blob: 3d932e64125af3a296d0a58c05de3c8b8178aeed [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>
Eric Dumazet56b17422014-11-03 08:19:53 -080023#include <linux/rbtree.h>
David S. Miller51f3d022014-11-05 16:46:40 -050024#include <linux/socket.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025
Arun Sharma600634972011-07-26 16:09:06 -070026#include <linux/atomic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <asm/types.h>
28#include <linux/spinlock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070029#include <linux/net.h>
Thomas Graf3fc7e8a2005-06-23 21:00:17 -070030#include <linux/textsearch.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070031#include <net/checksum.h>
Al Viroa80958f2006-12-04 20:41:19 +000032#include <linux/rcupdate.h>
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -070033#include <linux/hrtimer.h>
Ian Campbell131ea662011-08-19 06:25:00 +000034#include <linux/dma-mapping.h>
Michał Mirosławc8f44af2011-11-15 15:29:55 +000035#include <linux/netdev_features.h>
Eric Dumazet363ec39232014-02-26 14:02:11 -080036#include <linux/sched.h>
Jason Wang5203cd22013-03-26 23:11:21 +000037#include <net/flow_keys.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070038
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010039/* A. Checksumming of received packets by device.
40 *
41 * CHECKSUM_NONE:
42 *
43 * Device failed to checksum this packet e.g. due to lack of capabilities.
44 * The packet contains full (though not verified) checksum in packet but
45 * not in skb->csum. Thus, skb->csum is undefined in this case.
46 *
47 * CHECKSUM_UNNECESSARY:
48 *
49 * The hardware you're dealing with doesn't calculate the full checksum
50 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
Tom Herbert77cffe22014-08-27 21:26:46 -070051 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
52 * if their checksums are okay. skb->csum is still undefined in this case
53 * though. It is a bad option, but, unfortunately, nowadays most vendors do
54 * this. Apparently with the secret goal to sell you new devices, when you
55 * will add new protocol to your host, f.e. IPv6 8)
56 *
57 * CHECKSUM_UNNECESSARY is applicable to following protocols:
58 * TCP: IPv6 and IPv4.
59 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
60 * zero UDP checksum for either IPv4 or IPv6, the networking stack
61 * may perform further validation in this case.
62 * GRE: only if the checksum is present in the header.
63 * SCTP: indicates the CRC in SCTP header has been validated.
64 *
65 * skb->csum_level indicates the number of consecutive checksums found in
66 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
67 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
68 * and a device is able to verify the checksums for UDP (possibly zero),
69 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
70 * two. If the device were only able to verify the UDP checksum and not
71 * GRE, either because it doesn't support GRE checksum of because GRE
72 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
73 * not considered in this case).
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010074 *
75 * CHECKSUM_COMPLETE:
76 *
77 * This is the most generic way. The device supplied checksum of the _whole_
78 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
79 * hardware doesn't need to parse L3/L4 headers to implement this.
80 *
81 * Note: Even if device supports only some protocols, but is able to produce
82 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
83 *
84 * CHECKSUM_PARTIAL:
85 *
Tom Herbert6edec0e2015-02-10 16:30:28 -080086 * A checksum is set up to be offloaded to a device as described in the
87 * output description for CHECKSUM_PARTIAL. This may occur on a packet
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010088 * received directly from another Linux OS, e.g., a virtualized Linux kernel
Tom Herbert6edec0e2015-02-10 16:30:28 -080089 * on the same host, or it may be set in the input path in GRO or remote
90 * checksum offload. For the purposes of checksum verification, the checksum
91 * referred to by skb->csum_start + skb->csum_offset and any preceding
92 * checksums in the packet are considered verified. Any checksums in the
93 * packet that are after the checksum being offloaded are not considered to
94 * be verified.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010095 *
96 * B. Checksumming on output.
97 *
98 * CHECKSUM_NONE:
99 *
100 * The skb was already checksummed by the protocol, or a checksum is not
101 * required.
102 *
103 * CHECKSUM_PARTIAL:
104 *
105 * The device is required to checksum the packet as seen by hard_start_xmit()
106 * from skb->csum_start up to the end, and to record/write the checksum at
107 * offset skb->csum_start + skb->csum_offset.
108 *
109 * The device must show its capabilities in dev->features, set up at device
110 * setup time, e.g. netdev_features.h:
111 *
112 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
113 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
114 * IPv4. Sigh. Vendors like this way for an unknown reason.
115 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
116 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
117 * NETIF_F_... - Well, you get the picture.
118 *
119 * CHECKSUM_UNNECESSARY:
120 *
121 * Normally, the device will do per protocol specific checksumming. Protocol
122 * implementations that do not want the NIC to perform the checksum
123 * calculation should use this flag in their outgoing skbs.
124 *
125 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
126 * offload. Correspondingly, the FCoE protocol driver
127 * stack should use CHECKSUM_UNNECESSARY.
128 *
129 * Any questions? No questions, good. --ANK
130 */
131
Herbert Xu60476372007-04-09 11:59:39 -0700132/* Don't change this without changing skb_csum_unnecessary! */
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100133#define CHECKSUM_NONE 0
134#define CHECKSUM_UNNECESSARY 1
135#define CHECKSUM_COMPLETE 2
136#define CHECKSUM_PARTIAL 3
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137
Tom Herbert77cffe22014-08-27 21:26:46 -0700138/* Maximum value in skb->csum_level */
139#define SKB_MAX_CSUM_LEVEL 3
140
Tobias Klauser0bec8c82014-07-22 12:06:23 +0200141#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
David S. Millerfc910a22007-03-25 20:27:59 -0700142#define SKB_WITH_OVERHEAD(X) \
Herbert Xudeea84b2007-10-21 16:27:46 -0700143 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
David S. Millerfc910a22007-03-25 20:27:59 -0700144#define SKB_MAX_ORDER(X, ORDER) \
145 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
147#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
148
Eric Dumazet87fb4b72011-10-13 07:28:54 +0000149/* return minimum truesize of one skb containing X bytes of data */
150#define SKB_TRUESIZE(X) ((X) + \
151 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
152 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
153
Linus Torvalds1da177e2005-04-16 15:20:36 -0700154struct net_device;
David Howells716ea3a2007-04-02 20:19:53 -0700155struct scatterlist;
Jens Axboe9c55e012007-11-06 23:30:13 -0800156struct pipe_inode_info;
Herbert Xua8f820aa2014-11-07 21:22:22 +0800157struct iov_iter;
Alexander Duyckfd11a832014-12-09 19:40:49 -0800158struct napi_struct;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700160#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161struct nf_conntrack {
162 atomic_t use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700163};
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700164#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +0200166#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167struct nf_bridge_info {
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000168 atomic_t use;
Florian Westphal3eaf4022015-04-02 14:31:44 +0200169 enum {
170 BRNF_PROTO_UNCHANGED,
171 BRNF_PROTO_8021Q,
172 BRNF_PROTO_PPPOE
Florian Westphal7fb48c52015-05-03 22:05:28 +0200173 } orig_proto:8;
Florian Westphala1e67952015-04-02 14:31:45 +0200174 bool pkt_otherhost;
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000175 unsigned int mask;
176 struct net_device *physindev;
Florian Westphal7fb48c52015-05-03 22:05:28 +0200177 union {
178 struct net_device *physoutdev;
179 char neigh_header[8];
180 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700181};
182#endif
183
Linus Torvalds1da177e2005-04-16 15:20:36 -0700184struct sk_buff_head {
185 /* These two members must be first. */
186 struct sk_buff *next;
187 struct sk_buff *prev;
188
189 __u32 qlen;
190 spinlock_t lock;
191};
192
193struct sk_buff;
194
Ian Campbell9d4dde52011-12-22 23:39:14 +0000195/* To allow 64K frame to be packed as single skb without frag_list we
196 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
197 * buffers which do not start on a page boundary.
198 *
199 * Since GRO uses frags we allocate at least 16 regardless of page
200 * size.
Anton Blancharda715dea2011-03-27 14:57:26 +0000201 */
Ian Campbell9d4dde52011-12-22 23:39:14 +0000202#if (65536/PAGE_SIZE + 1) < 16
David S. Millereec00952011-03-29 23:34:08 -0700203#define MAX_SKB_FRAGS 16UL
Anton Blancharda715dea2011-03-27 14:57:26 +0000204#else
Ian Campbell9d4dde52011-12-22 23:39:14 +0000205#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
Anton Blancharda715dea2011-03-27 14:57:26 +0000206#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700207
208typedef struct skb_frag_struct skb_frag_t;
209
210struct skb_frag_struct {
Ian Campbella8605c62011-10-19 23:01:49 +0000211 struct {
212 struct page *p;
213 } page;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000214#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
David S. Millera309bb02007-07-30 18:47:03 -0700215 __u32 page_offset;
216 __u32 size;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000217#else
218 __u16 page_offset;
219 __u16 size;
220#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221};
222
Eric Dumazet9e903e02011-10-18 21:00:24 +0000223static inline unsigned int skb_frag_size(const skb_frag_t *frag)
224{
225 return frag->size;
226}
227
228static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
229{
230 frag->size = size;
231}
232
233static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
234{
235 frag->size += delta;
236}
237
238static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
239{
240 frag->size -= delta;
241}
242
Patrick Ohlyac45f602009-02-12 05:03:37 +0000243#define HAVE_HW_TIME_STAMP
244
245/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800246 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000247 * @hwtstamp: hardware time stamp transformed into duration
248 * since arbitrary point in time
Patrick Ohlyac45f602009-02-12 05:03:37 +0000249 *
250 * Software time stamps generated by ktime_get_real() are stored in
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400251 * skb->tstamp.
Patrick Ohlyac45f602009-02-12 05:03:37 +0000252 *
253 * hwtstamps can only be compared against other hwtstamps from
254 * the same device.
255 *
256 * This structure is attached to packets as part of the
257 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
258 */
259struct skb_shared_hwtstamps {
260 ktime_t hwtstamp;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000261};
262
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000263/* Definitions for tx_flags in struct skb_shared_info */
264enum {
265 /* generate hardware time stamp */
266 SKBTX_HW_TSTAMP = 1 << 0,
267
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400268 /* generate software time stamp when queueing packet to NIC */
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000269 SKBTX_SW_TSTAMP = 1 << 1,
270
271 /* device driver is going to provide hardware time stamp */
272 SKBTX_IN_PROGRESS = 1 << 2,
273
Shirley Maa6686f22011-07-06 12:22:12 +0000274 /* device driver supports TX zero-copy buffers */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000275 SKBTX_DEV_ZEROCOPY = 1 << 3,
Johannes Berg6e3e9392011-11-09 10:15:42 +0100276
277 /* generate wifi status information (where possible) */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000278 SKBTX_WIFI_STATUS = 1 << 4,
Pravin B Shelarc9af6db2013-02-11 09:27:41 +0000279
280 /* This indicates at least one fragment might be overwritten
281 * (as in vmsplice(), sendfile() ...)
282 * If we need to compute a TX checksum, we'll need to copy
283 * all frags to avoid possible bad checksum
284 */
285 SKBTX_SHARED_FRAG = 1 << 5,
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400286
287 /* generate software time stamp when entering packet scheduling */
288 SKBTX_SCHED_TSTAMP = 1 << 6,
Willem de Bruijne1c8a602014-08-04 22:11:50 -0400289
290 /* generate software timestamp on peer data acknowledgment */
291 SKBTX_ACK_TSTAMP = 1 << 7,
Shirley Maa6686f22011-07-06 12:22:12 +0000292};
293
Willem de Bruijne1c8a602014-08-04 22:11:50 -0400294#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
295 SKBTX_SCHED_TSTAMP | \
296 SKBTX_ACK_TSTAMP)
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400297#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
298
Shirley Maa6686f22011-07-06 12:22:12 +0000299/*
300 * The callback notifies userspace to release buffers when skb DMA is done in
301 * lower device, the skb last reference should be 0 when calling this.
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000302 * The zerocopy_success argument is true if zero copy transmit occurred,
303 * false on data copy or out of memory error caused by data copy attempt.
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000304 * The ctx field is used to track device context.
305 * The desc field is used to track userspace buffer index.
Shirley Maa6686f22011-07-06 12:22:12 +0000306 */
307struct ubuf_info {
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000308 void (*callback)(struct ubuf_info *, bool zerocopy_success);
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000309 void *ctx;
Shirley Maa6686f22011-07-06 12:22:12 +0000310 unsigned long desc;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000311};
312
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313/* This data is invariant across clones and lives at
314 * the end of the header data, ie. at skb->end.
315 */
316struct skb_shared_info {
Ian Campbell9f42f122012-01-05 07:13:39 +0000317 unsigned char nr_frags;
318 __u8 tx_flags;
Herbert Xu79671682006-06-22 02:40:14 -0700319 unsigned short gso_size;
320 /* Warning: this field is not always filled in (UFO)! */
321 unsigned short gso_segs;
322 unsigned short gso_type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000324 struct skb_shared_hwtstamps hwtstamps;
Willem de Bruijn09c2d252014-08-04 22:11:47 -0400325 u32 tskey;
Ian Campbell9f42f122012-01-05 07:13:39 +0000326 __be32 ip6_frag_id;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700327
328 /*
329 * Warning : all fields before dataref are cleared in __alloc_skb()
330 */
331 atomic_t dataref;
332
Johann Baudy69e3c752009-05-18 22:11:22 -0700333 /* Intermediate layers must ensure that destructor_arg
334 * remains valid until skb destructor */
335 void * destructor_arg;
Shirley Maa6686f22011-07-06 12:22:12 +0000336
Eric Dumazetfed66382010-07-22 19:09:08 +0000337 /* must be last field, see pskb_expand_head() */
338 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700339};
340
341/* We divide dataref into two halves. The higher 16 bits hold references
342 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700343 * the entire skb->data. A clone of a headerless skb holds the length of
344 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700345 *
346 * All users must obey the rule that the skb->data reference count must be
347 * greater than or equal to the payload reference count.
348 *
349 * Holding a reference to the payload part means that the user does not
350 * care about modifications to the header part of skb->data.
351 */
352#define SKB_DATAREF_SHIFT 16
353#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
354
David S. Millerd179cd12005-08-17 14:57:30 -0700355
356enum {
Vijay Subramanianc8753d52014-10-02 10:00:43 -0700357 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
358 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
359 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
David S. Millerd179cd12005-08-17 14:57:30 -0700360};
361
Herbert Xu79671682006-06-22 02:40:14 -0700362enum {
363 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700364 SKB_GSO_UDP = 1 << 1,
Herbert Xu576a30e2006-06-27 13:22:38 -0700365
366 /* This indicates the skb is from an untrusted source. */
367 SKB_GSO_DODGY = 1 << 2,
Michael Chanb0da85372006-06-29 12:30:00 -0700368
369 /* This indicates the tcp segment has CWR set. */
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700370 SKB_GSO_TCP_ECN = 1 << 3,
371
372 SKB_GSO_TCPV6 = 1 << 4,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800373
374 SKB_GSO_FCOE = 1 << 5,
Pravin B Shelar68c33162013-02-14 14:02:41 +0000375
376 SKB_GSO_GRE = 1 << 6,
Pravin B Shelar73136262013-03-07 13:21:51 +0000377
Tom Herbert4b282522014-06-14 23:23:52 -0700378 SKB_GSO_GRE_CSUM = 1 << 7,
Simon Horman0d89d202013-05-23 21:02:52 +0000379
Tom Herbert4b282522014-06-14 23:23:52 -0700380 SKB_GSO_IPIP = 1 << 8,
Eric Dumazetcb32f512013-10-19 11:42:57 -0700381
Tom Herbert4b282522014-06-14 23:23:52 -0700382 SKB_GSO_SIT = 1 << 9,
Eric Dumazet61c1db72013-10-20 20:47:30 -0700383
Tom Herbert4b282522014-06-14 23:23:52 -0700384 SKB_GSO_UDP_TUNNEL = 1 << 10,
Tom Herbert0f4f4ff2014-06-04 17:20:16 -0700385
386 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
Tom Herbert4749c092014-06-04 17:20:23 -0700387
Pravin B Shelar59b93b42014-11-05 15:27:48 -0800388 SKB_GSO_TUNNEL_REMCSUM = 1 << 12,
Herbert Xu79671682006-06-22 02:40:14 -0700389};
390
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700391#if BITS_PER_LONG > 32
392#define NET_SKBUFF_DATA_USES_OFFSET 1
393#endif
394
395#ifdef NET_SKBUFF_DATA_USES_OFFSET
396typedef unsigned int sk_buff_data_t;
397#else
398typedef unsigned char *sk_buff_data_t;
399#endif
400
Eric Dumazet363ec39232014-02-26 14:02:11 -0800401/**
402 * struct skb_mstamp - multi resolution time stamps
403 * @stamp_us: timestamp in us resolution
404 * @stamp_jiffies: timestamp in jiffies
405 */
406struct skb_mstamp {
407 union {
408 u64 v64;
409 struct {
410 u32 stamp_us;
411 u32 stamp_jiffies;
412 };
413 };
414};
415
416/**
417 * skb_mstamp_get - get current timestamp
418 * @cl: place to store timestamps
419 */
420static inline void skb_mstamp_get(struct skb_mstamp *cl)
421{
422 u64 val = local_clock();
423
424 do_div(val, NSEC_PER_USEC);
425 cl->stamp_us = (u32)val;
426 cl->stamp_jiffies = (u32)jiffies;
427}
428
429/**
430 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
431 * @t1: pointer to newest sample
432 * @t0: pointer to oldest sample
433 */
434static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
435 const struct skb_mstamp *t0)
436{
437 s32 delta_us = t1->stamp_us - t0->stamp_us;
438 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
439
440 /* If delta_us is negative, this might be because interval is too big,
441 * or local_clock() drift is too big : fallback using jiffies.
442 */
443 if (delta_us <= 0 ||
444 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
445
446 delta_us = jiffies_to_usecs(delta_jiffies);
447
448 return delta_us;
449}
450
451
Linus Torvalds1da177e2005-04-16 15:20:36 -0700452/**
453 * struct sk_buff - socket buffer
454 * @next: Next buffer in list
455 * @prev: Previous buffer in list
Eric Dumazet363ec39232014-02-26 14:02:11 -0800456 * @tstamp: Time we arrived/left
Eric Dumazet56b17422014-11-03 08:19:53 -0800457 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700458 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700460 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000461 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700462 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700463 * @len: Length of actual data
464 * @data_len: Data length
465 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700466 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700467 * @csum: Checksum (must include start/offset pair)
468 * @csum_start: Offset from skb->head where checksumming should start
469 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700470 * @priority: Packet queueing priority
WANG Cong60ff7462014-05-04 16:39:18 -0700471 * @ignore_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700472 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700473 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474 * @nohdr: Payload reference only, must not modify header
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700475 * @nfctinfo: Relationship of this skb to the connection
Linus Torvalds1da177e2005-04-16 15:20:36 -0700476 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700477 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700478 * @ipvs_property: skbuff is owned by ipvs
Randy Dunlap31729362008-02-18 20:52:13 -0800479 * @peeked: this packet has been seen already, so stats have been
480 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700481 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700482 * @protocol: Packet protocol from driver
483 * @destructor: Destruct function
484 * @nfct: Associated connection, if any
Linus Torvalds1da177e2005-04-16 15:20:36 -0700485 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800486 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700487 * @tc_index: Traffic control index
488 * @tc_verd: traffic control verdict
Tom Herbert61b905d2014-03-24 15:34:47 -0700489 * @hash: the packet hash
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700490 * @queue_mapping: Queue mapping for multiqueue devices
David S. Miller0b725a22014-08-25 15:51:53 -0700491 * @xmit_more: More SKBs are pending for this queue
Randy Dunlap553a5672008-04-20 10:51:01 -0700492 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700493 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Tom Herbert61b905d2014-03-24 15:34:47 -0700494 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
Changli Gao4ca24622011-08-19 07:26:44 -0700495 * ports.
Tom Herberta3b18dd2014-07-01 21:33:17 -0700496 * @sw_hash: indicates hash was computed in software stack
Johannes Berg6e3e9392011-11-09 10:15:42 +0100497 * @wifi_acked_valid: wifi_acked was set
498 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000499 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Eliezer Tamir06021292013-06-10 11:39:50 +0300500 * @napi_id: id of the NAPI struct this skb came from
James Morris984bc162006-06-09 00:29:17 -0700501 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700502 * @mark: Generic packet mark
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000503 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b2008-07-14 22:49:06 -0700504 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000505 * @inner_protocol: Protocol (encapsulation)
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000506 * @inner_transport_header: Inner transport layer header (encapsulation)
507 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000508 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700509 * @transport_header: Transport layer header
510 * @network_header: Network layer header
511 * @mac_header: Link layer header
512 * @tail: Tail pointer
513 * @end: End pointer
514 * @head: Head of buffer
515 * @data: Data head pointer
516 * @truesize: Buffer size
517 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518 */
519
520struct sk_buff {
Eric Dumazet363ec39232014-02-26 14:02:11 -0800521 union {
Eric Dumazet56b17422014-11-03 08:19:53 -0800522 struct {
523 /* These two members must be first. */
524 struct sk_buff *next;
525 struct sk_buff *prev;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000526
Eric Dumazet56b17422014-11-03 08:19:53 -0800527 union {
528 ktime_t tstamp;
529 struct skb_mstamp skb_mstamp;
530 };
531 };
532 struct rb_node rbnode; /* used in netem & tcp stack */
533 };
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000534 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700535 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537 /*
538 * This is the control buffer. It is free to use for every
539 * layer. Please put your private variables there. If you
540 * want to keep them across layers you have to do a skb_clone()
541 * first. This is owned by whoever has the skb queued ATM.
542 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000543 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544
Eric Dumazet7fee2262010-05-11 23:19:48 +0000545 unsigned long _skb_refdst;
Eric Dumazetb1937222014-09-28 22:18:47 -0700546 void (*destructor)(struct sk_buff *skb);
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000547#ifdef CONFIG_XFRM
548 struct sec_path *sp;
549#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700550#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
551 struct nf_conntrack *nfct;
552#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +0200553#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Eric Dumazetb1937222014-09-28 22:18:47 -0700554 struct nf_bridge_info *nf_bridge;
555#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700556 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700557 data_len;
558 __u16 mac_len,
559 hdr_len;
Eric Dumazetb1937222014-09-28 22:18:47 -0700560
561 /* Following fields are _not_ copied in __copy_skb_header()
562 * Note that queue_mapping is here mostly to fill a hole.
563 */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200564 kmemcheck_bitfield_begin(flags1);
Eric Dumazetb1937222014-09-28 22:18:47 -0700565 __u16 queue_mapping;
566 __u8 cloned:1,
Harald Welte6869c4d2005-08-09 19:24:19 -0700567 nohdr:1,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800568 fclone:2,
Herbert Xua59322b2007-12-05 01:53:40 -0800569 peeked:1,
Eric Dumazetb1937222014-09-28 22:18:47 -0700570 head_frag:1,
571 xmit_more:1;
572 /* one bit hole */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200573 kmemcheck_bitfield_end(flags1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700574
Eric Dumazetb1937222014-09-28 22:18:47 -0700575 /* fields enclosed in headers_start/headers_end are copied
576 * using a single memcpy() in __copy_skb_header()
577 */
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700578 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700579 __u32 headers_start[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700580 /* public: */
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200581
582/* if you move pkt_type around you also must adapt those constants */
583#ifdef __BIG_ENDIAN_BITFIELD
584#define PKT_TYPE_MAX (7 << 5)
585#else
586#define PKT_TYPE_MAX 7
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100587#endif
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200588#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
589
590 __u8 __pkt_type_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700591 __u8 pkt_type:3;
592 __u8 pfmemalloc:1;
593 __u8 ignore_df:1;
594 __u8 nfctinfo:3;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595
Eric Dumazetb1937222014-09-28 22:18:47 -0700596 __u8 nf_trace:1;
597 __u8 ip_summed:2;
598 __u8 ooo_okay:1;
599 __u8 l4_hash:1;
600 __u8 sw_hash:1;
601 __u8 wifi_acked_valid:1;
602 __u8 wifi_acked:1;
603
604 __u8 no_fcs:1;
605 /* Indicates the inner headers are valid in the skbuff. */
606 __u8 encapsulation:1;
607 __u8 encap_hdr_csum:1;
608 __u8 csum_valid:1;
609 __u8 csum_complete_sw:1;
610 __u8 csum_level:2;
611 __u8 csum_bad:1;
612
613#ifdef CONFIG_IPV6_NDISC_NODETYPE
614 __u8 ndisc_nodetype:2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700616 __u8 ipvs_property:1;
Tom Herbert8bce6d72014-09-29 20:22:29 -0700617 __u8 inner_protocol_type:1;
Tom Herberte585f232014-11-04 09:06:54 -0800618 __u8 remcsum_offload:1;
619 /* 3 or 5 bit hole */
Alexander Duyck4031ae62012-01-27 06:22:53 +0000620
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700622 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700624 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700625#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200627
Eric Dumazetb1937222014-09-28 22:18:47 -0700628 union {
629 __wsum csum;
630 struct {
631 __u16 csum_start;
632 __u16 csum_offset;
633 };
634 };
635 __u32 priority;
636 int skb_iif;
637 __u32 hash;
638 __be16 vlan_proto;
639 __u16 vlan_tci;
Eric Dumazet2bd82482015-02-03 23:48:24 -0800640#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
641 union {
642 unsigned int napi_id;
643 unsigned int sender_cpu;
644 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700645#endif
James Morris984bc162006-06-09 00:29:17 -0700646#ifdef CONFIG_NETWORK_SECMARK
647 __u32 secmark;
648#endif
Neil Horman3b885782009-10-12 13:26:31 -0700649 union {
650 __u32 mark;
Eric Dumazet16fad692013-03-14 05:40:32 +0000651 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700652 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700653
Tom Herbert8bce6d72014-09-29 20:22:29 -0700654 union {
655 __be16 inner_protocol;
656 __u8 inner_ipproto;
657 };
658
Simon Horman1a37e412013-05-23 21:02:51 +0000659 __u16 inner_transport_header;
660 __u16 inner_network_header;
661 __u16 inner_mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700662
663 __be16 protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000664 __u16 transport_header;
665 __u16 network_header;
666 __u16 mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700667
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700668 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700669 __u32 headers_end[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700670 /* public: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700671
Linus Torvalds1da177e2005-04-16 15:20:36 -0700672 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700673 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700674 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700675 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700676 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700677 unsigned int truesize;
678 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679};
680
681#ifdef __KERNEL__
682/*
683 * Handling routines are only of interest to the kernel
684 */
685#include <linux/slab.h>
686
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687
Mel Gormanc93bdd02012-07-31 16:44:19 -0700688#define SKB_ALLOC_FCLONE 0x01
689#define SKB_ALLOC_RX 0x02
Alexander Duyckfd11a832014-12-09 19:40:49 -0800690#define SKB_ALLOC_NAPI 0x04
Mel Gormanc93bdd02012-07-31 16:44:19 -0700691
692/* Returns true if the skb was allocated from PFMEMALLOC reserves */
693static inline bool skb_pfmemalloc(const struct sk_buff *skb)
694{
695 return unlikely(skb->pfmemalloc);
696}
697
Eric Dumazet7fee2262010-05-11 23:19:48 +0000698/*
699 * skb might have a dst pointer attached, refcounted or not.
700 * _skb_refdst low order bit is set if refcount was _not_ taken
701 */
702#define SKB_DST_NOREF 1UL
703#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
704
705/**
706 * skb_dst - returns skb dst_entry
707 * @skb: buffer
708 *
709 * Returns skb dst_entry, regardless of reference taken or not.
710 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000711static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
712{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000713 /* If refdst was not refcounted, check we still are in a
714 * rcu_read_lock section
715 */
716 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
717 !rcu_read_lock_held() &&
718 !rcu_read_lock_bh_held());
719 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000720}
721
Eric Dumazet7fee2262010-05-11 23:19:48 +0000722/**
723 * skb_dst_set - sets skb dst
724 * @skb: buffer
725 * @dst: dst entry
726 *
727 * Sets skb dst, assuming a reference was taken on dst and should
728 * be released by skb_dst_drop()
729 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000730static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
731{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000732 skb->_skb_refdst = (unsigned long)dst;
733}
734
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200735/**
736 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
737 * @skb: buffer
738 * @dst: dst entry
739 *
740 * Sets skb dst, assuming a reference was not taken on dst.
741 * If dst entry is cached, we do not take reference and dst_release
742 * will be avoided by refdst_drop. If dst entry is not cached, we take
743 * reference, so that last dst_release can destroy the dst immediately.
744 */
745static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
746{
Hannes Frederic Sowadbfc4fb2014-12-06 19:19:42 +0100747 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
748 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200749}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000750
751/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300752 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000753 * @skb: buffer
754 */
755static inline bool skb_dst_is_noref(const struct sk_buff *skb)
756{
757 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000758}
759
Eric Dumazet511c3f92009-06-02 05:14:27 +0000760static inline struct rtable *skb_rtable(const struct sk_buff *skb)
761{
Eric Dumazetadf30902009-06-02 05:19:30 +0000762 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000763}
764
Joe Perches7965bd42013-09-26 14:48:15 -0700765void kfree_skb(struct sk_buff *skb);
766void kfree_skb_list(struct sk_buff *segs);
767void skb_tx_error(struct sk_buff *skb);
768void consume_skb(struct sk_buff *skb);
769void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000770extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000771
Joe Perches7965bd42013-09-26 14:48:15 -0700772void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
773bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
774 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000775
Joe Perches7965bd42013-09-26 14:48:15 -0700776struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
777 int node);
Eric Dumazet2ea2f622015-04-24 16:05:01 -0700778struct sk_buff *__build_skb(void *data, unsigned int frag_size);
Joe Perches7965bd42013-09-26 14:48:15 -0700779struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700780static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100781 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700782{
Eric Dumazet564824b2010-10-11 19:05:25 +0000783 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700784}
785
Eric Dumazet2e4e4412014-09-17 04:49:49 -0700786struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
787 unsigned long data_len,
788 int max_page_order,
789 int *errcode,
790 gfp_t gfp_mask);
791
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700792/* Layout of fast clones : [skb1][skb2][fclone_ref] */
793struct sk_buff_fclones {
794 struct sk_buff skb1;
795
796 struct sk_buff skb2;
797
798 atomic_t fclone_ref;
799};
800
801/**
802 * skb_fclone_busy - check if fclone is busy
803 * @skb: buffer
804 *
805 * Returns true is skb is a fast clone, and its clone is not freed.
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700806 * Some drivers call skb_orphan() in their ndo_start_xmit(),
807 * so we also check that this didnt happen.
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700808 */
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700809static inline bool skb_fclone_busy(const struct sock *sk,
810 const struct sk_buff *skb)
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700811{
812 const struct sk_buff_fclones *fclones;
813
814 fclones = container_of(skb, struct sk_buff_fclones, skb1);
815
816 return skb->fclone == SKB_FCLONE_ORIG &&
Eric Dumazet6ffe75e2014-12-03 17:04:39 -0800817 atomic_read(&fclones->fclone_ref) > 1 &&
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700818 fclones->skb2.sk == sk;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700819}
820
David S. Millerd179cd12005-08-17 14:57:30 -0700821static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100822 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700823{
Mel Gormanc93bdd02012-07-31 16:44:19 -0700824 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700825}
826
Joe Perches7965bd42013-09-26 14:48:15 -0700827struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
Patrick McHardy0ebd0ac2013-04-17 06:46:58 +0000828static inline struct sk_buff *alloc_skb_head(gfp_t priority)
829{
830 return __alloc_skb_head(priority, -1);
831}
832
Joe Perches7965bd42013-09-26 14:48:15 -0700833struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
834int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
835struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
836struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
Octavian Purdilabad93e92014-06-12 01:36:26 +0300837struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
838 gfp_t gfp_mask, bool fclone);
839static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
840 gfp_t gfp_mask)
841{
842 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
843}
Eric Dumazet117632e2011-12-03 21:39:53 +0000844
Joe Perches7965bd42013-09-26 14:48:15 -0700845int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
846struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
847 unsigned int headroom);
848struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
849 int newtailroom, gfp_t priority);
Fan Du25a91d82014-01-18 09:54:23 +0800850int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
851 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -0700852int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
853 int len);
854int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
855int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000856#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700857
Joe Perches7965bd42013-09-26 14:48:15 -0700858int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
859 int getfrag(void *from, char *to, int offset,
860 int len, int odd, struct sk_buff *skb),
861 void *from, int length);
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700862
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800863struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700864 __u32 lower_offset;
865 __u32 upper_offset;
866 __u32 frag_idx;
867 __u32 stepped_offset;
868 struct sk_buff *root_skb;
869 struct sk_buff *cur_skb;
870 __u8 *frag_data;
871};
872
Joe Perches7965bd42013-09-26 14:48:15 -0700873void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
874 unsigned int to, struct skb_seq_state *st);
875unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
876 struct skb_seq_state *st);
877void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -0700878
Joe Perches7965bd42013-09-26 14:48:15 -0700879unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
Bojan Prtvar059a2442015-02-22 11:46:35 +0100880 unsigned int to, struct ts_config *config);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700881
Tom Herbert09323cc2013-12-15 22:16:19 -0800882/*
883 * Packet hash types specify the type of hash in skb_set_hash.
884 *
885 * Hash types refer to the protocol layer addresses which are used to
886 * construct a packet's hash. The hashes are used to differentiate or identify
887 * flows of the protocol layer for the hash type. Hash types are either
888 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
889 *
890 * Properties of hashes:
891 *
892 * 1) Two packets in different flows have different hash values
893 * 2) Two packets in the same flow should have the same hash value
894 *
895 * A hash at a higher layer is considered to be more specific. A driver should
896 * set the most specific hash possible.
897 *
898 * A driver cannot indicate a more specific hash than the layer at which a hash
899 * was computed. For instance an L3 hash cannot be set as an L4 hash.
900 *
901 * A driver may indicate a hash level which is less specific than the
902 * actual layer the hash was computed on. For instance, a hash computed
903 * at L4 may be considered an L3 hash. This should only be done if the
904 * driver can't unambiguously determine that the HW computed the hash at
905 * the higher layer. Note that the "should" in the second property above
906 * permits this.
907 */
908enum pkt_hash_types {
909 PKT_HASH_TYPE_NONE, /* Undefined type */
910 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
911 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
912 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
913};
914
915static inline void
916skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
917{
Tom Herbert61b905d2014-03-24 15:34:47 -0700918 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
Tom Herberta3b18dd2014-07-01 21:33:17 -0700919 skb->sw_hash = 0;
Tom Herbert61b905d2014-03-24 15:34:47 -0700920 skb->hash = hash;
Tom Herbert09323cc2013-12-15 22:16:19 -0800921}
922
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800923void __skb_get_hash(struct sk_buff *skb);
924static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000925{
Tom Herberta3b18dd2014-07-01 21:33:17 -0700926 if (!skb->l4_hash && !skb->sw_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800927 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000928
Tom Herbert61b905d2014-03-24 15:34:47 -0700929 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000930}
931
Tom Herbert50fb7992015-05-01 11:30:12 -0700932__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
933
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800934static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
935{
Tom Herbert61b905d2014-03-24 15:34:47 -0700936 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800937}
938
Tom Herbert7539fad2013-12-15 22:12:18 -0800939static inline void skb_clear_hash(struct sk_buff *skb)
940{
Tom Herbert61b905d2014-03-24 15:34:47 -0700941 skb->hash = 0;
Tom Herberta3b18dd2014-07-01 21:33:17 -0700942 skb->sw_hash = 0;
Tom Herbert61b905d2014-03-24 15:34:47 -0700943 skb->l4_hash = 0;
Tom Herbert7539fad2013-12-15 22:12:18 -0800944}
945
946static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
947{
Tom Herbert61b905d2014-03-24 15:34:47 -0700948 if (!skb->l4_hash)
Tom Herbert7539fad2013-12-15 22:12:18 -0800949 skb_clear_hash(skb);
950}
951
Tom Herbert3df7a742013-12-15 22:16:29 -0800952static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
953{
Tom Herbert61b905d2014-03-24 15:34:47 -0700954 to->hash = from->hash;
Tom Herberta3b18dd2014-07-01 21:33:17 -0700955 to->sw_hash = from->sw_hash;
Tom Herbert61b905d2014-03-24 15:34:47 -0700956 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -0800957};
958
Eric Dumazetc29390c2015-03-11 18:42:02 -0700959static inline void skb_sender_cpu_clear(struct sk_buff *skb)
960{
961#ifdef CONFIG_XPS
962 skb->sender_cpu = 0;
963#endif
964}
965
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700966#ifdef NET_SKBUFF_DATA_USES_OFFSET
967static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
968{
969 return skb->head + skb->end;
970}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000971
972static inline unsigned int skb_end_offset(const struct sk_buff *skb)
973{
974 return skb->end;
975}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700976#else
977static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
978{
979 return skb->end;
980}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000981
982static inline unsigned int skb_end_offset(const struct sk_buff *skb)
983{
984 return skb->end - skb->head;
985}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700986#endif
987
Linus Torvalds1da177e2005-04-16 15:20:36 -0700988/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700989#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990
Patrick Ohlyac45f602009-02-12 05:03:37 +0000991static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
992{
993 return &skb_shinfo(skb)->hwtstamps;
994}
995
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996/**
997 * skb_queue_empty - check if a queue is empty
998 * @list: queue head
999 *
1000 * Returns true if the queue is empty, false otherwise.
1001 */
1002static inline int skb_queue_empty(const struct sk_buff_head *list)
1003{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001004 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005}
1006
1007/**
David S. Millerfc7ebb22008-09-23 00:34:07 -07001008 * skb_queue_is_last - check if skb is the last entry in the queue
1009 * @list: queue head
1010 * @skb: buffer
1011 *
1012 * Returns true if @skb is the last buffer on the list.
1013 */
1014static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1015 const struct sk_buff *skb)
1016{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001017 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -07001018}
1019
1020/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001021 * skb_queue_is_first - check if skb is the first entry in the queue
1022 * @list: queue head
1023 * @skb: buffer
1024 *
1025 * Returns true if @skb is the first buffer on the list.
1026 */
1027static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1028 const struct sk_buff *skb)
1029{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001030 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001031}
1032
1033/**
David S. Miller249c8b42008-09-23 00:44:42 -07001034 * skb_queue_next - return the next packet in the queue
1035 * @list: queue head
1036 * @skb: current buffer
1037 *
1038 * Return the next packet in @list after @skb. It is only valid to
1039 * call this if skb_queue_is_last() evaluates to false.
1040 */
1041static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1042 const struct sk_buff *skb)
1043{
1044 /* This BUG_ON may seem severe, but if we just return then we
1045 * are going to dereference garbage.
1046 */
1047 BUG_ON(skb_queue_is_last(list, skb));
1048 return skb->next;
1049}
1050
1051/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001052 * skb_queue_prev - return the prev packet in the queue
1053 * @list: queue head
1054 * @skb: current buffer
1055 *
1056 * Return the prev packet in @list before @skb. It is only valid to
1057 * call this if skb_queue_is_first() evaluates to false.
1058 */
1059static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1060 const struct sk_buff *skb)
1061{
1062 /* This BUG_ON may seem severe, but if we just return then we
1063 * are going to dereference garbage.
1064 */
1065 BUG_ON(skb_queue_is_first(list, skb));
1066 return skb->prev;
1067}
1068
1069/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070 * skb_get - reference buffer
1071 * @skb: buffer to reference
1072 *
1073 * Makes another reference to a socket buffer and returns a pointer
1074 * to the buffer.
1075 */
1076static inline struct sk_buff *skb_get(struct sk_buff *skb)
1077{
1078 atomic_inc(&skb->users);
1079 return skb;
1080}
1081
1082/*
1083 * If users == 1, we are the only owner and are can avoid redundant
1084 * atomic change.
1085 */
1086
1087/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001088 * skb_cloned - is the buffer a clone
1089 * @skb: buffer to check
1090 *
1091 * Returns true if the buffer was generated with skb_clone() and is
1092 * one of multiple shared copies of the buffer. Cloned buffers are
1093 * shared data so must not be written to under normal circumstances.
1094 */
1095static inline int skb_cloned(const struct sk_buff *skb)
1096{
1097 return skb->cloned &&
1098 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1099}
1100
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001101static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1102{
1103 might_sleep_if(pri & __GFP_WAIT);
1104
1105 if (skb_cloned(skb))
1106 return pskb_expand_head(skb, 0, 0, pri);
1107
1108 return 0;
1109}
1110
Linus Torvalds1da177e2005-04-16 15:20:36 -07001111/**
1112 * skb_header_cloned - is the header a clone
1113 * @skb: buffer to check
1114 *
1115 * Returns true if modifying the header part of the buffer requires
1116 * the data to be copied.
1117 */
1118static inline int skb_header_cloned(const struct sk_buff *skb)
1119{
1120 int dataref;
1121
1122 if (!skb->cloned)
1123 return 0;
1124
1125 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1126 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1127 return dataref != 1;
1128}
1129
1130/**
1131 * skb_header_release - release reference to header
1132 * @skb: buffer to operate on
1133 *
1134 * Drop a reference to the header part of the buffer. This is done
1135 * by acquiring a payload reference. You must not read from the header
1136 * part of skb->data after this.
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001137 * Note : Check if you can use __skb_header_release() instead.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138 */
1139static inline void skb_header_release(struct sk_buff *skb)
1140{
1141 BUG_ON(skb->nohdr);
1142 skb->nohdr = 1;
1143 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1144}
1145
1146/**
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001147 * __skb_header_release - release reference to header
1148 * @skb: buffer to operate on
1149 *
1150 * Variant of skb_header_release() assuming skb is private to caller.
1151 * We can avoid one atomic operation.
1152 */
1153static inline void __skb_header_release(struct sk_buff *skb)
1154{
1155 skb->nohdr = 1;
1156 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1157}
1158
1159
1160/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161 * skb_shared - is the buffer shared
1162 * @skb: buffer to check
1163 *
1164 * Returns true if more than one person has a reference to this
1165 * buffer.
1166 */
1167static inline int skb_shared(const struct sk_buff *skb)
1168{
1169 return atomic_read(&skb->users) != 1;
1170}
1171
1172/**
1173 * skb_share_check - check if buffer is shared and if so clone it
1174 * @skb: buffer to check
1175 * @pri: priority for memory allocation
1176 *
1177 * If the buffer is shared the buffer is cloned and the old copy
1178 * drops a reference. A new clone with a single reference is returned.
1179 * If the buffer is not shared the original buffer is returned. When
1180 * being called from interrupt status or with spinlocks held pri must
1181 * be GFP_ATOMIC.
1182 *
1183 * NULL is returned on a memory allocation failure.
1184 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001185static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001186{
1187 might_sleep_if(pri & __GFP_WAIT);
1188 if (skb_shared(skb)) {
1189 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001190
1191 if (likely(nskb))
1192 consume_skb(skb);
1193 else
1194 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001195 skb = nskb;
1196 }
1197 return skb;
1198}
1199
1200/*
1201 * Copy shared buffers into a new sk_buff. We effectively do COW on
1202 * packets to handle cases where we have a local reader and forward
1203 * and a couple of other messy ones. The normal one is tcpdumping
1204 * a packet thats being forwarded.
1205 */
1206
1207/**
1208 * skb_unshare - make a copy of a shared buffer
1209 * @skb: buffer to check
1210 * @pri: priority for memory allocation
1211 *
1212 * If the socket buffer is a clone then this function creates a new
1213 * copy of the data, drops a reference count on the old copy and returns
1214 * the new copy with the reference count at 1. If the buffer is not a clone
1215 * the original buffer is returned. When called with a spinlock held or
1216 * from interrupt state @pri must be %GFP_ATOMIC
1217 *
1218 * %NULL is returned on a memory allocation failure.
1219 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001220static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001221 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001222{
1223 might_sleep_if(pri & __GFP_WAIT);
1224 if (skb_cloned(skb)) {
1225 struct sk_buff *nskb = skb_copy(skb, pri);
Alexander Aring31eff812014-10-10 23:10:47 +02001226
1227 /* Free our shared copy */
1228 if (likely(nskb))
1229 consume_skb(skb);
1230 else
1231 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001232 skb = nskb;
1233 }
1234 return skb;
1235}
1236
1237/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001238 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001239 * @list_: list to peek at
1240 *
1241 * Peek an &sk_buff. Unlike most other operations you _MUST_
1242 * be careful with this one. A peek leaves the buffer on the
1243 * list and someone else may run off with it. You must hold
1244 * the appropriate locks or have a private queue to do this.
1245 *
1246 * Returns %NULL for an empty list or a pointer to the head element.
1247 * The reference count is not incremented and the reference is therefore
1248 * volatile. Use with caution.
1249 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001250static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001251{
Eric Dumazet18d07002012-04-30 16:31:46 +00001252 struct sk_buff *skb = list_->next;
1253
1254 if (skb == (struct sk_buff *)list_)
1255 skb = NULL;
1256 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001257}
1258
1259/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001260 * skb_peek_next - peek skb following the given one from a queue
1261 * @skb: skb to start from
1262 * @list_: list to peek at
1263 *
1264 * Returns %NULL when the end of the list is met or a pointer to the
1265 * next element. The reference count is not incremented and the
1266 * reference is therefore volatile. Use with caution.
1267 */
1268static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1269 const struct sk_buff_head *list_)
1270{
1271 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001272
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001273 if (next == (struct sk_buff *)list_)
1274 next = NULL;
1275 return next;
1276}
1277
1278/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001279 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001280 * @list_: list to peek at
1281 *
1282 * Peek an &sk_buff. Unlike most other operations you _MUST_
1283 * be careful with this one. A peek leaves the buffer on the
1284 * list and someone else may run off with it. You must hold
1285 * the appropriate locks or have a private queue to do this.
1286 *
1287 * Returns %NULL for an empty list or a pointer to the tail element.
1288 * The reference count is not incremented and the reference is therefore
1289 * volatile. Use with caution.
1290 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001291static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292{
Eric Dumazet18d07002012-04-30 16:31:46 +00001293 struct sk_buff *skb = list_->prev;
1294
1295 if (skb == (struct sk_buff *)list_)
1296 skb = NULL;
1297 return skb;
1298
Linus Torvalds1da177e2005-04-16 15:20:36 -07001299}
1300
1301/**
1302 * skb_queue_len - get queue length
1303 * @list_: list to measure
1304 *
1305 * Return the length of an &sk_buff queue.
1306 */
1307static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1308{
1309 return list_->qlen;
1310}
1311
David S. Miller67fed452008-09-21 22:36:24 -07001312/**
1313 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1314 * @list: queue to initialize
1315 *
1316 * This initializes only the list and queue length aspects of
1317 * an sk_buff_head object. This allows to initialize the list
1318 * aspects of an sk_buff_head without reinitializing things like
1319 * the spinlock. It can also be used for on-stack sk_buff_head
1320 * objects where the spinlock is known to not be used.
1321 */
1322static inline void __skb_queue_head_init(struct sk_buff_head *list)
1323{
1324 list->prev = list->next = (struct sk_buff *)list;
1325 list->qlen = 0;
1326}
1327
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001328/*
1329 * This function creates a split out lock class for each invocation;
1330 * this is needed for now since a whole lot of users of the skb-queue
1331 * infrastructure in drivers have different locking usage (in hardirq)
1332 * than the networking core (in softirq only). In the long run either the
1333 * network layer or drivers should need annotation to consolidate the
1334 * main types of usage into 3 classes.
1335 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336static inline void skb_queue_head_init(struct sk_buff_head *list)
1337{
1338 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001339 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001340}
1341
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001342static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1343 struct lock_class_key *class)
1344{
1345 skb_queue_head_init(list);
1346 lockdep_set_class(&list->lock, class);
1347}
1348
Linus Torvalds1da177e2005-04-16 15:20:36 -07001349/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001350 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001351 *
1352 * The "__skb_xxxx()" functions are the non-atomic ones that
1353 * can only be called with interrupts disabled.
1354 */
Joe Perches7965bd42013-09-26 14:48:15 -07001355void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1356 struct sk_buff_head *list);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001357static inline void __skb_insert(struct sk_buff *newsk,
1358 struct sk_buff *prev, struct sk_buff *next,
1359 struct sk_buff_head *list)
1360{
1361 newsk->next = next;
1362 newsk->prev = prev;
1363 next->prev = prev->next = newsk;
1364 list->qlen++;
1365}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001366
David S. Miller67fed452008-09-21 22:36:24 -07001367static inline void __skb_queue_splice(const struct sk_buff_head *list,
1368 struct sk_buff *prev,
1369 struct sk_buff *next)
1370{
1371 struct sk_buff *first = list->next;
1372 struct sk_buff *last = list->prev;
1373
1374 first->prev = prev;
1375 prev->next = first;
1376
1377 last->next = next;
1378 next->prev = last;
1379}
1380
1381/**
1382 * skb_queue_splice - join two skb lists, this is designed for stacks
1383 * @list: the new list to add
1384 * @head: the place to add it in the first list
1385 */
1386static inline void skb_queue_splice(const struct sk_buff_head *list,
1387 struct sk_buff_head *head)
1388{
1389 if (!skb_queue_empty(list)) {
1390 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001391 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001392 }
1393}
1394
1395/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001396 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001397 * @list: the new list to add
1398 * @head: the place to add it in the first list
1399 *
1400 * The list at @list is reinitialised
1401 */
1402static inline void skb_queue_splice_init(struct sk_buff_head *list,
1403 struct sk_buff_head *head)
1404{
1405 if (!skb_queue_empty(list)) {
1406 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001407 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001408 __skb_queue_head_init(list);
1409 }
1410}
1411
1412/**
1413 * skb_queue_splice_tail - join two skb lists, each list being a queue
1414 * @list: the new list to add
1415 * @head: the place to add it in the first list
1416 */
1417static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1418 struct sk_buff_head *head)
1419{
1420 if (!skb_queue_empty(list)) {
1421 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001422 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001423 }
1424}
1425
1426/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001427 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001428 * @list: the new list to add
1429 * @head: the place to add it in the first list
1430 *
1431 * Each of the lists is a queue.
1432 * The list at @list is reinitialised
1433 */
1434static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1435 struct sk_buff_head *head)
1436{
1437 if (!skb_queue_empty(list)) {
1438 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001439 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001440 __skb_queue_head_init(list);
1441 }
1442}
1443
Linus Torvalds1da177e2005-04-16 15:20:36 -07001444/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001445 * __skb_queue_after - queue a buffer at the list head
1446 * @list: list to use
1447 * @prev: place after this buffer
1448 * @newsk: buffer to queue
1449 *
1450 * Queue a buffer int the middle of a list. This function takes no locks
1451 * and you must therefore hold required locks before calling it.
1452 *
1453 * A buffer cannot be placed on two lists at the same time.
1454 */
1455static inline void __skb_queue_after(struct sk_buff_head *list,
1456 struct sk_buff *prev,
1457 struct sk_buff *newsk)
1458{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001459 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001460}
1461
Joe Perches7965bd42013-09-26 14:48:15 -07001462void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1463 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001464
Gerrit Renkerf5572852008-04-14 00:05:28 -07001465static inline void __skb_queue_before(struct sk_buff_head *list,
1466 struct sk_buff *next,
1467 struct sk_buff *newsk)
1468{
1469 __skb_insert(newsk, next->prev, next, list);
1470}
1471
Stephen Hemminger300ce172005-10-30 13:47:34 -08001472/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001473 * __skb_queue_head - queue a buffer at the list head
1474 * @list: list to use
1475 * @newsk: buffer to queue
1476 *
1477 * Queue a buffer at the start of a list. This function takes no locks
1478 * and you must therefore hold required locks before calling it.
1479 *
1480 * A buffer cannot be placed on two lists at the same time.
1481 */
Joe Perches7965bd42013-09-26 14:48:15 -07001482void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001483static inline void __skb_queue_head(struct sk_buff_head *list,
1484 struct sk_buff *newsk)
1485{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001486 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001487}
1488
1489/**
1490 * __skb_queue_tail - queue a buffer at the list tail
1491 * @list: list to use
1492 * @newsk: buffer to queue
1493 *
1494 * Queue a buffer at the end of a list. This function takes no locks
1495 * and you must therefore hold required locks before calling it.
1496 *
1497 * A buffer cannot be placed on two lists at the same time.
1498 */
Joe Perches7965bd42013-09-26 14:48:15 -07001499void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001500static inline void __skb_queue_tail(struct sk_buff_head *list,
1501 struct sk_buff *newsk)
1502{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001503 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001504}
1505
Linus Torvalds1da177e2005-04-16 15:20:36 -07001506/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507 * remove sk_buff from list. _Must_ be called atomically, and with
1508 * the list known..
1509 */
Joe Perches7965bd42013-09-26 14:48:15 -07001510void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001511static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1512{
1513 struct sk_buff *next, *prev;
1514
1515 list->qlen--;
1516 next = skb->next;
1517 prev = skb->prev;
1518 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519 next->prev = prev;
1520 prev->next = next;
1521}
1522
Gerrit Renkerf525c062008-04-14 00:04:12 -07001523/**
1524 * __skb_dequeue - remove from the head of the queue
1525 * @list: list to dequeue from
1526 *
1527 * Remove the head of the list. This function does not take any locks
1528 * so must be used with appropriate locks held only. The head item is
1529 * returned or %NULL if the list is empty.
1530 */
Joe Perches7965bd42013-09-26 14:48:15 -07001531struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Gerrit Renkerf525c062008-04-14 00:04:12 -07001532static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1533{
1534 struct sk_buff *skb = skb_peek(list);
1535 if (skb)
1536 __skb_unlink(skb, list);
1537 return skb;
1538}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539
1540/**
1541 * __skb_dequeue_tail - remove from the tail of the queue
1542 * @list: list to dequeue from
1543 *
1544 * Remove the tail of the list. This function does not take any locks
1545 * so must be used with appropriate locks held only. The tail item is
1546 * returned or %NULL if the list is empty.
1547 */
Joe Perches7965bd42013-09-26 14:48:15 -07001548struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001549static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1550{
1551 struct sk_buff *skb = skb_peek_tail(list);
1552 if (skb)
1553 __skb_unlink(skb, list);
1554 return skb;
1555}
1556
1557
David S. Millerbdcc0922012-03-07 20:53:36 -05001558static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001559{
1560 return skb->data_len;
1561}
1562
1563static inline unsigned int skb_headlen(const struct sk_buff *skb)
1564{
1565 return skb->len - skb->data_len;
1566}
1567
1568static inline int skb_pagelen(const struct sk_buff *skb)
1569{
1570 int i, len = 0;
1571
1572 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001573 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001574 return len + skb_headlen(skb);
1575}
1576
Ian Campbell131ea662011-08-19 06:25:00 +00001577/**
1578 * __skb_fill_page_desc - initialise a paged fragment in an skb
1579 * @skb: buffer containing fragment to be initialised
1580 * @i: paged fragment index to initialise
1581 * @page: the page to use for this fragment
1582 * @off: the offset to the data with @page
1583 * @size: the length of the data
1584 *
1585 * Initialises the @i'th fragment of @skb to point to &size bytes at
1586 * offset @off within @page.
1587 *
1588 * Does not take any additional reference on the fragment.
1589 */
1590static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1591 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001592{
1593 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1594
Mel Gormanc48a11c2012-07-31 16:44:23 -07001595 /*
1596 * Propagate page->pfmemalloc to the skb if we can. The problem is
1597 * that not all callers have unique ownership of the page. If
1598 * pfmemalloc is set, we check the mapping as a mapping implies
1599 * page->index is set (index and pfmemalloc share space).
1600 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1601 * do not lose pfmemalloc information as the pages would not be
1602 * allocated using __GFP_MEMALLOC.
1603 */
Ian Campbella8605c62011-10-19 23:01:49 +00001604 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001605 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001606 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001607
1608 page = compound_head(page);
1609 if (page->pfmemalloc && !page->mapping)
1610 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00001611}
1612
1613/**
1614 * skb_fill_page_desc - initialise a paged fragment in an skb
1615 * @skb: buffer containing fragment to be initialised
1616 * @i: paged fragment index to initialise
1617 * @page: the page to use for this fragment
1618 * @off: the offset to the data with @page
1619 * @size: the length of the data
1620 *
1621 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01001622 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00001623 * addition updates @skb such that @i is the last fragment.
1624 *
1625 * Does not take any additional reference on the fragment.
1626 */
1627static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1628 struct page *page, int off, int size)
1629{
1630 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001631 skb_shinfo(skb)->nr_frags = i + 1;
1632}
1633
Joe Perches7965bd42013-09-26 14:48:15 -07001634void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1635 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001636
Jason Wangf8e617e2013-11-01 14:07:47 +08001637void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1638 unsigned int truesize);
1639
Linus Torvalds1da177e2005-04-16 15:20:36 -07001640#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001641#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001642#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1643
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001644#ifdef NET_SKBUFF_DATA_USES_OFFSET
1645static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1646{
1647 return skb->head + skb->tail;
1648}
1649
1650static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1651{
1652 skb->tail = skb->data - skb->head;
1653}
1654
1655static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1656{
1657 skb_reset_tail_pointer(skb);
1658 skb->tail += offset;
1659}
Simon Horman7cc46192013-05-28 20:34:29 +00001660
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001661#else /* NET_SKBUFF_DATA_USES_OFFSET */
1662static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1663{
1664 return skb->tail;
1665}
1666
1667static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1668{
1669 skb->tail = skb->data;
1670}
1671
1672static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1673{
1674 skb->tail = skb->data + offset;
1675}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001676
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001677#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1678
Linus Torvalds1da177e2005-04-16 15:20:36 -07001679/*
1680 * Add data to an sk_buff
1681 */
Mathias Krause0c7ddf32013-11-07 14:18:24 +01001682unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001683unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001684static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1685{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001686 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687 SKB_LINEAR_ASSERT(skb);
1688 skb->tail += len;
1689 skb->len += len;
1690 return tmp;
1691}
1692
Joe Perches7965bd42013-09-26 14:48:15 -07001693unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001694static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1695{
1696 skb->data -= len;
1697 skb->len += len;
1698 return skb->data;
1699}
1700
Joe Perches7965bd42013-09-26 14:48:15 -07001701unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001702static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1703{
1704 skb->len -= len;
1705 BUG_ON(skb->len < skb->data_len);
1706 return skb->data += len;
1707}
1708
David S. Miller47d29642010-05-02 02:21:44 -07001709static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1710{
1711 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1712}
1713
Joe Perches7965bd42013-09-26 14:48:15 -07001714unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001715
1716static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1717{
1718 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001719 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001720 return NULL;
1721 skb->len -= len;
1722 return skb->data += len;
1723}
1724
1725static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1726{
1727 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1728}
1729
1730static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1731{
1732 if (likely(len <= skb_headlen(skb)))
1733 return 1;
1734 if (unlikely(len > skb->len))
1735 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001736 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001737}
1738
1739/**
1740 * skb_headroom - bytes at buffer head
1741 * @skb: buffer to check
1742 *
1743 * Return the number of bytes of free space at the head of an &sk_buff.
1744 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001745static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746{
1747 return skb->data - skb->head;
1748}
1749
1750/**
1751 * skb_tailroom - bytes at buffer end
1752 * @skb: buffer to check
1753 *
1754 * Return the number of bytes of free space at the tail of an sk_buff
1755 */
1756static inline int skb_tailroom(const struct sk_buff *skb)
1757{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001758 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001759}
1760
1761/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00001762 * skb_availroom - bytes at buffer end
1763 * @skb: buffer to check
1764 *
1765 * Return the number of bytes of free space at the tail of an sk_buff
1766 * allocated by sk_stream_alloc()
1767 */
1768static inline int skb_availroom(const struct sk_buff *skb)
1769{
Eric Dumazet16fad692013-03-14 05:40:32 +00001770 if (skb_is_nonlinear(skb))
1771 return 0;
1772
1773 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00001774}
1775
1776/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001777 * skb_reserve - adjust headroom
1778 * @skb: buffer to alter
1779 * @len: bytes to move
1780 *
1781 * Increase the headroom of an empty &sk_buff by reducing the tail
1782 * room. This is only allowed for an empty buffer.
1783 */
David S. Miller8243126c2006-01-17 02:54:21 -08001784static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001785{
1786 skb->data += len;
1787 skb->tail += len;
1788}
1789
Tom Herbert8bce6d72014-09-29 20:22:29 -07001790#define ENCAP_TYPE_ETHER 0
1791#define ENCAP_TYPE_IPPROTO 1
1792
1793static inline void skb_set_inner_protocol(struct sk_buff *skb,
1794 __be16 protocol)
1795{
1796 skb->inner_protocol = protocol;
1797 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
1798}
1799
1800static inline void skb_set_inner_ipproto(struct sk_buff *skb,
1801 __u8 ipproto)
1802{
1803 skb->inner_ipproto = ipproto;
1804 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
1805}
1806
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001807static inline void skb_reset_inner_headers(struct sk_buff *skb)
1808{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001809 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001810 skb->inner_network_header = skb->network_header;
1811 skb->inner_transport_header = skb->transport_header;
1812}
1813
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00001814static inline void skb_reset_mac_len(struct sk_buff *skb)
1815{
1816 skb->mac_len = skb->network_header - skb->mac_header;
1817}
1818
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001819static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1820 *skb)
1821{
1822 return skb->head + skb->inner_transport_header;
1823}
1824
1825static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1826{
1827 skb->inner_transport_header = skb->data - skb->head;
1828}
1829
1830static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1831 const int offset)
1832{
1833 skb_reset_inner_transport_header(skb);
1834 skb->inner_transport_header += offset;
1835}
1836
1837static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1838{
1839 return skb->head + skb->inner_network_header;
1840}
1841
1842static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1843{
1844 skb->inner_network_header = skb->data - skb->head;
1845}
1846
1847static inline void skb_set_inner_network_header(struct sk_buff *skb,
1848 const int offset)
1849{
1850 skb_reset_inner_network_header(skb);
1851 skb->inner_network_header += offset;
1852}
1853
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001854static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1855{
1856 return skb->head + skb->inner_mac_header;
1857}
1858
1859static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1860{
1861 skb->inner_mac_header = skb->data - skb->head;
1862}
1863
1864static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1865 const int offset)
1866{
1867 skb_reset_inner_mac_header(skb);
1868 skb->inner_mac_header += offset;
1869}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001870static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1871{
Cong Wang35d04612013-05-29 15:16:05 +08001872 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001873}
1874
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001875static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1876{
1877 return skb->head + skb->transport_header;
1878}
1879
1880static inline void skb_reset_transport_header(struct sk_buff *skb)
1881{
1882 skb->transport_header = skb->data - skb->head;
1883}
1884
1885static inline void skb_set_transport_header(struct sk_buff *skb,
1886 const int offset)
1887{
1888 skb_reset_transport_header(skb);
1889 skb->transport_header += offset;
1890}
1891
1892static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1893{
1894 return skb->head + skb->network_header;
1895}
1896
1897static inline void skb_reset_network_header(struct sk_buff *skb)
1898{
1899 skb->network_header = skb->data - skb->head;
1900}
1901
1902static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1903{
1904 skb_reset_network_header(skb);
1905 skb->network_header += offset;
1906}
1907
1908static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1909{
1910 return skb->head + skb->mac_header;
1911}
1912
1913static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1914{
Cong Wang35d04612013-05-29 15:16:05 +08001915 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001916}
1917
1918static inline void skb_reset_mac_header(struct sk_buff *skb)
1919{
1920 skb->mac_header = skb->data - skb->head;
1921}
1922
1923static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1924{
1925 skb_reset_mac_header(skb);
1926 skb->mac_header += offset;
1927}
1928
Timo Teräs0e3da5b2013-12-16 11:02:09 +02001929static inline void skb_pop_mac_header(struct sk_buff *skb)
1930{
1931 skb->mac_header = skb->network_header;
1932}
1933
Ying Xuefbbdb8f2013-03-27 16:46:06 +00001934static inline void skb_probe_transport_header(struct sk_buff *skb,
1935 const int offset_hint)
1936{
1937 struct flow_keys keys;
1938
1939 if (skb_transport_header_was_set(skb))
1940 return;
1941 else if (skb_flow_dissect(skb, &keys))
1942 skb_set_transport_header(skb, keys.thoff);
1943 else
1944 skb_set_transport_header(skb, offset_hint);
1945}
1946
Eric Dumazet03606892012-02-23 10:55:02 +00001947static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1948{
1949 if (skb_mac_header_was_set(skb)) {
1950 const unsigned char *old_mac = skb_mac_header(skb);
1951
1952 skb_set_mac_header(skb, -skb->mac_len);
1953 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1954 }
1955}
1956
Michał Mirosław04fb4512010-12-14 15:24:08 +00001957static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1958{
1959 return skb->csum_start - skb_headroom(skb);
1960}
1961
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001962static inline int skb_transport_offset(const struct sk_buff *skb)
1963{
1964 return skb_transport_header(skb) - skb->data;
1965}
1966
1967static inline u32 skb_network_header_len(const struct sk_buff *skb)
1968{
1969 return skb->transport_header - skb->network_header;
1970}
1971
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001972static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1973{
1974 return skb->inner_transport_header - skb->inner_network_header;
1975}
1976
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001977static inline int skb_network_offset(const struct sk_buff *skb)
1978{
1979 return skb_network_header(skb) - skb->data;
1980}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001981
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001982static inline int skb_inner_network_offset(const struct sk_buff *skb)
1983{
1984 return skb_inner_network_header(skb) - skb->data;
1985}
1986
Changli Gaof9599ce2010-08-04 04:43:44 +00001987static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1988{
1989 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1990}
1991
Linus Torvalds1da177e2005-04-16 15:20:36 -07001992/*
1993 * CPUs often take a performance hit when accessing unaligned memory
1994 * locations. The actual performance hit varies, it can be small if the
1995 * hardware handles it or large if we have to take an exception and fix it
1996 * in software.
1997 *
1998 * Since an ethernet header is 14 bytes network drivers often end up with
1999 * the IP header at an unaligned offset. The IP header can be aligned by
2000 * shifting the start of the packet by 2 bytes. Drivers should do this
2001 * with:
2002 *
Tobias Klauser8660c122009-07-13 22:48:16 +00002003 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002004 *
2005 * The downside to this alignment of the IP header is that the DMA is now
2006 * unaligned. On some architectures the cost of an unaligned DMA is high
2007 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00002008 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2010 * to be overridden.
2011 */
2012#ifndef NET_IP_ALIGN
2013#define NET_IP_ALIGN 2
2014#endif
2015
Anton Blanchard025be812006-03-31 02:27:06 -08002016/*
2017 * The networking layer reserves some headroom in skb data (via
2018 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2019 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08002020 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08002021 *
2022 * Unfortunately this headroom changes the DMA alignment of the resulting
2023 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2024 * on some architectures. An architecture can override this value,
2025 * perhaps setting it to a cacheline in size (since that will maintain
2026 * cacheline alignment of the DMA). It must be a power of 2.
2027 *
David S. Millerd6301d32009-02-08 19:24:13 -08002028 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08002029 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07002030 *
2031 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2032 * to reduce average number of cache lines per packet.
2033 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07002034 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08002035 */
2036#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07002037#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08002038#endif
2039
Joe Perches7965bd42013-09-26 14:48:15 -07002040int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002041
2042static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2043{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00002044 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07002045 WARN_ON(1);
2046 return;
2047 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002048 skb->len = len;
2049 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002050}
2051
Joe Perches7965bd42013-09-26 14:48:15 -07002052void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002053
2054static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2055{
Herbert Xu3cc0e872006-06-09 16:13:38 -07002056 if (skb->data_len)
2057 return ___pskb_trim(skb, len);
2058 __skb_trim(skb, len);
2059 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002060}
2061
2062static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2063{
2064 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2065}
2066
2067/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07002068 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2069 * @skb: buffer to alter
2070 * @len: new length
2071 *
2072 * This is identical to pskb_trim except that the caller knows that
2073 * the skb is not cloned so we should never get an error due to out-
2074 * of-memory.
2075 */
2076static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2077{
2078 int err = pskb_trim(skb, len);
2079 BUG_ON(err);
2080}
2081
2082/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002083 * skb_orphan - orphan a buffer
2084 * @skb: buffer to orphan
2085 *
2086 * If a buffer currently has an owner then we call the owner's
2087 * destructor function and make the @skb unowned. The buffer continues
2088 * to exist but is no longer charged to its former owner.
2089 */
2090static inline void skb_orphan(struct sk_buff *skb)
2091{
Eric Dumazetc34a7612013-07-30 16:11:15 -07002092 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002093 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002094 skb->destructor = NULL;
2095 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07002096 } else {
2097 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002098 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002099}
2100
2101/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002102 * skb_orphan_frags - orphan the frags contained in a buffer
2103 * @skb: buffer to orphan frags from
2104 * @gfp_mask: allocation mask for replacement pages
2105 *
2106 * For each frag in the SKB which needs a destructor (i.e. has an
2107 * owner) create a copy of that frag and release the original
2108 * page by calling the destructor.
2109 */
2110static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2111{
2112 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2113 return 0;
2114 return skb_copy_ubufs(skb, gfp_mask);
2115}
2116
2117/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002118 * __skb_queue_purge - empty a list
2119 * @list: list to empty
2120 *
2121 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2122 * the list and one reference dropped. This function does not take the
2123 * list lock and the caller must hold the relevant locks to use it.
2124 */
Joe Perches7965bd42013-09-26 14:48:15 -07002125void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002126static inline void __skb_queue_purge(struct sk_buff_head *list)
2127{
2128 struct sk_buff *skb;
2129 while ((skb = __skb_dequeue(list)) != NULL)
2130 kfree_skb(skb);
2131}
2132
Joe Perches7965bd42013-09-26 14:48:15 -07002133void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002134
Joe Perches7965bd42013-09-26 14:48:15 -07002135struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2136 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07002137
2138/**
2139 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2140 * @dev: network device to receive on
2141 * @length: length to allocate
2142 *
2143 * Allocate a new &sk_buff and assign it a usage count of one. The
2144 * buffer has unspecified headroom built in. Users should allocate
2145 * the headroom they think they need without accounting for the
2146 * built in space. The built in space is used for optimisations.
2147 *
2148 * %NULL is returned if there is no free memory. Although this function
2149 * allocates memory it can be called from an interrupt.
2150 */
2151static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002152 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002153{
2154 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2155}
2156
Eric Dumazet6f532612012-05-18 05:12:12 +00002157/* legacy helper around __netdev_alloc_skb() */
2158static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2159 gfp_t gfp_mask)
2160{
2161 return __netdev_alloc_skb(NULL, length, gfp_mask);
2162}
2163
2164/* legacy helper around netdev_alloc_skb() */
2165static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2166{
2167 return netdev_alloc_skb(NULL, length);
2168}
2169
2170
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002171static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2172 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002173{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002174 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002175
2176 if (NET_IP_ALIGN && skb)
2177 skb_reserve(skb, NET_IP_ALIGN);
2178 return skb;
2179}
2180
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002181static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2182 unsigned int length)
2183{
2184 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2185}
2186
Alexander Duyck181edb22015-05-06 21:12:03 -07002187static inline void skb_free_frag(void *addr)
2188{
2189 __free_page_frag(addr);
2190}
2191
Alexander Duyckffde7322014-12-09 19:40:42 -08002192void *napi_alloc_frag(unsigned int fragsz);
Alexander Duyckfd11a832014-12-09 19:40:49 -08002193struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2194 unsigned int length, gfp_t gfp_mask);
2195static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2196 unsigned int length)
2197{
2198 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2199}
Alexander Duyckffde7322014-12-09 19:40:42 -08002200
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002201/**
Alexander Duyck71dfda52014-11-11 09:26:34 -08002202 * __dev_alloc_pages - allocate page for network Rx
2203 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2204 * @order: size of the allocation
2205 *
2206 * Allocate a new page.
2207 *
2208 * %NULL is returned if there is no free memory.
2209*/
2210static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2211 unsigned int order)
2212{
2213 /* This piece of code contains several assumptions.
2214 * 1. This is for device Rx, therefor a cold page is preferred.
2215 * 2. The expectation is the user wants a compound page.
2216 * 3. If requesting a order 0 page it will not be compound
2217 * due to the check to see if order has a value in prep_new_page
2218 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2219 * code in gfp_to_alloc_flags that should be enforcing this.
2220 */
2221 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2222
2223 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2224}
2225
2226static inline struct page *dev_alloc_pages(unsigned int order)
2227{
2228 return __dev_alloc_pages(GFP_ATOMIC, order);
2229}
2230
2231/**
2232 * __dev_alloc_page - allocate a page for network Rx
2233 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2234 *
2235 * Allocate a new page.
2236 *
2237 * %NULL is returned if there is no free memory.
2238 */
2239static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2240{
2241 return __dev_alloc_pages(gfp_mask, 0);
2242}
2243
2244static inline struct page *dev_alloc_page(void)
2245{
2246 return __dev_alloc_page(GFP_ATOMIC);
2247}
2248
2249/**
Mel Gorman06140022012-07-31 16:44:24 -07002250 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2251 * @page: The page that was allocated from skb_alloc_page
2252 * @skb: The skb that may need pfmemalloc set
2253 */
2254static inline void skb_propagate_pfmemalloc(struct page *page,
2255 struct sk_buff *skb)
2256{
2257 if (page && page->pfmemalloc)
2258 skb->pfmemalloc = true;
2259}
2260
Eric Dumazet564824b2010-10-11 19:05:25 +00002261/**
Masanari Iidae2278672014-02-18 22:54:36 +09002262 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002263 * @frag: the paged fragment
2264 *
2265 * Returns the &struct page associated with @frag.
2266 */
2267static inline struct page *skb_frag_page(const skb_frag_t *frag)
2268{
Ian Campbella8605c62011-10-19 23:01:49 +00002269 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00002270}
2271
2272/**
2273 * __skb_frag_ref - take an addition reference on a paged fragment.
2274 * @frag: the paged fragment
2275 *
2276 * Takes an additional reference on the paged fragment @frag.
2277 */
2278static inline void __skb_frag_ref(skb_frag_t *frag)
2279{
2280 get_page(skb_frag_page(frag));
2281}
2282
2283/**
2284 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2285 * @skb: the buffer
2286 * @f: the fragment offset.
2287 *
2288 * Takes an additional reference on the @f'th paged fragment of @skb.
2289 */
2290static inline void skb_frag_ref(struct sk_buff *skb, int f)
2291{
2292 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2293}
2294
2295/**
2296 * __skb_frag_unref - release a reference on a paged fragment.
2297 * @frag: the paged fragment
2298 *
2299 * Releases a reference on the paged fragment @frag.
2300 */
2301static inline void __skb_frag_unref(skb_frag_t *frag)
2302{
2303 put_page(skb_frag_page(frag));
2304}
2305
2306/**
2307 * skb_frag_unref - release a reference on a paged fragment of an skb.
2308 * @skb: the buffer
2309 * @f: the fragment offset
2310 *
2311 * Releases a reference on the @f'th paged fragment of @skb.
2312 */
2313static inline void skb_frag_unref(struct sk_buff *skb, int f)
2314{
2315 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2316}
2317
2318/**
2319 * skb_frag_address - gets the address of the data contained in a paged fragment
2320 * @frag: the paged fragment buffer
2321 *
2322 * Returns the address of the data within @frag. The page must already
2323 * be mapped.
2324 */
2325static inline void *skb_frag_address(const skb_frag_t *frag)
2326{
2327 return page_address(skb_frag_page(frag)) + frag->page_offset;
2328}
2329
2330/**
2331 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2332 * @frag: the paged fragment buffer
2333 *
2334 * Returns the address of the data within @frag. Checks that the page
2335 * is mapped and returns %NULL otherwise.
2336 */
2337static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2338{
2339 void *ptr = page_address(skb_frag_page(frag));
2340 if (unlikely(!ptr))
2341 return NULL;
2342
2343 return ptr + frag->page_offset;
2344}
2345
2346/**
2347 * __skb_frag_set_page - sets the page contained in a paged fragment
2348 * @frag: the paged fragment
2349 * @page: the page to set
2350 *
2351 * Sets the fragment @frag to contain @page.
2352 */
2353static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2354{
Ian Campbella8605c62011-10-19 23:01:49 +00002355 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002356}
2357
2358/**
2359 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2360 * @skb: the buffer
2361 * @f: the fragment offset
2362 * @page: the page to set
2363 *
2364 * Sets the @f'th fragment of @skb to contain @page.
2365 */
2366static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2367 struct page *page)
2368{
2369 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2370}
2371
Eric Dumazet400dfd32013-10-17 16:27:07 -07002372bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2373
Ian Campbell131ea662011-08-19 06:25:00 +00002374/**
2375 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002376 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002377 * @frag: the paged fragment to map
2378 * @offset: the offset within the fragment (starting at the
2379 * fragment's own offset)
2380 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002381 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00002382 *
2383 * Maps the page associated with @frag to @device.
2384 */
2385static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2386 const skb_frag_t *frag,
2387 size_t offset, size_t size,
2388 enum dma_data_direction dir)
2389{
2390 return dma_map_page(dev, skb_frag_page(frag),
2391 frag->page_offset + offset, size, dir);
2392}
2393
Eric Dumazet117632e2011-12-03 21:39:53 +00002394static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2395 gfp_t gfp_mask)
2396{
2397 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2398}
2399
Octavian Purdilabad93e92014-06-12 01:36:26 +03002400
2401static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2402 gfp_t gfp_mask)
2403{
2404 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2405}
2406
2407
Ian Campbell131ea662011-08-19 06:25:00 +00002408/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002409 * skb_clone_writable - is the header of a clone writable
2410 * @skb: buffer to check
2411 * @len: length up to which to write
2412 *
2413 * Returns true if modifying the header part of the cloned buffer
2414 * does not requires the data to be copied.
2415 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002416static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002417{
2418 return !skb_header_cloned(skb) &&
2419 skb_headroom(skb) + len <= skb->hdr_len;
2420}
2421
Herbert Xud9cc2042007-09-16 16:21:16 -07002422static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2423 int cloned)
2424{
2425 int delta = 0;
2426
Herbert Xud9cc2042007-09-16 16:21:16 -07002427 if (headroom > skb_headroom(skb))
2428 delta = headroom - skb_headroom(skb);
2429
2430 if (delta || cloned)
2431 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2432 GFP_ATOMIC);
2433 return 0;
2434}
2435
Patrick McHardy334a8132007-06-25 04:35:20 -07002436/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002437 * skb_cow - copy header of skb when it is required
2438 * @skb: buffer to cow
2439 * @headroom: needed headroom
2440 *
2441 * If the skb passed lacks sufficient headroom or its data part
2442 * is shared, data is reallocated. If reallocation fails, an error
2443 * is returned and original skb is not changed.
2444 *
2445 * The result is skb with writable area skb->head...skb->tail
2446 * and at least @headroom of space at head.
2447 */
2448static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2449{
Herbert Xud9cc2042007-09-16 16:21:16 -07002450 return __skb_cow(skb, headroom, skb_cloned(skb));
2451}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002452
Herbert Xud9cc2042007-09-16 16:21:16 -07002453/**
2454 * skb_cow_head - skb_cow but only making the head writable
2455 * @skb: buffer to cow
2456 * @headroom: needed headroom
2457 *
2458 * This function is identical to skb_cow except that we replace the
2459 * skb_cloned check by skb_header_cloned. It should be used when
2460 * you only need to push on some header and do not need to modify
2461 * the data.
2462 */
2463static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2464{
2465 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002466}
2467
2468/**
2469 * skb_padto - pad an skbuff up to a minimal size
2470 * @skb: buffer to pad
2471 * @len: minimal length
2472 *
2473 * Pads up a buffer to ensure the trailing bytes exist and are
2474 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002475 * is untouched. Otherwise it is extended. Returns zero on
2476 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002477 */
Herbert Xu5b057c62006-06-23 02:06:41 -07002478static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002479{
2480 unsigned int size = skb->len;
2481 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002482 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002483 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002484}
2485
Alexander Duyck9c0c1122014-12-03 08:17:33 -08002486/**
2487 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2488 * @skb: buffer to pad
2489 * @len: minimal length
2490 *
2491 * Pads up a buffer to ensure the trailing bytes exist and are
2492 * blanked. If the buffer already contains sufficient data it
2493 * is untouched. Otherwise it is extended. Returns zero on
2494 * success. The skb is freed on error.
2495 */
2496static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2497{
2498 unsigned int size = skb->len;
2499
2500 if (unlikely(size < len)) {
2501 len -= size;
2502 if (skb_pad(skb, len))
2503 return -ENOMEM;
2504 __skb_put(skb, len);
2505 }
2506 return 0;
2507}
2508
Linus Torvalds1da177e2005-04-16 15:20:36 -07002509static inline int skb_add_data(struct sk_buff *skb,
Al Viroaf2b0402014-11-27 21:44:24 -05002510 struct iov_iter *from, int copy)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002511{
2512 const int off = skb->len;
2513
2514 if (skb->ip_summed == CHECKSUM_NONE) {
Al Viroaf2b0402014-11-27 21:44:24 -05002515 __wsum csum = 0;
2516 if (csum_and_copy_from_iter(skb_put(skb, copy), copy,
2517 &csum, from) == copy) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002518 skb->csum = csum_block_add(skb->csum, csum, off);
2519 return 0;
2520 }
Al Viroaf2b0402014-11-27 21:44:24 -05002521 } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002522 return 0;
2523
2524 __skb_trim(skb, off);
2525 return -EFAULT;
2526}
2527
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002528static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2529 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002530{
2531 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002532 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002533
Ian Campbellea2ab692011-08-22 23:44:58 +00002534 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002535 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002536 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002537 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002538}
2539
Herbert Xu364c6ba2006-06-09 16:10:40 -07002540static inline int __skb_linearize(struct sk_buff *skb)
2541{
2542 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2543}
2544
Linus Torvalds1da177e2005-04-16 15:20:36 -07002545/**
2546 * skb_linearize - convert paged skb to linear one
2547 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002548 *
2549 * If there is no free memory -ENOMEM is returned, otherwise zero
2550 * is returned and the old skb data released.
2551 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002552static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002553{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002554 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2555}
2556
2557/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00002558 * skb_has_shared_frag - can any frag be overwritten
2559 * @skb: buffer to test
2560 *
2561 * Return true if the skb has at least one frag that might be modified
2562 * by an external entity (as in vmsplice()/sendfile())
2563 */
2564static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2565{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00002566 return skb_is_nonlinear(skb) &&
2567 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00002568}
2569
2570/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07002571 * skb_linearize_cow - make sure skb is linear and writable
2572 * @skb: buffer to process
2573 *
2574 * If there is no free memory -ENOMEM is returned, otherwise zero
2575 * is returned and the old skb data released.
2576 */
2577static inline int skb_linearize_cow(struct sk_buff *skb)
2578{
2579 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2580 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002581}
2582
2583/**
2584 * skb_postpull_rcsum - update checksum for received skb after pull
2585 * @skb: buffer to update
2586 * @start: start of data before pull
2587 * @len: length of data pulled
2588 *
2589 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002590 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2591 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002592 */
2593
2594static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002595 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002596{
Patrick McHardy84fa7932006-08-29 16:44:56 -07002597 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002598 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2599}
2600
Herbert Xucbb042f2006-03-20 22:43:56 -08002601unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2602
David S. Miller7ce5a272013-12-02 17:26:05 -05002603/**
2604 * pskb_trim_rcsum - trim received skb and update checksum
2605 * @skb: buffer to trim
2606 * @len: new length
2607 *
2608 * This is exactly the same as pskb_trim except that it ensures the
2609 * checksum of received packets are still valid after the operation.
2610 */
2611
2612static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2613{
2614 if (likely(len >= skb->len))
2615 return 0;
2616 if (skb->ip_summed == CHECKSUM_COMPLETE)
2617 skb->ip_summed = CHECKSUM_NONE;
2618 return __pskb_trim(skb, len);
2619}
2620
Linus Torvalds1da177e2005-04-16 15:20:36 -07002621#define skb_queue_walk(queue, skb) \
2622 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002623 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002624 skb = skb->next)
2625
James Chapman46f89142007-04-30 00:07:31 -07002626#define skb_queue_walk_safe(queue, skb, tmp) \
2627 for (skb = (queue)->next, tmp = skb->next; \
2628 skb != (struct sk_buff *)(queue); \
2629 skb = tmp, tmp = skb->next)
2630
David S. Miller1164f522008-09-23 00:49:44 -07002631#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002632 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002633 skb = skb->next)
2634
2635#define skb_queue_walk_from_safe(queue, skb, tmp) \
2636 for (tmp = skb->next; \
2637 skb != (struct sk_buff *)(queue); \
2638 skb = tmp, tmp = skb->next)
2639
Stephen Hemminger300ce172005-10-30 13:47:34 -08002640#define skb_queue_reverse_walk(queue, skb) \
2641 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002642 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08002643 skb = skb->prev)
2644
David S. Miller686a2952011-01-20 22:47:32 -08002645#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2646 for (skb = (queue)->prev, tmp = skb->prev; \
2647 skb != (struct sk_buff *)(queue); \
2648 skb = tmp, tmp = skb->prev)
2649
2650#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2651 for (tmp = skb->prev; \
2652 skb != (struct sk_buff *)(queue); \
2653 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002654
David S. Miller21dc3302010-08-23 00:13:46 -07002655static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07002656{
2657 return skb_shinfo(skb)->frag_list != NULL;
2658}
2659
2660static inline void skb_frag_list_init(struct sk_buff *skb)
2661{
2662 skb_shinfo(skb)->frag_list = NULL;
2663}
2664
2665static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2666{
2667 frag->next = skb_shinfo(skb)->frag_list;
2668 skb_shinfo(skb)->frag_list = frag;
2669}
2670
2671#define skb_walk_frags(skb, iter) \
2672 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2673
Joe Perches7965bd42013-09-26 14:48:15 -07002674struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2675 int *peeked, int *off, int *err);
2676struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2677 int *err);
2678unsigned int datagram_poll(struct file *file, struct socket *sock,
2679 struct poll_table_struct *wait);
Al Viroc0371da2014-11-24 10:42:55 -05002680int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
2681 struct iov_iter *to, int size);
David S. Miller51f3d022014-11-05 16:46:40 -05002682static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
2683 struct msghdr *msg, int size)
2684{
Al Viroe5a4b0b2014-11-24 18:17:55 -05002685 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
David S. Miller51f3d022014-11-05 16:46:40 -05002686}
Al Viroe5a4b0b2014-11-24 18:17:55 -05002687int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
2688 struct msghdr *msg);
Al Viro3a654f92014-06-19 14:15:22 -04002689int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
2690 struct iov_iter *from, int len);
Al Viro3a654f92014-06-19 14:15:22 -04002691int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
Joe Perches7965bd42013-09-26 14:48:15 -07002692void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2693void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2694int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07002695int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2696int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2697__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2698 int len, __wsum csum);
2699int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2700 struct pipe_inode_info *pipe, unsigned int len,
2701 unsigned int flags);
2702void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01002703unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00002704int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2705 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07002706void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2707int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2708void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Florian Westphalde960aa2014-01-26 10:58:16 +01002709unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -07002710struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Vlad Yasevich0d5501c2014-08-08 14:42:13 -04002711struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
Jiri Pirkoe2195122014-11-19 14:05:01 +01002712int skb_ensure_writable(struct sk_buff *skb, int write_len);
Jiri Pirko93515d52014-11-19 14:05:02 +01002713int skb_vlan_pop(struct sk_buff *skb);
2714int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002715
Al Viro6ce8e9c2014-04-06 21:25:44 -04002716static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
2717{
Al Viro21226ab2014-11-28 15:48:29 -05002718 return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro6ce8e9c2014-04-06 21:25:44 -04002719}
2720
Al Viro7eab8d92014-04-06 21:51:23 -04002721static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
2722{
Al Viroe5a4b0b2014-11-24 18:17:55 -05002723 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro7eab8d92014-04-06 21:51:23 -04002724}
2725
Daniel Borkmann2817a332013-10-30 11:50:51 +01002726struct skb_checksum_ops {
2727 __wsum (*update)(const void *mem, int len, __wsum wsum);
2728 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2729};
2730
2731__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2732 __wsum csum, const struct skb_checksum_ops *ops);
2733__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2734 __wsum csum);
2735
David S. Miller690e36e2014-08-23 12:13:41 -07002736static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset,
2737 int len, void *data, int hlen, void *buffer)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002738{
Patrick McHardy55820ee2005-07-05 14:08:10 -07002739 if (hlen - offset >= len)
David S. Miller690e36e2014-08-23 12:13:41 -07002740 return data + offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002741
David S. Miller690e36e2014-08-23 12:13:41 -07002742 if (!skb ||
2743 skb_copy_bits(skb, offset, buffer, len) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002744 return NULL;
2745
2746 return buffer;
2747}
2748
David S. Miller690e36e2014-08-23 12:13:41 -07002749static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2750 int len, void *buffer)
2751{
2752 return __skb_header_pointer(skb, offset, len, skb->data,
2753 skb_headlen(skb), buffer);
2754}
2755
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01002756/**
2757 * skb_needs_linearize - check if we need to linearize a given skb
2758 * depending on the given device features.
2759 * @skb: socket buffer to check
2760 * @features: net device features
2761 *
2762 * Returns true if either:
2763 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2764 * 2. skb is fragmented and the device does not support SG.
2765 */
2766static inline bool skb_needs_linearize(struct sk_buff *skb,
2767 netdev_features_t features)
2768{
2769 return skb_is_nonlinear(skb) &&
2770 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2771 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2772}
2773
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03002774static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2775 void *to,
2776 const unsigned int len)
2777{
2778 memcpy(to, skb->data, len);
2779}
2780
2781static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2782 const int offset, void *to,
2783 const unsigned int len)
2784{
2785 memcpy(to, skb->data + offset, len);
2786}
2787
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03002788static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2789 const void *from,
2790 const unsigned int len)
2791{
2792 memcpy(skb->data, from, len);
2793}
2794
2795static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2796 const int offset,
2797 const void *from,
2798 const unsigned int len)
2799{
2800 memcpy(skb->data + offset, from, len);
2801}
2802
Joe Perches7965bd42013-09-26 14:48:15 -07002803void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002804
Patrick Ohlyac45f602009-02-12 05:03:37 +00002805static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2806{
2807 return skb->tstamp;
2808}
2809
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002810/**
2811 * skb_get_timestamp - get timestamp from a skb
2812 * @skb: skb to get stamp from
2813 * @stamp: pointer to struct timeval to store stamp in
2814 *
2815 * Timestamps are stored in the skb as offsets to a base timestamp.
2816 * This function converts the offset back to a struct timeval and stores
2817 * it in stamp.
2818 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00002819static inline void skb_get_timestamp(const struct sk_buff *skb,
2820 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002821{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002822 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002823}
2824
Patrick Ohlyac45f602009-02-12 05:03:37 +00002825static inline void skb_get_timestampns(const struct sk_buff *skb,
2826 struct timespec *stamp)
2827{
2828 *stamp = ktime_to_timespec(skb->tstamp);
2829}
2830
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002831static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002832{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002833 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002834}
2835
Stephen Hemminger164891a2007-04-23 22:26:16 -07002836static inline ktime_t net_timedelta(ktime_t t)
2837{
2838 return ktime_sub(ktime_get_real(), t);
2839}
2840
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07002841static inline ktime_t net_invalid_timestamp(void)
2842{
2843 return ktime_set(0, 0);
2844}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002845
Alexander Duyck62bccb82014-09-04 13:31:35 -04002846struct sk_buff *skb_clone_sk(struct sk_buff *skb);
2847
Richard Cochranc1f19b52010-07-17 08:49:36 +00002848#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2849
Joe Perches7965bd42013-09-26 14:48:15 -07002850void skb_clone_tx_timestamp(struct sk_buff *skb);
2851bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00002852
2853#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2854
2855static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2856{
2857}
2858
2859static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2860{
2861 return false;
2862}
2863
2864#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2865
2866/**
2867 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2868 *
Richard Cochranda92b192011-10-21 00:49:15 +00002869 * PHY drivers may accept clones of transmitted packets for
2870 * timestamping via their phy_driver.txtstamp method. These drivers
2871 * must call this function to return the skb back to the stack, with
2872 * or without a timestamp.
2873 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00002874 * @skb: clone of the the original outgoing packet
Richard Cochranda92b192011-10-21 00:49:15 +00002875 * @hwtstamps: hardware time stamps, may be NULL if not available
Richard Cochranc1f19b52010-07-17 08:49:36 +00002876 *
2877 */
2878void skb_complete_tx_timestamp(struct sk_buff *skb,
2879 struct skb_shared_hwtstamps *hwtstamps);
2880
Willem de Bruijne7fd2882014-08-04 22:11:48 -04002881void __skb_tstamp_tx(struct sk_buff *orig_skb,
2882 struct skb_shared_hwtstamps *hwtstamps,
2883 struct sock *sk, int tstype);
2884
Patrick Ohlyac45f602009-02-12 05:03:37 +00002885/**
2886 * skb_tstamp_tx - queue clone of skb with send time stamps
2887 * @orig_skb: the original outgoing packet
2888 * @hwtstamps: hardware time stamps, may be NULL if not available
2889 *
2890 * If the skb has a socket associated, then this function clones the
2891 * skb (thus sharing the actual data and optional structures), stores
2892 * the optional hardware time stamping information (if non NULL) or
2893 * generates a software time stamp (otherwise), then queues the clone
2894 * to the error queue of the socket. Errors are silently ignored.
2895 */
Joe Perches7965bd42013-09-26 14:48:15 -07002896void skb_tstamp_tx(struct sk_buff *orig_skb,
2897 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00002898
Richard Cochran4507a712010-07-17 08:48:28 +00002899static inline void sw_tx_timestamp(struct sk_buff *skb)
2900{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002901 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2902 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00002903 skb_tstamp_tx(skb, NULL);
2904}
2905
2906/**
2907 * skb_tx_timestamp() - Driver hook for transmit timestamping
2908 *
2909 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00002910 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00002911 *
David S. Miller73409f32013-12-27 13:04:33 -05002912 * Specifically, one should make absolutely sure that this function is
2913 * called before TX completion of this packet can trigger. Otherwise
2914 * the packet could potentially already be freed.
2915 *
Richard Cochran4507a712010-07-17 08:48:28 +00002916 * @skb: A socket buffer.
2917 */
2918static inline void skb_tx_timestamp(struct sk_buff *skb)
2919{
Richard Cochranc1f19b52010-07-17 08:49:36 +00002920 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002921 sw_tx_timestamp(skb);
2922}
2923
Johannes Berg6e3e9392011-11-09 10:15:42 +01002924/**
2925 * skb_complete_wifi_ack - deliver skb with wifi status
2926 *
2927 * @skb: the original outgoing packet
2928 * @acked: ack status
2929 *
2930 */
2931void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2932
Joe Perches7965bd42013-09-26 14:48:15 -07002933__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2934__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002935
Herbert Xu60476372007-04-09 11:59:39 -07002936static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2937{
Tom Herbert6edec0e2015-02-10 16:30:28 -08002938 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
2939 skb->csum_valid ||
2940 (skb->ip_summed == CHECKSUM_PARTIAL &&
2941 skb_checksum_start_offset(skb) >= 0));
Herbert Xu60476372007-04-09 11:59:39 -07002942}
2943
Herbert Xufb286bb2005-11-10 13:01:24 -08002944/**
2945 * skb_checksum_complete - Calculate checksum of an entire packet
2946 * @skb: packet to process
2947 *
2948 * This function calculates the checksum over the entire packet plus
2949 * the value of skb->csum. The latter can be used to supply the
2950 * checksum of a pseudo header as used by TCP/UDP. It returns the
2951 * checksum.
2952 *
2953 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2954 * this function can be used to verify that checksum on received
2955 * packets. In that case the function should return zero if the
2956 * checksum is correct. In particular, this function will return zero
2957 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2958 * hardware has already verified the correctness of the checksum.
2959 */
Al Viro4381ca32007-07-15 21:00:11 +01002960static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002961{
Herbert Xu60476372007-04-09 11:59:39 -07002962 return skb_csum_unnecessary(skb) ?
2963 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002964}
2965
Tom Herbert77cffe22014-08-27 21:26:46 -07002966static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
2967{
2968 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2969 if (skb->csum_level == 0)
2970 skb->ip_summed = CHECKSUM_NONE;
2971 else
2972 skb->csum_level--;
2973 }
2974}
2975
2976static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
2977{
2978 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2979 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
2980 skb->csum_level++;
2981 } else if (skb->ip_summed == CHECKSUM_NONE) {
2982 skb->ip_summed = CHECKSUM_UNNECESSARY;
2983 skb->csum_level = 0;
2984 }
2985}
2986
Tom Herbert5a212322014-08-31 15:12:41 -07002987static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
2988{
2989 /* Mark current checksum as bad (typically called from GRO
2990 * path). In the case that ip_summed is CHECKSUM_NONE
2991 * this must be the first checksum encountered in the packet.
2992 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
2993 * checksum after the last one validated. For UDP, a zero
2994 * checksum can not be marked as bad.
2995 */
2996
2997 if (skb->ip_summed == CHECKSUM_NONE ||
2998 skb->ip_summed == CHECKSUM_UNNECESSARY)
2999 skb->csum_bad = 1;
3000}
3001
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003002/* Check if we need to perform checksum complete validation.
3003 *
3004 * Returns true if checksum complete is needed, false otherwise
3005 * (either checksum is unnecessary or zero checksum is allowed).
3006 */
3007static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3008 bool zero_okay,
3009 __sum16 check)
3010{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003011 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3012 skb->csum_valid = 1;
Tom Herbert77cffe22014-08-27 21:26:46 -07003013 __skb_decr_checksum_unnecessary(skb);
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003014 return false;
3015 }
3016
3017 return true;
3018}
3019
3020/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3021 * in checksum_init.
3022 */
3023#define CHECKSUM_BREAK 76
3024
Tom Herbert4e18b9a2015-04-20 14:10:04 -07003025/* Unset checksum-complete
3026 *
3027 * Unset checksum complete can be done when packet is being modified
3028 * (uncompressed for instance) and checksum-complete value is
3029 * invalidated.
3030 */
3031static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3032{
3033 if (skb->ip_summed == CHECKSUM_COMPLETE)
3034 skb->ip_summed = CHECKSUM_NONE;
3035}
3036
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003037/* Validate (init) checksum based on checksum complete.
3038 *
3039 * Return values:
3040 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3041 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3042 * checksum is stored in skb->csum for use in __skb_checksum_complete
3043 * non-zero: value of invalid checksum
3044 *
3045 */
3046static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3047 bool complete,
3048 __wsum psum)
3049{
3050 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3051 if (!csum_fold(csum_add(psum, skb->csum))) {
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003052 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003053 return 0;
3054 }
Tom Herbert5a212322014-08-31 15:12:41 -07003055 } else if (skb->csum_bad) {
3056 /* ip_summed == CHECKSUM_NONE in this case */
3057 return 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003058 }
3059
3060 skb->csum = psum;
3061
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003062 if (complete || skb->len <= CHECKSUM_BREAK) {
3063 __sum16 csum;
3064
3065 csum = __skb_checksum_complete(skb);
3066 skb->csum_valid = !csum;
3067 return csum;
3068 }
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003069
3070 return 0;
3071}
3072
3073static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3074{
3075 return 0;
3076}
3077
3078/* Perform checksum validate (init). Note that this is a macro since we only
3079 * want to calculate the pseudo header which is an input function if necessary.
3080 * First we try to validate without any computation (checksum unnecessary) and
3081 * then calculate based on checksum complete calling the function to compute
3082 * pseudo header.
3083 *
3084 * Return values:
3085 * 0: checksum is validated or try to in skb_checksum_complete
3086 * non-zero: value of invalid checksum
3087 */
3088#define __skb_checksum_validate(skb, proto, complete, \
3089 zero_okay, check, compute_pseudo) \
3090({ \
3091 __sum16 __ret = 0; \
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003092 skb->csum_valid = 0; \
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003093 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3094 __ret = __skb_checksum_validate_complete(skb, \
3095 complete, compute_pseudo(skb, proto)); \
3096 __ret; \
3097})
3098
3099#define skb_checksum_init(skb, proto, compute_pseudo) \
3100 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3101
3102#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3103 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3104
3105#define skb_checksum_validate(skb, proto, compute_pseudo) \
3106 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3107
3108#define skb_checksum_validate_zero_check(skb, proto, check, \
3109 compute_pseudo) \
Sabrina Dubroca096a4cf2015-02-06 18:54:19 +01003110 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003111
3112#define skb_checksum_simple_validate(skb) \
3113 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3114
Tom Herbertd96535a2014-08-31 15:12:42 -07003115static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3116{
3117 return (skb->ip_summed == CHECKSUM_NONE &&
3118 skb->csum_valid && !skb->csum_bad);
3119}
3120
3121static inline void __skb_checksum_convert(struct sk_buff *skb,
3122 __sum16 check, __wsum pseudo)
3123{
3124 skb->csum = ~pseudo;
3125 skb->ip_summed = CHECKSUM_COMPLETE;
3126}
3127
3128#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3129do { \
3130 if (__skb_checksum_convert_check(skb)) \
3131 __skb_checksum_convert(skb, check, \
3132 compute_pseudo(skb, proto)); \
3133} while (0)
3134
Tom Herbert15e23962015-02-10 16:30:31 -08003135static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3136 u16 start, u16 offset)
3137{
3138 skb->ip_summed = CHECKSUM_PARTIAL;
3139 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3140 skb->csum_offset = offset - start;
3141}
3142
Tom Herbertdcdc8992015-02-02 16:07:34 -08003143/* Update skbuf and packet to reflect the remote checksum offload operation.
3144 * When called, ptr indicates the starting point for skb->csum when
3145 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3146 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3147 */
3148static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
Tom Herbert15e23962015-02-10 16:30:31 -08003149 int start, int offset, bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08003150{
3151 __wsum delta;
3152
Tom Herbert15e23962015-02-10 16:30:31 -08003153 if (!nopartial) {
3154 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3155 return;
3156 }
3157
Tom Herbertdcdc8992015-02-02 16:07:34 -08003158 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3159 __skb_checksum_complete(skb);
3160 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3161 }
3162
3163 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3164
3165 /* Adjust skb->csum since we changed the packet */
3166 skb->csum = csum_add(skb->csum, delta);
3167}
3168
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003169#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07003170void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003171static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3172{
3173 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07003174 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003175}
3176static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3177{
3178 if (nfct)
3179 atomic_inc(&nfct->use);
3180}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003181#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003182#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003183static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3184{
3185 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3186 kfree(nf_bridge);
3187}
3188static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3189{
3190 if (nf_bridge)
3191 atomic_inc(&nf_bridge->use);
3192}
3193#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08003194static inline void nf_reset(struct sk_buff *skb)
3195{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003196#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08003197 nf_conntrack_put(skb->nfct);
3198 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003199#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003200#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Patrick McHardya193a4a2006-03-20 19:23:05 -08003201 nf_bridge_put(skb->nf_bridge);
3202 skb->nf_bridge = NULL;
3203#endif
3204}
3205
Patrick McHardy124dff02013-04-05 20:42:05 +02003206static inline void nf_reset_trace(struct sk_buff *skb)
3207{
Florian Westphal478b3602014-02-15 23:48:45 +01003208#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00003209 skb->nf_trace = 0;
3210#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003211}
3212
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003213/* Note: This doesn't put any conntrack and bridge info in dst. */
Eric Dumazetb1937222014-09-28 22:18:47 -07003214static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3215 bool copy)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003216{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003217#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003218 dst->nfct = src->nfct;
3219 nf_conntrack_get(src->nfct);
Eric Dumazetb1937222014-09-28 22:18:47 -07003220 if (copy)
3221 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003222#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003223#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003224 dst->nf_bridge = src->nf_bridge;
3225 nf_bridge_get(src->nf_bridge);
3226#endif
Florian Westphal478b3602014-02-15 23:48:45 +01003227#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Eric Dumazetb1937222014-09-28 22:18:47 -07003228 if (copy)
3229 dst->nf_trace = src->nf_trace;
Florian Westphal478b3602014-02-15 23:48:45 +01003230#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003231}
3232
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003233static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3234{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003235#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003236 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003237#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003238#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003239 nf_bridge_put(dst->nf_bridge);
3240#endif
Eric Dumazetb1937222014-09-28 22:18:47 -07003241 __nf_copy(dst, src, true);
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003242}
3243
James Morris984bc162006-06-09 00:29:17 -07003244#ifdef CONFIG_NETWORK_SECMARK
3245static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3246{
3247 to->secmark = from->secmark;
3248}
3249
3250static inline void skb_init_secmark(struct sk_buff *skb)
3251{
3252 skb->secmark = 0;
3253}
3254#else
3255static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3256{ }
3257
3258static inline void skb_init_secmark(struct sk_buff *skb)
3259{ }
3260#endif
3261
Eric W. Biederman574f7192014-04-01 12:20:24 -07003262static inline bool skb_irq_freeable(const struct sk_buff *skb)
3263{
3264 return !skb->destructor &&
3265#if IS_ENABLED(CONFIG_XFRM)
3266 !skb->sp &&
3267#endif
3268#if IS_ENABLED(CONFIG_NF_CONNTRACK)
3269 !skb->nfct &&
3270#endif
3271 !skb->_skb_refdst &&
3272 !skb_has_frag_list(skb);
3273}
3274
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003275static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3276{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003277 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003278}
3279
Stephen Hemminger92477442009-03-21 13:39:26 -07003280static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003281{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003282 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003283}
3284
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003285static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3286{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003287 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003288}
3289
David S. Millerd5a9e242009-01-27 16:22:11 -08003290static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3291{
3292 skb->queue_mapping = rx_queue + 1;
3293}
3294
Stephen Hemminger92477442009-03-21 13:39:26 -07003295static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003296{
3297 return skb->queue_mapping - 1;
3298}
3299
Stephen Hemminger92477442009-03-21 13:39:26 -07003300static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003301{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003302 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08003303}
3304
Tom Herbert0e001612014-07-01 21:32:27 -07003305u16 __skb_tx_hash(const struct net_device *dev, struct sk_buff *skb,
Joe Perches7965bd42013-09-26 14:48:15 -07003306 unsigned int num_tx_queues);
Stephen Hemminger92477442009-03-21 13:39:26 -07003307
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003308static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3309{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003310#ifdef CONFIG_XFRM
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003311 return skb->sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003312#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003313 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003314#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003315}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003316
Pravin B Shelar68c33162013-02-14 14:02:41 +00003317/* Keeps track of mac header offset relative to skb->head.
3318 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3319 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07003320 * tunnel skb it points to outer mac header.
3321 * Keeps track of level of encapsulation of network headers.
3322 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00003323struct skb_gso_cb {
Eric Dumazet3347c962013-10-19 11:42:56 -07003324 int mac_offset;
3325 int encap_level;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003326 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00003327};
3328#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3329
3330static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3331{
3332 return (skb_mac_header(inner_skb) - inner_skb->head) -
3333 SKB_GSO_CB(inner_skb)->mac_offset;
3334}
3335
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00003336static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3337{
3338 int new_headroom, headroom;
3339 int ret;
3340
3341 headroom = skb_headroom(skb);
3342 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3343 if (ret)
3344 return ret;
3345
3346 new_headroom = skb_headroom(skb);
3347 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3348 return 0;
3349}
3350
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003351/* Compute the checksum for a gso segment. First compute the checksum value
3352 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3353 * then add in skb->csum (checksum from csum_start to end of packet).
3354 * skb->csum and csum_start are then updated to reflect the checksum of the
3355 * resultant packet starting from the transport header-- the resultant checksum
3356 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3357 * header.
3358 */
3359static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3360{
3361 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3362 skb_transport_offset(skb);
3363 __u16 csum;
3364
3365 csum = csum_fold(csum_partial(skb_transport_header(skb),
3366 plen, skb->csum));
3367 skb->csum = res;
3368 SKB_GSO_CB(skb)->csum_start -= plen;
3369
3370 return csum;
3371}
3372
David S. Millerbdcc0922012-03-07 20:53:36 -05003373static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07003374{
3375 return skb_shinfo(skb)->gso_size;
3376}
3377
Eric Dumazet36a8f392013-09-29 01:21:32 -07003378/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05003379static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02003380{
3381 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3382}
3383
Joe Perches7965bd42013-09-26 14:48:15 -07003384void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07003385
3386static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3387{
3388 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3389 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003390 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3391
Alexander Duyckb78462e2010-06-02 12:24:37 +00003392 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3393 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07003394 __skb_warn_lro_forwarding(skb);
3395 return true;
3396 }
3397 return false;
3398}
3399
Herbert Xu35fc92a2007-03-26 23:22:20 -07003400static inline void skb_forward_csum(struct sk_buff *skb)
3401{
3402 /* Unfortunately we don't support this one. Any brave souls? */
3403 if (skb->ip_summed == CHECKSUM_COMPLETE)
3404 skb->ip_summed = CHECKSUM_NONE;
3405}
3406
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003407/**
3408 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3409 * @skb: skb to check
3410 *
3411 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3412 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3413 * use this helper, to document places where we make this assertion.
3414 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003415static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003416{
3417#ifdef DEBUG
3418 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3419#endif
3420}
3421
Rusty Russellf35d9d82008-02-04 23:49:54 -05003422bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00003423
Paul Durranted1f50c2014-01-09 10:02:46 +00003424int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
Linus Lüssing9afd85c2015-05-02 14:01:07 +02003425struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
3426 unsigned int transport_len,
3427 __sum16(*skb_chkf)(struct sk_buff *skb));
Paul Durranted1f50c2014-01-09 10:02:46 +00003428
Alexander Duyck56193d12014-09-05 19:20:26 -04003429u32 skb_get_poff(const struct sk_buff *skb);
3430u32 __skb_get_poff(const struct sk_buff *skb, void *data,
3431 const struct flow_keys *keys, int hlen);
Daniel Borkmannf77668d2013-03-19 06:39:30 +00003432
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00003433/**
3434 * skb_head_is_locked - Determine if the skb->head is locked down
3435 * @skb: skb to check
3436 *
3437 * The head on skbs build around a head frag can be removed if they are
3438 * not cloned. This function returns true if the skb head is locked down
3439 * due to either being allocated via kmalloc, or by being a clone with
3440 * multiple references to the head.
3441 */
3442static inline bool skb_head_is_locked(const struct sk_buff *skb)
3443{
3444 return !skb->head_frag || skb_cloned(skb);
3445}
Florian Westphalfe6cc552014-02-13 23:09:12 +01003446
3447/**
3448 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3449 *
3450 * @skb: GSO skb
3451 *
3452 * skb_gso_network_seglen is used to determine the real size of the
3453 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3454 *
3455 * The MAC/L2 header is not accounted for.
3456 */
3457static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3458{
3459 unsigned int hdr_len = skb_transport_header(skb) -
3460 skb_network_header(skb);
3461 return hdr_len + skb_gso_transport_seglen(skb);
3462}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003463#endif /* __KERNEL__ */
3464#endif /* _LINUX_SKBUFF_H */