14:48:15 -07001342void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1343 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001344
Gerrit Renkerf5572852008-04-14 00:05:28 -07001345static inline void __skb_queue_before(struct sk_buff_head *list,
1346 struct sk_buff *next,
1347 struct sk_buff *newsk)
1348{
1349 __skb_insert(newsk, next->prev, next, list);
1350}
1351
Stephen Hemminger300ce172005-10-30 13:47:34 -08001352/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001353 * __skb_queue_head - queue a buffer at the list head
1354 * @list: list to use
1355 * @newsk: buffer to queue
1356 *
1357 * Queue a buffer at the start of a list. This function takes no locks
1358 * and you must therefore hold required locks before calling it.
1359 *
1360 * A buffer cannot be placed on two lists at the same time.
1361 */
Joe Perches7965bd42013-09-26 14:48:15 -07001362void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363static inline void __skb_queue_head(struct sk_buff_head *list,
1364 struct sk_buff *newsk)
1365{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001366 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001367}
1368
1369/**
1370 * __skb_queue_tail - queue a buffer at the list tail
1371 * @list: list to use
1372 * @newsk: buffer to queue
1373 *
1374 * Queue a buffer at the end of a list. This function takes no locks
1375 * and you must therefore hold required locks before calling it.
1376 *
1377 * A buffer cannot be placed on two lists at the same time.
1378 */
Joe Perches7965bd42013-09-26 14:48:15 -07001379void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001380static inline void __skb_queue_tail(struct sk_buff_head *list,
1381 struct sk_buff *newsk)
1382{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001383 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001384}
1385
Linus Torvalds1da177e2005-04-16 15:20:36 -07001386/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001387 * remove sk_buff from list. _Must_ be called atomically, and with
1388 * the list known..
1389 */
Joe Perches7965bd42013-09-26 14:48:15 -07001390void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001391static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1392{
1393 struct sk_buff *next, *prev;
1394
1395 list->qlen--;
1396 next = skb->next;
1397 prev = skb->prev;
1398 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001399 next->prev = prev;
1400 prev->next = next;
1401}
1402
Gerrit Renkerf525c062008-04-14 00:04:12 -07001403/**
1404 * __skb_dequeue - remove from the head of the queue
1405 * @list: list to dequeue from
1406 *
1407 * Remove the head of the list. This function does not take any locks
1408 * so must be used with appropriate locks held only. The head item is
1409 * returned or %NULL if the list is empty.
1410 */
Joe Perches7965bd42013-09-26 14:48:15 -07001411struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Gerrit Renkerf525c062008-04-14 00:04:12 -07001412static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1413{
1414 struct sk_buff *skb = skb_peek(list);
1415 if (skb)
1416 __skb_unlink(skb, list);
1417 return skb;
1418}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001419
1420/**
1421 * __skb_dequeue_tail - remove from the tail of the queue
1422 * @list: list to dequeue from
1423 *
1424 * Remove the tail of the list. This function does not take any locks
1425 * so must be used with appropriate locks held only. The tail item is
1426 * returned or %NULL if the list is empty.
1427 */
Joe Perches7965bd42013-09-26 14:48:15 -07001428struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001429static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1430{
1431 struct sk_buff *skb = skb_peek_tail(list);
1432 if (skb)
1433 __skb_unlink(skb, list);
1434 return skb;
1435}
1436
1437
David S. Millerbdcc0922012-03-07 20:53:36 -05001438static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001439{
1440 return skb->data_len;
1441}
1442
1443static inline unsigned int skb_headlen(const struct sk_buff *skb)
1444{
1445 return skb->len - skb->data_len;
1446}
1447
1448static inline int skb_pagelen(const struct sk_buff *skb)
1449{
1450 int i, len = 0;
1451
1452 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001453 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001454 return len + skb_headlen(skb);
1455}
1456
Ian Campbell131ea662011-08-19 06:25:00 +00001457/**
1458 * __skb_fill_page_desc - initialise a paged fragment in an skb
1459 * @skb: buffer containing fragment to be initialised
1460 * @i: paged fragment index to initialise
1461 * @page: the page to use for this fragment
1462 * @off: the offset to the data with @page
1463 * @size: the length of the data
1464 *
1465 * Initialises the @i'th fragment of @skb to point to &size bytes at
1466 * offset @off within @page.
1467 *
1468 * Does not take any additional reference on the fragment.
1469 */
1470static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1471 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472{
1473 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1474
Mel Gormanc48a11c2012-07-31 16:44:23 -07001475 /*
1476 * Propagate page->pfmemalloc to the skb if we can. The problem is
1477 * that not all callers have unique ownership of the page. If
1478 * pfmemalloc is set, we check the mapping as a mapping implies
1479 * page->index is set (index and pfmemalloc share space).
1480 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1481 * do not lose pfmemalloc information as the pages would not be
1482 * allocated using __GFP_MEMALLOC.
1483 */
Ian Campbella8605c62011-10-19 23:01:49 +00001484 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001485 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001486 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001487
1488 page = compound_head(page);
1489 if (page->pfmemalloc && !page->mapping)
1490 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00001491}
1492
1493/**
1494 * skb_fill_page_desc - initialise a paged fragment in an skb
1495 * @skb: buffer containing fragment to be initialised
1496 * @i: paged fragment index to initialise
1497 * @page: the page to use for this fragment
1498 * @off: the offset to the data with @page
1499 * @size: the length of the data
1500 *
1501 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01001502 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00001503 * addition updates @skb such that @i is the last fragment.
1504 *
1505 * Does not take any additional reference on the fragment.
1506 */
1507static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1508 struct page *page, int off, int size)
1509{
1510 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001511 skb_shinfo(skb)->nr_frags = i + 1;
1512}
1513
Joe Perches7965bd42013-09-26 14:48:15 -07001514void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1515 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001516
Jason Wangf8e617e2013-11-01 14:07:47 +08001517void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1518 unsigned int truesize);
1519
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001521#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001522#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1523
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001524#ifdef NET_SKBUFF_DATA_USES_OFFSET
1525static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1526{
1527 return skb->head + skb->tail;
1528}
1529
1530static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1531{
1532 skb->tail = skb->data - skb->head;
1533}
1534
1535static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1536{
1537 skb_reset_tail_pointer(skb);
1538 skb->tail += offset;
1539}
Simon Horman7cc46192013-05-28 20:34:29 +00001540
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001541#else /* NET_SKBUFF_DATA_USES_OFFSET */
1542static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1543{
1544 return skb->tail;
1545}
1546
1547static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1548{
1549 skb->tail = skb->data;
1550}
1551
1552static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1553{
1554 skb->tail = skb->data + offset;
1555}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001556
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001557#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1558
Linus Torvalds1da177e2005-04-16 15:20:36 -07001559/*
1560 * Add data to an sk_buff
1561 */
Mathias Krause0c7ddf32013-11-07 14:18:24 +01001562unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001563unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001564static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1565{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001566 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567 SKB_LINEAR_ASSERT(skb);
1568 skb->tail += len;
1569 skb->len += len;
1570 return tmp;
1571}
1572
Joe Perches7965bd42013-09-26 14:48:15 -07001573unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1575{
1576 skb->data -= len;
1577 skb->len += len;
1578 return skb->data;
1579}
1580
Joe Perches7965bd42013-09-26 14:48:15 -07001581unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1583{
1584 skb->len -= len;
1585 BUG_ON(skb->len < skb->data_len);
1586 return skb->data += len;
1587}
1588
David S. Miller47d29642010-05-02 02:21:44 -07001589static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1590{
1591 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1592}
1593
Joe Perches7965bd42013-09-26 14:48:15 -07001594unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001595
1596static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1597{
1598 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001599 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001600 return NULL;
1601 skb->len -= len;
1602 return skb->data += len;
1603}
1604
1605static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1606{
1607 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1608}
1609
1610static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1611{
1612 if (likely(len <= skb_headlen(skb)))
1613 return 1;
1614 if (unlikely(len > skb->len))
1615 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001616 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001617}
1618
1619/**
1620 * skb_headroom - bytes at buffer head
1621 * @skb: buffer to check
1622 *
1623 * Return the number of bytes of free space at the head of an &sk_buff.
1624 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001625static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001626{
1627 return skb->data - skb->head;
1628}
1629
1630/**
1631 * skb_tailroom - bytes at buffer end
1632 * @skb: buffer to check
1633 *
1634 * Return the number of bytes of free space at the tail of an sk_buff
1635 */
1636static inline int skb_tailroom(const struct sk_buff *skb)
1637{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001638 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001639}
1640
1641/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00001642 * skb_availroom - bytes at buffer end
1643 * @skb: buffer to check
1644 *
1645 * Return the number of bytes of free space at the tail of an sk_buff
1646 * allocated by sk_stream_alloc()
1647 */
1648static inline int skb_availroom(const struct sk_buff *skb)
1649{
Eric Dumazet16fad692013-03-14 05:40:32 +00001650 if (skb_is_nonlinear(skb))
1651 return 0;
1652
1653 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00001654}
1655
1656/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001657 * skb_reserve - adjust headroom
1658 * @skb: buffer to alter
1659 * @len: bytes to move
1660 *
1661 * Increase the headroom of an empty &sk_buff by reducing the tail
1662 * room. This is only allowed for an empty buffer.
1663 */
David S. Miller8243126c2006-01-17 02:54:21 -08001664static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001665{
1666 skb->data += len;
1667 skb->tail += len;
1668}
1669
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001670static inline void skb_reset_inner_headers(struct sk_buff *skb)
1671{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001672 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001673 skb->inner_network_header = skb->network_header;
1674 skb->inner_transport_header = skb->transport_header;
1675}
1676
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00001677static inline void skb_reset_mac_len(struct sk_buff *skb)
1678{
1679 skb->mac_len = skb->network_header - skb->mac_header;
1680}
1681
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001682static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1683 *skb)
1684{
1685 return skb->head + skb->inner_transport_header;
1686}
1687
1688static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1689{
1690 skb->inner_transport_header = skb->data - skb->head;
1691}
1692
1693static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1694 const int offset)
1695{
1696 skb_reset_inner_transport_header(skb);
1697 skb->inner_transport_header += offset;
1698}
1699
1700static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1701{
1702 return skb->head + skb->inner_network_header;
1703}
1704
1705static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1706{
1707 skb->inner_network_header = skb->data - skb->head;
1708}
1709
1710static inline void skb_set_inner_network_header(struct sk_buff *skb,
1711 const int offset)
1712{
1713 skb_reset_inner_network_header(skb);
1714 skb->inner_network_header += offset;
1715}
1716
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001717static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1718{
1719 return skb->head + skb->inner_mac_header;
1720}
1721
1722static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1723{
1724 skb->inner_mac_header = skb->data - skb->head;
1725}
1726
1727static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1728 const int offset)
1729{
1730 skb_reset_inner_mac_header(skb);
1731 skb->inner_mac_header += offset;
1732}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001733static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1734{
Cong Wang35d04612013-05-29 15:16:05 +08001735 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001736}
1737
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001738static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1739{
1740 return skb->head + skb->transport_header;
1741}
1742
1743static inline void skb_reset_transport_header(struct sk_buff *skb)
1744{
1745 skb->transport_header = skb->data - skb->head;
1746}
1747
1748static inline void skb_set_transport_header(struct sk_buff *skb,
1749 const int offset)
1750{
1751 skb_reset_transport_header(skb);
1752 skb->transport_header += offset;
1753}
1754
1755static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1756{
1757 return skb->head + skb->network_header;
1758}
1759
1760static inline void skb_reset_network_header(struct sk_buff *skb)
1761{
1762 skb->network_header = skb->data - skb->head;
1763}
1764
1765static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1766{
1767 skb_reset_network_header(skb);
1768 skb->network_header += offset;
1769}
1770
1771static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1772{
1773 return skb->head + skb->mac_header;
1774}
1775
1776static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1777{
Cong Wang35d04612013-05-29 15:16:05 +08001778 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001779}
1780
1781static inline void skb_reset_mac_header(struct sk_buff *skb)
1782{
1783 skb->mac_header = skb->data - skb->head;
1784}
1785
1786static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1787{
1788 skb_reset_mac_header(skb);
1789 skb->mac_header += offset;
1790}
1791
Timo Teräs0e3da5b2013-12-16 11:02:09 +02001792static inline void skb_pop_mac_header(struct sk_buff *skb)
1793{
1794 skb->mac_header = skb->network_header;
1795}
1796
Ying Xuefbbdb8f2013-03-27 16:46:06 +00001797static inline void skb_probe_transport_header(struct sk_buff *skb,
1798 const int offset_hint)
1799{
1800 struct flow_keys keys;
1801
1802 if (skb_transport_header_was_set(skb))
1803 return;
1804 else if (skb_flow_dissect(skb, &keys))
1805 skb_set_transport_header(skb, keys.thoff);
1806 else
1807 skb_set_transport_header(skb, offset_hint);
1808}
1809
Eric Dumazet03606892012-02-23 10:55:02 +00001810static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1811{
1812 if (skb_mac_header_was_set(skb)) {
1813 const unsigned char *old_mac = skb_mac_header(skb);
1814
1815 skb_set_mac_header(skb, -skb->mac_len);
1816 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1817 }
1818}
1819
Michał Mirosław04fb4512010-12-14 15:24:08 +00001820static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1821{
1822 return skb->csum_start - skb_headroom(skb);
1823}
1824
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001825static inline int skb_transport_offset(const struct sk_buff *skb)
1826{
1827 return skb_transport_header(skb) - skb->data;
1828}
1829
1830static inline u32 skb_network_header_len(const struct sk_buff *skb)
1831{
1832 return skb->transport_header - skb->network_header;
1833}
1834
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001835static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1836{
1837 return skb->inner_transport_header - skb->inner_network_header;
1838}
1839
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001840static inline int skb_network_offset(const struct sk_buff *skb)
1841{
1842 return skb_network_header(skb) - skb->data;
1843}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001844
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001845static inline int skb_inner_network_offset(const struct sk_buff *skb)
1846{
1847 return skb_inner_network_header(skb) - skb->data;
1848}
1849
Changli Gaof9599ce2010-08-04 04:43:44 +00001850static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1851{
1852 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1853}
1854
Tom Herberte5eb4e32014-06-14 23:24:28 -07001855static inline void skb_pop_rcv_encapsulation(struct sk_buff *skb)
1856{
1857 /* Only continue with checksum unnecessary if device indicated
1858 * it is valid across encapsulation (skb->encapsulation was set).
1859 */
1860 if (skb->ip_summed == CHECKSUM_UNNECESSARY && !skb->encapsulation)
1861 skb->ip_summed = CHECKSUM_NONE;
1862
1863 skb->encapsulation = 0;
1864 skb->csum_valid = 0;
1865}
1866
Linus Torvalds1da177e2005-04-16 15:20:36 -07001867/*
1868 * CPUs often take a performance hit when accessing unaligned memory
1869 * locations. The actual performance hit varies, it can be small if the
1870 * hardware handles it or large if we have to take an exception and fix it
1871 * in software.
1872 *
1873 * Since an ethernet header is 14 bytes network drivers often end up with
1874 * the IP header at an unaligned offset. The IP header can be aligned by
1875 * shifting the start of the packet by 2 bytes. Drivers should do this
1876 * with:
1877 *
Tobias Klauser8660c122009-07-13 22:48:16 +00001878 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001879 *
1880 * The downside to this alignment of the IP header is that the DMA is now
1881 * unaligned. On some architectures the cost of an unaligned DMA is high
1882 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00001883 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001884 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1885 * to be overridden.
1886 */
1887#ifndef NET_IP_ALIGN
1888#define NET_IP_ALIGN 2
1889#endif
1890
Anton Blanchard025be812006-03-31 02:27:06 -08001891/*
1892 * The networking layer reserves some headroom in skb data (via
1893 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1894 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08001895 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08001896 *
1897 * Unfortunately this headroom changes the DMA alignment of the resulting
1898 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1899 * on some architectures. An architecture can override this value,
1900 * perhaps setting it to a cacheline in size (since that will maintain
1901 * cacheline alignment of the DMA). It must be a power of 2.
1902 *
David S. Millerd6301d32009-02-08 19:24:13 -08001903 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08001904 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07001905 *
1906 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1907 * to reduce average number of cache lines per packet.
1908 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07001909 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08001910 */
1911#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07001912#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08001913#endif
1914
Joe Perches7965bd42013-09-26 14:48:15 -07001915int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001916
1917static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1918{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00001919 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07001920 WARN_ON(1);
1921 return;
1922 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001923 skb->len = len;
1924 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001925}
1926
Joe Perches7965bd42013-09-26 14:48:15 -07001927void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001928
1929static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1930{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001931 if (skb->data_len)
1932 return ___pskb_trim(skb, len);
1933 __skb_trim(skb, len);
1934 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001935}
1936
1937static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1938{
1939 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1940}
1941
1942/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07001943 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1944 * @skb: buffer to alter
1945 * @len: new length
1946 *
1947 * This is identical to pskb_trim except that the caller knows that
1948 * the skb is not cloned so we should never get an error due to out-
1949 * of-memory.
1950 */
1951static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1952{
1953 int err = pskb_trim(skb, len);
1954 BUG_ON(err);
1955}
1956
1957/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001958 * skb_orphan - orphan a buffer
1959 * @skb: buffer to orphan
1960 *
1961 * If a buffer currently has an owner then we call the owner's
1962 * destructor function and make the @skb unowned. The buffer continues
1963 * to exist but is no longer charged to its former owner.
1964 */
1965static inline void skb_orphan(struct sk_buff *skb)
1966{
Eric Dumazetc34a7612013-07-30 16:11:15 -07001967 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001968 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07001969 skb->destructor = NULL;
1970 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07001971 } else {
1972 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07001973 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001974}
1975
1976/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00001977 * skb_orphan_frags - orphan the frags contained in a buffer
1978 * @skb: buffer to orphan frags from
1979 * @gfp_mask: allocation mask for replacement pages
1980 *
1981 * For each frag in the SKB which needs a destructor (i.e. has an
1982 * owner) create a copy of that frag and release the original
1983 * page by calling the destructor.
1984 */
1985static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1986{
1987 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1988 return 0;
1989 return skb_copy_ubufs(skb, gfp_mask);
1990}
1991
1992/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001993 * __skb_queue_purge - empty a list
1994 * @list: list to empty
1995 *
1996 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1997 * the list and one reference dropped. This function does not take the
1998 * list lock and the caller must hold the relevant locks to use it.
1999 */
Joe Perches7965bd42013-09-26 14:48:15 -07002000void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002001static inline void __skb_queue_purge(struct sk_buff_head *list)
2002{
2003 struct sk_buff *skb;
2004 while ((skb = __skb_dequeue(list)) != NULL)
2005 kfree_skb(skb);
2006}
2007
Alexander Duycke5e67302013-02-08 10:17:15 +00002008#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
2009#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
2010#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
2011
Joe Perches7965bd42013-09-26 14:48:15 -07002012void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002013
Joe Perches7965bd42013-09-26 14:48:15 -07002014struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2015 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07002016
2017/**
2018 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2019 * @dev: network device to receive on
2020 * @length: length to allocate
2021 *
2022 * Allocate a new &sk_buff and assign it a usage count of one. The
2023 * buffer has unspecified headroom built in. Users should allocate
2024 * the headroom they think they need without accounting for the
2025 * built in space. The built in space is used for optimisations.
2026 *
2027 * %NULL is returned if there is no free memory. Although this function
2028 * allocates memory it can be called from an interrupt.
2029 */
2030static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002031 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002032{
2033 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2034}
2035
Eric Dumazet6f532612012-05-18 05:12:12 +00002036/* legacy helper around __netdev_alloc_skb() */
2037static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2038 gfp_t gfp_mask)
2039{
2040 return __netdev_alloc_skb(NULL, length, gfp_mask);
2041}
2042
2043/* legacy helper around netdev_alloc_skb() */
2044static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2045{
2046 return netdev_alloc_skb(NULL, length);
2047}
2048
2049
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002050static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2051 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002052{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002053 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002054
2055 if (NET_IP_ALIGN && skb)
2056 skb_reserve(skb, NET_IP_ALIGN);
2057 return skb;
2058}
2059
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002060static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2061 unsigned int length)
2062{
2063 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2064}
2065
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002066/**
2067 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
Mel Gorman06140022012-07-31 16:44:24 -07002068 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2069 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2070 * @order: size of the allocation
2071 *
2072 * Allocate a new page.
2073 *
2074 * %NULL is returned if there is no free memory.
2075*/
2076static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
2077 struct sk_buff *skb,
2078 unsigned int order)
2079{
2080 struct page *page;
2081
2082 gfp_mask |= __GFP_COLD;
2083
2084 if (!(gfp_mask & __GFP_NOMEMALLOC))
2085 gfp_mask |= __GFP_MEMALLOC;
2086
2087 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2088 if (skb && page && page->pfmemalloc)
2089 skb->pfmemalloc = true;
2090
2091 return page;
2092}
2093
2094/**
2095 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2096 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2097 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2098 *
2099 * Allocate a new page.
2100 *
2101 * %NULL is returned if there is no free memory.
2102 */
2103static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
2104 struct sk_buff *skb)
2105{
2106 return __skb_alloc_pages(gfp_mask, skb, 0);
2107}
2108
2109/**
2110 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2111 * @page: The page that was allocated from skb_alloc_page
2112 * @skb: The skb that may need pfmemalloc set
2113 */
2114static inline void skb_propagate_pfmemalloc(struct page *page,
2115 struct sk_buff *skb)
2116{
2117 if (page && page->pfmemalloc)
2118 skb->pfmemalloc = true;
2119}
2120
Eric Dumazet564824b2010-10-11 19:05:25 +00002121/**
Masanari Iidae2278672014-02-18 22:54:36 +09002122 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002123 * @frag: the paged fragment
2124 *
2125 * Returns the &struct page associated with @frag.
2126 */
2127static inline struct page *skb_frag_page(const skb_frag_t *frag)
2128{
Ian Campbella8605c62011-10-19 23:01:49 +00002129 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00002130}
2131
2132/**
2133 * __skb_frag_ref - take an addition reference on a paged fragment.
2134 * @frag: the paged fragment
2135 *
2136 * Takes an additional reference on the paged fragment @frag.
2137 */
2138static inline void __skb_frag_ref(skb_frag_t *frag)
2139{
2140 get_page(skb_frag_page(frag));
2141}
2142
2143/**
2144 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2145 * @skb: the buffer
2146 * @f: the fragment offset.
2147 *
2148 * Takes an additional reference on the @f'th paged fragment of @skb.
2149 */
2150static inline void skb_frag_ref(struct sk_buff *skb, int f)
2151{
2152 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2153}
2154
2155/**
2156 * __skb_frag_unref - release a reference on a paged fragment.
2157 * @frag: the paged fragment
2158 *
2159 * Releases a reference on the paged fragment @frag.
2160 */
2161static inline void __skb_frag_unref(skb_frag_t *frag)
2162{
2163 put_page(skb_frag_page(frag));
2164}
2165
2166/**
2167 * skb_frag_unref - release a reference on a paged fragment of an skb.
2168 * @skb: the buffer
2169 * @f: the fragment offset
2170 *
2171 * Releases a reference on the @f'th paged fragment of @skb.
2172 */
2173static inline void skb_frag_unref(struct sk_buff *skb, int f)
2174{
2175 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2176}
2177
2178/**
2179 * skb_frag_address - gets the address of the data contained in a paged fragment
2180 * @frag: the paged fragment buffer
2181 *
2182 * Returns the address of the data within @frag. The page must already
2183 * be mapped.
2184 */
2185static inline void *skb_frag_address(const skb_frag_t *frag)
2186{
2187 return page_address(skb_frag_page(frag)) + frag->page_offset;
2188}
2189
2190/**
2191 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2192 * @frag: the paged fragment buffer
2193 *
2194 * Returns the address of the data within @frag. Checks that the page
2195 * is mapped and returns %NULL otherwise.
2196 */
2197static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2198{
2199 void *ptr = page_address(skb_frag_page(frag));
2200 if (unlikely(!ptr))
2201 return NULL;
2202
2203 return ptr + frag->page_offset;
2204}
2205
2206/**
2207 * __skb_frag_set_page - sets the page contained in a paged fragment
2208 * @frag: the paged fragment
2209 * @page: the page to set
2210 *
2211 * Sets the fragment @frag to contain @page.
2212 */
2213static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2214{
Ian Campbella8605c62011-10-19 23:01:49 +00002215 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002216}
2217
2218/**
2219 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2220 * @skb: the buffer
2221 * @f: the fragment offset
2222 * @page: the page to set
2223 *
2224 * Sets the @f'th fragment of @skb to contain @page.
2225 */
2226static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2227 struct page *page)
2228{
2229 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2230}
2231
Eric Dumazet400dfd32013-10-17 16:27:07 -07002232bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2233
Ian Campbell131ea662011-08-19 06:25:00 +00002234/**
2235 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002236 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002237 * @frag: the paged fragment to map
2238 * @offset: the offset within the fragment (starting at the
2239 * fragment's own offset)
2240 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002241 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00002242 *
2243 * Maps the page associated with @frag to @device.
2244 */
2245static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2246 const skb_frag_t *frag,
2247 size_t offset, size_t size,
2248 enum dma_data_direction dir)
2249{
2250 return dma_map_page(dev, skb_frag_page(frag),
2251 frag->page_offset + offset, size, dir);
2252}
2253
Eric Dumazet117632e2011-12-03 21:39:53 +00002254static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2255 gfp_t gfp_mask)
2256{
2257 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2258}
2259
Octavian Purdilabad93e92014-06-12 01:36:26 +03002260
2261static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2262 gfp_t gfp_mask)
2263{
2264 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2265}
2266
2267
Ian Campbell131ea662011-08-19 06:25:00 +00002268/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002269 * skb_clone_writable - is the header of a clone writable
2270 * @skb: buffer to check
2271 * @len: length up to which to write
2272 *
2273 * Returns true if modifying the header part of the cloned buffer
2274 * does not requires the data to be copied.
2275 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002276static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002277{
2278 return !skb_header_cloned(skb) &&
2279 skb_headroom(skb) + len <= skb->hdr_len;
2280}
2281
Herbert Xud9cc2042007-09-16 16:21:16 -07002282static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2283 int cloned)
2284{
2285 int delta = 0;
2286
Herbert Xud9cc2042007-09-16 16:21:16 -07002287 if (headroom > skb_headroom(skb))
2288 delta = headroom - skb_headroom(skb);
2289
2290 if (delta || cloned)
2291 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2292 GFP_ATOMIC);
2293 return 0;
2294}
2295
Patrick McHardy334a8132007-06-25 04:35:20 -07002296/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002297 * skb_cow - copy header of skb when it is required
2298 * @skb: buffer to cow
2299 * @headroom: needed headroom
2300 *
2301 * If the skb passed lacks sufficient headroom or its data part
2302 * is shared, data is reallocated. If reallocation fails, an error
2303 * is returned and original skb is not changed.
2304 *
2305 * The result is skb with writable area skb->head...skb->tail
2306 * and at least @headroom of space at head.
2307 */
2308static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2309{
Herbert Xud9cc2042007-09-16 16:21:16 -07002310 return __skb_cow(skb, headroom, skb_cloned(skb));
2311}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002312
Herbert Xud9cc2042007-09-16 16:21:16 -07002313/**
2314 * skb_cow_head - skb_cow but only making the head writable
2315 * @skb: buffer to cow
2316 * @headroom: needed headroom
2317 *
2318 * This function is identical to skb_cow except that we replace the
2319 * skb_cloned check by skb_header_cloned. It should be used when
2320 * you only need to push on some header and do not need to modify
2321 * the data.
2322 */
2323static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2324{
2325 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002326}
2327
2328/**
2329 * skb_padto - pad an skbuff up to a minimal size
2330 * @skb: buffer to pad
2331 * @len: minimal length
2332 *
2333 * Pads up a buffer to ensure the trailing bytes exist and are
2334 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002335 * is untouched. Otherwise it is extended. Returns zero on
2336 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002337 */
2338
Herbert Xu5b057c62006-06-23 02:06:41 -07002339static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002340{
2341 unsigned int size = skb->len;
2342 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002343 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002344 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002345}
2346
2347static inline int skb_add_data(struct sk_buff *skb,
2348 char __user *from, int copy)
2349{
2350 const int off = skb->len;
2351
2352 if (skb->ip_summed == CHECKSUM_NONE) {
2353 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08002354 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
Linus Torvalds1da177e2005-04-16 15:20:36 -07002355 copy, 0, &err);
2356 if (!err) {
2357 skb->csum = csum_block_add(skb->csum, csum, off);
2358 return 0;
2359 }
2360 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2361 return 0;
2362
2363 __skb_trim(skb, off);
2364 return -EFAULT;
2365}
2366
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002367static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2368 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002369{
2370 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002371 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002372
Ian Campbellea2ab692011-08-22 23:44:58 +00002373 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002374 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002375 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002376 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002377}
2378
Herbert Xu364c6ba2006-06-09 16:10:40 -07002379static inline int __skb_linearize(struct sk_buff *skb)
2380{
2381 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2382}
2383
Linus Torvalds1da177e2005-04-16 15:20:36 -07002384/**
2385 * skb_linearize - convert paged skb to linear one
2386 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002387 *
2388 * If there is no free memory -ENOMEM is returned, otherwise zero
2389 * is returned and the old skb data released.
2390 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002391static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002392{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002393 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2394}
2395
2396/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00002397 * skb_has_shared_frag - can any frag be overwritten
2398 * @skb: buffer to test
2399 *
2400 * Return true if the skb has at least one frag that might be modified
2401 * by an external entity (as in vmsplice()/sendfile())
2402 */
2403static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2404{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00002405 return skb_is_nonlinear(skb) &&
2406 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00002407}
2408
2409/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07002410 * skb_linearize_cow - make sure skb is linear and writable
2411 * @skb: buffer to process
2412 *
2413 * If there is no free memory -ENOMEM is returned, otherwise zero
2414 * is returned and the old skb data released.
2415 */
2416static inline int skb_linearize_cow(struct sk_buff *skb)
2417{
2418 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2419 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002420}
2421
2422/**
2423 * skb_postpull_rcsum - update checksum for received skb after pull
2424 * @skb: buffer to update
2425 * @start: start of data before pull
2426 * @len: length of data pulled
2427 *
2428 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002429 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2430 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002431 */
2432
2433static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002434 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002435{
Patrick McHardy84fa7932006-08-29 16:44:56 -07002436 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002437 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2438}
2439
Herbert Xucbb042f2006-03-20 22:43:56 -08002440unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2441
David S. Miller7ce5a272013-12-02 17:26:05 -05002442/**
2443 * pskb_trim_rcsum - trim received skb and update checksum
2444 * @skb: buffer to trim
2445 * @len: new length
2446 *
2447 * This is exactly the same as pskb_trim except that it ensures the
2448 * checksum of received packets are still valid after the operation.
2449 */
2450
2451static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2452{
2453 if (likely(len >= skb->len))
2454 return 0;
2455 if (skb->ip_summed == CHECKSUM_COMPLETE)
2456 skb->ip_summed = CHECKSUM_NONE;
2457 return __pskb_trim(skb, len);
2458}
2459
Linus Torvalds1da177e2005-04-16 15:20:36 -07002460#define skb_queue_walk(queue, skb) \
2461 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002462 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002463 skb = skb->next)
2464
James Chapman46f89142007-04-30 00:07:31 -07002465#define skb_queue_walk_safe(queue, skb, tmp) \
2466 for (skb = (queue)->next, tmp = skb->next; \
2467 skb != (struct sk_buff *)(queue); \
2468 skb = tmp, tmp = skb->next)
2469
David S. Miller1164f522008-09-23 00:49:44 -07002470#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002471 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002472 skb = skb->next)
2473
2474#define skb_queue_walk_from_safe(queue, skb, tmp) \
2475 for (tmp = skb->next; \
2476 skb != (struct sk_buff *)(queue); \
2477 skb = tmp, tmp = skb->next)
2478
Stephen Hemminger300ce172005-10-30 13:47:34 -08002479#define skb_queue_reverse_walk(queue, skb) \
2480 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002481 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08002482 skb = skb->prev)
2483
David S. Miller686a2952011-01-20 22:47:32 -08002484#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2485 for (skb = (queue)->prev, tmp = skb->prev; \
2486 skb != (struct sk_buff *)(queue); \
2487 skb = tmp, tmp = skb->prev)
2488
2489#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2490 for (tmp = skb->prev; \
2491 skb != (struct sk_buff *)(queue); \
2492 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002493
David S. Miller21dc3302010-08-23 00:13:46 -07002494static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07002495{
2496 return skb_shinfo(skb)->frag_list != NULL;
2497}
2498
2499static inline void skb_frag_list_init(struct sk_buff *skb)
2500{
2501 skb_shinfo(skb)->frag_list = NULL;
2502}
2503
2504static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2505{
2506 frag->next = skb_shinfo(skb)->frag_list;
2507 skb_shinfo(skb)->frag_list = frag;
2508}
2509
2510#define skb_walk_frags(skb, iter) \
2511 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2512
Joe Perches7965bd42013-09-26 14:48:15 -07002513struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2514 int *peeked, int *off, int *err);
2515struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2516 int *err);
2517unsigned int datagram_poll(struct file *file, struct socket *sock,
2518 struct poll_table_struct *wait);
2519int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2520 struct iovec *to, int size);
2521int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2522 struct iovec *iov);
2523int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2524 const struct iovec *from, int from_offset,
2525 int len);
2526int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2527 int offset, size_t count);
2528int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2529 const struct iovec *to, int to_offset,
2530 int size);
2531void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2532void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2533int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07002534int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2535int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2536__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2537 int len, __wsum csum);
2538int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2539 struct pipe_inode_info *pipe, unsigned int len,
2540 unsigned int flags);
2541void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01002542unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00002543int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2544 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07002545void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2546int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2547void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Florian Westphalde960aa2014-01-26 10:58:16 +01002548unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -07002549struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002550
Daniel Borkmann2817a332013-10-30 11:50:51 +01002551struct skb_checksum_ops {
2552 __wsum (*update)(const void *mem, int len, __wsum wsum);
2553 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2554};
2555
2556__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2557 __wsum csum, const struct skb_checksum_ops *ops);
2558__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2559 __wsum csum);
2560
Linus Torvalds1da177e2005-04-16 15:20:36 -07002561static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2562 int len, void *buffer)
2563{
2564 int hlen = skb_headlen(skb);
2565
Patrick McHardy55820ee2005-07-05 14:08:10 -07002566 if (hlen - offset >= len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002567 return skb->data + offset;
2568
2569 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2570 return NULL;
2571
2572 return buffer;
2573}
2574
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01002575/**
2576 * skb_needs_linearize - check if we need to linearize a given skb
2577 * depending on the given device features.
2578 * @skb: socket buffer to check
2579 * @features: net device features
2580 *
2581 * Returns true if either:
2582 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2583 * 2. skb is fragmented and the device does not support SG.
2584 */
2585static inline bool skb_needs_linearize(struct sk_buff *skb,
2586 netdev_features_t features)
2587{
2588 return skb_is_nonlinear(skb) &&
2589 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2590 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2591}
2592
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03002593static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2594 void *to,
2595 const unsigned int len)
2596{
2597 memcpy(to, skb->data, len);
2598}
2599
2600static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2601 const int offset, void *to,
2602 const unsigned int len)
2603{
2604 memcpy(to, skb->data + offset, len);
2605}
2606
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03002607static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2608 const void *from,
2609 const unsigned int len)
2610{
2611 memcpy(skb->data, from, len);
2612}
2613
2614static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2615 const int offset,
2616 const void *from,
2617 const unsigned int len)
2618{
2619 memcpy(skb->data + offset, from, len);
2620}
2621
Joe Perches7965bd42013-09-26 14:48:15 -07002622void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002623
Patrick Ohlyac45f602009-02-12 05:03:37 +00002624static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2625{
2626 return skb->tstamp;
2627}
2628
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002629/**
2630 * skb_get_timestamp - get timestamp from a skb
2631 * @skb: skb to get stamp from
2632 * @stamp: pointer to struct timeval to store stamp in
2633 *
2634 * Timestamps are stored in the skb as offsets to a base timestamp.
2635 * This function converts the offset back to a struct timeval and stores
2636 * it in stamp.
2637 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00002638static inline void skb_get_timestamp(const struct sk_buff *skb,
2639 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002640{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002641 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002642}
2643
Patrick Ohlyac45f602009-02-12 05:03:37 +00002644static inline void skb_get_timestampns(const struct sk_buff *skb,
2645 struct timespec *stamp)
2646{
2647 *stamp = ktime_to_timespec(skb->tstamp);
2648}
2649
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002650static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002651{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002652 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002653}
2654
Stephen Hemminger164891a2007-04-23 22:26:16 -07002655static inline ktime_t net_timedelta(ktime_t t)
2656{
2657 return ktime_sub(ktime_get_real(), t);
2658}
2659
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07002660static inline ktime_t net_invalid_timestamp(void)
2661{
2662 return ktime_set(0, 0);
2663}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002664
Richard Cochranc1f19b52010-07-17 08:49:36 +00002665#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2666
Joe Perches7965bd42013-09-26 14:48:15 -07002667void skb_clone_tx_timestamp(struct sk_buff *skb);
2668bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00002669
2670#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2671
2672static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2673{
2674}
2675
2676static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2677{
2678 return false;
2679}
2680
2681#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2682
2683/**
2684 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2685 *
Richard Cochranda92b192011-10-21 00:49:15 +00002686 * PHY drivers may accept clones of transmitted packets for
2687 * timestamping via their phy_driver.txtstamp method. These drivers
2688 * must call this function to return the skb back to the stack, with
2689 * or without a timestamp.
2690 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00002691 * @skb: clone of the the original outgoing packet
Richard Cochranda92b192011-10-21 00:49:15 +00002692 * @hwtstamps: hardware time stamps, may be NULL if not available
Richard Cochranc1f19b52010-07-17 08:49:36 +00002693 *
2694 */
2695void skb_complete_tx_timestamp(struct sk_buff *skb,
2696 struct skb_shared_hwtstamps *hwtstamps);
2697
Patrick Ohlyac45f602009-02-12 05:03:37 +00002698/**
2699 * skb_tstamp_tx - queue clone of skb with send time stamps
2700 * @orig_skb: the original outgoing packet
2701 * @hwtstamps: hardware time stamps, may be NULL if not available
2702 *
2703 * If the skb has a socket associated, then this function clones the
2704 * skb (thus sharing the actual data and optional structures), stores
2705 * the optional hardware time stamping information (if non NULL) or
2706 * generates a software time stamp (otherwise), then queues the clone
2707 * to the error queue of the socket. Errors are silently ignored.
2708 */
Joe Perches7965bd42013-09-26 14:48:15 -07002709void skb_tstamp_tx(struct sk_buff *orig_skb,
2710 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00002711
Richard Cochran4507a712010-07-17 08:48:28 +00002712static inline void sw_tx_timestamp(struct sk_buff *skb)
2713{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002714 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2715 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00002716 skb_tstamp_tx(skb, NULL);
2717}
2718
2719/**
2720 * skb_tx_timestamp() - Driver hook for transmit timestamping
2721 *
2722 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00002723 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00002724 *
David S. Miller73409f32013-12-27 13:04:33 -05002725 * Specifically, one should make absolutely sure that this function is
2726 * called before TX completion of this packet can trigger. Otherwise
2727 * the packet could potentially already be freed.
2728 *
Richard Cochran4507a712010-07-17 08:48:28 +00002729 * @skb: A socket buffer.
2730 */
2731static inline void skb_tx_timestamp(struct sk_buff *skb)
2732{
Richard Cochranc1f19b52010-07-17 08:49:36 +00002733 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002734 sw_tx_timestamp(skb);
2735}
2736
Johannes Berg6e3e9392011-11-09 10:15:42 +01002737/**
2738 * skb_complete_wifi_ack - deliver skb with wifi status
2739 *
2740 * @skb: the original outgoing packet
2741 * @acked: ack status
2742 *
2743 */
2744void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2745
Joe Perches7965bd42013-09-26 14:48:15 -07002746__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2747__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002748
Herbert Xu60476372007-04-09 11:59:39 -07002749static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2750{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002751 return ((skb->ip_summed & CHECKSUM_UNNECESSARY) || skb->csum_valid);
Herbert Xu60476372007-04-09 11:59:39 -07002752}
2753
Herbert Xufb286bb2005-11-10 13:01:24 -08002754/**
2755 * skb_checksum_complete - Calculate checksum of an entire packet
2756 * @skb: packet to process
2757 *
2758 * This function calculates the checksum over the entire packet plus
2759 * the value of skb->csum. The latter can be used to supply the
2760 * checksum of a pseudo header as used by TCP/UDP. It returns the
2761 * checksum.
2762 *
2763 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2764 * this function can be used to verify that checksum on received
2765 * packets. In that case the function should return zero if the
2766 * checksum is correct. In particular, this function will return zero
2767 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2768 * hardware has already verified the correctness of the checksum.
2769 */
Al Viro4381ca32007-07-15 21:00:11 +01002770static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002771{
Herbert Xu60476372007-04-09 11:59:39 -07002772 return skb_csum_unnecessary(skb) ?
2773 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002774}
2775
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002776/* Check if we need to perform checksum complete validation.
2777 *
2778 * Returns true if checksum complete is needed, false otherwise
2779 * (either checksum is unnecessary or zero checksum is allowed).
2780 */
2781static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
2782 bool zero_okay,
2783 __sum16 check)
2784{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002785 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
2786 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002787 return false;
2788 }
2789
2790 return true;
2791}
2792
2793/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
2794 * in checksum_init.
2795 */
2796#define CHECKSUM_BREAK 76
2797
2798/* Validate (init) checksum based on checksum complete.
2799 *
2800 * Return values:
2801 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
2802 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
2803 * checksum is stored in skb->csum for use in __skb_checksum_complete
2804 * non-zero: value of invalid checksum
2805 *
2806 */
2807static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
2808 bool complete,
2809 __wsum psum)
2810{
2811 if (skb->ip_summed == CHECKSUM_COMPLETE) {
2812 if (!csum_fold(csum_add(psum, skb->csum))) {
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002813 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002814 return 0;
2815 }
2816 }
2817
2818 skb->csum = psum;
2819
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002820 if (complete || skb->len <= CHECKSUM_BREAK) {
2821 __sum16 csum;
2822
2823 csum = __skb_checksum_complete(skb);
2824 skb->csum_valid = !csum;
2825 return csum;
2826 }
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002827
2828 return 0;
2829}
2830
2831static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
2832{
2833 return 0;
2834}
2835
2836/* Perform checksum validate (init). Note that this is a macro since we only
2837 * want to calculate the pseudo header which is an input function if necessary.
2838 * First we try to validate without any computation (checksum unnecessary) and
2839 * then calculate based on checksum complete calling the function to compute
2840 * pseudo header.
2841 *
2842 * Return values:
2843 * 0: checksum is validated or try to in skb_checksum_complete
2844 * non-zero: value of invalid checksum
2845 */
2846#define __skb_checksum_validate(skb, proto, complete, \
2847 zero_okay, check, compute_pseudo) \
2848({ \
2849 __sum16 __ret = 0; \
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002850 skb->csum_valid = 0; \
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002851 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
2852 __ret = __skb_checksum_validate_complete(skb, \
2853 complete, compute_pseudo(skb, proto)); \
2854 __ret; \
2855})
2856
2857#define skb_checksum_init(skb, proto, compute_pseudo) \
2858 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
2859
2860#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
2861 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
2862
2863#define skb_checksum_validate(skb, proto, compute_pseudo) \
2864 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
2865
2866#define skb_checksum_validate_zero_check(skb, proto, check, \
2867 compute_pseudo) \
2868 __skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
2869
2870#define skb_checksum_simple_validate(skb) \
2871 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
2872
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002873#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07002874void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002875static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2876{
2877 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002878 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002879}
2880static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2881{
2882 if (nfct)
2883 atomic_inc(&nfct->use);
2884}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002885#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002886#ifdef CONFIG_BRIDGE_NETFILTER
2887static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2888{
2889 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2890 kfree(nf_bridge);
2891}
2892static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2893{
2894 if (nf_bridge)
2895 atomic_inc(&nf_bridge->use);
2896}
2897#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08002898static inline void nf_reset(struct sk_buff *skb)
2899{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002900#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08002901 nf_conntrack_put(skb->nfct);
2902 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002903#endif
Patrick McHardya193a4a2006-03-20 19:23:05 -08002904#ifdef CONFIG_BRIDGE_NETFILTER
2905 nf_bridge_put(skb->nf_bridge);
2906 skb->nf_bridge = NULL;
2907#endif
2908}
2909
Patrick McHardy124dff02013-04-05 20:42:05 +02002910static inline void nf_reset_trace(struct sk_buff *skb)
2911{
Florian Westphal478b3602014-02-15 23:48:45 +01002912#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00002913 skb->nf_trace = 0;
2914#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002915}
2916
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002917/* Note: This doesn't put any conntrack and bridge info in dst. */
2918static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2919{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002920#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002921 dst->nfct = src->nfct;
2922 nf_conntrack_get(src->nfct);
2923 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002924#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002925#ifdef CONFIG_BRIDGE_NETFILTER
2926 dst->nf_bridge = src->nf_bridge;
2927 nf_bridge_get(src->nf_bridge);
2928#endif
Florian Westphal478b3602014-02-15 23:48:45 +01002929#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
2930 dst->nf_trace = src->nf_trace;
2931#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002932}
2933
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002934static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2935{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002936#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002937 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002938#endif
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002939#ifdef CONFIG_BRIDGE_NETFILTER
2940 nf_bridge_put(dst->nf_bridge);
2941#endif
2942 __nf_copy(dst, src);
2943}
2944
James Morris984bc162006-06-09 00:29:17 -07002945#ifdef CONFIG_NETWORK_SECMARK
2946static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2947{
2948 to->secmark = from->secmark;
2949}
2950
2951static inline void skb_init_secmark(struct sk_buff *skb)
2952{
2953 skb->secmark = 0;
2954}
2955#else
2956static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2957{ }
2958
2959static inline void skb_init_secmark(struct sk_buff *skb)
2960{ }
2961#endif
2962
Eric W. Biederman574f7192014-04-01 12:20:24 -07002963static inline bool skb_irq_freeable(const struct sk_buff *skb)
2964{
2965 return !skb->destructor &&
2966#if IS_ENABLED(CONFIG_XFRM)
2967 !skb->sp &&
2968#endif
2969#if IS_ENABLED(CONFIG_NF_CONNTRACK)
2970 !skb->nfct &&
2971#endif
2972 !skb->_skb_refdst &&
2973 !skb_has_frag_list(skb);
2974}
2975
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002976static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2977{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002978 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002979}
2980
Stephen Hemminger92477442009-03-21 13:39:26 -07002981static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002982{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002983 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002984}
2985
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002986static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2987{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002988 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002989}
2990
David S. Millerd5a9e242009-01-27 16:22:11 -08002991static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2992{
2993 skb->queue_mapping = rx_queue + 1;
2994}
2995
Stephen Hemminger92477442009-03-21 13:39:26 -07002996static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002997{
2998 return skb->queue_mapping - 1;
2999}
3000
Stephen Hemminger92477442009-03-21 13:39:26 -07003001static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003002{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003003 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08003004}
3005
Tom Herbert0e001612014-07-01 21:32:27 -07003006u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
Joe Perches7965bd42013-09-26 14:48:15 -07003007 unsigned int num_tx_queues);
Stephen Hemminger92477442009-03-21 13:39:26 -07003008
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003009static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3010{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003011#ifdef CONFIG_XFRM
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003012 return skb->sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003013#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003014 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003015#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003016}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003017
Pravin B Shelar68c33162013-02-14 14:02:41 +00003018/* Keeps track of mac header offset relative to skb->head.
3019 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3020 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07003021 * tunnel skb it points to outer mac header.
3022 * Keeps track of level of encapsulation of network headers.
3023 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00003024struct skb_gso_cb {
Eric Dumazet3347c962013-10-19 11:42:56 -07003025 int mac_offset;
3026 int encap_level;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003027 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00003028};
3029#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3030
3031static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3032{
3033 return (skb_mac_header(inner_skb) - inner_skb->head) -
3034 SKB_GSO_CB(inner_skb)->mac_offset;
3035}
3036
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00003037static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3038{
3039 int new_headroom, headroom;
3040 int ret;
3041
3042 headroom = skb_headroom(skb);
3043 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3044 if (ret)
3045 return ret;
3046
3047 new_headroom = skb_headroom(skb);
3048 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3049 return 0;
3050}
3051
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003052/* Compute the checksum for a gso segment. First compute the checksum value
3053 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3054 * then add in skb->csum (checksum from csum_start to end of packet).
3055 * skb->csum and csum_start are then updated to reflect the checksum of the
3056 * resultant packet starting from the transport header-- the resultant checksum
3057 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3058 * header.
3059 */
3060static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3061{
3062 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3063 skb_transport_offset(skb);
3064 __u16 csum;
3065
3066 csum = csum_fold(csum_partial(skb_transport_header(skb),
3067 plen, skb->csum));
3068 skb->csum = res;
3069 SKB_GSO_CB(skb)->csum_start -= plen;
3070
3071 return csum;
3072}
3073
David S. Millerbdcc0922012-03-07 20:53:36 -05003074static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07003075{
3076 return skb_shinfo(skb)->gso_size;
3077}
3078
Eric Dumazet36a8f392013-09-29 01:21:32 -07003079/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05003080static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02003081{
3082 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3083}
3084
Joe Perches7965bd42013-09-26 14:48:15 -07003085void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07003086
3087static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3088{
3089 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3090 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003091 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3092
Alexander Duyckb78462e2010-06-02 12:24:37 +00003093 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3094 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07003095 __skb_warn_lro_forwarding(skb);
3096 return true;
3097 }
3098 return false;
3099}
3100
Herbert Xu35fc92a2007-03-26 23:22:20 -07003101static inline void skb_forward_csum(struct sk_buff *skb)
3102{
3103 /* Unfortunately we don't support this one. Any brave souls? */
3104 if (skb->ip_summed == CHECKSUM_COMPLETE)
3105 skb->ip_summed = CHECKSUM_NONE;
3106}
3107
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003108/**
3109 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3110 * @skb: skb to check
3111 *
3112 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3113 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3114 * use this helper, to document places where we make this assertion.
3115 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003116static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003117{
3118#ifdef DEBUG
3119 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3120#endif
3121}
3122
Rusty Russellf35d9d82008-02-04 23:49:54 -05003123bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00003124
Paul Durranted1f50c2014-01-09 10:02:46 +00003125int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3126
Daniel Borkmannf77668d2013-03-19 06:39:30 +00003127u32 __skb_get_poff(const struct sk_buff *skb);
3128
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00003129/**
3130 * skb_head_is_locked - Determine if the skb->head is locked down
3131 * @skb: skb to check
3132 *
3133 * The head on skbs build around a head frag can be removed if they are
3134 * not cloned. This function returns true if the skb head is locked down
3135 * due to either being allocated via kmalloc, or by being a clone with
3136 * multiple references to the head.
3137 */
3138static inline bool skb_head_is_locked(const struct sk_buff *skb)
3139{
3140 return !skb->head_frag || skb_cloned(skb);
3141}
Florian Westphalfe6cc552014-02-13 23:09:12 +01003142
3143/**
3144 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3145 *
3146 * @skb: GSO skb
3147 *
3148 * skb_gso_network_seglen is used to determine the real size of the
3149 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3150 *
3151 * The MAC/L2 header is not accounted for.
3152 */
3153static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3154{
3155 unsigned int hdr_len = skb_transport_header(skb) -
3156 skb_network_header(skb);
3157 return hdr_len + skb_gso_transport_seglen(skb);
3158}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003159#endif /* __KERNEL__ */
3160#endif /* _LINUX_SKBUFF_H */