blob: af7a09650fa24b0181de169bc94f8341aae72e2b [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>
Jiri Pirko1bd758e2015-05-12 14:56:07 +020037#include <net/flow_dissector.h>
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +020038#include <linux/splice.h>
Bernhard Thaler72b31f72015-05-30 15:27:40 +020039#include <linux/in6.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070040
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010041/* A. Checksumming of received packets by device.
42 *
43 * CHECKSUM_NONE:
44 *
45 * Device failed to checksum this packet e.g. due to lack of capabilities.
46 * The packet contains full (though not verified) checksum in packet but
47 * not in skb->csum. Thus, skb->csum is undefined in this case.
48 *
49 * CHECKSUM_UNNECESSARY:
50 *
51 * The hardware you're dealing with doesn't calculate the full checksum
52 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
Tom Herbert77cffe22014-08-27 21:26:46 -070053 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
54 * if their checksums are okay. skb->csum is still undefined in this case
55 * though. It is a bad option, but, unfortunately, nowadays most vendors do
56 * this. Apparently with the secret goal to sell you new devices, when you
57 * will add new protocol to your host, f.e. IPv6 8)
58 *
59 * CHECKSUM_UNNECESSARY is applicable to following protocols:
60 * TCP: IPv6 and IPv4.
61 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
62 * zero UDP checksum for either IPv4 or IPv6, the networking stack
63 * may perform further validation in this case.
64 * GRE: only if the checksum is present in the header.
65 * SCTP: indicates the CRC in SCTP header has been validated.
66 *
67 * skb->csum_level indicates the number of consecutive checksums found in
68 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
69 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
70 * and a device is able to verify the checksums for UDP (possibly zero),
71 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
72 * two. If the device were only able to verify the UDP checksum and not
73 * GRE, either because it doesn't support GRE checksum of because GRE
74 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
75 * not considered in this case).
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010076 *
77 * CHECKSUM_COMPLETE:
78 *
79 * This is the most generic way. The device supplied checksum of the _whole_
80 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
81 * hardware doesn't need to parse L3/L4 headers to implement this.
82 *
83 * Note: Even if device supports only some protocols, but is able to produce
84 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
85 *
86 * CHECKSUM_PARTIAL:
87 *
Tom Herbert6edec0e2015-02-10 16:30:28 -080088 * A checksum is set up to be offloaded to a device as described in the
89 * output description for CHECKSUM_PARTIAL. This may occur on a packet
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010090 * received directly from another Linux OS, e.g., a virtualized Linux kernel
Tom Herbert6edec0e2015-02-10 16:30:28 -080091 * on the same host, or it may be set in the input path in GRO or remote
92 * checksum offload. For the purposes of checksum verification, the checksum
93 * referred to by skb->csum_start + skb->csum_offset and any preceding
94 * checksums in the packet are considered verified. Any checksums in the
95 * packet that are after the checksum being offloaded are not considered to
96 * be verified.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010097 *
98 * B. Checksumming on output.
99 *
100 * CHECKSUM_NONE:
101 *
102 * The skb was already checksummed by the protocol, or a checksum is not
103 * required.
104 *
105 * CHECKSUM_PARTIAL:
106 *
107 * The device is required to checksum the packet as seen by hard_start_xmit()
108 * from skb->csum_start up to the end, and to record/write the checksum at
109 * offset skb->csum_start + skb->csum_offset.
110 *
111 * The device must show its capabilities in dev->features, set up at device
112 * setup time, e.g. netdev_features.h:
113 *
114 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
115 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
116 * IPv4. Sigh. Vendors like this way for an unknown reason.
117 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
118 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
119 * NETIF_F_... - Well, you get the picture.
120 *
121 * CHECKSUM_UNNECESSARY:
122 *
123 * Normally, the device will do per protocol specific checksumming. Protocol
124 * implementations that do not want the NIC to perform the checksum
125 * calculation should use this flag in their outgoing skbs.
126 *
127 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
128 * offload. Correspondingly, the FCoE protocol driver
129 * stack should use CHECKSUM_UNNECESSARY.
130 *
131 * Any questions? No questions, good. --ANK
132 */
133
Herbert Xu60476372007-04-09 11:59:39 -0700134/* Don't change this without changing skb_csum_unnecessary! */
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100135#define CHECKSUM_NONE 0
136#define CHECKSUM_UNNECESSARY 1
137#define CHECKSUM_COMPLETE 2
138#define CHECKSUM_PARTIAL 3
Linus Torvalds1da177e2005-04-16 15:20:36 -0700139
Tom Herbert77cffe22014-08-27 21:26:46 -0700140/* Maximum value in skb->csum_level */
141#define SKB_MAX_CSUM_LEVEL 3
142
Tobias Klauser0bec8c82014-07-22 12:06:23 +0200143#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
David S. Millerfc910a22007-03-25 20:27:59 -0700144#define SKB_WITH_OVERHEAD(X) \
Herbert Xudeea84b2007-10-21 16:27:46 -0700145 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
David S. Millerfc910a22007-03-25 20:27:59 -0700146#define SKB_MAX_ORDER(X, ORDER) \
147 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700148#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
149#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
150
Eric Dumazet87fb4b72011-10-13 07:28:54 +0000151/* return minimum truesize of one skb containing X bytes of data */
152#define SKB_TRUESIZE(X) ((X) + \
153 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
154 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
155
Linus Torvalds1da177e2005-04-16 15:20:36 -0700156struct net_device;
David Howells716ea3a2007-04-02 20:19:53 -0700157struct scatterlist;
Jens Axboe9c55e012007-11-06 23:30:13 -0800158struct pipe_inode_info;
Herbert Xua8f820aa2014-11-07 21:22:22 +0800159struct iov_iter;
Alexander Duyckfd11a832014-12-09 19:40:49 -0800160struct napi_struct;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700162#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700163struct nf_conntrack {
164 atomic_t use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165};
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700166#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +0200168#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700169struct nf_bridge_info {
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000170 atomic_t use;
Florian Westphal3eaf4022015-04-02 14:31:44 +0200171 enum {
172 BRNF_PROTO_UNCHANGED,
173 BRNF_PROTO_8021Q,
174 BRNF_PROTO_PPPOE
Florian Westphal7fb48c52015-05-03 22:05:28 +0200175 } orig_proto:8;
Florian Westphala1e67952015-04-02 14:31:45 +0200176 bool pkt_otherhost;
Bernhard Thaler411ffb42015-05-30 15:28:28 +0200177 __u16 frag_max_size;
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000178 unsigned int mask;
179 struct net_device *physindev;
Florian Westphal7fb48c52015-05-03 22:05:28 +0200180 union {
181 struct net_device *physoutdev;
182 char neigh_header[8];
183 };
Bernhard Thaler72b31f72015-05-30 15:27:40 +0200184 union {
185 __be32 ipv4_daddr;
186 struct in6_addr ipv6_daddr;
187 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700188};
189#endif
190
Linus Torvalds1da177e2005-04-16 15:20:36 -0700191struct sk_buff_head {
192 /* These two members must be first. */
193 struct sk_buff *next;
194 struct sk_buff *prev;
195
196 __u32 qlen;
197 spinlock_t lock;
198};
199
200struct sk_buff;
201
Ian Campbell9d4dde52011-12-22 23:39:14 +0000202/* To allow 64K frame to be packed as single skb without frag_list we
203 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
204 * buffers which do not start on a page boundary.
205 *
206 * Since GRO uses frags we allocate at least 16 regardless of page
207 * size.
Anton Blancharda715dea2011-03-27 14:57:26 +0000208 */
Ian Campbell9d4dde52011-12-22 23:39:14 +0000209#if (65536/PAGE_SIZE + 1) < 16
David S. Millereec00952011-03-29 23:34:08 -0700210#define MAX_SKB_FRAGS 16UL
Anton Blancharda715dea2011-03-27 14:57:26 +0000211#else
Ian Campbell9d4dde52011-12-22 23:39:14 +0000212#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
Anton Blancharda715dea2011-03-27 14:57:26 +0000213#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700214
215typedef struct skb_frag_struct skb_frag_t;
216
217struct skb_frag_struct {
Ian Campbella8605c62011-10-19 23:01:49 +0000218 struct {
219 struct page *p;
220 } page;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000221#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
David S. Millera309bb02007-07-30 18:47:03 -0700222 __u32 page_offset;
223 __u32 size;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000224#else
225 __u16 page_offset;
226 __u16 size;
227#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700228};
229
Eric Dumazet9e903e02011-10-18 21:00:24 +0000230static inline unsigned int skb_frag_size(const skb_frag_t *frag)
231{
232 return frag->size;
233}
234
235static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
236{
237 frag->size = size;
238}
239
240static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
241{
242 frag->size += delta;
243}
244
245static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
246{
247 frag->size -= delta;
248}
249
Patrick Ohlyac45f602009-02-12 05:03:37 +0000250#define HAVE_HW_TIME_STAMP
251
252/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800253 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000254 * @hwtstamp: hardware time stamp transformed into duration
255 * since arbitrary point in time
Patrick Ohlyac45f602009-02-12 05:03:37 +0000256 *
257 * Software time stamps generated by ktime_get_real() are stored in
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400258 * skb->tstamp.
Patrick Ohlyac45f602009-02-12 05:03:37 +0000259 *
260 * hwtstamps can only be compared against other hwtstamps from
261 * the same device.
262 *
263 * This structure is attached to packets as part of the
264 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
265 */
266struct skb_shared_hwtstamps {
267 ktime_t hwtstamp;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000268};
269
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000270/* Definitions for tx_flags in struct skb_shared_info */
271enum {
272 /* generate hardware time stamp */
273 SKBTX_HW_TSTAMP = 1 << 0,
274
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400275 /* generate software time stamp when queueing packet to NIC */
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000276 SKBTX_SW_TSTAMP = 1 << 1,
277
278 /* device driver is going to provide hardware time stamp */
279 SKBTX_IN_PROGRESS = 1 << 2,
280
Shirley Maa6686f22011-07-06 12:22:12 +0000281 /* device driver supports TX zero-copy buffers */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000282 SKBTX_DEV_ZEROCOPY = 1 << 3,
Johannes Berg6e3e9392011-11-09 10:15:42 +0100283
284 /* generate wifi status information (where possible) */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000285 SKBTX_WIFI_STATUS = 1 << 4,
Pravin B Shelarc9af6db2013-02-11 09:27:41 +0000286
287 /* This indicates at least one fragment might be overwritten
288 * (as in vmsplice(), sendfile() ...)
289 * If we need to compute a TX checksum, we'll need to copy
290 * all frags to avoid possible bad checksum
291 */
292 SKBTX_SHARED_FRAG = 1 << 5,
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400293
294 /* generate software time stamp when entering packet scheduling */
295 SKBTX_SCHED_TSTAMP = 1 << 6,
Willem de Bruijne1c8a602014-08-04 22:11:50 -0400296
297 /* generate software timestamp on peer data acknowledgment */
298 SKBTX_ACK_TSTAMP = 1 << 7,
Shirley Maa6686f22011-07-06 12:22:12 +0000299};
300
Willem de Bruijne1c8a602014-08-04 22:11:50 -0400301#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
302 SKBTX_SCHED_TSTAMP | \
303 SKBTX_ACK_TSTAMP)
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400304#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
305
Shirley Maa6686f22011-07-06 12:22:12 +0000306/*
307 * The callback notifies userspace to release buffers when skb DMA is done in
308 * lower device, the skb last reference should be 0 when calling this.
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000309 * The zerocopy_success argument is true if zero copy transmit occurred,
310 * false on data copy or out of memory error caused by data copy attempt.
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000311 * The ctx field is used to track device context.
312 * The desc field is used to track userspace buffer index.
Shirley Maa6686f22011-07-06 12:22:12 +0000313 */
314struct ubuf_info {
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000315 void (*callback)(struct ubuf_info *, bool zerocopy_success);
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000316 void *ctx;
Shirley Maa6686f22011-07-06 12:22:12 +0000317 unsigned long desc;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000318};
319
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320/* This data is invariant across clones and lives at
321 * the end of the header data, ie. at skb->end.
322 */
323struct skb_shared_info {
Ian Campbell9f42f122012-01-05 07:13:39 +0000324 unsigned char nr_frags;
325 __u8 tx_flags;
Herbert Xu79671682006-06-22 02:40:14 -0700326 unsigned short gso_size;
327 /* Warning: this field is not always filled in (UFO)! */
328 unsigned short gso_segs;
329 unsigned short gso_type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700330 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000331 struct skb_shared_hwtstamps hwtstamps;
Willem de Bruijn09c2d252014-08-04 22:11:47 -0400332 u32 tskey;
Ian Campbell9f42f122012-01-05 07:13:39 +0000333 __be32 ip6_frag_id;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700334
335 /*
336 * Warning : all fields before dataref are cleared in __alloc_skb()
337 */
338 atomic_t dataref;
339
Johann Baudy69e3c752009-05-18 22:11:22 -0700340 /* Intermediate layers must ensure that destructor_arg
341 * remains valid until skb destructor */
342 void * destructor_arg;
Shirley Maa6686f22011-07-06 12:22:12 +0000343
Eric Dumazetfed66382010-07-22 19:09:08 +0000344 /* must be last field, see pskb_expand_head() */
345 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346};
347
348/* We divide dataref into two halves. The higher 16 bits hold references
349 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700350 * the entire skb->data. A clone of a headerless skb holds the length of
351 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700352 *
353 * All users must obey the rule that the skb->data reference count must be
354 * greater than or equal to the payload reference count.
355 *
356 * Holding a reference to the payload part means that the user does not
357 * care about modifications to the header part of skb->data.
358 */
359#define SKB_DATAREF_SHIFT 16
360#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
361
David S. Millerd179cd12005-08-17 14:57:30 -0700362
363enum {
Vijay Subramanianc8753d52014-10-02 10:00:43 -0700364 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
365 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
366 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
David S. Millerd179cd12005-08-17 14:57:30 -0700367};
368
Herbert Xu79671682006-06-22 02:40:14 -0700369enum {
370 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700371 SKB_GSO_UDP = 1 << 1,
Herbert Xu576a30e2006-06-27 13:22:38 -0700372
373 /* This indicates the skb is from an untrusted source. */
374 SKB_GSO_DODGY = 1 << 2,
Michael Chanb0da85372006-06-29 12:30:00 -0700375
376 /* This indicates the tcp segment has CWR set. */
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700377 SKB_GSO_TCP_ECN = 1 << 3,
378
379 SKB_GSO_TCPV6 = 1 << 4,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800380
381 SKB_GSO_FCOE = 1 << 5,
Pravin B Shelar68c33162013-02-14 14:02:41 +0000382
383 SKB_GSO_GRE = 1 << 6,
Pravin B Shelar73136262013-03-07 13:21:51 +0000384
Tom Herbert4b282522014-06-14 23:23:52 -0700385 SKB_GSO_GRE_CSUM = 1 << 7,
Simon Horman0d89d202013-05-23 21:02:52 +0000386
Tom Herbert4b282522014-06-14 23:23:52 -0700387 SKB_GSO_IPIP = 1 << 8,
Eric Dumazetcb32f512013-10-19 11:42:57 -0700388
Tom Herbert4b282522014-06-14 23:23:52 -0700389 SKB_GSO_SIT = 1 << 9,
Eric Dumazet61c1db72013-10-20 20:47:30 -0700390
Tom Herbert4b282522014-06-14 23:23:52 -0700391 SKB_GSO_UDP_TUNNEL = 1 << 10,
Tom Herbert0f4f4ff2014-06-04 17:20:16 -0700392
393 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
Tom Herbert4749c092014-06-04 17:20:23 -0700394
Pravin B Shelar59b93b42014-11-05 15:27:48 -0800395 SKB_GSO_TUNNEL_REMCSUM = 1 << 12,
Herbert Xu79671682006-06-22 02:40:14 -0700396};
397
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700398#if BITS_PER_LONG > 32
399#define NET_SKBUFF_DATA_USES_OFFSET 1
400#endif
401
402#ifdef NET_SKBUFF_DATA_USES_OFFSET
403typedef unsigned int sk_buff_data_t;
404#else
405typedef unsigned char *sk_buff_data_t;
406#endif
407
Eric Dumazet363ec39232014-02-26 14:02:11 -0800408/**
409 * struct skb_mstamp - multi resolution time stamps
410 * @stamp_us: timestamp in us resolution
411 * @stamp_jiffies: timestamp in jiffies
412 */
413struct skb_mstamp {
414 union {
415 u64 v64;
416 struct {
417 u32 stamp_us;
418 u32 stamp_jiffies;
419 };
420 };
421};
422
423/**
424 * skb_mstamp_get - get current timestamp
425 * @cl: place to store timestamps
426 */
427static inline void skb_mstamp_get(struct skb_mstamp *cl)
428{
429 u64 val = local_clock();
430
431 do_div(val, NSEC_PER_USEC);
432 cl->stamp_us = (u32)val;
433 cl->stamp_jiffies = (u32)jiffies;
434}
435
436/**
437 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
438 * @t1: pointer to newest sample
439 * @t0: pointer to oldest sample
440 */
441static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
442 const struct skb_mstamp *t0)
443{
444 s32 delta_us = t1->stamp_us - t0->stamp_us;
445 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
446
447 /* If delta_us is negative, this might be because interval is too big,
448 * or local_clock() drift is too big : fallback using jiffies.
449 */
450 if (delta_us <= 0 ||
451 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
452
453 delta_us = jiffies_to_usecs(delta_jiffies);
454
455 return delta_us;
456}
457
458
Linus Torvalds1da177e2005-04-16 15:20:36 -0700459/**
460 * struct sk_buff - socket buffer
461 * @next: Next buffer in list
462 * @prev: Previous buffer in list
Eric Dumazet363ec39232014-02-26 14:02:11 -0800463 * @tstamp: Time we arrived/left
Eric Dumazet56b17422014-11-03 08:19:53 -0800464 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700465 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700466 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700467 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000468 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700469 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700470 * @len: Length of actual data
471 * @data_len: Data length
472 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700473 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700474 * @csum: Checksum (must include start/offset pair)
475 * @csum_start: Offset from skb->head where checksumming should start
476 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700477 * @priority: Packet queueing priority
WANG Cong60ff7462014-05-04 16:39:18 -0700478 * @ignore_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700480 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481 * @nohdr: Payload reference only, must not modify header
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700482 * @nfctinfo: Relationship of this skb to the connection
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700484 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700485 * @ipvs_property: skbuff is owned by ipvs
Randy Dunlap31729362008-02-18 20:52:13 -0800486 * @peeked: this packet has been seen already, so stats have been
487 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700488 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700489 * @protocol: Packet protocol from driver
490 * @destructor: Destruct function
491 * @nfct: Associated connection, if any
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800493 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 * @tc_index: Traffic control index
495 * @tc_verd: traffic control verdict
Tom Herbert61b905d2014-03-24 15:34:47 -0700496 * @hash: the packet hash
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700497 * @queue_mapping: Queue mapping for multiqueue devices
David S. Miller0b725a22014-08-25 15:51:53 -0700498 * @xmit_more: More SKBs are pending for this queue
Randy Dunlap553a5672008-04-20 10:51:01 -0700499 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700500 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Tom Herbert61b905d2014-03-24 15:34:47 -0700501 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
Changli Gao4ca24622011-08-19 07:26:44 -0700502 * ports.
Tom Herberta3b18dd2014-07-01 21:33:17 -0700503 * @sw_hash: indicates hash was computed in software stack
Johannes Berg6e3e9392011-11-09 10:15:42 +0100504 * @wifi_acked_valid: wifi_acked was set
505 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000506 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Eliezer Tamir06021292013-06-10 11:39:50 +0300507 * @napi_id: id of the NAPI struct this skb came from
James Morris984bc162006-06-09 00:29:17 -0700508 * @secmark: security marking
Scott Feldman0c4f6912015-07-18 18:24:48 -0700509 * @offload_fwd_mark: fwding offload mark
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700510 * @mark: Generic packet mark
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000511 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b2008-07-14 22:49:06 -0700512 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000513 * @inner_protocol: Protocol (encapsulation)
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000514 * @inner_transport_header: Inner transport layer header (encapsulation)
515 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000516 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700517 * @transport_header: Transport layer header
518 * @network_header: Network layer header
519 * @mac_header: Link layer header
520 * @tail: Tail pointer
521 * @end: End pointer
522 * @head: Head of buffer
523 * @data: Data head pointer
524 * @truesize: Buffer size
525 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700526 */
527
528struct sk_buff {
Eric Dumazet363ec39232014-02-26 14:02:11 -0800529 union {
Eric Dumazet56b17422014-11-03 08:19:53 -0800530 struct {
531 /* These two members must be first. */
532 struct sk_buff *next;
533 struct sk_buff *prev;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000534
Eric Dumazet56b17422014-11-03 08:19:53 -0800535 union {
536 ktime_t tstamp;
537 struct skb_mstamp skb_mstamp;
538 };
539 };
540 struct rb_node rbnode; /* used in netem & tcp stack */
541 };
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000542 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545 /*
546 * This is the control buffer. It is free to use for every
547 * layer. Please put your private variables there. If you
548 * want to keep them across layers you have to do a skb_clone()
549 * first. This is owned by whoever has the skb queued ATM.
550 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000551 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552
Eric Dumazet7fee2262010-05-11 23:19:48 +0000553 unsigned long _skb_refdst;
Eric Dumazetb1937222014-09-28 22:18:47 -0700554 void (*destructor)(struct sk_buff *skb);
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000555#ifdef CONFIG_XFRM
556 struct sec_path *sp;
557#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700558#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
559 struct nf_conntrack *nfct;
560#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +0200561#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Eric Dumazetb1937222014-09-28 22:18:47 -0700562 struct nf_bridge_info *nf_bridge;
563#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700564 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700565 data_len;
566 __u16 mac_len,
567 hdr_len;
Eric Dumazetb1937222014-09-28 22:18:47 -0700568
569 /* Following fields are _not_ copied in __copy_skb_header()
570 * Note that queue_mapping is here mostly to fill a hole.
571 */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200572 kmemcheck_bitfield_begin(flags1);
Eric Dumazetb1937222014-09-28 22:18:47 -0700573 __u16 queue_mapping;
574 __u8 cloned:1,
Harald Welte6869c4d2005-08-09 19:24:19 -0700575 nohdr:1,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800576 fclone:2,
Herbert Xua59322b2007-12-05 01:53:40 -0800577 peeked:1,
Eric Dumazetb1937222014-09-28 22:18:47 -0700578 head_frag:1,
579 xmit_more:1;
580 /* one bit hole */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200581 kmemcheck_bitfield_end(flags1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582
Eric Dumazetb1937222014-09-28 22:18:47 -0700583 /* fields enclosed in headers_start/headers_end are copied
584 * using a single memcpy() in __copy_skb_header()
585 */
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700586 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700587 __u32 headers_start[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700588 /* public: */
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200589
590/* if you move pkt_type around you also must adapt those constants */
591#ifdef __BIG_ENDIAN_BITFIELD
592#define PKT_TYPE_MAX (7 << 5)
593#else
594#define PKT_TYPE_MAX 7
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100595#endif
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200596#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
597
598 __u8 __pkt_type_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700599 __u8 pkt_type:3;
600 __u8 pfmemalloc:1;
601 __u8 ignore_df:1;
602 __u8 nfctinfo:3;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603
Eric Dumazetb1937222014-09-28 22:18:47 -0700604 __u8 nf_trace:1;
605 __u8 ip_summed:2;
606 __u8 ooo_okay:1;
607 __u8 l4_hash:1;
608 __u8 sw_hash:1;
609 __u8 wifi_acked_valid:1;
610 __u8 wifi_acked:1;
611
612 __u8 no_fcs:1;
613 /* Indicates the inner headers are valid in the skbuff. */
614 __u8 encapsulation:1;
615 __u8 encap_hdr_csum:1;
616 __u8 csum_valid:1;
617 __u8 csum_complete_sw:1;
618 __u8 csum_level:2;
619 __u8 csum_bad:1;
620
621#ifdef CONFIG_IPV6_NDISC_NODETYPE
622 __u8 ndisc_nodetype:2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700624 __u8 ipvs_property:1;
Tom Herbert8bce6d72014-09-29 20:22:29 -0700625 __u8 inner_protocol_type:1;
Tom Herberte585f232014-11-04 09:06:54 -0800626 __u8 remcsum_offload:1;
627 /* 3 or 5 bit hole */
Alexander Duyck4031ae62012-01-27 06:22:53 +0000628
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700630 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700632 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200635
Eric Dumazetb1937222014-09-28 22:18:47 -0700636 union {
637 __wsum csum;
638 struct {
639 __u16 csum_start;
640 __u16 csum_offset;
641 };
642 };
643 __u32 priority;
644 int skb_iif;
645 __u32 hash;
646 __be16 vlan_proto;
647 __u16 vlan_tci;
Eric Dumazet2bd82482015-02-03 23:48:24 -0800648#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
649 union {
650 unsigned int napi_id;
651 unsigned int sender_cpu;
652 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700653#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -0700654 union {
James Morris984bc162006-06-09 00:29:17 -0700655#ifdef CONFIG_NETWORK_SECMARK
Scott Feldman0c4f6912015-07-18 18:24:48 -0700656 __u32 secmark;
James Morris984bc162006-06-09 00:29:17 -0700657#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -0700658#ifdef CONFIG_NET_SWITCHDEV
659 __u32 offload_fwd_mark;
660#endif
661 };
662
Neil Horman3b885782009-10-12 13:26:31 -0700663 union {
664 __u32 mark;
Eric Dumazet16fad692013-03-14 05:40:32 +0000665 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700666 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700667
Tom Herbert8bce6d72014-09-29 20:22:29 -0700668 union {
669 __be16 inner_protocol;
670 __u8 inner_ipproto;
671 };
672
Simon Horman1a37e412013-05-23 21:02:51 +0000673 __u16 inner_transport_header;
674 __u16 inner_network_header;
675 __u16 inner_mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700676
677 __be16 protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000678 __u16 transport_header;
679 __u16 network_header;
680 __u16 mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700681
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700682 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700683 __u32 headers_end[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700684 /* public: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700685
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700687 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700688 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700689 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700690 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700691 unsigned int truesize;
692 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700693};
694
695#ifdef __KERNEL__
696/*
697 * Handling routines are only of interest to the kernel
698 */
699#include <linux/slab.h>
700
Linus Torvalds1da177e2005-04-16 15:20:36 -0700701
Mel Gormanc93bdd02012-07-31 16:44:19 -0700702#define SKB_ALLOC_FCLONE 0x01
703#define SKB_ALLOC_RX 0x02
Alexander Duyckfd11a832014-12-09 19:40:49 -0800704#define SKB_ALLOC_NAPI 0x04
Mel Gormanc93bdd02012-07-31 16:44:19 -0700705
706/* Returns true if the skb was allocated from PFMEMALLOC reserves */
707static inline bool skb_pfmemalloc(const struct sk_buff *skb)
708{
709 return unlikely(skb->pfmemalloc);
710}
711
Eric Dumazet7fee2262010-05-11 23:19:48 +0000712/*
713 * skb might have a dst pointer attached, refcounted or not.
714 * _skb_refdst low order bit is set if refcount was _not_ taken
715 */
716#define SKB_DST_NOREF 1UL
717#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
718
719/**
720 * skb_dst - returns skb dst_entry
721 * @skb: buffer
722 *
723 * Returns skb dst_entry, regardless of reference taken or not.
724 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000725static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
726{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000727 /* If refdst was not refcounted, check we still are in a
728 * rcu_read_lock section
729 */
730 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
731 !rcu_read_lock_held() &&
732 !rcu_read_lock_bh_held());
733 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000734}
735
Eric Dumazet7fee2262010-05-11 23:19:48 +0000736/**
737 * skb_dst_set - sets skb dst
738 * @skb: buffer
739 * @dst: dst entry
740 *
741 * Sets skb dst, assuming a reference was taken on dst and should
742 * be released by skb_dst_drop()
743 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000744static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
745{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000746 skb->_skb_refdst = (unsigned long)dst;
747}
748
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200749/**
750 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
751 * @skb: buffer
752 * @dst: dst entry
753 *
754 * Sets skb dst, assuming a reference was not taken on dst.
755 * If dst entry is cached, we do not take reference and dst_release
756 * will be avoided by refdst_drop. If dst entry is not cached, we take
757 * reference, so that last dst_release can destroy the dst immediately.
758 */
759static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
760{
Hannes Frederic Sowadbfc4fb2014-12-06 19:19:42 +0100761 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
762 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200763}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000764
765/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300766 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000767 * @skb: buffer
768 */
769static inline bool skb_dst_is_noref(const struct sk_buff *skb)
770{
771 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000772}
773
Eric Dumazet511c3f92009-06-02 05:14:27 +0000774static inline struct rtable *skb_rtable(const struct sk_buff *skb)
775{
Eric Dumazetadf30902009-06-02 05:19:30 +0000776 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000777}
778
Joe Perches7965bd42013-09-26 14:48:15 -0700779void kfree_skb(struct sk_buff *skb);
780void kfree_skb_list(struct sk_buff *segs);
781void skb_tx_error(struct sk_buff *skb);
782void consume_skb(struct sk_buff *skb);
783void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000784extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000785
Joe Perches7965bd42013-09-26 14:48:15 -0700786void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
787bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
788 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000789
Joe Perches7965bd42013-09-26 14:48:15 -0700790struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
791 int node);
Eric Dumazet2ea2f622015-04-24 16:05:01 -0700792struct sk_buff *__build_skb(void *data, unsigned int frag_size);
Joe Perches7965bd42013-09-26 14:48:15 -0700793struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700794static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100795 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700796{
Eric Dumazet564824b2010-10-11 19:05:25 +0000797 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700798}
799
Eric Dumazet2e4e4412014-09-17 04:49:49 -0700800struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
801 unsigned long data_len,
802 int max_page_order,
803 int *errcode,
804 gfp_t gfp_mask);
805
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700806/* Layout of fast clones : [skb1][skb2][fclone_ref] */
807struct sk_buff_fclones {
808 struct sk_buff skb1;
809
810 struct sk_buff skb2;
811
812 atomic_t fclone_ref;
813};
814
815/**
816 * skb_fclone_busy - check if fclone is busy
817 * @skb: buffer
818 *
819 * Returns true is skb is a fast clone, and its clone is not freed.
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700820 * Some drivers call skb_orphan() in their ndo_start_xmit(),
821 * so we also check that this didnt happen.
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700822 */
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700823static inline bool skb_fclone_busy(const struct sock *sk,
824 const struct sk_buff *skb)
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700825{
826 const struct sk_buff_fclones *fclones;
827
828 fclones = container_of(skb, struct sk_buff_fclones, skb1);
829
830 return skb->fclone == SKB_FCLONE_ORIG &&
Eric Dumazet6ffe75e2014-12-03 17:04:39 -0800831 atomic_read(&fclones->fclone_ref) > 1 &&
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700832 fclones->skb2.sk == sk;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700833}
834
David S. Millerd179cd12005-08-17 14:57:30 -0700835static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100836 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700837{
Mel Gormanc93bdd02012-07-31 16:44:19 -0700838 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700839}
840
Joe Perches7965bd42013-09-26 14:48:15 -0700841struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
Patrick McHardy0ebd0ac2013-04-17 06:46:58 +0000842static inline struct sk_buff *alloc_skb_head(gfp_t priority)
843{
844 return __alloc_skb_head(priority, -1);
845}
846
Joe Perches7965bd42013-09-26 14:48:15 -0700847struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
848int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
849struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
850struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
Octavian Purdilabad93e92014-06-12 01:36:26 +0300851struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
852 gfp_t gfp_mask, bool fclone);
853static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
854 gfp_t gfp_mask)
855{
856 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
857}
Eric Dumazet117632e2011-12-03 21:39:53 +0000858
Joe Perches7965bd42013-09-26 14:48:15 -0700859int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
860struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
861 unsigned int headroom);
862struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
863 int newtailroom, gfp_t priority);
Fan Du25a91d82014-01-18 09:54:23 +0800864int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
865 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -0700866int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
867 int len);
868int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
869int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000870#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700871
Joe Perches7965bd42013-09-26 14:48:15 -0700872int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
873 int getfrag(void *from, char *to, int offset,
874 int len, int odd, struct sk_buff *skb),
875 void *from, int length);
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700876
Hannes Frederic Sowabe12a1f2015-05-21 16:59:58 +0200877int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
878 int offset, size_t size);
879
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800880struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700881 __u32 lower_offset;
882 __u32 upper_offset;
883 __u32 frag_idx;
884 __u32 stepped_offset;
885 struct sk_buff *root_skb;
886 struct sk_buff *cur_skb;
887 __u8 *frag_data;
888};
889
Joe Perches7965bd42013-09-26 14:48:15 -0700890void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
891 unsigned int to, struct skb_seq_state *st);
892unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
893 struct skb_seq_state *st);
894void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -0700895
Joe Perches7965bd42013-09-26 14:48:15 -0700896unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
Bojan Prtvar059a2442015-02-22 11:46:35 +0100897 unsigned int to, struct ts_config *config);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700898
Tom Herbert09323cc2013-12-15 22:16:19 -0800899/*
900 * Packet hash types specify the type of hash in skb_set_hash.
901 *
902 * Hash types refer to the protocol layer addresses which are used to
903 * construct a packet's hash. The hashes are used to differentiate or identify
904 * flows of the protocol layer for the hash type. Hash types are either
905 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
906 *
907 * Properties of hashes:
908 *
909 * 1) Two packets in different flows have different hash values
910 * 2) Two packets in the same flow should have the same hash value
911 *
912 * A hash at a higher layer is considered to be more specific. A driver should
913 * set the most specific hash possible.
914 *
915 * A driver cannot indicate a more specific hash than the layer at which a hash
916 * was computed. For instance an L3 hash cannot be set as an L4 hash.
917 *
918 * A driver may indicate a hash level which is less specific than the
919 * actual layer the hash was computed on. For instance, a hash computed
920 * at L4 may be considered an L3 hash. This should only be done if the
921 * driver can't unambiguously determine that the HW computed the hash at
922 * the higher layer. Note that the "should" in the second property above
923 * permits this.
924 */
925enum pkt_hash_types {
926 PKT_HASH_TYPE_NONE, /* Undefined type */
927 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
928 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
929 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
930};
931
932static inline void
933skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
934{
Tom Herbert61b905d2014-03-24 15:34:47 -0700935 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
Tom Herberta3b18dd2014-07-01 21:33:17 -0700936 skb->sw_hash = 0;
Tom Herbert61b905d2014-03-24 15:34:47 -0700937 skb->hash = hash;
Tom Herbert09323cc2013-12-15 22:16:19 -0800938}
939
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800940static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000941{
Tom Herberta3b18dd2014-07-01 21:33:17 -0700942 if (!skb->l4_hash && !skb->sw_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800943 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000944
Tom Herbert61b905d2014-03-24 15:34:47 -0700945 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000946}
947
Tom Herbert50fb7992015-05-01 11:30:12 -0700948__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
949
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800950static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
951{
Tom Herbert61b905d2014-03-24 15:34:47 -0700952 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800953}
954
Tom Herbert7539fad2013-12-15 22:12:18 -0800955static inline void skb_clear_hash(struct sk_buff *skb)
956{
Tom Herbert61b905d2014-03-24 15:34:47 -0700957 skb->hash = 0;
Tom Herberta3b18dd2014-07-01 21:33:17 -0700958 skb->sw_hash = 0;
Tom Herbert61b905d2014-03-24 15:34:47 -0700959 skb->l4_hash = 0;
Tom Herbert7539fad2013-12-15 22:12:18 -0800960}
961
962static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
963{
Tom Herbert61b905d2014-03-24 15:34:47 -0700964 if (!skb->l4_hash)
Tom Herbert7539fad2013-12-15 22:12:18 -0800965 skb_clear_hash(skb);
966}
967
Tom Herbert3df7a742013-12-15 22:16:29 -0800968static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
969{
Tom Herbert61b905d2014-03-24 15:34:47 -0700970 to->hash = from->hash;
Tom Herberta3b18dd2014-07-01 21:33:17 -0700971 to->sw_hash = from->sw_hash;
Tom Herbert61b905d2014-03-24 15:34:47 -0700972 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -0800973};
974
Eric Dumazetc29390c2015-03-11 18:42:02 -0700975static inline void skb_sender_cpu_clear(struct sk_buff *skb)
976{
977#ifdef CONFIG_XPS
978 skb->sender_cpu = 0;
979#endif
980}
981
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700982#ifdef NET_SKBUFF_DATA_USES_OFFSET
983static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
984{
985 return skb->head + skb->end;
986}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000987
988static inline unsigned int skb_end_offset(const struct sk_buff *skb)
989{
990 return skb->end;
991}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700992#else
993static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
994{
995 return skb->end;
996}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000997
998static inline unsigned int skb_end_offset(const struct sk_buff *skb)
999{
1000 return skb->end - skb->head;
1001}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001002#endif
1003
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001005#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006
Patrick Ohlyac45f602009-02-12 05:03:37 +00001007static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1008{
1009 return &skb_shinfo(skb)->hwtstamps;
1010}
1011
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012/**
1013 * skb_queue_empty - check if a queue is empty
1014 * @list: queue head
1015 *
1016 * Returns true if the queue is empty, false otherwise.
1017 */
1018static inline int skb_queue_empty(const struct sk_buff_head *list)
1019{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001020 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021}
1022
1023/**
David S. Millerfc7ebb22008-09-23 00:34:07 -07001024 * skb_queue_is_last - check if skb is the last entry in the queue
1025 * @list: queue head
1026 * @skb: buffer
1027 *
1028 * Returns true if @skb is the last buffer on the list.
1029 */
1030static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1031 const struct sk_buff *skb)
1032{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001033 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -07001034}
1035
1036/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001037 * skb_queue_is_first - check if skb is the first entry in the queue
1038 * @list: queue head
1039 * @skb: buffer
1040 *
1041 * Returns true if @skb is the first buffer on the list.
1042 */
1043static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1044 const struct sk_buff *skb)
1045{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001046 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001047}
1048
1049/**
David S. Miller249c8b42008-09-23 00:44:42 -07001050 * skb_queue_next - return the next packet in the queue
1051 * @list: queue head
1052 * @skb: current buffer
1053 *
1054 * Return the next packet in @list after @skb. It is only valid to
1055 * call this if skb_queue_is_last() evaluates to false.
1056 */
1057static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1058 const struct sk_buff *skb)
1059{
1060 /* This BUG_ON may seem severe, but if we just return then we
1061 * are going to dereference garbage.
1062 */
1063 BUG_ON(skb_queue_is_last(list, skb));
1064 return skb->next;
1065}
1066
1067/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001068 * skb_queue_prev - return the prev packet in the queue
1069 * @list: queue head
1070 * @skb: current buffer
1071 *
1072 * Return the prev packet in @list before @skb. It is only valid to
1073 * call this if skb_queue_is_first() evaluates to false.
1074 */
1075static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1076 const struct sk_buff *skb)
1077{
1078 /* This BUG_ON may seem severe, but if we just return then we
1079 * are going to dereference garbage.
1080 */
1081 BUG_ON(skb_queue_is_first(list, skb));
1082 return skb->prev;
1083}
1084
1085/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001086 * skb_get - reference buffer
1087 * @skb: buffer to reference
1088 *
1089 * Makes another reference to a socket buffer and returns a pointer
1090 * to the buffer.
1091 */
1092static inline struct sk_buff *skb_get(struct sk_buff *skb)
1093{
1094 atomic_inc(&skb->users);
1095 return skb;
1096}
1097
1098/*
1099 * If users == 1, we are the only owner and are can avoid redundant
1100 * atomic change.
1101 */
1102
1103/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001104 * skb_cloned - is the buffer a clone
1105 * @skb: buffer to check
1106 *
1107 * Returns true if the buffer was generated with skb_clone() and is
1108 * one of multiple shared copies of the buffer. Cloned buffers are
1109 * shared data so must not be written to under normal circumstances.
1110 */
1111static inline int skb_cloned(const struct sk_buff *skb)
1112{
1113 return skb->cloned &&
1114 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1115}
1116
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001117static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1118{
1119 might_sleep_if(pri & __GFP_WAIT);
1120
1121 if (skb_cloned(skb))
1122 return pskb_expand_head(skb, 0, 0, pri);
1123
1124 return 0;
1125}
1126
Linus Torvalds1da177e2005-04-16 15:20:36 -07001127/**
1128 * skb_header_cloned - is the header a clone
1129 * @skb: buffer to check
1130 *
1131 * Returns true if modifying the header part of the buffer requires
1132 * the data to be copied.
1133 */
1134static inline int skb_header_cloned(const struct sk_buff *skb)
1135{
1136 int dataref;
1137
1138 if (!skb->cloned)
1139 return 0;
1140
1141 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1142 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1143 return dataref != 1;
1144}
1145
1146/**
1147 * skb_header_release - release reference to header
1148 * @skb: buffer to operate on
1149 *
1150 * Drop a reference to the header part of the buffer. This is done
1151 * by acquiring a payload reference. You must not read from the header
1152 * part of skb->data after this.
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001153 * Note : Check if you can use __skb_header_release() instead.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001154 */
1155static inline void skb_header_release(struct sk_buff *skb)
1156{
1157 BUG_ON(skb->nohdr);
1158 skb->nohdr = 1;
1159 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1160}
1161
1162/**
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001163 * __skb_header_release - release reference to header
1164 * @skb: buffer to operate on
1165 *
1166 * Variant of skb_header_release() assuming skb is private to caller.
1167 * We can avoid one atomic operation.
1168 */
1169static inline void __skb_header_release(struct sk_buff *skb)
1170{
1171 skb->nohdr = 1;
1172 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1173}
1174
1175
1176/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001177 * skb_shared - is the buffer shared
1178 * @skb: buffer to check
1179 *
1180 * Returns true if more than one person has a reference to this
1181 * buffer.
1182 */
1183static inline int skb_shared(const struct sk_buff *skb)
1184{
1185 return atomic_read(&skb->users) != 1;
1186}
1187
1188/**
1189 * skb_share_check - check if buffer is shared and if so clone it
1190 * @skb: buffer to check
1191 * @pri: priority for memory allocation
1192 *
1193 * If the buffer is shared the buffer is cloned and the old copy
1194 * drops a reference. A new clone with a single reference is returned.
1195 * If the buffer is not shared the original buffer is returned. When
1196 * being called from interrupt status or with spinlocks held pri must
1197 * be GFP_ATOMIC.
1198 *
1199 * NULL is returned on a memory allocation failure.
1200 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001201static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001202{
1203 might_sleep_if(pri & __GFP_WAIT);
1204 if (skb_shared(skb)) {
1205 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001206
1207 if (likely(nskb))
1208 consume_skb(skb);
1209 else
1210 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211 skb = nskb;
1212 }
1213 return skb;
1214}
1215
1216/*
1217 * Copy shared buffers into a new sk_buff. We effectively do COW on
1218 * packets to handle cases where we have a local reader and forward
1219 * and a couple of other messy ones. The normal one is tcpdumping
1220 * a packet thats being forwarded.
1221 */
1222
1223/**
1224 * skb_unshare - make a copy of a shared buffer
1225 * @skb: buffer to check
1226 * @pri: priority for memory allocation
1227 *
1228 * If the socket buffer is a clone then this function creates a new
1229 * copy of the data, drops a reference count on the old copy and returns
1230 * the new copy with the reference count at 1. If the buffer is not a clone
1231 * the original buffer is returned. When called with a spinlock held or
1232 * from interrupt state @pri must be %GFP_ATOMIC
1233 *
1234 * %NULL is returned on a memory allocation failure.
1235 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001236static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001237 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001238{
1239 might_sleep_if(pri & __GFP_WAIT);
1240 if (skb_cloned(skb)) {
1241 struct sk_buff *nskb = skb_copy(skb, pri);
Alexander Aring31eff812014-10-10 23:10:47 +02001242
1243 /* Free our shared copy */
1244 if (likely(nskb))
1245 consume_skb(skb);
1246 else
1247 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248 skb = nskb;
1249 }
1250 return skb;
1251}
1252
1253/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001254 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001255 * @list_: list to peek at
1256 *
1257 * Peek an &sk_buff. Unlike most other operations you _MUST_
1258 * be careful with this one. A peek leaves the buffer on the
1259 * list and someone else may run off with it. You must hold
1260 * the appropriate locks or have a private queue to do this.
1261 *
1262 * Returns %NULL for an empty list or a pointer to the head element.
1263 * The reference count is not incremented and the reference is therefore
1264 * volatile. Use with caution.
1265 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001266static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001267{
Eric Dumazet18d07002012-04-30 16:31:46 +00001268 struct sk_buff *skb = list_->next;
1269
1270 if (skb == (struct sk_buff *)list_)
1271 skb = NULL;
1272 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001273}
1274
1275/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001276 * skb_peek_next - peek skb following the given one from a queue
1277 * @skb: skb to start from
1278 * @list_: list to peek at
1279 *
1280 * Returns %NULL when the end of the list is met or a pointer to the
1281 * next element. The reference count is not incremented and the
1282 * reference is therefore volatile. Use with caution.
1283 */
1284static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1285 const struct sk_buff_head *list_)
1286{
1287 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001288
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001289 if (next == (struct sk_buff *)list_)
1290 next = NULL;
1291 return next;
1292}
1293
1294/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001295 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001296 * @list_: list to peek at
1297 *
1298 * Peek an &sk_buff. Unlike most other operations you _MUST_
1299 * be careful with this one. A peek leaves the buffer on the
1300 * list and someone else may run off with it. You must hold
1301 * the appropriate locks or have a private queue to do this.
1302 *
1303 * Returns %NULL for an empty list or a pointer to the tail element.
1304 * The reference count is not incremented and the reference is therefore
1305 * volatile. Use with caution.
1306 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001307static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001308{
Eric Dumazet18d07002012-04-30 16:31:46 +00001309 struct sk_buff *skb = list_->prev;
1310
1311 if (skb == (struct sk_buff *)list_)
1312 skb = NULL;
1313 return skb;
1314
Linus Torvalds1da177e2005-04-16 15:20:36 -07001315}
1316
1317/**
1318 * skb_queue_len - get queue length
1319 * @list_: list to measure
1320 *
1321 * Return the length of an &sk_buff queue.
1322 */
1323static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1324{
1325 return list_->qlen;
1326}
1327
David S. Miller67fed452008-09-21 22:36:24 -07001328/**
1329 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1330 * @list: queue to initialize
1331 *
1332 * This initializes only the list and queue length aspects of
1333 * an sk_buff_head object. This allows to initialize the list
1334 * aspects of an sk_buff_head without reinitializing things like
1335 * the spinlock. It can also be used for on-stack sk_buff_head
1336 * objects where the spinlock is known to not be used.
1337 */
1338static inline void __skb_queue_head_init(struct sk_buff_head *list)
1339{
1340 list->prev = list->next = (struct sk_buff *)list;
1341 list->qlen = 0;
1342}
1343
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001344/*
1345 * This function creates a split out lock class for each invocation;
1346 * this is needed for now since a whole lot of users of the skb-queue
1347 * infrastructure in drivers have different locking usage (in hardirq)
1348 * than the networking core (in softirq only). In the long run either the
1349 * network layer or drivers should need annotation to consolidate the
1350 * main types of usage into 3 classes.
1351 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352static inline void skb_queue_head_init(struct sk_buff_head *list)
1353{
1354 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001355 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356}
1357
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001358static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1359 struct lock_class_key *class)
1360{
1361 skb_queue_head_init(list);
1362 lockdep_set_class(&list->lock, class);
1363}
1364
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001366 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001367 *
1368 * The "__skb_xxxx()" functions are the non-atomic ones that
1369 * can only be called with interrupts disabled.
1370 */
Joe Perches7965bd42013-09-26 14:48:15 -07001371void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1372 struct sk_buff_head *list);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001373static inline void __skb_insert(struct sk_buff *newsk,
1374 struct sk_buff *prev, struct sk_buff *next,
1375 struct sk_buff_head *list)
1376{
1377 newsk->next = next;
1378 newsk->prev = prev;
1379 next->prev = prev->next = newsk;
1380 list->qlen++;
1381}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001382
David S. Miller67fed452008-09-21 22:36:24 -07001383static inline void __skb_queue_splice(const struct sk_buff_head *list,
1384 struct sk_buff *prev,
1385 struct sk_buff *next)
1386{
1387 struct sk_buff *first = list->next;
1388 struct sk_buff *last = list->prev;
1389
1390 first->prev = prev;
1391 prev->next = first;
1392
1393 last->next = next;
1394 next->prev = last;
1395}
1396
1397/**
1398 * skb_queue_splice - join two skb lists, this is designed for stacks
1399 * @list: the new list to add
1400 * @head: the place to add it in the first list
1401 */
1402static inline void skb_queue_splice(const 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 }
1409}
1410
1411/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001412 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001413 * @list: the new list to add
1414 * @head: the place to add it in the first list
1415 *
1416 * The list at @list is reinitialised
1417 */
1418static inline void skb_queue_splice_init(struct sk_buff_head *list,
1419 struct sk_buff_head *head)
1420{
1421 if (!skb_queue_empty(list)) {
1422 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001423 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001424 __skb_queue_head_init(list);
1425 }
1426}
1427
1428/**
1429 * skb_queue_splice_tail - join two skb lists, each list being a queue
1430 * @list: the new list to add
1431 * @head: the place to add it in the first list
1432 */
1433static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1434 struct sk_buff_head *head)
1435{
1436 if (!skb_queue_empty(list)) {
1437 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001438 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001439 }
1440}
1441
1442/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001443 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001444 * @list: the new list to add
1445 * @head: the place to add it in the first list
1446 *
1447 * Each of the lists is a queue.
1448 * The list at @list is reinitialised
1449 */
1450static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1451 struct sk_buff_head *head)
1452{
1453 if (!skb_queue_empty(list)) {
1454 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001455 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001456 __skb_queue_head_init(list);
1457 }
1458}
1459
Linus Torvalds1da177e2005-04-16 15:20:36 -07001460/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001461 * __skb_queue_after - queue a buffer at the list head
1462 * @list: list to use
1463 * @prev: place after this buffer
1464 * @newsk: buffer to queue
1465 *
1466 * Queue a buffer int the middle of a list. This function takes no locks
1467 * and you must therefore hold required locks before calling it.
1468 *
1469 * A buffer cannot be placed on two lists at the same time.
1470 */
1471static inline void __skb_queue_after(struct sk_buff_head *list,
1472 struct sk_buff *prev,
1473 struct sk_buff *newsk)
1474{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001475 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001476}
1477
Joe Perches7965bd42013-09-26 14:48:15 -07001478void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1479 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001480
Gerrit Renkerf5572852008-04-14 00:05:28 -07001481static inline void __skb_queue_before(struct sk_buff_head *list,
1482 struct sk_buff *next,
1483 struct sk_buff *newsk)
1484{
1485 __skb_insert(newsk, next->prev, next, list);
1486}
1487
Stephen Hemminger300ce172005-10-30 13:47:34 -08001488/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001489 * __skb_queue_head - queue a buffer at the list head
1490 * @list: list to use
1491 * @newsk: buffer to queue
1492 *
1493 * Queue a buffer at the start of a list. This function takes no locks
1494 * and you must therefore hold required locks before calling it.
1495 *
1496 * A buffer cannot be placed on two lists at the same time.
1497 */
Joe Perches7965bd42013-09-26 14:48:15 -07001498void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499static inline void __skb_queue_head(struct sk_buff_head *list,
1500 struct sk_buff *newsk)
1501{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001502 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001503}
1504
1505/**
1506 * __skb_queue_tail - queue a buffer at the list tail
1507 * @list: list to use
1508 * @newsk: buffer to queue
1509 *
1510 * Queue a buffer at the end of a list. This function takes no locks
1511 * and you must therefore hold required locks before calling it.
1512 *
1513 * A buffer cannot be placed on two lists at the same time.
1514 */
Joe Perches7965bd42013-09-26 14:48:15 -07001515void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001516static inline void __skb_queue_tail(struct sk_buff_head *list,
1517 struct sk_buff *newsk)
1518{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001519 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520}
1521
Linus Torvalds1da177e2005-04-16 15:20:36 -07001522/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523 * remove sk_buff from list. _Must_ be called atomically, and with
1524 * the list known..
1525 */
Joe Perches7965bd42013-09-26 14:48:15 -07001526void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001527static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1528{
1529 struct sk_buff *next, *prev;
1530
1531 list->qlen--;
1532 next = skb->next;
1533 prev = skb->prev;
1534 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001535 next->prev = prev;
1536 prev->next = next;
1537}
1538
Gerrit Renkerf525c062008-04-14 00:04:12 -07001539/**
1540 * __skb_dequeue - remove from the head of the queue
1541 * @list: list to dequeue from
1542 *
1543 * Remove the head of the list. This function does not take any locks
1544 * so must be used with appropriate locks held only. The head item is
1545 * returned or %NULL if the list is empty.
1546 */
Joe Perches7965bd42013-09-26 14:48:15 -07001547struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Gerrit Renkerf525c062008-04-14 00:04:12 -07001548static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1549{
1550 struct sk_buff *skb = skb_peek(list);
1551 if (skb)
1552 __skb_unlink(skb, list);
1553 return skb;
1554}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001555
1556/**
1557 * __skb_dequeue_tail - remove from the tail of the queue
1558 * @list: list to dequeue from
1559 *
1560 * Remove the tail of the list. This function does not take any locks
1561 * so must be used with appropriate locks held only. The tail item is
1562 * returned or %NULL if the list is empty.
1563 */
Joe Perches7965bd42013-09-26 14:48:15 -07001564struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1566{
1567 struct sk_buff *skb = skb_peek_tail(list);
1568 if (skb)
1569 __skb_unlink(skb, list);
1570 return skb;
1571}
1572
1573
David S. Millerbdcc0922012-03-07 20:53:36 -05001574static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001575{
1576 return skb->data_len;
1577}
1578
1579static inline unsigned int skb_headlen(const struct sk_buff *skb)
1580{
1581 return skb->len - skb->data_len;
1582}
1583
1584static inline int skb_pagelen(const struct sk_buff *skb)
1585{
1586 int i, len = 0;
1587
1588 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001589 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001590 return len + skb_headlen(skb);
1591}
1592
Ian Campbell131ea662011-08-19 06:25:00 +00001593/**
1594 * __skb_fill_page_desc - initialise a paged fragment in an skb
1595 * @skb: buffer containing fragment to be initialised
1596 * @i: paged fragment index to initialise
1597 * @page: the page to use for this fragment
1598 * @off: the offset to the data with @page
1599 * @size: the length of the data
1600 *
1601 * Initialises the @i'th fragment of @skb to point to &size bytes at
1602 * offset @off within @page.
1603 *
1604 * Does not take any additional reference on the fragment.
1605 */
1606static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1607 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001608{
1609 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1610
Mel Gormanc48a11c2012-07-31 16:44:23 -07001611 /*
1612 * Propagate page->pfmemalloc to the skb if we can. The problem is
1613 * that not all callers have unique ownership of the page. If
1614 * pfmemalloc is set, we check the mapping as a mapping implies
1615 * page->index is set (index and pfmemalloc share space).
1616 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1617 * do not lose pfmemalloc information as the pages would not be
1618 * allocated using __GFP_MEMALLOC.
1619 */
Ian Campbella8605c62011-10-19 23:01:49 +00001620 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001621 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001622 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001623
1624 page = compound_head(page);
1625 if (page->pfmemalloc && !page->mapping)
1626 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00001627}
1628
1629/**
1630 * skb_fill_page_desc - initialise a paged fragment in an skb
1631 * @skb: buffer containing fragment to be initialised
1632 * @i: paged fragment index to initialise
1633 * @page: the page to use for this fragment
1634 * @off: the offset to the data with @page
1635 * @size: the length of the data
1636 *
1637 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01001638 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00001639 * addition updates @skb such that @i is the last fragment.
1640 *
1641 * Does not take any additional reference on the fragment.
1642 */
1643static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1644 struct page *page, int off, int size)
1645{
1646 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647 skb_shinfo(skb)->nr_frags = i + 1;
1648}
1649
Joe Perches7965bd42013-09-26 14:48:15 -07001650void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1651 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001652
Jason Wangf8e617e2013-11-01 14:07:47 +08001653void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1654 unsigned int truesize);
1655
Linus Torvalds1da177e2005-04-16 15:20:36 -07001656#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001657#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001658#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1659
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001660#ifdef NET_SKBUFF_DATA_USES_OFFSET
1661static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1662{
1663 return skb->head + skb->tail;
1664}
1665
1666static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1667{
1668 skb->tail = skb->data - skb->head;
1669}
1670
1671static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1672{
1673 skb_reset_tail_pointer(skb);
1674 skb->tail += offset;
1675}
Simon Horman7cc46192013-05-28 20:34:29 +00001676
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001677#else /* NET_SKBUFF_DATA_USES_OFFSET */
1678static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1679{
1680 return skb->tail;
1681}
1682
1683static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1684{
1685 skb->tail = skb->data;
1686}
1687
1688static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1689{
1690 skb->tail = skb->data + offset;
1691}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001692
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001693#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1694
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695/*
1696 * Add data to an sk_buff
1697 */
Mathias Krause0c7ddf32013-11-07 14:18:24 +01001698unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001699unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001700static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1701{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001702 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001703 SKB_LINEAR_ASSERT(skb);
1704 skb->tail += len;
1705 skb->len += len;
1706 return tmp;
1707}
1708
Joe Perches7965bd42013-09-26 14:48:15 -07001709unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1711{
1712 skb->data -= len;
1713 skb->len += len;
1714 return skb->data;
1715}
1716
Joe Perches7965bd42013-09-26 14:48:15 -07001717unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001718static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1719{
1720 skb->len -= len;
1721 BUG_ON(skb->len < skb->data_len);
1722 return skb->data += len;
1723}
1724
David S. Miller47d29642010-05-02 02:21:44 -07001725static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1726{
1727 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1728}
1729
Joe Perches7965bd42013-09-26 14:48:15 -07001730unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731
1732static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1733{
1734 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001735 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001736 return NULL;
1737 skb->len -= len;
1738 return skb->data += len;
1739}
1740
1741static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1742{
1743 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1744}
1745
1746static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1747{
1748 if (likely(len <= skb_headlen(skb)))
1749 return 1;
1750 if (unlikely(len > skb->len))
1751 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001752 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001753}
1754
1755/**
1756 * skb_headroom - bytes at buffer head
1757 * @skb: buffer to check
1758 *
1759 * Return the number of bytes of free space at the head of an &sk_buff.
1760 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001761static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001762{
1763 return skb->data - skb->head;
1764}
1765
1766/**
1767 * skb_tailroom - bytes at buffer end
1768 * @skb: buffer to check
1769 *
1770 * Return the number of bytes of free space at the tail of an sk_buff
1771 */
1772static inline int skb_tailroom(const struct sk_buff *skb)
1773{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001774 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001775}
1776
1777/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00001778 * skb_availroom - bytes at buffer end
1779 * @skb: buffer to check
1780 *
1781 * Return the number of bytes of free space at the tail of an sk_buff
1782 * allocated by sk_stream_alloc()
1783 */
1784static inline int skb_availroom(const struct sk_buff *skb)
1785{
Eric Dumazet16fad692013-03-14 05:40:32 +00001786 if (skb_is_nonlinear(skb))
1787 return 0;
1788
1789 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00001790}
1791
1792/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001793 * skb_reserve - adjust headroom
1794 * @skb: buffer to alter
1795 * @len: bytes to move
1796 *
1797 * Increase the headroom of an empty &sk_buff by reducing the tail
1798 * room. This is only allowed for an empty buffer.
1799 */
David S. Miller8243126c2006-01-17 02:54:21 -08001800static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001801{
1802 skb->data += len;
1803 skb->tail += len;
1804}
1805
Tom Herbert8bce6d72014-09-29 20:22:29 -07001806#define ENCAP_TYPE_ETHER 0
1807#define ENCAP_TYPE_IPPROTO 1
1808
1809static inline void skb_set_inner_protocol(struct sk_buff *skb,
1810 __be16 protocol)
1811{
1812 skb->inner_protocol = protocol;
1813 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
1814}
1815
1816static inline void skb_set_inner_ipproto(struct sk_buff *skb,
1817 __u8 ipproto)
1818{
1819 skb->inner_ipproto = ipproto;
1820 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
1821}
1822
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001823static inline void skb_reset_inner_headers(struct sk_buff *skb)
1824{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001825 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001826 skb->inner_network_header = skb->network_header;
1827 skb->inner_transport_header = skb->transport_header;
1828}
1829
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00001830static inline void skb_reset_mac_len(struct sk_buff *skb)
1831{
1832 skb->mac_len = skb->network_header - skb->mac_header;
1833}
1834
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001835static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1836 *skb)
1837{
1838 return skb->head + skb->inner_transport_header;
1839}
1840
1841static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1842{
1843 skb->inner_transport_header = skb->data - skb->head;
1844}
1845
1846static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1847 const int offset)
1848{
1849 skb_reset_inner_transport_header(skb);
1850 skb->inner_transport_header += offset;
1851}
1852
1853static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1854{
1855 return skb->head + skb->inner_network_header;
1856}
1857
1858static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1859{
1860 skb->inner_network_header = skb->data - skb->head;
1861}
1862
1863static inline void skb_set_inner_network_header(struct sk_buff *skb,
1864 const int offset)
1865{
1866 skb_reset_inner_network_header(skb);
1867 skb->inner_network_header += offset;
1868}
1869
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001870static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1871{
1872 return skb->head + skb->inner_mac_header;
1873}
1874
1875static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1876{
1877 skb->inner_mac_header = skb->data - skb->head;
1878}
1879
1880static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1881 const int offset)
1882{
1883 skb_reset_inner_mac_header(skb);
1884 skb->inner_mac_header += offset;
1885}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001886static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1887{
Cong Wang35d04612013-05-29 15:16:05 +08001888 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001889}
1890
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001891static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1892{
1893 return skb->head + skb->transport_header;
1894}
1895
1896static inline void skb_reset_transport_header(struct sk_buff *skb)
1897{
1898 skb->transport_header = skb->data - skb->head;
1899}
1900
1901static inline void skb_set_transport_header(struct sk_buff *skb,
1902 const int offset)
1903{
1904 skb_reset_transport_header(skb);
1905 skb->transport_header += offset;
1906}
1907
1908static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1909{
1910 return skb->head + skb->network_header;
1911}
1912
1913static inline void skb_reset_network_header(struct sk_buff *skb)
1914{
1915 skb->network_header = skb->data - skb->head;
1916}
1917
1918static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1919{
1920 skb_reset_network_header(skb);
1921 skb->network_header += offset;
1922}
1923
1924static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1925{
1926 return skb->head + skb->mac_header;
1927}
1928
1929static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1930{
Cong Wang35d04612013-05-29 15:16:05 +08001931 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001932}
1933
1934static inline void skb_reset_mac_header(struct sk_buff *skb)
1935{
1936 skb->mac_header = skb->data - skb->head;
1937}
1938
1939static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1940{
1941 skb_reset_mac_header(skb);
1942 skb->mac_header += offset;
1943}
1944
Timo Teräs0e3da5b2013-12-16 11:02:09 +02001945static inline void skb_pop_mac_header(struct sk_buff *skb)
1946{
1947 skb->mac_header = skb->network_header;
1948}
1949
Ying Xuefbbdb8f2013-03-27 16:46:06 +00001950static inline void skb_probe_transport_header(struct sk_buff *skb,
1951 const int offset_hint)
1952{
1953 struct flow_keys keys;
1954
1955 if (skb_transport_header_was_set(skb))
1956 return;
Jiri Pirko06635a32015-05-12 14:56:16 +02001957 else if (skb_flow_dissect_flow_keys(skb, &keys))
Tom Herbert42aecaa2015-06-04 09:16:39 -07001958 skb_set_transport_header(skb, keys.control.thoff);
Ying Xuefbbdb8f2013-03-27 16:46:06 +00001959 else
1960 skb_set_transport_header(skb, offset_hint);
1961}
1962
Eric Dumazet03606892012-02-23 10:55:02 +00001963static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1964{
1965 if (skb_mac_header_was_set(skb)) {
1966 const unsigned char *old_mac = skb_mac_header(skb);
1967
1968 skb_set_mac_header(skb, -skb->mac_len);
1969 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1970 }
1971}
1972
Michał Mirosław04fb4512010-12-14 15:24:08 +00001973static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1974{
1975 return skb->csum_start - skb_headroom(skb);
1976}
1977
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001978static inline int skb_transport_offset(const struct sk_buff *skb)
1979{
1980 return skb_transport_header(skb) - skb->data;
1981}
1982
1983static inline u32 skb_network_header_len(const struct sk_buff *skb)
1984{
1985 return skb->transport_header - skb->network_header;
1986}
1987
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001988static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1989{
1990 return skb->inner_transport_header - skb->inner_network_header;
1991}
1992
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001993static inline int skb_network_offset(const struct sk_buff *skb)
1994{
1995 return skb_network_header(skb) - skb->data;
1996}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001997
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001998static inline int skb_inner_network_offset(const struct sk_buff *skb)
1999{
2000 return skb_inner_network_header(skb) - skb->data;
2001}
2002
Changli Gaof9599ce2010-08-04 04:43:44 +00002003static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2004{
2005 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2006}
2007
Linus Torvalds1da177e2005-04-16 15:20:36 -07002008/*
2009 * CPUs often take a performance hit when accessing unaligned memory
2010 * locations. The actual performance hit varies, it can be small if the
2011 * hardware handles it or large if we have to take an exception and fix it
2012 * in software.
2013 *
2014 * Since an ethernet header is 14 bytes network drivers often end up with
2015 * the IP header at an unaligned offset. The IP header can be aligned by
2016 * shifting the start of the packet by 2 bytes. Drivers should do this
2017 * with:
2018 *
Tobias Klauser8660c122009-07-13 22:48:16 +00002019 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002020 *
2021 * The downside to this alignment of the IP header is that the DMA is now
2022 * unaligned. On some architectures the cost of an unaligned DMA is high
2023 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00002024 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002025 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2026 * to be overridden.
2027 */
2028#ifndef NET_IP_ALIGN
2029#define NET_IP_ALIGN 2
2030#endif
2031
Anton Blanchard025be812006-03-31 02:27:06 -08002032/*
2033 * The networking layer reserves some headroom in skb data (via
2034 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2035 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08002036 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08002037 *
2038 * Unfortunately this headroom changes the DMA alignment of the resulting
2039 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2040 * on some architectures. An architecture can override this value,
2041 * perhaps setting it to a cacheline in size (since that will maintain
2042 * cacheline alignment of the DMA). It must be a power of 2.
2043 *
David S. Millerd6301d32009-02-08 19:24:13 -08002044 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08002045 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07002046 *
2047 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2048 * to reduce average number of cache lines per packet.
2049 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07002050 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08002051 */
2052#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07002053#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08002054#endif
2055
Joe Perches7965bd42013-09-26 14:48:15 -07002056int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002057
2058static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2059{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00002060 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07002061 WARN_ON(1);
2062 return;
2063 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002064 skb->len = len;
2065 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002066}
2067
Joe Perches7965bd42013-09-26 14:48:15 -07002068void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002069
2070static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2071{
Herbert Xu3cc0e872006-06-09 16:13:38 -07002072 if (skb->data_len)
2073 return ___pskb_trim(skb, len);
2074 __skb_trim(skb, len);
2075 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002076}
2077
2078static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2079{
2080 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2081}
2082
2083/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07002084 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2085 * @skb: buffer to alter
2086 * @len: new length
2087 *
2088 * This is identical to pskb_trim except that the caller knows that
2089 * the skb is not cloned so we should never get an error due to out-
2090 * of-memory.
2091 */
2092static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2093{
2094 int err = pskb_trim(skb, len);
2095 BUG_ON(err);
2096}
2097
2098/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002099 * skb_orphan - orphan a buffer
2100 * @skb: buffer to orphan
2101 *
2102 * If a buffer currently has an owner then we call the owner's
2103 * destructor function and make the @skb unowned. The buffer continues
2104 * to exist but is no longer charged to its former owner.
2105 */
2106static inline void skb_orphan(struct sk_buff *skb)
2107{
Eric Dumazetc34a7612013-07-30 16:11:15 -07002108 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002109 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002110 skb->destructor = NULL;
2111 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07002112 } else {
2113 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002114 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002115}
2116
2117/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002118 * skb_orphan_frags - orphan the frags contained in a buffer
2119 * @skb: buffer to orphan frags from
2120 * @gfp_mask: allocation mask for replacement pages
2121 *
2122 * For each frag in the SKB which needs a destructor (i.e. has an
2123 * owner) create a copy of that frag and release the original
2124 * page by calling the destructor.
2125 */
2126static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2127{
2128 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2129 return 0;
2130 return skb_copy_ubufs(skb, gfp_mask);
2131}
2132
2133/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002134 * __skb_queue_purge - empty a list
2135 * @list: list to empty
2136 *
2137 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2138 * the list and one reference dropped. This function does not take the
2139 * list lock and the caller must hold the relevant locks to use it.
2140 */
Joe Perches7965bd42013-09-26 14:48:15 -07002141void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002142static inline void __skb_queue_purge(struct sk_buff_head *list)
2143{
2144 struct sk_buff *skb;
2145 while ((skb = __skb_dequeue(list)) != NULL)
2146 kfree_skb(skb);
2147}
2148
Joe Perches7965bd42013-09-26 14:48:15 -07002149void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002150
Joe Perches7965bd42013-09-26 14:48:15 -07002151struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2152 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07002153
2154/**
2155 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2156 * @dev: network device to receive on
2157 * @length: length to allocate
2158 *
2159 * Allocate a new &sk_buff and assign it a usage count of one. The
2160 * buffer has unspecified headroom built in. Users should allocate
2161 * the headroom they think they need without accounting for the
2162 * built in space. The built in space is used for optimisations.
2163 *
2164 * %NULL is returned if there is no free memory. Although this function
2165 * allocates memory it can be called from an interrupt.
2166 */
2167static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002168 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002169{
2170 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2171}
2172
Eric Dumazet6f532612012-05-18 05:12:12 +00002173/* legacy helper around __netdev_alloc_skb() */
2174static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2175 gfp_t gfp_mask)
2176{
2177 return __netdev_alloc_skb(NULL, length, gfp_mask);
2178}
2179
2180/* legacy helper around netdev_alloc_skb() */
2181static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2182{
2183 return netdev_alloc_skb(NULL, length);
2184}
2185
2186
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002187static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2188 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002189{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002190 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002191
2192 if (NET_IP_ALIGN && skb)
2193 skb_reserve(skb, NET_IP_ALIGN);
2194 return skb;
2195}
2196
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002197static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2198 unsigned int length)
2199{
2200 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2201}
2202
Alexander Duyck181edb22015-05-06 21:12:03 -07002203static inline void skb_free_frag(void *addr)
2204{
2205 __free_page_frag(addr);
2206}
2207
Alexander Duyckffde7322014-12-09 19:40:42 -08002208void *napi_alloc_frag(unsigned int fragsz);
Alexander Duyckfd11a832014-12-09 19:40:49 -08002209struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2210 unsigned int length, gfp_t gfp_mask);
2211static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2212 unsigned int length)
2213{
2214 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2215}
Alexander Duyckffde7322014-12-09 19:40:42 -08002216
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002217/**
Alexander Duyck71dfda52014-11-11 09:26:34 -08002218 * __dev_alloc_pages - allocate page for network Rx
2219 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2220 * @order: size of the allocation
2221 *
2222 * Allocate a new page.
2223 *
2224 * %NULL is returned if there is no free memory.
2225*/
2226static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2227 unsigned int order)
2228{
2229 /* This piece of code contains several assumptions.
2230 * 1. This is for device Rx, therefor a cold page is preferred.
2231 * 2. The expectation is the user wants a compound page.
2232 * 3. If requesting a order 0 page it will not be compound
2233 * due to the check to see if order has a value in prep_new_page
2234 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2235 * code in gfp_to_alloc_flags that should be enforcing this.
2236 */
2237 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2238
2239 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2240}
2241
2242static inline struct page *dev_alloc_pages(unsigned int order)
2243{
2244 return __dev_alloc_pages(GFP_ATOMIC, order);
2245}
2246
2247/**
2248 * __dev_alloc_page - allocate a page for network Rx
2249 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2250 *
2251 * Allocate a new page.
2252 *
2253 * %NULL is returned if there is no free memory.
2254 */
2255static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2256{
2257 return __dev_alloc_pages(gfp_mask, 0);
2258}
2259
2260static inline struct page *dev_alloc_page(void)
2261{
2262 return __dev_alloc_page(GFP_ATOMIC);
2263}
2264
2265/**
Mel Gorman06140022012-07-31 16:44:24 -07002266 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2267 * @page: The page that was allocated from skb_alloc_page
2268 * @skb: The skb that may need pfmemalloc set
2269 */
2270static inline void skb_propagate_pfmemalloc(struct page *page,
2271 struct sk_buff *skb)
2272{
2273 if (page && page->pfmemalloc)
2274 skb->pfmemalloc = true;
2275}
2276
Eric Dumazet564824b2010-10-11 19:05:25 +00002277/**
Masanari Iidae2278672014-02-18 22:54:36 +09002278 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002279 * @frag: the paged fragment
2280 *
2281 * Returns the &struct page associated with @frag.
2282 */
2283static inline struct page *skb_frag_page(const skb_frag_t *frag)
2284{
Ian Campbella8605c62011-10-19 23:01:49 +00002285 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00002286}
2287
2288/**
2289 * __skb_frag_ref - take an addition reference on a paged fragment.
2290 * @frag: the paged fragment
2291 *
2292 * Takes an additional reference on the paged fragment @frag.
2293 */
2294static inline void __skb_frag_ref(skb_frag_t *frag)
2295{
2296 get_page(skb_frag_page(frag));
2297}
2298
2299/**
2300 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2301 * @skb: the buffer
2302 * @f: the fragment offset.
2303 *
2304 * Takes an additional reference on the @f'th paged fragment of @skb.
2305 */
2306static inline void skb_frag_ref(struct sk_buff *skb, int f)
2307{
2308 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2309}
2310
2311/**
2312 * __skb_frag_unref - release a reference on a paged fragment.
2313 * @frag: the paged fragment
2314 *
2315 * Releases a reference on the paged fragment @frag.
2316 */
2317static inline void __skb_frag_unref(skb_frag_t *frag)
2318{
2319 put_page(skb_frag_page(frag));
2320}
2321
2322/**
2323 * skb_frag_unref - release a reference on a paged fragment of an skb.
2324 * @skb: the buffer
2325 * @f: the fragment offset
2326 *
2327 * Releases a reference on the @f'th paged fragment of @skb.
2328 */
2329static inline void skb_frag_unref(struct sk_buff *skb, int f)
2330{
2331 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2332}
2333
2334/**
2335 * skb_frag_address - gets the address of the data contained in a paged fragment
2336 * @frag: the paged fragment buffer
2337 *
2338 * Returns the address of the data within @frag. The page must already
2339 * be mapped.
2340 */
2341static inline void *skb_frag_address(const skb_frag_t *frag)
2342{
2343 return page_address(skb_frag_page(frag)) + frag->page_offset;
2344}
2345
2346/**
2347 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2348 * @frag: the paged fragment buffer
2349 *
2350 * Returns the address of the data within @frag. Checks that the page
2351 * is mapped and returns %NULL otherwise.
2352 */
2353static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2354{
2355 void *ptr = page_address(skb_frag_page(frag));
2356 if (unlikely(!ptr))
2357 return NULL;
2358
2359 return ptr + frag->page_offset;
2360}
2361
2362/**
2363 * __skb_frag_set_page - sets the page contained in a paged fragment
2364 * @frag: the paged fragment
2365 * @page: the page to set
2366 *
2367 * Sets the fragment @frag to contain @page.
2368 */
2369static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2370{
Ian Campbella8605c62011-10-19 23:01:49 +00002371 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002372}
2373
2374/**
2375 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2376 * @skb: the buffer
2377 * @f: the fragment offset
2378 * @page: the page to set
2379 *
2380 * Sets the @f'th fragment of @skb to contain @page.
2381 */
2382static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2383 struct page *page)
2384{
2385 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2386}
2387
Eric Dumazet400dfd32013-10-17 16:27:07 -07002388bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2389
Ian Campbell131ea662011-08-19 06:25:00 +00002390/**
2391 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002392 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002393 * @frag: the paged fragment to map
2394 * @offset: the offset within the fragment (starting at the
2395 * fragment's own offset)
2396 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002397 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00002398 *
2399 * Maps the page associated with @frag to @device.
2400 */
2401static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2402 const skb_frag_t *frag,
2403 size_t offset, size_t size,
2404 enum dma_data_direction dir)
2405{
2406 return dma_map_page(dev, skb_frag_page(frag),
2407 frag->page_offset + offset, size, dir);
2408}
2409
Eric Dumazet117632e2011-12-03 21:39:53 +00002410static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2411 gfp_t gfp_mask)
2412{
2413 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2414}
2415
Octavian Purdilabad93e92014-06-12 01:36:26 +03002416
2417static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2418 gfp_t gfp_mask)
2419{
2420 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2421}
2422
2423
Ian Campbell131ea662011-08-19 06:25:00 +00002424/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002425 * skb_clone_writable - is the header of a clone writable
2426 * @skb: buffer to check
2427 * @len: length up to which to write
2428 *
2429 * Returns true if modifying the header part of the cloned buffer
2430 * does not requires the data to be copied.
2431 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002432static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002433{
2434 return !skb_header_cloned(skb) &&
2435 skb_headroom(skb) + len <= skb->hdr_len;
2436}
2437
Herbert Xud9cc2042007-09-16 16:21:16 -07002438static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2439 int cloned)
2440{
2441 int delta = 0;
2442
Herbert Xud9cc2042007-09-16 16:21:16 -07002443 if (headroom > skb_headroom(skb))
2444 delta = headroom - skb_headroom(skb);
2445
2446 if (delta || cloned)
2447 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2448 GFP_ATOMIC);
2449 return 0;
2450}
2451
Patrick McHardy334a8132007-06-25 04:35:20 -07002452/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002453 * skb_cow - copy header of skb when it is required
2454 * @skb: buffer to cow
2455 * @headroom: needed headroom
2456 *
2457 * If the skb passed lacks sufficient headroom or its data part
2458 * is shared, data is reallocated. If reallocation fails, an error
2459 * is returned and original skb is not changed.
2460 *
2461 * The result is skb with writable area skb->head...skb->tail
2462 * and at least @headroom of space at head.
2463 */
2464static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2465{
Herbert Xud9cc2042007-09-16 16:21:16 -07002466 return __skb_cow(skb, headroom, skb_cloned(skb));
2467}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002468
Herbert Xud9cc2042007-09-16 16:21:16 -07002469/**
2470 * skb_cow_head - skb_cow but only making the head writable
2471 * @skb: buffer to cow
2472 * @headroom: needed headroom
2473 *
2474 * This function is identical to skb_cow except that we replace the
2475 * skb_cloned check by skb_header_cloned. It should be used when
2476 * you only need to push on some header and do not need to modify
2477 * the data.
2478 */
2479static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2480{
2481 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002482}
2483
2484/**
2485 * skb_padto - pad an skbuff up to a minimal size
2486 * @skb: buffer to pad
2487 * @len: minimal length
2488 *
2489 * Pads up a buffer to ensure the trailing bytes exist and are
2490 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002491 * is untouched. Otherwise it is extended. Returns zero on
2492 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002493 */
Herbert Xu5b057c62006-06-23 02:06:41 -07002494static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002495{
2496 unsigned int size = skb->len;
2497 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002498 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002499 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002500}
2501
Alexander Duyck9c0c1122014-12-03 08:17:33 -08002502/**
2503 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2504 * @skb: buffer to pad
2505 * @len: minimal length
2506 *
2507 * Pads up a buffer to ensure the trailing bytes exist and are
2508 * blanked. If the buffer already contains sufficient data it
2509 * is untouched. Otherwise it is extended. Returns zero on
2510 * success. The skb is freed on error.
2511 */
2512static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2513{
2514 unsigned int size = skb->len;
2515
2516 if (unlikely(size < len)) {
2517 len -= size;
2518 if (skb_pad(skb, len))
2519 return -ENOMEM;
2520 __skb_put(skb, len);
2521 }
2522 return 0;
2523}
2524
Linus Torvalds1da177e2005-04-16 15:20:36 -07002525static inline int skb_add_data(struct sk_buff *skb,
Al Viroaf2b0402014-11-27 21:44:24 -05002526 struct iov_iter *from, int copy)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002527{
2528 const int off = skb->len;
2529
2530 if (skb->ip_summed == CHECKSUM_NONE) {
Al Viroaf2b0402014-11-27 21:44:24 -05002531 __wsum csum = 0;
2532 if (csum_and_copy_from_iter(skb_put(skb, copy), copy,
2533 &csum, from) == copy) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002534 skb->csum = csum_block_add(skb->csum, csum, off);
2535 return 0;
2536 }
Al Viroaf2b0402014-11-27 21:44:24 -05002537 } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002538 return 0;
2539
2540 __skb_trim(skb, off);
2541 return -EFAULT;
2542}
2543
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002544static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2545 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002546{
2547 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002548 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002549
Ian Campbellea2ab692011-08-22 23:44:58 +00002550 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002551 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002552 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002553 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002554}
2555
Herbert Xu364c6ba2006-06-09 16:10:40 -07002556static inline int __skb_linearize(struct sk_buff *skb)
2557{
2558 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2559}
2560
Linus Torvalds1da177e2005-04-16 15:20:36 -07002561/**
2562 * skb_linearize - convert paged skb to linear one
2563 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002564 *
2565 * If there is no free memory -ENOMEM is returned, otherwise zero
2566 * is returned and the old skb data released.
2567 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002568static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002569{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002570 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2571}
2572
2573/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00002574 * skb_has_shared_frag - can any frag be overwritten
2575 * @skb: buffer to test
2576 *
2577 * Return true if the skb has at least one frag that might be modified
2578 * by an external entity (as in vmsplice()/sendfile())
2579 */
2580static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2581{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00002582 return skb_is_nonlinear(skb) &&
2583 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00002584}
2585
2586/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07002587 * skb_linearize_cow - make sure skb is linear and writable
2588 * @skb: buffer to process
2589 *
2590 * If there is no free memory -ENOMEM is returned, otherwise zero
2591 * is returned and the old skb data released.
2592 */
2593static inline int skb_linearize_cow(struct sk_buff *skb)
2594{
2595 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2596 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002597}
2598
2599/**
2600 * skb_postpull_rcsum - update checksum for received skb after pull
2601 * @skb: buffer to update
2602 * @start: start of data before pull
2603 * @len: length of data pulled
2604 *
2605 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002606 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2607 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002608 */
2609
2610static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002611 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002612{
Patrick McHardy84fa7932006-08-29 16:44:56 -07002613 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002614 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2615}
2616
Herbert Xucbb042f2006-03-20 22:43:56 -08002617unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2618
David S. Miller7ce5a272013-12-02 17:26:05 -05002619/**
2620 * pskb_trim_rcsum - trim received skb and update checksum
2621 * @skb: buffer to trim
2622 * @len: new length
2623 *
2624 * This is exactly the same as pskb_trim except that it ensures the
2625 * checksum of received packets are still valid after the operation.
2626 */
2627
2628static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2629{
2630 if (likely(len >= skb->len))
2631 return 0;
2632 if (skb->ip_summed == CHECKSUM_COMPLETE)
2633 skb->ip_summed = CHECKSUM_NONE;
2634 return __pskb_trim(skb, len);
2635}
2636
Linus Torvalds1da177e2005-04-16 15:20:36 -07002637#define skb_queue_walk(queue, skb) \
2638 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002639 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002640 skb = skb->next)
2641
James Chapman46f89142007-04-30 00:07:31 -07002642#define skb_queue_walk_safe(queue, skb, tmp) \
2643 for (skb = (queue)->next, tmp = skb->next; \
2644 skb != (struct sk_buff *)(queue); \
2645 skb = tmp, tmp = skb->next)
2646
David S. Miller1164f522008-09-23 00:49:44 -07002647#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002648 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002649 skb = skb->next)
2650
2651#define skb_queue_walk_from_safe(queue, skb, tmp) \
2652 for (tmp = skb->next; \
2653 skb != (struct sk_buff *)(queue); \
2654 skb = tmp, tmp = skb->next)
2655
Stephen Hemminger300ce172005-10-30 13:47:34 -08002656#define skb_queue_reverse_walk(queue, skb) \
2657 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002658 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08002659 skb = skb->prev)
2660
David S. Miller686a2952011-01-20 22:47:32 -08002661#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2662 for (skb = (queue)->prev, tmp = skb->prev; \
2663 skb != (struct sk_buff *)(queue); \
2664 skb = tmp, tmp = skb->prev)
2665
2666#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2667 for (tmp = skb->prev; \
2668 skb != (struct sk_buff *)(queue); \
2669 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002670
David S. Miller21dc3302010-08-23 00:13:46 -07002671static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07002672{
2673 return skb_shinfo(skb)->frag_list != NULL;
2674}
2675
2676static inline void skb_frag_list_init(struct sk_buff *skb)
2677{
2678 skb_shinfo(skb)->frag_list = NULL;
2679}
2680
2681static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2682{
2683 frag->next = skb_shinfo(skb)->frag_list;
2684 skb_shinfo(skb)->frag_list = frag;
2685}
2686
2687#define skb_walk_frags(skb, iter) \
2688 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2689
Joe Perches7965bd42013-09-26 14:48:15 -07002690struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2691 int *peeked, int *off, int *err);
2692struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2693 int *err);
2694unsigned int datagram_poll(struct file *file, struct socket *sock,
2695 struct poll_table_struct *wait);
Al Viroc0371da2014-11-24 10:42:55 -05002696int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
2697 struct iov_iter *to, int size);
David S. Miller51f3d022014-11-05 16:46:40 -05002698static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
2699 struct msghdr *msg, int size)
2700{
Al Viroe5a4b0b2014-11-24 18:17:55 -05002701 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
David S. Miller51f3d022014-11-05 16:46:40 -05002702}
Al Viroe5a4b0b2014-11-24 18:17:55 -05002703int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
2704 struct msghdr *msg);
Al Viro3a654f92014-06-19 14:15:22 -04002705int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
2706 struct iov_iter *from, int len);
Al Viro3a654f92014-06-19 14:15:22 -04002707int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
Joe Perches7965bd42013-09-26 14:48:15 -07002708void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2709void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2710int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07002711int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2712int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2713__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2714 int len, __wsum csum);
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +02002715ssize_t skb_socket_splice(struct sock *sk,
2716 struct pipe_inode_info *pipe,
2717 struct splice_pipe_desc *spd);
2718int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
Joe Perches7965bd42013-09-26 14:48:15 -07002719 struct pipe_inode_info *pipe, unsigned int len,
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +02002720 unsigned int flags,
2721 ssize_t (*splice_cb)(struct sock *,
2722 struct pipe_inode_info *,
2723 struct splice_pipe_desc *));
Joe Perches7965bd42013-09-26 14:48:15 -07002724void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01002725unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00002726int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2727 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07002728void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2729int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2730void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Florian Westphalde960aa2014-01-26 10:58:16 +01002731unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -07002732struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Vlad Yasevich0d5501c2014-08-08 14:42:13 -04002733struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
Jiri Pirkoe2195122014-11-19 14:05:01 +01002734int skb_ensure_writable(struct sk_buff *skb, int write_len);
Jiri Pirko93515d52014-11-19 14:05:02 +01002735int skb_vlan_pop(struct sk_buff *skb);
2736int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002737
Al Viro6ce8e9c2014-04-06 21:25:44 -04002738static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
2739{
Al Viro21226ab2014-11-28 15:48:29 -05002740 return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro6ce8e9c2014-04-06 21:25:44 -04002741}
2742
Al Viro7eab8d92014-04-06 21:51:23 -04002743static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
2744{
Al Viroe5a4b0b2014-11-24 18:17:55 -05002745 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro7eab8d92014-04-06 21:51:23 -04002746}
2747
Daniel Borkmann2817a332013-10-30 11:50:51 +01002748struct skb_checksum_ops {
2749 __wsum (*update)(const void *mem, int len, __wsum wsum);
2750 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2751};
2752
2753__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2754 __wsum csum, const struct skb_checksum_ops *ops);
2755__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2756 __wsum csum);
2757
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07002758static inline void * __must_check
2759__skb_header_pointer(const struct sk_buff *skb, int offset,
2760 int len, void *data, int hlen, void *buffer)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002761{
Patrick McHardy55820ee2005-07-05 14:08:10 -07002762 if (hlen - offset >= len)
David S. Miller690e36e2014-08-23 12:13:41 -07002763 return data + offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002764
David S. Miller690e36e2014-08-23 12:13:41 -07002765 if (!skb ||
2766 skb_copy_bits(skb, offset, buffer, len) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002767 return NULL;
2768
2769 return buffer;
2770}
2771
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07002772static inline void * __must_check
2773skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
David S. Miller690e36e2014-08-23 12:13:41 -07002774{
2775 return __skb_header_pointer(skb, offset, len, skb->data,
2776 skb_headlen(skb), buffer);
2777}
2778
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01002779/**
2780 * skb_needs_linearize - check if we need to linearize a given skb
2781 * depending on the given device features.
2782 * @skb: socket buffer to check
2783 * @features: net device features
2784 *
2785 * Returns true if either:
2786 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2787 * 2. skb is fragmented and the device does not support SG.
2788 */
2789static inline bool skb_needs_linearize(struct sk_buff *skb,
2790 netdev_features_t features)
2791{
2792 return skb_is_nonlinear(skb) &&
2793 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2794 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2795}
2796
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03002797static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2798 void *to,
2799 const unsigned int len)
2800{
2801 memcpy(to, skb->data, len);
2802}
2803
2804static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2805 const int offset, void *to,
2806 const unsigned int len)
2807{
2808 memcpy(to, skb->data + offset, len);
2809}
2810
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03002811static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2812 const void *from,
2813 const unsigned int len)
2814{
2815 memcpy(skb->data, from, len);
2816}
2817
2818static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2819 const int offset,
2820 const void *from,
2821 const unsigned int len)
2822{
2823 memcpy(skb->data + offset, from, len);
2824}
2825
Joe Perches7965bd42013-09-26 14:48:15 -07002826void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002827
Patrick Ohlyac45f602009-02-12 05:03:37 +00002828static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2829{
2830 return skb->tstamp;
2831}
2832
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002833/**
2834 * skb_get_timestamp - get timestamp from a skb
2835 * @skb: skb to get stamp from
2836 * @stamp: pointer to struct timeval to store stamp in
2837 *
2838 * Timestamps are stored in the skb as offsets to a base timestamp.
2839 * This function converts the offset back to a struct timeval and stores
2840 * it in stamp.
2841 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00002842static inline void skb_get_timestamp(const struct sk_buff *skb,
2843 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002844{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002845 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002846}
2847
Patrick Ohlyac45f602009-02-12 05:03:37 +00002848static inline void skb_get_timestampns(const struct sk_buff *skb,
2849 struct timespec *stamp)
2850{
2851 *stamp = ktime_to_timespec(skb->tstamp);
2852}
2853
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002854static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002855{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002856 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002857}
2858
Stephen Hemminger164891a2007-04-23 22:26:16 -07002859static inline ktime_t net_timedelta(ktime_t t)
2860{
2861 return ktime_sub(ktime_get_real(), t);
2862}
2863
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07002864static inline ktime_t net_invalid_timestamp(void)
2865{
2866 return ktime_set(0, 0);
2867}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002868
Alexander Duyck62bccb82014-09-04 13:31:35 -04002869struct sk_buff *skb_clone_sk(struct sk_buff *skb);
2870
Richard Cochranc1f19b52010-07-17 08:49:36 +00002871#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2872
Joe Perches7965bd42013-09-26 14:48:15 -07002873void skb_clone_tx_timestamp(struct sk_buff *skb);
2874bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00002875
2876#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2877
2878static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2879{
2880}
2881
2882static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2883{
2884 return false;
2885}
2886
2887#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2888
2889/**
2890 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2891 *
Richard Cochranda92b192011-10-21 00:49:15 +00002892 * PHY drivers may accept clones of transmitted packets for
2893 * timestamping via their phy_driver.txtstamp method. These drivers
2894 * must call this function to return the skb back to the stack, with
2895 * or without a timestamp.
2896 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00002897 * @skb: clone of the the original outgoing packet
Richard Cochranda92b192011-10-21 00:49:15 +00002898 * @hwtstamps: hardware time stamps, may be NULL if not available
Richard Cochranc1f19b52010-07-17 08:49:36 +00002899 *
2900 */
2901void skb_complete_tx_timestamp(struct sk_buff *skb,
2902 struct skb_shared_hwtstamps *hwtstamps);
2903
Willem de Bruijne7fd2882014-08-04 22:11:48 -04002904void __skb_tstamp_tx(struct sk_buff *orig_skb,
2905 struct skb_shared_hwtstamps *hwtstamps,
2906 struct sock *sk, int tstype);
2907
Patrick Ohlyac45f602009-02-12 05:03:37 +00002908/**
2909 * skb_tstamp_tx - queue clone of skb with send time stamps
2910 * @orig_skb: the original outgoing packet
2911 * @hwtstamps: hardware time stamps, may be NULL if not available
2912 *
2913 * If the skb has a socket associated, then this function clones the
2914 * skb (thus sharing the actual data and optional structures), stores
2915 * the optional hardware time stamping information (if non NULL) or
2916 * generates a software time stamp (otherwise), then queues the clone
2917 * to the error queue of the socket. Errors are silently ignored.
2918 */
Joe Perches7965bd42013-09-26 14:48:15 -07002919void skb_tstamp_tx(struct sk_buff *orig_skb,
2920 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00002921
Richard Cochran4507a712010-07-17 08:48:28 +00002922static inline void sw_tx_timestamp(struct sk_buff *skb)
2923{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002924 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2925 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00002926 skb_tstamp_tx(skb, NULL);
2927}
2928
2929/**
2930 * skb_tx_timestamp() - Driver hook for transmit timestamping
2931 *
2932 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00002933 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00002934 *
David S. Miller73409f32013-12-27 13:04:33 -05002935 * Specifically, one should make absolutely sure that this function is
2936 * called before TX completion of this packet can trigger. Otherwise
2937 * the packet could potentially already be freed.
2938 *
Richard Cochran4507a712010-07-17 08:48:28 +00002939 * @skb: A socket buffer.
2940 */
2941static inline void skb_tx_timestamp(struct sk_buff *skb)
2942{
Richard Cochranc1f19b52010-07-17 08:49:36 +00002943 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002944 sw_tx_timestamp(skb);
2945}
2946
Johannes Berg6e3e9392011-11-09 10:15:42 +01002947/**
2948 * skb_complete_wifi_ack - deliver skb with wifi status
2949 *
2950 * @skb: the original outgoing packet
2951 * @acked: ack status
2952 *
2953 */
2954void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2955
Joe Perches7965bd42013-09-26 14:48:15 -07002956__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2957__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002958
Herbert Xu60476372007-04-09 11:59:39 -07002959static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2960{
Tom Herbert6edec0e2015-02-10 16:30:28 -08002961 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
2962 skb->csum_valid ||
2963 (skb->ip_summed == CHECKSUM_PARTIAL &&
2964 skb_checksum_start_offset(skb) >= 0));
Herbert Xu60476372007-04-09 11:59:39 -07002965}
2966
Herbert Xufb286bb2005-11-10 13:01:24 -08002967/**
2968 * skb_checksum_complete - Calculate checksum of an entire packet
2969 * @skb: packet to process
2970 *
2971 * This function calculates the checksum over the entire packet plus
2972 * the value of skb->csum. The latter can be used to supply the
2973 * checksum of a pseudo header as used by TCP/UDP. It returns the
2974 * checksum.
2975 *
2976 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2977 * this function can be used to verify that checksum on received
2978 * packets. In that case the function should return zero if the
2979 * checksum is correct. In particular, this function will return zero
2980 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2981 * hardware has already verified the correctness of the checksum.
2982 */
Al Viro4381ca32007-07-15 21:00:11 +01002983static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002984{
Herbert Xu60476372007-04-09 11:59:39 -07002985 return skb_csum_unnecessary(skb) ?
2986 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002987}
2988
Tom Herbert77cffe22014-08-27 21:26:46 -07002989static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
2990{
2991 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2992 if (skb->csum_level == 0)
2993 skb->ip_summed = CHECKSUM_NONE;
2994 else
2995 skb->csum_level--;
2996 }
2997}
2998
2999static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3000{
3001 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3002 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3003 skb->csum_level++;
3004 } else if (skb->ip_summed == CHECKSUM_NONE) {
3005 skb->ip_summed = CHECKSUM_UNNECESSARY;
3006 skb->csum_level = 0;
3007 }
3008}
3009
Tom Herbert5a212322014-08-31 15:12:41 -07003010static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
3011{
3012 /* Mark current checksum as bad (typically called from GRO
3013 * path). In the case that ip_summed is CHECKSUM_NONE
3014 * this must be the first checksum encountered in the packet.
3015 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
3016 * checksum after the last one validated. For UDP, a zero
3017 * checksum can not be marked as bad.
3018 */
3019
3020 if (skb->ip_summed == CHECKSUM_NONE ||
3021 skb->ip_summed == CHECKSUM_UNNECESSARY)
3022 skb->csum_bad = 1;
3023}
3024
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003025/* Check if we need to perform checksum complete validation.
3026 *
3027 * Returns true if checksum complete is needed, false otherwise
3028 * (either checksum is unnecessary or zero checksum is allowed).
3029 */
3030static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3031 bool zero_okay,
3032 __sum16 check)
3033{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003034 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3035 skb->csum_valid = 1;
Tom Herbert77cffe22014-08-27 21:26:46 -07003036 __skb_decr_checksum_unnecessary(skb);
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003037 return false;
3038 }
3039
3040 return true;
3041}
3042
3043/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3044 * in checksum_init.
3045 */
3046#define CHECKSUM_BREAK 76
3047
Tom Herbert4e18b9a2015-04-20 14:10:04 -07003048/* Unset checksum-complete
3049 *
3050 * Unset checksum complete can be done when packet is being modified
3051 * (uncompressed for instance) and checksum-complete value is
3052 * invalidated.
3053 */
3054static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3055{
3056 if (skb->ip_summed == CHECKSUM_COMPLETE)
3057 skb->ip_summed = CHECKSUM_NONE;
3058}
3059
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003060/* Validate (init) checksum based on checksum complete.
3061 *
3062 * Return values:
3063 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3064 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3065 * checksum is stored in skb->csum for use in __skb_checksum_complete
3066 * non-zero: value of invalid checksum
3067 *
3068 */
3069static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3070 bool complete,
3071 __wsum psum)
3072{
3073 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3074 if (!csum_fold(csum_add(psum, skb->csum))) {
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003075 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003076 return 0;
3077 }
Tom Herbert5a212322014-08-31 15:12:41 -07003078 } else if (skb->csum_bad) {
3079 /* ip_summed == CHECKSUM_NONE in this case */
Eric Dumazetc91d4602015-05-15 05:48:07 -07003080 return (__force __sum16)1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003081 }
3082
3083 skb->csum = psum;
3084
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003085 if (complete || skb->len <= CHECKSUM_BREAK) {
3086 __sum16 csum;
3087
3088 csum = __skb_checksum_complete(skb);
3089 skb->csum_valid = !csum;
3090 return csum;
3091 }
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003092
3093 return 0;
3094}
3095
3096static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3097{
3098 return 0;
3099}
3100
3101/* Perform checksum validate (init). Note that this is a macro since we only
3102 * want to calculate the pseudo header which is an input function if necessary.
3103 * First we try to validate without any computation (checksum unnecessary) and
3104 * then calculate based on checksum complete calling the function to compute
3105 * pseudo header.
3106 *
3107 * Return values:
3108 * 0: checksum is validated or try to in skb_checksum_complete
3109 * non-zero: value of invalid checksum
3110 */
3111#define __skb_checksum_validate(skb, proto, complete, \
3112 zero_okay, check, compute_pseudo) \
3113({ \
3114 __sum16 __ret = 0; \
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003115 skb->csum_valid = 0; \
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003116 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3117 __ret = __skb_checksum_validate_complete(skb, \
3118 complete, compute_pseudo(skb, proto)); \
3119 __ret; \
3120})
3121
3122#define skb_checksum_init(skb, proto, compute_pseudo) \
3123 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3124
3125#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3126 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3127
3128#define skb_checksum_validate(skb, proto, compute_pseudo) \
3129 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3130
3131#define skb_checksum_validate_zero_check(skb, proto, check, \
3132 compute_pseudo) \
Sabrina Dubroca096a4cf2015-02-06 18:54:19 +01003133 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003134
3135#define skb_checksum_simple_validate(skb) \
3136 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3137
Tom Herbertd96535a2014-08-31 15:12:42 -07003138static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3139{
3140 return (skb->ip_summed == CHECKSUM_NONE &&
3141 skb->csum_valid && !skb->csum_bad);
3142}
3143
3144static inline void __skb_checksum_convert(struct sk_buff *skb,
3145 __sum16 check, __wsum pseudo)
3146{
3147 skb->csum = ~pseudo;
3148 skb->ip_summed = CHECKSUM_COMPLETE;
3149}
3150
3151#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3152do { \
3153 if (__skb_checksum_convert_check(skb)) \
3154 __skb_checksum_convert(skb, check, \
3155 compute_pseudo(skb, proto)); \
3156} while (0)
3157
Tom Herbert15e23962015-02-10 16:30:31 -08003158static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3159 u16 start, u16 offset)
3160{
3161 skb->ip_summed = CHECKSUM_PARTIAL;
3162 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3163 skb->csum_offset = offset - start;
3164}
3165
Tom Herbertdcdc8992015-02-02 16:07:34 -08003166/* Update skbuf and packet to reflect the remote checksum offload operation.
3167 * When called, ptr indicates the starting point for skb->csum when
3168 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3169 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3170 */
3171static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
Tom Herbert15e23962015-02-10 16:30:31 -08003172 int start, int offset, bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08003173{
3174 __wsum delta;
3175
Tom Herbert15e23962015-02-10 16:30:31 -08003176 if (!nopartial) {
3177 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3178 return;
3179 }
3180
Tom Herbertdcdc8992015-02-02 16:07:34 -08003181 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3182 __skb_checksum_complete(skb);
3183 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3184 }
3185
3186 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3187
3188 /* Adjust skb->csum since we changed the packet */
3189 skb->csum = csum_add(skb->csum, delta);
3190}
3191
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003192#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07003193void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003194static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3195{
3196 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07003197 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003198}
3199static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3200{
3201 if (nfct)
3202 atomic_inc(&nfct->use);
3203}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003204#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003205#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003206static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3207{
3208 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3209 kfree(nf_bridge);
3210}
3211static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3212{
3213 if (nf_bridge)
3214 atomic_inc(&nf_bridge->use);
3215}
3216#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08003217static inline void nf_reset(struct sk_buff *skb)
3218{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003219#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08003220 nf_conntrack_put(skb->nfct);
3221 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003222#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003223#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Patrick McHardya193a4a2006-03-20 19:23:05 -08003224 nf_bridge_put(skb->nf_bridge);
3225 skb->nf_bridge = NULL;
3226#endif
3227}
3228
Patrick McHardy124dff02013-04-05 20:42:05 +02003229static inline void nf_reset_trace(struct sk_buff *skb)
3230{
Florian Westphal478b3602014-02-15 23:48:45 +01003231#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00003232 skb->nf_trace = 0;
3233#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003234}
3235
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003236/* Note: This doesn't put any conntrack and bridge info in dst. */
Eric Dumazetb1937222014-09-28 22:18:47 -07003237static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3238 bool copy)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003239{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003240#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003241 dst->nfct = src->nfct;
3242 nf_conntrack_get(src->nfct);
Eric Dumazetb1937222014-09-28 22:18:47 -07003243 if (copy)
3244 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003245#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003246#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003247 dst->nf_bridge = src->nf_bridge;
3248 nf_bridge_get(src->nf_bridge);
3249#endif
Florian Westphal478b3602014-02-15 23:48:45 +01003250#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Eric Dumazetb1937222014-09-28 22:18:47 -07003251 if (copy)
3252 dst->nf_trace = src->nf_trace;
Florian Westphal478b3602014-02-15 23:48:45 +01003253#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003254}
3255
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003256static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3257{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003258#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003259 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003260#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003261#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003262 nf_bridge_put(dst->nf_bridge);
3263#endif
Eric Dumazetb1937222014-09-28 22:18:47 -07003264 __nf_copy(dst, src, true);
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003265}
3266
James Morris984bc162006-06-09 00:29:17 -07003267#ifdef CONFIG_NETWORK_SECMARK
3268static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3269{
3270 to->secmark = from->secmark;
3271}
3272
3273static inline void skb_init_secmark(struct sk_buff *skb)
3274{
3275 skb->secmark = 0;
3276}
3277#else
3278static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3279{ }
3280
3281static inline void skb_init_secmark(struct sk_buff *skb)
3282{ }
3283#endif
3284
Eric W. Biederman574f7192014-04-01 12:20:24 -07003285static inline bool skb_irq_freeable(const struct sk_buff *skb)
3286{
3287 return !skb->destructor &&
3288#if IS_ENABLED(CONFIG_XFRM)
3289 !skb->sp &&
3290#endif
3291#if IS_ENABLED(CONFIG_NF_CONNTRACK)
3292 !skb->nfct &&
3293#endif
3294 !skb->_skb_refdst &&
3295 !skb_has_frag_list(skb);
3296}
3297
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003298static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3299{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003300 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003301}
3302
Stephen Hemminger92477442009-03-21 13:39:26 -07003303static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003304{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003305 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003306}
3307
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003308static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3309{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003310 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003311}
3312
David S. Millerd5a9e242009-01-27 16:22:11 -08003313static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3314{
3315 skb->queue_mapping = rx_queue + 1;
3316}
3317
Stephen Hemminger92477442009-03-21 13:39:26 -07003318static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003319{
3320 return skb->queue_mapping - 1;
3321}
3322
Stephen Hemminger92477442009-03-21 13:39:26 -07003323static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003324{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003325 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08003326}
3327
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003328static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3329{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003330#ifdef CONFIG_XFRM
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003331 return skb->sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003332#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003333 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003334#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003335}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003336
Pravin B Shelar68c33162013-02-14 14:02:41 +00003337/* Keeps track of mac header offset relative to skb->head.
3338 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3339 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07003340 * tunnel skb it points to outer mac header.
3341 * Keeps track of level of encapsulation of network headers.
3342 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00003343struct skb_gso_cb {
Eric Dumazet3347c962013-10-19 11:42:56 -07003344 int mac_offset;
3345 int encap_level;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003346 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00003347};
3348#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3349
3350static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3351{
3352 return (skb_mac_header(inner_skb) - inner_skb->head) -
3353 SKB_GSO_CB(inner_skb)->mac_offset;
3354}
3355
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00003356static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3357{
3358 int new_headroom, headroom;
3359 int ret;
3360
3361 headroom = skb_headroom(skb);
3362 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3363 if (ret)
3364 return ret;
3365
3366 new_headroom = skb_headroom(skb);
3367 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3368 return 0;
3369}
3370
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003371/* Compute the checksum for a gso segment. First compute the checksum value
3372 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3373 * then add in skb->csum (checksum from csum_start to end of packet).
3374 * skb->csum and csum_start are then updated to reflect the checksum of the
3375 * resultant packet starting from the transport header-- the resultant checksum
3376 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3377 * header.
3378 */
3379static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3380{
3381 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
Eric Dumazetc91d4602015-05-15 05:48:07 -07003382 skb_transport_offset(skb);
3383 __wsum partial;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003384
Eric Dumazetc91d4602015-05-15 05:48:07 -07003385 partial = csum_partial(skb_transport_header(skb), plen, skb->csum);
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003386 skb->csum = res;
3387 SKB_GSO_CB(skb)->csum_start -= plen;
3388
Eric Dumazetc91d4602015-05-15 05:48:07 -07003389 return csum_fold(partial);
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003390}
3391
David S. Millerbdcc0922012-03-07 20:53:36 -05003392static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07003393{
3394 return skb_shinfo(skb)->gso_size;
3395}
3396
Eric Dumazet36a8f392013-09-29 01:21:32 -07003397/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05003398static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02003399{
3400 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3401}
3402
Joe Perches7965bd42013-09-26 14:48:15 -07003403void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07003404
3405static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3406{
3407 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3408 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003409 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3410
Alexander Duyckb78462e2010-06-02 12:24:37 +00003411 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3412 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07003413 __skb_warn_lro_forwarding(skb);
3414 return true;
3415 }
3416 return false;
3417}
3418
Herbert Xu35fc92a2007-03-26 23:22:20 -07003419static inline void skb_forward_csum(struct sk_buff *skb)
3420{
3421 /* Unfortunately we don't support this one. Any brave souls? */
3422 if (skb->ip_summed == CHECKSUM_COMPLETE)
3423 skb->ip_summed = CHECKSUM_NONE;
3424}
3425
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003426/**
3427 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3428 * @skb: skb to check
3429 *
3430 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3431 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3432 * use this helper, to document places where we make this assertion.
3433 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003434static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003435{
3436#ifdef DEBUG
3437 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3438#endif
3439}
3440
Rusty Russellf35d9d82008-02-04 23:49:54 -05003441bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00003442
Paul Durranted1f50c2014-01-09 10:02:46 +00003443int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
Linus Lüssing9afd85c2015-05-02 14:01:07 +02003444struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
3445 unsigned int transport_len,
3446 __sum16(*skb_chkf)(struct sk_buff *skb));
Paul Durranted1f50c2014-01-09 10:02:46 +00003447
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00003448/**
3449 * skb_head_is_locked - Determine if the skb->head is locked down
3450 * @skb: skb to check
3451 *
3452 * The head on skbs build around a head frag can be removed if they are
3453 * not cloned. This function returns true if the skb head is locked down
3454 * due to either being allocated via kmalloc, or by being a clone with
3455 * multiple references to the head.
3456 */
3457static inline bool skb_head_is_locked(const struct sk_buff *skb)
3458{
3459 return !skb->head_frag || skb_cloned(skb);
3460}
Florian Westphalfe6cc552014-02-13 23:09:12 +01003461
3462/**
3463 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3464 *
3465 * @skb: GSO skb
3466 *
3467 * skb_gso_network_seglen is used to determine the real size of the
3468 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3469 *
3470 * The MAC/L2 header is not accounted for.
3471 */
3472static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3473{
3474 unsigned int hdr_len = skb_transport_header(skb) -
3475 skb_network_header(skb);
3476 return hdr_len + skb_gso_transport_seglen(skb);
3477}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003478#endif /* __KERNEL__ */
3479#endif /* _LINUX_SKBUFF_H */