blob: 9b88536487e667b8727414e4131188db869711d4 [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
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700509 * @mark: Generic packet mark
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000510 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b2008-07-14 22:49:06 -0700511 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000512 * @inner_protocol: Protocol (encapsulation)
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000513 * @inner_transport_header: Inner transport layer header (encapsulation)
514 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000515 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700516 * @transport_header: Transport layer header
517 * @network_header: Network layer header
518 * @mac_header: Link layer header
519 * @tail: Tail pointer
520 * @end: End pointer
521 * @head: Head of buffer
522 * @data: Data head pointer
523 * @truesize: Buffer size
524 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700525 */
526
527struct sk_buff {
Eric Dumazet363ec39232014-02-26 14:02:11 -0800528 union {
Eric Dumazet56b17422014-11-03 08:19:53 -0800529 struct {
530 /* These two members must be first. */
531 struct sk_buff *next;
532 struct sk_buff *prev;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000533
Eric Dumazet56b17422014-11-03 08:19:53 -0800534 union {
535 ktime_t tstamp;
536 struct skb_mstamp skb_mstamp;
537 };
538 };
539 struct rb_node rbnode; /* used in netem & tcp stack */
540 };
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000541 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700542 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544 /*
545 * This is the control buffer. It is free to use for every
546 * layer. Please put your private variables there. If you
547 * want to keep them across layers you have to do a skb_clone()
548 * first. This is owned by whoever has the skb queued ATM.
549 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000550 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551
Eric Dumazet7fee2262010-05-11 23:19:48 +0000552 unsigned long _skb_refdst;
Eric Dumazetb1937222014-09-28 22:18:47 -0700553 void (*destructor)(struct sk_buff *skb);
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000554#ifdef CONFIG_XFRM
555 struct sec_path *sp;
556#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700557#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
558 struct nf_conntrack *nfct;
559#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +0200560#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Eric Dumazetb1937222014-09-28 22:18:47 -0700561 struct nf_bridge_info *nf_bridge;
562#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700563 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700564 data_len;
565 __u16 mac_len,
566 hdr_len;
Eric Dumazetb1937222014-09-28 22:18:47 -0700567
568 /* Following fields are _not_ copied in __copy_skb_header()
569 * Note that queue_mapping is here mostly to fill a hole.
570 */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200571 kmemcheck_bitfield_begin(flags1);
Eric Dumazetb1937222014-09-28 22:18:47 -0700572 __u16 queue_mapping;
573 __u8 cloned:1,
Harald Welte6869c4d2005-08-09 19:24:19 -0700574 nohdr:1,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800575 fclone:2,
Herbert Xua59322b2007-12-05 01:53:40 -0800576 peeked:1,
Eric Dumazetb1937222014-09-28 22:18:47 -0700577 head_frag:1,
578 xmit_more:1;
579 /* one bit hole */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200580 kmemcheck_bitfield_end(flags1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581
Eric Dumazetb1937222014-09-28 22:18:47 -0700582 /* fields enclosed in headers_start/headers_end are copied
583 * using a single memcpy() in __copy_skb_header()
584 */
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700585 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700586 __u32 headers_start[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700587 /* public: */
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200588
589/* if you move pkt_type around you also must adapt those constants */
590#ifdef __BIG_ENDIAN_BITFIELD
591#define PKT_TYPE_MAX (7 << 5)
592#else
593#define PKT_TYPE_MAX 7
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100594#endif
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200595#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
596
597 __u8 __pkt_type_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700598 __u8 pkt_type:3;
599 __u8 pfmemalloc:1;
600 __u8 ignore_df:1;
601 __u8 nfctinfo:3;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602
Eric Dumazetb1937222014-09-28 22:18:47 -0700603 __u8 nf_trace:1;
604 __u8 ip_summed:2;
605 __u8 ooo_okay:1;
606 __u8 l4_hash:1;
607 __u8 sw_hash:1;
608 __u8 wifi_acked_valid:1;
609 __u8 wifi_acked:1;
610
611 __u8 no_fcs:1;
612 /* Indicates the inner headers are valid in the skbuff. */
613 __u8 encapsulation:1;
614 __u8 encap_hdr_csum:1;
615 __u8 csum_valid:1;
616 __u8 csum_complete_sw:1;
617 __u8 csum_level:2;
618 __u8 csum_bad:1;
619
620#ifdef CONFIG_IPV6_NDISC_NODETYPE
621 __u8 ndisc_nodetype:2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700623 __u8 ipvs_property:1;
Tom Herbert8bce6d72014-09-29 20:22:29 -0700624 __u8 inner_protocol_type:1;
Tom Herberte585f232014-11-04 09:06:54 -0800625 __u8 remcsum_offload:1;
626 /* 3 or 5 bit hole */
Alexander Duyck4031ae62012-01-27 06:22:53 +0000627
Linus Torvalds1da177e2005-04-16 15:20:36 -0700628#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700629 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700631 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200634
Eric Dumazetb1937222014-09-28 22:18:47 -0700635 union {
636 __wsum csum;
637 struct {
638 __u16 csum_start;
639 __u16 csum_offset;
640 };
641 };
642 __u32 priority;
643 int skb_iif;
644 __u32 hash;
645 __be16 vlan_proto;
646 __u16 vlan_tci;
Eric Dumazet2bd82482015-02-03 23:48:24 -0800647#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
648 union {
649 unsigned int napi_id;
650 unsigned int sender_cpu;
651 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700652#endif
James Morris984bc162006-06-09 00:29:17 -0700653#ifdef CONFIG_NETWORK_SECMARK
654 __u32 secmark;
655#endif
Neil Horman3b885782009-10-12 13:26:31 -0700656 union {
657 __u32 mark;
Eric Dumazet16fad692013-03-14 05:40:32 +0000658 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700659 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660
Tom Herbert8bce6d72014-09-29 20:22:29 -0700661 union {
662 __be16 inner_protocol;
663 __u8 inner_ipproto;
664 };
665
Simon Horman1a37e412013-05-23 21:02:51 +0000666 __u16 inner_transport_header;
667 __u16 inner_network_header;
668 __u16 inner_mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700669
670 __be16 protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000671 __u16 transport_header;
672 __u16 network_header;
673 __u16 mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700674
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700675 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700676 __u32 headers_end[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700677 /* public: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700678
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700680 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700681 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700683 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700684 unsigned int truesize;
685 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700686};
687
688#ifdef __KERNEL__
689/*
690 * Handling routines are only of interest to the kernel
691 */
692#include <linux/slab.h>
693
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694
Mel Gormanc93bdd02012-07-31 16:44:19 -0700695#define SKB_ALLOC_FCLONE 0x01
696#define SKB_ALLOC_RX 0x02
Alexander Duyckfd11a832014-12-09 19:40:49 -0800697#define SKB_ALLOC_NAPI 0x04
Mel Gormanc93bdd02012-07-31 16:44:19 -0700698
699/* Returns true if the skb was allocated from PFMEMALLOC reserves */
700static inline bool skb_pfmemalloc(const struct sk_buff *skb)
701{
702 return unlikely(skb->pfmemalloc);
703}
704
Eric Dumazet7fee2262010-05-11 23:19:48 +0000705/*
706 * skb might have a dst pointer attached, refcounted or not.
707 * _skb_refdst low order bit is set if refcount was _not_ taken
708 */
709#define SKB_DST_NOREF 1UL
710#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
711
712/**
713 * skb_dst - returns skb dst_entry
714 * @skb: buffer
715 *
716 * Returns skb dst_entry, regardless of reference taken or not.
717 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000718static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
719{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000720 /* If refdst was not refcounted, check we still are in a
721 * rcu_read_lock section
722 */
723 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
724 !rcu_read_lock_held() &&
725 !rcu_read_lock_bh_held());
726 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000727}
728
Eric Dumazet7fee2262010-05-11 23:19:48 +0000729/**
730 * skb_dst_set - sets skb dst
731 * @skb: buffer
732 * @dst: dst entry
733 *
734 * Sets skb dst, assuming a reference was taken on dst and should
735 * be released by skb_dst_drop()
736 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000737static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
738{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000739 skb->_skb_refdst = (unsigned long)dst;
740}
741
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200742/**
743 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
744 * @skb: buffer
745 * @dst: dst entry
746 *
747 * Sets skb dst, assuming a reference was not taken on dst.
748 * If dst entry is cached, we do not take reference and dst_release
749 * will be avoided by refdst_drop. If dst entry is not cached, we take
750 * reference, so that last dst_release can destroy the dst immediately.
751 */
752static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
753{
Hannes Frederic Sowadbfc4fb2014-12-06 19:19:42 +0100754 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
755 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200756}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000757
758/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300759 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000760 * @skb: buffer
761 */
762static inline bool skb_dst_is_noref(const struct sk_buff *skb)
763{
764 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000765}
766
Eric Dumazet511c3f92009-06-02 05:14:27 +0000767static inline struct rtable *skb_rtable(const struct sk_buff *skb)
768{
Eric Dumazetadf30902009-06-02 05:19:30 +0000769 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000770}
771
Joe Perches7965bd42013-09-26 14:48:15 -0700772void kfree_skb(struct sk_buff *skb);
773void kfree_skb_list(struct sk_buff *segs);
774void skb_tx_error(struct sk_buff *skb);
775void consume_skb(struct sk_buff *skb);
776void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000777extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000778
Joe Perches7965bd42013-09-26 14:48:15 -0700779void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
780bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
781 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000782
Joe Perches7965bd42013-09-26 14:48:15 -0700783struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
784 int node);
Eric Dumazet2ea2f622015-04-24 16:05:01 -0700785struct sk_buff *__build_skb(void *data, unsigned int frag_size);
Joe Perches7965bd42013-09-26 14:48:15 -0700786struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700787static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100788 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700789{
Eric Dumazet564824b2010-10-11 19:05:25 +0000790 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700791}
792
Eric Dumazet2e4e4412014-09-17 04:49:49 -0700793struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
794 unsigned long data_len,
795 int max_page_order,
796 int *errcode,
797 gfp_t gfp_mask);
798
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700799/* Layout of fast clones : [skb1][skb2][fclone_ref] */
800struct sk_buff_fclones {
801 struct sk_buff skb1;
802
803 struct sk_buff skb2;
804
805 atomic_t fclone_ref;
806};
807
808/**
809 * skb_fclone_busy - check if fclone is busy
810 * @skb: buffer
811 *
812 * Returns true is skb is a fast clone, and its clone is not freed.
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700813 * Some drivers call skb_orphan() in their ndo_start_xmit(),
814 * so we also check that this didnt happen.
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700815 */
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700816static inline bool skb_fclone_busy(const struct sock *sk,
817 const struct sk_buff *skb)
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700818{
819 const struct sk_buff_fclones *fclones;
820
821 fclones = container_of(skb, struct sk_buff_fclones, skb1);
822
823 return skb->fclone == SKB_FCLONE_ORIG &&
Eric Dumazet6ffe75e2014-12-03 17:04:39 -0800824 atomic_read(&fclones->fclone_ref) > 1 &&
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700825 fclones->skb2.sk == sk;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700826}
827
David S. Millerd179cd12005-08-17 14:57:30 -0700828static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100829 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700830{
Mel Gormanc93bdd02012-07-31 16:44:19 -0700831 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700832}
833
Joe Perches7965bd42013-09-26 14:48:15 -0700834struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
Patrick McHardy0ebd0ac2013-04-17 06:46:58 +0000835static inline struct sk_buff *alloc_skb_head(gfp_t priority)
836{
837 return __alloc_skb_head(priority, -1);
838}
839
Joe Perches7965bd42013-09-26 14:48:15 -0700840struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
841int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
842struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
843struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
Octavian Purdilabad93e92014-06-12 01:36:26 +0300844struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
845 gfp_t gfp_mask, bool fclone);
846static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
847 gfp_t gfp_mask)
848{
849 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
850}
Eric Dumazet117632e2011-12-03 21:39:53 +0000851
Joe Perches7965bd42013-09-26 14:48:15 -0700852int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
853struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
854 unsigned int headroom);
855struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
856 int newtailroom, gfp_t priority);
Fan Du25a91d82014-01-18 09:54:23 +0800857int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
858 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -0700859int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
860 int len);
861int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
862int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000863#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700864
Joe Perches7965bd42013-09-26 14:48:15 -0700865int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
866 int getfrag(void *from, char *to, int offset,
867 int len, int odd, struct sk_buff *skb),
868 void *from, int length);
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700869
Hannes Frederic Sowabe12a1f2015-05-21 16:59:58 +0200870int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
871 int offset, size_t size);
872
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800873struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700874 __u32 lower_offset;
875 __u32 upper_offset;
876 __u32 frag_idx;
877 __u32 stepped_offset;
878 struct sk_buff *root_skb;
879 struct sk_buff *cur_skb;
880 __u8 *frag_data;
881};
882
Joe Perches7965bd42013-09-26 14:48:15 -0700883void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
884 unsigned int to, struct skb_seq_state *st);
885unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
886 struct skb_seq_state *st);
887void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -0700888
Joe Perches7965bd42013-09-26 14:48:15 -0700889unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
Bojan Prtvar059a2442015-02-22 11:46:35 +0100890 unsigned int to, struct ts_config *config);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700891
Tom Herbert09323cc2013-12-15 22:16:19 -0800892/*
893 * Packet hash types specify the type of hash in skb_set_hash.
894 *
895 * Hash types refer to the protocol layer addresses which are used to
896 * construct a packet's hash. The hashes are used to differentiate or identify
897 * flows of the protocol layer for the hash type. Hash types are either
898 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
899 *
900 * Properties of hashes:
901 *
902 * 1) Two packets in different flows have different hash values
903 * 2) Two packets in the same flow should have the same hash value
904 *
905 * A hash at a higher layer is considered to be more specific. A driver should
906 * set the most specific hash possible.
907 *
908 * A driver cannot indicate a more specific hash than the layer at which a hash
909 * was computed. For instance an L3 hash cannot be set as an L4 hash.
910 *
911 * A driver may indicate a hash level which is less specific than the
912 * actual layer the hash was computed on. For instance, a hash computed
913 * at L4 may be considered an L3 hash. This should only be done if the
914 * driver can't unambiguously determine that the HW computed the hash at
915 * the higher layer. Note that the "should" in the second property above
916 * permits this.
917 */
918enum pkt_hash_types {
919 PKT_HASH_TYPE_NONE, /* Undefined type */
920 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
921 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
922 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
923};
924
925static inline void
926skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
927{
Tom Herbert61b905d2014-03-24 15:34:47 -0700928 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
Tom Herberta3b18dd2014-07-01 21:33:17 -0700929 skb->sw_hash = 0;
Tom Herbert61b905d2014-03-24 15:34:47 -0700930 skb->hash = hash;
Tom Herbert09323cc2013-12-15 22:16:19 -0800931}
932
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800933static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000934{
Tom Herberta3b18dd2014-07-01 21:33:17 -0700935 if (!skb->l4_hash && !skb->sw_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800936 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000937
Tom Herbert61b905d2014-03-24 15:34:47 -0700938 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000939}
940
Tom Herbert50fb7992015-05-01 11:30:12 -0700941__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
942
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800943static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
944{
Tom Herbert61b905d2014-03-24 15:34:47 -0700945 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800946}
947
Tom Herbert7539fad2013-12-15 22:12:18 -0800948static inline void skb_clear_hash(struct sk_buff *skb)
949{
Tom Herbert61b905d2014-03-24 15:34:47 -0700950 skb->hash = 0;
Tom Herberta3b18dd2014-07-01 21:33:17 -0700951 skb->sw_hash = 0;
Tom Herbert61b905d2014-03-24 15:34:47 -0700952 skb->l4_hash = 0;
Tom Herbert7539fad2013-12-15 22:12:18 -0800953}
954
955static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
956{
Tom Herbert61b905d2014-03-24 15:34:47 -0700957 if (!skb->l4_hash)
Tom Herbert7539fad2013-12-15 22:12:18 -0800958 skb_clear_hash(skb);
959}
960
Tom Herbert3df7a742013-12-15 22:16:29 -0800961static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
962{
Tom Herbert61b905d2014-03-24 15:34:47 -0700963 to->hash = from->hash;
Tom Herberta3b18dd2014-07-01 21:33:17 -0700964 to->sw_hash = from->sw_hash;
Tom Herbert61b905d2014-03-24 15:34:47 -0700965 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -0800966};
967
Eric Dumazetc29390c2015-03-11 18:42:02 -0700968static inline void skb_sender_cpu_clear(struct sk_buff *skb)
969{
970#ifdef CONFIG_XPS
971 skb->sender_cpu = 0;
972#endif
973}
974
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700975#ifdef NET_SKBUFF_DATA_USES_OFFSET
976static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
977{
978 return skb->head + skb->end;
979}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000980
981static inline unsigned int skb_end_offset(const struct sk_buff *skb)
982{
983 return skb->end;
984}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700985#else
986static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
987{
988 return skb->end;
989}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000990
991static inline unsigned int skb_end_offset(const struct sk_buff *skb)
992{
993 return skb->end - skb->head;
994}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700995#endif
996
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700998#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999
Patrick Ohlyac45f602009-02-12 05:03:37 +00001000static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1001{
1002 return &skb_shinfo(skb)->hwtstamps;
1003}
1004
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005/**
1006 * skb_queue_empty - check if a queue is empty
1007 * @list: queue head
1008 *
1009 * Returns true if the queue is empty, false otherwise.
1010 */
1011static inline int skb_queue_empty(const struct sk_buff_head *list)
1012{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001013 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001014}
1015
1016/**
David S. Millerfc7ebb22008-09-23 00:34:07 -07001017 * skb_queue_is_last - check if skb is the last entry in the queue
1018 * @list: queue head
1019 * @skb: buffer
1020 *
1021 * Returns true if @skb is the last buffer on the list.
1022 */
1023static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1024 const struct sk_buff *skb)
1025{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001026 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -07001027}
1028
1029/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001030 * skb_queue_is_first - check if skb is the first entry in the queue
1031 * @list: queue head
1032 * @skb: buffer
1033 *
1034 * Returns true if @skb is the first buffer on the list.
1035 */
1036static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1037 const struct sk_buff *skb)
1038{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001039 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001040}
1041
1042/**
David S. Miller249c8b42008-09-23 00:44:42 -07001043 * skb_queue_next - return the next packet in the queue
1044 * @list: queue head
1045 * @skb: current buffer
1046 *
1047 * Return the next packet in @list after @skb. It is only valid to
1048 * call this if skb_queue_is_last() evaluates to false.
1049 */
1050static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1051 const struct sk_buff *skb)
1052{
1053 /* This BUG_ON may seem severe, but if we just return then we
1054 * are going to dereference garbage.
1055 */
1056 BUG_ON(skb_queue_is_last(list, skb));
1057 return skb->next;
1058}
1059
1060/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001061 * skb_queue_prev - return the prev packet in the queue
1062 * @list: queue head
1063 * @skb: current buffer
1064 *
1065 * Return the prev packet in @list before @skb. It is only valid to
1066 * call this if skb_queue_is_first() evaluates to false.
1067 */
1068static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1069 const struct sk_buff *skb)
1070{
1071 /* This BUG_ON may seem severe, but if we just return then we
1072 * are going to dereference garbage.
1073 */
1074 BUG_ON(skb_queue_is_first(list, skb));
1075 return skb->prev;
1076}
1077
1078/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079 * skb_get - reference buffer
1080 * @skb: buffer to reference
1081 *
1082 * Makes another reference to a socket buffer and returns a pointer
1083 * to the buffer.
1084 */
1085static inline struct sk_buff *skb_get(struct sk_buff *skb)
1086{
1087 atomic_inc(&skb->users);
1088 return skb;
1089}
1090
1091/*
1092 * If users == 1, we are the only owner and are can avoid redundant
1093 * atomic change.
1094 */
1095
1096/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001097 * skb_cloned - is the buffer a clone
1098 * @skb: buffer to check
1099 *
1100 * Returns true if the buffer was generated with skb_clone() and is
1101 * one of multiple shared copies of the buffer. Cloned buffers are
1102 * shared data so must not be written to under normal circumstances.
1103 */
1104static inline int skb_cloned(const struct sk_buff *skb)
1105{
1106 return skb->cloned &&
1107 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1108}
1109
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001110static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1111{
1112 might_sleep_if(pri & __GFP_WAIT);
1113
1114 if (skb_cloned(skb))
1115 return pskb_expand_head(skb, 0, 0, pri);
1116
1117 return 0;
1118}
1119
Linus Torvalds1da177e2005-04-16 15:20:36 -07001120/**
1121 * skb_header_cloned - is the header a clone
1122 * @skb: buffer to check
1123 *
1124 * Returns true if modifying the header part of the buffer requires
1125 * the data to be copied.
1126 */
1127static inline int skb_header_cloned(const struct sk_buff *skb)
1128{
1129 int dataref;
1130
1131 if (!skb->cloned)
1132 return 0;
1133
1134 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1135 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1136 return dataref != 1;
1137}
1138
1139/**
1140 * skb_header_release - release reference to header
1141 * @skb: buffer to operate on
1142 *
1143 * Drop a reference to the header part of the buffer. This is done
1144 * by acquiring a payload reference. You must not read from the header
1145 * part of skb->data after this.
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001146 * Note : Check if you can use __skb_header_release() instead.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001147 */
1148static inline void skb_header_release(struct sk_buff *skb)
1149{
1150 BUG_ON(skb->nohdr);
1151 skb->nohdr = 1;
1152 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1153}
1154
1155/**
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001156 * __skb_header_release - release reference to header
1157 * @skb: buffer to operate on
1158 *
1159 * Variant of skb_header_release() assuming skb is private to caller.
1160 * We can avoid one atomic operation.
1161 */
1162static inline void __skb_header_release(struct sk_buff *skb)
1163{
1164 skb->nohdr = 1;
1165 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1166}
1167
1168
1169/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001170 * skb_shared - is the buffer shared
1171 * @skb: buffer to check
1172 *
1173 * Returns true if more than one person has a reference to this
1174 * buffer.
1175 */
1176static inline int skb_shared(const struct sk_buff *skb)
1177{
1178 return atomic_read(&skb->users) != 1;
1179}
1180
1181/**
1182 * skb_share_check - check if buffer is shared and if so clone it
1183 * @skb: buffer to check
1184 * @pri: priority for memory allocation
1185 *
1186 * If the buffer is shared the buffer is cloned and the old copy
1187 * drops a reference. A new clone with a single reference is returned.
1188 * If the buffer is not shared the original buffer is returned. When
1189 * being called from interrupt status or with spinlocks held pri must
1190 * be GFP_ATOMIC.
1191 *
1192 * NULL is returned on a memory allocation failure.
1193 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001194static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001195{
1196 might_sleep_if(pri & __GFP_WAIT);
1197 if (skb_shared(skb)) {
1198 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001199
1200 if (likely(nskb))
1201 consume_skb(skb);
1202 else
1203 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001204 skb = nskb;
1205 }
1206 return skb;
1207}
1208
1209/*
1210 * Copy shared buffers into a new sk_buff. We effectively do COW on
1211 * packets to handle cases where we have a local reader and forward
1212 * and a couple of other messy ones. The normal one is tcpdumping
1213 * a packet thats being forwarded.
1214 */
1215
1216/**
1217 * skb_unshare - make a copy of a shared buffer
1218 * @skb: buffer to check
1219 * @pri: priority for memory allocation
1220 *
1221 * If the socket buffer is a clone then this function creates a new
1222 * copy of the data, drops a reference count on the old copy and returns
1223 * the new copy with the reference count at 1. If the buffer is not a clone
1224 * the original buffer is returned. When called with a spinlock held or
1225 * from interrupt state @pri must be %GFP_ATOMIC
1226 *
1227 * %NULL is returned on a memory allocation failure.
1228 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001229static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001230 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001231{
1232 might_sleep_if(pri & __GFP_WAIT);
1233 if (skb_cloned(skb)) {
1234 struct sk_buff *nskb = skb_copy(skb, pri);
Alexander Aring31eff812014-10-10 23:10:47 +02001235
1236 /* Free our shared copy */
1237 if (likely(nskb))
1238 consume_skb(skb);
1239 else
1240 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001241 skb = nskb;
1242 }
1243 return skb;
1244}
1245
1246/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001247 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248 * @list_: list to peek at
1249 *
1250 * Peek an &sk_buff. Unlike most other operations you _MUST_
1251 * be careful with this one. A peek leaves the buffer on the
1252 * list and someone else may run off with it. You must hold
1253 * the appropriate locks or have a private queue to do this.
1254 *
1255 * Returns %NULL for an empty list or a pointer to the head element.
1256 * The reference count is not incremented and the reference is therefore
1257 * volatile. Use with caution.
1258 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001259static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001260{
Eric Dumazet18d07002012-04-30 16:31:46 +00001261 struct sk_buff *skb = list_->next;
1262
1263 if (skb == (struct sk_buff *)list_)
1264 skb = NULL;
1265 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001266}
1267
1268/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001269 * skb_peek_next - peek skb following the given one from a queue
1270 * @skb: skb to start from
1271 * @list_: list to peek at
1272 *
1273 * Returns %NULL when the end of the list is met or a pointer to the
1274 * next element. The reference count is not incremented and the
1275 * reference is therefore volatile. Use with caution.
1276 */
1277static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1278 const struct sk_buff_head *list_)
1279{
1280 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001281
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001282 if (next == (struct sk_buff *)list_)
1283 next = NULL;
1284 return next;
1285}
1286
1287/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001288 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289 * @list_: list to peek at
1290 *
1291 * Peek an &sk_buff. Unlike most other operations you _MUST_
1292 * be careful with this one. A peek leaves the buffer on the
1293 * list and someone else may run off with it. You must hold
1294 * the appropriate locks or have a private queue to do this.
1295 *
1296 * Returns %NULL for an empty list or a pointer to the tail element.
1297 * The reference count is not incremented and the reference is therefore
1298 * volatile. Use with caution.
1299 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001300static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001301{
Eric Dumazet18d07002012-04-30 16:31:46 +00001302 struct sk_buff *skb = list_->prev;
1303
1304 if (skb == (struct sk_buff *)list_)
1305 skb = NULL;
1306 return skb;
1307
Linus Torvalds1da177e2005-04-16 15:20:36 -07001308}
1309
1310/**
1311 * skb_queue_len - get queue length
1312 * @list_: list to measure
1313 *
1314 * Return the length of an &sk_buff queue.
1315 */
1316static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1317{
1318 return list_->qlen;
1319}
1320
David S. Miller67fed452008-09-21 22:36:24 -07001321/**
1322 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1323 * @list: queue to initialize
1324 *
1325 * This initializes only the list and queue length aspects of
1326 * an sk_buff_head object. This allows to initialize the list
1327 * aspects of an sk_buff_head without reinitializing things like
1328 * the spinlock. It can also be used for on-stack sk_buff_head
1329 * objects where the spinlock is known to not be used.
1330 */
1331static inline void __skb_queue_head_init(struct sk_buff_head *list)
1332{
1333 list->prev = list->next = (struct sk_buff *)list;
1334 list->qlen = 0;
1335}
1336
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001337/*
1338 * This function creates a split out lock class for each invocation;
1339 * this is needed for now since a whole lot of users of the skb-queue
1340 * infrastructure in drivers have different locking usage (in hardirq)
1341 * than the networking core (in softirq only). In the long run either the
1342 * network layer or drivers should need annotation to consolidate the
1343 * main types of usage into 3 classes.
1344 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345static inline void skb_queue_head_init(struct sk_buff_head *list)
1346{
1347 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001348 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001349}
1350
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001351static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1352 struct lock_class_key *class)
1353{
1354 skb_queue_head_init(list);
1355 lockdep_set_class(&list->lock, class);
1356}
1357
Linus Torvalds1da177e2005-04-16 15:20:36 -07001358/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001359 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360 *
1361 * The "__skb_xxxx()" functions are the non-atomic ones that
1362 * can only be called with interrupts disabled.
1363 */
Joe Perches7965bd42013-09-26 14:48:15 -07001364void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1365 struct sk_buff_head *list);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001366static inline void __skb_insert(struct sk_buff *newsk,
1367 struct sk_buff *prev, struct sk_buff *next,
1368 struct sk_buff_head *list)
1369{
1370 newsk->next = next;
1371 newsk->prev = prev;
1372 next->prev = prev->next = newsk;
1373 list->qlen++;
1374}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375
David S. Miller67fed452008-09-21 22:36:24 -07001376static inline void __skb_queue_splice(const struct sk_buff_head *list,
1377 struct sk_buff *prev,
1378 struct sk_buff *next)
1379{
1380 struct sk_buff *first = list->next;
1381 struct sk_buff *last = list->prev;
1382
1383 first->prev = prev;
1384 prev->next = first;
1385
1386 last->next = next;
1387 next->prev = last;
1388}
1389
1390/**
1391 * skb_queue_splice - join two skb lists, this is designed for stacks
1392 * @list: the new list to add
1393 * @head: the place to add it in the first list
1394 */
1395static inline void skb_queue_splice(const struct sk_buff_head *list,
1396 struct sk_buff_head *head)
1397{
1398 if (!skb_queue_empty(list)) {
1399 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001400 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001401 }
1402}
1403
1404/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001405 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001406 * @list: the new list to add
1407 * @head: the place to add it in the first list
1408 *
1409 * The list at @list is reinitialised
1410 */
1411static inline void skb_queue_splice_init(struct sk_buff_head *list,
1412 struct sk_buff_head *head)
1413{
1414 if (!skb_queue_empty(list)) {
1415 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001416 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001417 __skb_queue_head_init(list);
1418 }
1419}
1420
1421/**
1422 * skb_queue_splice_tail - join two skb lists, each list being a queue
1423 * @list: the new list to add
1424 * @head: the place to add it in the first list
1425 */
1426static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1427 struct sk_buff_head *head)
1428{
1429 if (!skb_queue_empty(list)) {
1430 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001431 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001432 }
1433}
1434
1435/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001436 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001437 * @list: the new list to add
1438 * @head: the place to add it in the first list
1439 *
1440 * Each of the lists is a queue.
1441 * The list at @list is reinitialised
1442 */
1443static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1444 struct sk_buff_head *head)
1445{
1446 if (!skb_queue_empty(list)) {
1447 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001448 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001449 __skb_queue_head_init(list);
1450 }
1451}
1452
Linus Torvalds1da177e2005-04-16 15:20:36 -07001453/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001454 * __skb_queue_after - queue a buffer at the list head
1455 * @list: list to use
1456 * @prev: place after this buffer
1457 * @newsk: buffer to queue
1458 *
1459 * Queue a buffer int the middle of a list. This function takes no locks
1460 * and you must therefore hold required locks before calling it.
1461 *
1462 * A buffer cannot be placed on two lists at the same time.
1463 */
1464static inline void __skb_queue_after(struct sk_buff_head *list,
1465 struct sk_buff *prev,
1466 struct sk_buff *newsk)
1467{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001468 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001469}
1470
Joe Perches7965bd42013-09-26 14:48:15 -07001471void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1472 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001473
Gerrit Renkerf5572852008-04-14 00:05:28 -07001474static inline void __skb_queue_before(struct sk_buff_head *list,
1475 struct sk_buff *next,
1476 struct sk_buff *newsk)
1477{
1478 __skb_insert(newsk, next->prev, next, list);
1479}
1480
Stephen Hemminger300ce172005-10-30 13:47:34 -08001481/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001482 * __skb_queue_head - queue a buffer at the list head
1483 * @list: list to use
1484 * @newsk: buffer to queue
1485 *
1486 * Queue a buffer at the start of a list. This function takes no locks
1487 * and you must therefore hold required locks before calling it.
1488 *
1489 * A buffer cannot be placed on two lists at the same time.
1490 */
Joe Perches7965bd42013-09-26 14:48:15 -07001491void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001492static inline void __skb_queue_head(struct sk_buff_head *list,
1493 struct sk_buff *newsk)
1494{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001495 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001496}
1497
1498/**
1499 * __skb_queue_tail - queue a buffer at the list tail
1500 * @list: list to use
1501 * @newsk: buffer to queue
1502 *
1503 * Queue a buffer at the end of a list. This function takes no locks
1504 * and you must therefore hold required locks before calling it.
1505 *
1506 * A buffer cannot be placed on two lists at the same time.
1507 */
Joe Perches7965bd42013-09-26 14:48:15 -07001508void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001509static inline void __skb_queue_tail(struct sk_buff_head *list,
1510 struct sk_buff *newsk)
1511{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001512 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001513}
1514
Linus Torvalds1da177e2005-04-16 15:20:36 -07001515/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001516 * remove sk_buff from list. _Must_ be called atomically, and with
1517 * the list known..
1518 */
Joe Perches7965bd42013-09-26 14:48:15 -07001519void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1521{
1522 struct sk_buff *next, *prev;
1523
1524 list->qlen--;
1525 next = skb->next;
1526 prev = skb->prev;
1527 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001528 next->prev = prev;
1529 prev->next = next;
1530}
1531
Gerrit Renkerf525c062008-04-14 00:04:12 -07001532/**
1533 * __skb_dequeue - remove from the head of the queue
1534 * @list: list to dequeue from
1535 *
1536 * Remove the head of the list. This function does not take any locks
1537 * so must be used with appropriate locks held only. The head item is
1538 * returned or %NULL if the list is empty.
1539 */
Joe Perches7965bd42013-09-26 14:48:15 -07001540struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Gerrit Renkerf525c062008-04-14 00:04:12 -07001541static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1542{
1543 struct sk_buff *skb = skb_peek(list);
1544 if (skb)
1545 __skb_unlink(skb, list);
1546 return skb;
1547}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001548
1549/**
1550 * __skb_dequeue_tail - remove from the tail of the queue
1551 * @list: list to dequeue from
1552 *
1553 * Remove the tail of the list. This function does not take any locks
1554 * so must be used with appropriate locks held only. The tail item is
1555 * returned or %NULL if the list is empty.
1556 */
Joe Perches7965bd42013-09-26 14:48:15 -07001557struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1559{
1560 struct sk_buff *skb = skb_peek_tail(list);
1561 if (skb)
1562 __skb_unlink(skb, list);
1563 return skb;
1564}
1565
1566
David S. Millerbdcc0922012-03-07 20:53:36 -05001567static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001568{
1569 return skb->data_len;
1570}
1571
1572static inline unsigned int skb_headlen(const struct sk_buff *skb)
1573{
1574 return skb->len - skb->data_len;
1575}
1576
1577static inline int skb_pagelen(const struct sk_buff *skb)
1578{
1579 int i, len = 0;
1580
1581 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001582 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583 return len + skb_headlen(skb);
1584}
1585
Ian Campbell131ea662011-08-19 06:25:00 +00001586/**
1587 * __skb_fill_page_desc - initialise a paged fragment in an skb
1588 * @skb: buffer containing fragment to be initialised
1589 * @i: paged fragment index to initialise
1590 * @page: the page to use for this fragment
1591 * @off: the offset to the data with @page
1592 * @size: the length of the data
1593 *
1594 * Initialises the @i'th fragment of @skb to point to &size bytes at
1595 * offset @off within @page.
1596 *
1597 * Does not take any additional reference on the fragment.
1598 */
1599static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1600 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001601{
1602 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1603
Mel Gormanc48a11c2012-07-31 16:44:23 -07001604 /*
Michal Hocko2f064f32015-08-21 14:11:51 -07001605 * Propagate page pfmemalloc to the skb if we can. The problem is
1606 * that not all callers have unique ownership of the page but rely
1607 * on page_is_pfmemalloc doing the right thing(tm).
Mel Gormanc48a11c2012-07-31 16:44:23 -07001608 */
Ian Campbella8605c62011-10-19 23:01:49 +00001609 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001610 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001611 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001612
1613 page = compound_head(page);
Michal Hocko2f064f32015-08-21 14:11:51 -07001614 if (page_is_pfmemalloc(page))
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001615 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00001616}
1617
1618/**
1619 * skb_fill_page_desc - initialise a paged fragment in an skb
1620 * @skb: buffer containing fragment to be initialised
1621 * @i: paged fragment index to initialise
1622 * @page: the page to use for this fragment
1623 * @off: the offset to the data with @page
1624 * @size: the length of the data
1625 *
1626 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01001627 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00001628 * addition updates @skb such that @i is the last fragment.
1629 *
1630 * Does not take any additional reference on the fragment.
1631 */
1632static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1633 struct page *page, int off, int size)
1634{
1635 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001636 skb_shinfo(skb)->nr_frags = i + 1;
1637}
1638
Joe Perches7965bd42013-09-26 14:48:15 -07001639void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1640 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001641
Jason Wangf8e617e2013-11-01 14:07:47 +08001642void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1643 unsigned int truesize);
1644
Linus Torvalds1da177e2005-04-16 15:20:36 -07001645#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001646#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1648
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001649#ifdef NET_SKBUFF_DATA_USES_OFFSET
1650static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1651{
1652 return skb->head + skb->tail;
1653}
1654
1655static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1656{
1657 skb->tail = skb->data - skb->head;
1658}
1659
1660static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1661{
1662 skb_reset_tail_pointer(skb);
1663 skb->tail += offset;
1664}
Simon Horman7cc46192013-05-28 20:34:29 +00001665
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001666#else /* NET_SKBUFF_DATA_USES_OFFSET */
1667static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1668{
1669 return skb->tail;
1670}
1671
1672static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1673{
1674 skb->tail = skb->data;
1675}
1676
1677static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1678{
1679 skb->tail = skb->data + offset;
1680}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001681
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001682#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1683
Linus Torvalds1da177e2005-04-16 15:20:36 -07001684/*
1685 * Add data to an sk_buff
1686 */
Mathias Krause0c7ddf32013-11-07 14:18:24 +01001687unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001688unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001689static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1690{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001691 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001692 SKB_LINEAR_ASSERT(skb);
1693 skb->tail += len;
1694 skb->len += len;
1695 return tmp;
1696}
1697
Joe Perches7965bd42013-09-26 14:48:15 -07001698unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001699static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1700{
1701 skb->data -= len;
1702 skb->len += len;
1703 return skb->data;
1704}
1705
Joe Perches7965bd42013-09-26 14:48:15 -07001706unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001707static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1708{
1709 skb->len -= len;
1710 BUG_ON(skb->len < skb->data_len);
1711 return skb->data += len;
1712}
1713
David S. Miller47d29642010-05-02 02:21:44 -07001714static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1715{
1716 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1717}
1718
Joe Perches7965bd42013-09-26 14:48:15 -07001719unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001720
1721static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1722{
1723 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001724 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001725 return NULL;
1726 skb->len -= len;
1727 return skb->data += len;
1728}
1729
1730static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1731{
1732 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1733}
1734
1735static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1736{
1737 if (likely(len <= skb_headlen(skb)))
1738 return 1;
1739 if (unlikely(len > skb->len))
1740 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001741 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001742}
1743
1744/**
1745 * skb_headroom - bytes at buffer head
1746 * @skb: buffer to check
1747 *
1748 * Return the number of bytes of free space at the head of an &sk_buff.
1749 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001750static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751{
1752 return skb->data - skb->head;
1753}
1754
1755/**
1756 * skb_tailroom - bytes at buffer end
1757 * @skb: buffer to check
1758 *
1759 * Return the number of bytes of free space at the tail of an sk_buff
1760 */
1761static inline int skb_tailroom(const struct sk_buff *skb)
1762{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001763 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001764}
1765
1766/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00001767 * skb_availroom - 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 * allocated by sk_stream_alloc()
1772 */
1773static inline int skb_availroom(const struct sk_buff *skb)
1774{
Eric Dumazet16fad692013-03-14 05:40:32 +00001775 if (skb_is_nonlinear(skb))
1776 return 0;
1777
1778 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00001779}
1780
1781/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001782 * skb_reserve - adjust headroom
1783 * @skb: buffer to alter
1784 * @len: bytes to move
1785 *
1786 * Increase the headroom of an empty &sk_buff by reducing the tail
1787 * room. This is only allowed for an empty buffer.
1788 */
David S. Miller8243126c2006-01-17 02:54:21 -08001789static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001790{
1791 skb->data += len;
1792 skb->tail += len;
1793}
1794
Tom Herbert8bce6d72014-09-29 20:22:29 -07001795#define ENCAP_TYPE_ETHER 0
1796#define ENCAP_TYPE_IPPROTO 1
1797
1798static inline void skb_set_inner_protocol(struct sk_buff *skb,
1799 __be16 protocol)
1800{
1801 skb->inner_protocol = protocol;
1802 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
1803}
1804
1805static inline void skb_set_inner_ipproto(struct sk_buff *skb,
1806 __u8 ipproto)
1807{
1808 skb->inner_ipproto = ipproto;
1809 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
1810}
1811
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001812static inline void skb_reset_inner_headers(struct sk_buff *skb)
1813{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001814 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001815 skb->inner_network_header = skb->network_header;
1816 skb->inner_transport_header = skb->transport_header;
1817}
1818
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00001819static inline void skb_reset_mac_len(struct sk_buff *skb)
1820{
1821 skb->mac_len = skb->network_header - skb->mac_header;
1822}
1823
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001824static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1825 *skb)
1826{
1827 return skb->head + skb->inner_transport_header;
1828}
1829
1830static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1831{
1832 skb->inner_transport_header = skb->data - skb->head;
1833}
1834
1835static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1836 const int offset)
1837{
1838 skb_reset_inner_transport_header(skb);
1839 skb->inner_transport_header += offset;
1840}
1841
1842static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1843{
1844 return skb->head + skb->inner_network_header;
1845}
1846
1847static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1848{
1849 skb->inner_network_header = skb->data - skb->head;
1850}
1851
1852static inline void skb_set_inner_network_header(struct sk_buff *skb,
1853 const int offset)
1854{
1855 skb_reset_inner_network_header(skb);
1856 skb->inner_network_header += offset;
1857}
1858
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001859static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1860{
1861 return skb->head + skb->inner_mac_header;
1862}
1863
1864static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1865{
1866 skb->inner_mac_header = skb->data - skb->head;
1867}
1868
1869static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1870 const int offset)
1871{
1872 skb_reset_inner_mac_header(skb);
1873 skb->inner_mac_header += offset;
1874}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001875static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1876{
Cong Wang35d04612013-05-29 15:16:05 +08001877 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001878}
1879
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001880static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1881{
1882 return skb->head + skb->transport_header;
1883}
1884
1885static inline void skb_reset_transport_header(struct sk_buff *skb)
1886{
1887 skb->transport_header = skb->data - skb->head;
1888}
1889
1890static inline void skb_set_transport_header(struct sk_buff *skb,
1891 const int offset)
1892{
1893 skb_reset_transport_header(skb);
1894 skb->transport_header += offset;
1895}
1896
1897static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1898{
1899 return skb->head + skb->network_header;
1900}
1901
1902static inline void skb_reset_network_header(struct sk_buff *skb)
1903{
1904 skb->network_header = skb->data - skb->head;
1905}
1906
1907static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1908{
1909 skb_reset_network_header(skb);
1910 skb->network_header += offset;
1911}
1912
1913static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1914{
1915 return skb->head + skb->mac_header;
1916}
1917
1918static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1919{
Cong Wang35d04612013-05-29 15:16:05 +08001920 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001921}
1922
1923static inline void skb_reset_mac_header(struct sk_buff *skb)
1924{
1925 skb->mac_header = skb->data - skb->head;
1926}
1927
1928static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1929{
1930 skb_reset_mac_header(skb);
1931 skb->mac_header += offset;
1932}
1933
Timo Teräs0e3da5b2013-12-16 11:02:09 +02001934static inline void skb_pop_mac_header(struct sk_buff *skb)
1935{
1936 skb->mac_header = skb->network_header;
1937}
1938
Ying Xuefbbdb8f2013-03-27 16:46:06 +00001939static inline void skb_probe_transport_header(struct sk_buff *skb,
1940 const int offset_hint)
1941{
1942 struct flow_keys keys;
1943
1944 if (skb_transport_header_was_set(skb))
1945 return;
Jiri Pirko06635a32015-05-12 14:56:16 +02001946 else if (skb_flow_dissect_flow_keys(skb, &keys))
Tom Herbert42aecaa2015-06-04 09:16:39 -07001947 skb_set_transport_header(skb, keys.control.thoff);
Ying Xuefbbdb8f2013-03-27 16:46:06 +00001948 else
1949 skb_set_transport_header(skb, offset_hint);
1950}
1951
Eric Dumazet03606892012-02-23 10:55:02 +00001952static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1953{
1954 if (skb_mac_header_was_set(skb)) {
1955 const unsigned char *old_mac = skb_mac_header(skb);
1956
1957 skb_set_mac_header(skb, -skb->mac_len);
1958 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1959 }
1960}
1961
Michał Mirosław04fb4512010-12-14 15:24:08 +00001962static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1963{
1964 return skb->csum_start - skb_headroom(skb);
1965}
1966
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001967static inline int skb_transport_offset(const struct sk_buff *skb)
1968{
1969 return skb_transport_header(skb) - skb->data;
1970}
1971
1972static inline u32 skb_network_header_len(const struct sk_buff *skb)
1973{
1974 return skb->transport_header - skb->network_header;
1975}
1976
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001977static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1978{
1979 return skb->inner_transport_header - skb->inner_network_header;
1980}
1981
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001982static inline int skb_network_offset(const struct sk_buff *skb)
1983{
1984 return skb_network_header(skb) - skb->data;
1985}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001986
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001987static inline int skb_inner_network_offset(const struct sk_buff *skb)
1988{
1989 return skb_inner_network_header(skb) - skb->data;
1990}
1991
Changli Gaof9599ce2010-08-04 04:43:44 +00001992static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1993{
1994 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1995}
1996
Linus Torvalds1da177e2005-04-16 15:20:36 -07001997/*
1998 * CPUs often take a performance hit when accessing unaligned memory
1999 * locations. The actual performance hit varies, it can be small if the
2000 * hardware handles it or large if we have to take an exception and fix it
2001 * in software.
2002 *
2003 * Since an ethernet header is 14 bytes network drivers often end up with
2004 * the IP header at an unaligned offset. The IP header can be aligned by
2005 * shifting the start of the packet by 2 bytes. Drivers should do this
2006 * with:
2007 *
Tobias Klauser8660c122009-07-13 22:48:16 +00002008 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009 *
2010 * The downside to this alignment of the IP header is that the DMA is now
2011 * unaligned. On some architectures the cost of an unaligned DMA is high
2012 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00002013 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002014 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2015 * to be overridden.
2016 */
2017#ifndef NET_IP_ALIGN
2018#define NET_IP_ALIGN 2
2019#endif
2020
Anton Blanchard025be812006-03-31 02:27:06 -08002021/*
2022 * The networking layer reserves some headroom in skb data (via
2023 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2024 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08002025 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08002026 *
2027 * Unfortunately this headroom changes the DMA alignment of the resulting
2028 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2029 * on some architectures. An architecture can override this value,
2030 * perhaps setting it to a cacheline in size (since that will maintain
2031 * cacheline alignment of the DMA). It must be a power of 2.
2032 *
David S. Millerd6301d32009-02-08 19:24:13 -08002033 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08002034 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07002035 *
2036 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2037 * to reduce average number of cache lines per packet.
2038 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07002039 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08002040 */
2041#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07002042#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08002043#endif
2044
Joe Perches7965bd42013-09-26 14:48:15 -07002045int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002046
2047static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2048{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00002049 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07002050 WARN_ON(1);
2051 return;
2052 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002053 skb->len = len;
2054 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002055}
2056
Joe Perches7965bd42013-09-26 14:48:15 -07002057void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002058
2059static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2060{
Herbert Xu3cc0e872006-06-09 16:13:38 -07002061 if (skb->data_len)
2062 return ___pskb_trim(skb, len);
2063 __skb_trim(skb, len);
2064 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002065}
2066
2067static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2068{
2069 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2070}
2071
2072/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07002073 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2074 * @skb: buffer to alter
2075 * @len: new length
2076 *
2077 * This is identical to pskb_trim except that the caller knows that
2078 * the skb is not cloned so we should never get an error due to out-
2079 * of-memory.
2080 */
2081static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2082{
2083 int err = pskb_trim(skb, len);
2084 BUG_ON(err);
2085}
2086
2087/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002088 * skb_orphan - orphan a buffer
2089 * @skb: buffer to orphan
2090 *
2091 * If a buffer currently has an owner then we call the owner's
2092 * destructor function and make the @skb unowned. The buffer continues
2093 * to exist but is no longer charged to its former owner.
2094 */
2095static inline void skb_orphan(struct sk_buff *skb)
2096{
Eric Dumazetc34a7612013-07-30 16:11:15 -07002097 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002098 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002099 skb->destructor = NULL;
2100 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07002101 } else {
2102 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002103 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002104}
2105
2106/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002107 * skb_orphan_frags - orphan the frags contained in a buffer
2108 * @skb: buffer to orphan frags from
2109 * @gfp_mask: allocation mask for replacement pages
2110 *
2111 * For each frag in the SKB which needs a destructor (i.e. has an
2112 * owner) create a copy of that frag and release the original
2113 * page by calling the destructor.
2114 */
2115static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2116{
2117 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2118 return 0;
2119 return skb_copy_ubufs(skb, gfp_mask);
2120}
2121
2122/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002123 * __skb_queue_purge - empty a list
2124 * @list: list to empty
2125 *
2126 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2127 * the list and one reference dropped. This function does not take the
2128 * list lock and the caller must hold the relevant locks to use it.
2129 */
Joe Perches7965bd42013-09-26 14:48:15 -07002130void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002131static inline void __skb_queue_purge(struct sk_buff_head *list)
2132{
2133 struct sk_buff *skb;
2134 while ((skb = __skb_dequeue(list)) != NULL)
2135 kfree_skb(skb);
2136}
2137
Joe Perches7965bd42013-09-26 14:48:15 -07002138void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002139
Joe Perches7965bd42013-09-26 14:48:15 -07002140struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2141 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07002142
2143/**
2144 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2145 * @dev: network device to receive on
2146 * @length: length to allocate
2147 *
2148 * Allocate a new &sk_buff and assign it a usage count of one. The
2149 * buffer has unspecified headroom built in. Users should allocate
2150 * the headroom they think they need without accounting for the
2151 * built in space. The built in space is used for optimisations.
2152 *
2153 * %NULL is returned if there is no free memory. Although this function
2154 * allocates memory it can be called from an interrupt.
2155 */
2156static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002157 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002158{
2159 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2160}
2161
Eric Dumazet6f532612012-05-18 05:12:12 +00002162/* legacy helper around __netdev_alloc_skb() */
2163static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2164 gfp_t gfp_mask)
2165{
2166 return __netdev_alloc_skb(NULL, length, gfp_mask);
2167}
2168
2169/* legacy helper around netdev_alloc_skb() */
2170static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2171{
2172 return netdev_alloc_skb(NULL, length);
2173}
2174
2175
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002176static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2177 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002178{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002179 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002180
2181 if (NET_IP_ALIGN && skb)
2182 skb_reserve(skb, NET_IP_ALIGN);
2183 return skb;
2184}
2185
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002186static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2187 unsigned int length)
2188{
2189 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2190}
2191
Alexander Duyck181edb22015-05-06 21:12:03 -07002192static inline void skb_free_frag(void *addr)
2193{
2194 __free_page_frag(addr);
2195}
2196
Alexander Duyckffde7322014-12-09 19:40:42 -08002197void *napi_alloc_frag(unsigned int fragsz);
Alexander Duyckfd11a832014-12-09 19:40:49 -08002198struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2199 unsigned int length, gfp_t gfp_mask);
2200static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2201 unsigned int length)
2202{
2203 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2204}
Alexander Duyckffde7322014-12-09 19:40:42 -08002205
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002206/**
Alexander Duyck71dfda52014-11-11 09:26:34 -08002207 * __dev_alloc_pages - allocate page for network Rx
2208 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2209 * @order: size of the allocation
2210 *
2211 * Allocate a new page.
2212 *
2213 * %NULL is returned if there is no free memory.
2214*/
2215static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2216 unsigned int order)
2217{
2218 /* This piece of code contains several assumptions.
2219 * 1. This is for device Rx, therefor a cold page is preferred.
2220 * 2. The expectation is the user wants a compound page.
2221 * 3. If requesting a order 0 page it will not be compound
2222 * due to the check to see if order has a value in prep_new_page
2223 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2224 * code in gfp_to_alloc_flags that should be enforcing this.
2225 */
2226 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2227
2228 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2229}
2230
2231static inline struct page *dev_alloc_pages(unsigned int order)
2232{
2233 return __dev_alloc_pages(GFP_ATOMIC, order);
2234}
2235
2236/**
2237 * __dev_alloc_page - allocate a page for network Rx
2238 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2239 *
2240 * Allocate a new page.
2241 *
2242 * %NULL is returned if there is no free memory.
2243 */
2244static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2245{
2246 return __dev_alloc_pages(gfp_mask, 0);
2247}
2248
2249static inline struct page *dev_alloc_page(void)
2250{
2251 return __dev_alloc_page(GFP_ATOMIC);
2252}
2253
2254/**
Mel Gorman06140022012-07-31 16:44:24 -07002255 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2256 * @page: The page that was allocated from skb_alloc_page
2257 * @skb: The skb that may need pfmemalloc set
2258 */
2259static inline void skb_propagate_pfmemalloc(struct page *page,
2260 struct sk_buff *skb)
2261{
Michal Hocko2f064f32015-08-21 14:11:51 -07002262 if (page_is_pfmemalloc(page))
Mel Gorman06140022012-07-31 16:44:24 -07002263 skb->pfmemalloc = true;
2264}
2265
Eric Dumazet564824b2010-10-11 19:05:25 +00002266/**
Masanari Iidae2278672014-02-18 22:54:36 +09002267 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002268 * @frag: the paged fragment
2269 *
2270 * Returns the &struct page associated with @frag.
2271 */
2272static inline struct page *skb_frag_page(const skb_frag_t *frag)
2273{
Ian Campbella8605c62011-10-19 23:01:49 +00002274 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00002275}
2276
2277/**
2278 * __skb_frag_ref - take an addition reference on a paged fragment.
2279 * @frag: the paged fragment
2280 *
2281 * Takes an additional reference on the paged fragment @frag.
2282 */
2283static inline void __skb_frag_ref(skb_frag_t *frag)
2284{
2285 get_page(skb_frag_page(frag));
2286}
2287
2288/**
2289 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2290 * @skb: the buffer
2291 * @f: the fragment offset.
2292 *
2293 * Takes an additional reference on the @f'th paged fragment of @skb.
2294 */
2295static inline void skb_frag_ref(struct sk_buff *skb, int f)
2296{
2297 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2298}
2299
2300/**
2301 * __skb_frag_unref - release a reference on a paged fragment.
2302 * @frag: the paged fragment
2303 *
2304 * Releases a reference on the paged fragment @frag.
2305 */
2306static inline void __skb_frag_unref(skb_frag_t *frag)
2307{
2308 put_page(skb_frag_page(frag));
2309}
2310
2311/**
2312 * skb_frag_unref - release a reference on a paged fragment of an skb.
2313 * @skb: the buffer
2314 * @f: the fragment offset
2315 *
2316 * Releases a reference on the @f'th paged fragment of @skb.
2317 */
2318static inline void skb_frag_unref(struct sk_buff *skb, int f)
2319{
2320 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2321}
2322
2323/**
2324 * skb_frag_address - gets the address of the data contained in a paged fragment
2325 * @frag: the paged fragment buffer
2326 *
2327 * Returns the address of the data within @frag. The page must already
2328 * be mapped.
2329 */
2330static inline void *skb_frag_address(const skb_frag_t *frag)
2331{
2332 return page_address(skb_frag_page(frag)) + frag->page_offset;
2333}
2334
2335/**
2336 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2337 * @frag: the paged fragment buffer
2338 *
2339 * Returns the address of the data within @frag. Checks that the page
2340 * is mapped and returns %NULL otherwise.
2341 */
2342static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2343{
2344 void *ptr = page_address(skb_frag_page(frag));
2345 if (unlikely(!ptr))
2346 return NULL;
2347
2348 return ptr + frag->page_offset;
2349}
2350
2351/**
2352 * __skb_frag_set_page - sets the page contained in a paged fragment
2353 * @frag: the paged fragment
2354 * @page: the page to set
2355 *
2356 * Sets the fragment @frag to contain @page.
2357 */
2358static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2359{
Ian Campbella8605c62011-10-19 23:01:49 +00002360 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002361}
2362
2363/**
2364 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2365 * @skb: the buffer
2366 * @f: the fragment offset
2367 * @page: the page to set
2368 *
2369 * Sets the @f'th fragment of @skb to contain @page.
2370 */
2371static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2372 struct page *page)
2373{
2374 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2375}
2376
Eric Dumazet400dfd32013-10-17 16:27:07 -07002377bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2378
Ian Campbell131ea662011-08-19 06:25:00 +00002379/**
2380 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002381 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002382 * @frag: the paged fragment to map
2383 * @offset: the offset within the fragment (starting at the
2384 * fragment's own offset)
2385 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002386 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00002387 *
2388 * Maps the page associated with @frag to @device.
2389 */
2390static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2391 const skb_frag_t *frag,
2392 size_t offset, size_t size,
2393 enum dma_data_direction dir)
2394{
2395 return dma_map_page(dev, skb_frag_page(frag),
2396 frag->page_offset + offset, size, dir);
2397}
2398
Eric Dumazet117632e2011-12-03 21:39:53 +00002399static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2400 gfp_t gfp_mask)
2401{
2402 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2403}
2404
Octavian Purdilabad93e92014-06-12 01:36:26 +03002405
2406static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2407 gfp_t gfp_mask)
2408{
2409 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2410}
2411
2412
Ian Campbell131ea662011-08-19 06:25:00 +00002413/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002414 * skb_clone_writable - is the header of a clone writable
2415 * @skb: buffer to check
2416 * @len: length up to which to write
2417 *
2418 * Returns true if modifying the header part of the cloned buffer
2419 * does not requires the data to be copied.
2420 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002421static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002422{
2423 return !skb_header_cloned(skb) &&
2424 skb_headroom(skb) + len <= skb->hdr_len;
2425}
2426
Herbert Xud9cc2042007-09-16 16:21:16 -07002427static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2428 int cloned)
2429{
2430 int delta = 0;
2431
Herbert Xud9cc2042007-09-16 16:21:16 -07002432 if (headroom > skb_headroom(skb))
2433 delta = headroom - skb_headroom(skb);
2434
2435 if (delta || cloned)
2436 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2437 GFP_ATOMIC);
2438 return 0;
2439}
2440
Patrick McHardy334a8132007-06-25 04:35:20 -07002441/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002442 * skb_cow - copy header of skb when it is required
2443 * @skb: buffer to cow
2444 * @headroom: needed headroom
2445 *
2446 * If the skb passed lacks sufficient headroom or its data part
2447 * is shared, data is reallocated. If reallocation fails, an error
2448 * is returned and original skb is not changed.
2449 *
2450 * The result is skb with writable area skb->head...skb->tail
2451 * and at least @headroom of space at head.
2452 */
2453static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2454{
Herbert Xud9cc2042007-09-16 16:21:16 -07002455 return __skb_cow(skb, headroom, skb_cloned(skb));
2456}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002457
Herbert Xud9cc2042007-09-16 16:21:16 -07002458/**
2459 * skb_cow_head - skb_cow but only making the head writable
2460 * @skb: buffer to cow
2461 * @headroom: needed headroom
2462 *
2463 * This function is identical to skb_cow except that we replace the
2464 * skb_cloned check by skb_header_cloned. It should be used when
2465 * you only need to push on some header and do not need to modify
2466 * the data.
2467 */
2468static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2469{
2470 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002471}
2472
2473/**
2474 * skb_padto - pad an skbuff up to a minimal size
2475 * @skb: buffer to pad
2476 * @len: minimal length
2477 *
2478 * Pads up a buffer to ensure the trailing bytes exist and are
2479 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002480 * is untouched. Otherwise it is extended. Returns zero on
2481 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002482 */
Herbert Xu5b057c62006-06-23 02:06:41 -07002483static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002484{
2485 unsigned int size = skb->len;
2486 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002487 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002488 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002489}
2490
Alexander Duyck9c0c1122014-12-03 08:17:33 -08002491/**
2492 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2493 * @skb: buffer to pad
2494 * @len: minimal length
2495 *
2496 * Pads up a buffer to ensure the trailing bytes exist and are
2497 * blanked. If the buffer already contains sufficient data it
2498 * is untouched. Otherwise it is extended. Returns zero on
2499 * success. The skb is freed on error.
2500 */
2501static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2502{
2503 unsigned int size = skb->len;
2504
2505 if (unlikely(size < len)) {
2506 len -= size;
2507 if (skb_pad(skb, len))
2508 return -ENOMEM;
2509 __skb_put(skb, len);
2510 }
2511 return 0;
2512}
2513
Linus Torvalds1da177e2005-04-16 15:20:36 -07002514static inline int skb_add_data(struct sk_buff *skb,
Al Viroaf2b0402014-11-27 21:44:24 -05002515 struct iov_iter *from, int copy)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002516{
2517 const int off = skb->len;
2518
2519 if (skb->ip_summed == CHECKSUM_NONE) {
Al Viroaf2b0402014-11-27 21:44:24 -05002520 __wsum csum = 0;
2521 if (csum_and_copy_from_iter(skb_put(skb, copy), copy,
2522 &csum, from) == copy) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002523 skb->csum = csum_block_add(skb->csum, csum, off);
2524 return 0;
2525 }
Al Viroaf2b0402014-11-27 21:44:24 -05002526 } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002527 return 0;
2528
2529 __skb_trim(skb, off);
2530 return -EFAULT;
2531}
2532
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002533static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2534 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002535{
2536 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002537 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002538
Ian Campbellea2ab692011-08-22 23:44:58 +00002539 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002540 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002541 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002542 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002543}
2544
Herbert Xu364c6ba2006-06-09 16:10:40 -07002545static inline int __skb_linearize(struct sk_buff *skb)
2546{
2547 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2548}
2549
Linus Torvalds1da177e2005-04-16 15:20:36 -07002550/**
2551 * skb_linearize - convert paged skb to linear one
2552 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002553 *
2554 * If there is no free memory -ENOMEM is returned, otherwise zero
2555 * is returned and the old skb data released.
2556 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002557static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002558{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002559 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2560}
2561
2562/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00002563 * skb_has_shared_frag - can any frag be overwritten
2564 * @skb: buffer to test
2565 *
2566 * Return true if the skb has at least one frag that might be modified
2567 * by an external entity (as in vmsplice()/sendfile())
2568 */
2569static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2570{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00002571 return skb_is_nonlinear(skb) &&
2572 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00002573}
2574
2575/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07002576 * skb_linearize_cow - make sure skb is linear and writable
2577 * @skb: buffer to process
2578 *
2579 * If there is no free memory -ENOMEM is returned, otherwise zero
2580 * is returned and the old skb data released.
2581 */
2582static inline int skb_linearize_cow(struct sk_buff *skb)
2583{
2584 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2585 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002586}
2587
2588/**
2589 * skb_postpull_rcsum - update checksum for received skb after pull
2590 * @skb: buffer to update
2591 * @start: start of data before pull
2592 * @len: length of data pulled
2593 *
2594 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002595 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2596 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002597 */
2598
2599static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002600 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002601{
Patrick McHardy84fa7932006-08-29 16:44:56 -07002602 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002603 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2604}
2605
Herbert Xucbb042f2006-03-20 22:43:56 -08002606unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2607
David S. Miller7ce5a272013-12-02 17:26:05 -05002608/**
2609 * pskb_trim_rcsum - trim received skb and update checksum
2610 * @skb: buffer to trim
2611 * @len: new length
2612 *
2613 * This is exactly the same as pskb_trim except that it ensures the
2614 * checksum of received packets are still valid after the operation.
2615 */
2616
2617static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2618{
2619 if (likely(len >= skb->len))
2620 return 0;
2621 if (skb->ip_summed == CHECKSUM_COMPLETE)
2622 skb->ip_summed = CHECKSUM_NONE;
2623 return __pskb_trim(skb, len);
2624}
2625
Linus Torvalds1da177e2005-04-16 15:20:36 -07002626#define skb_queue_walk(queue, skb) \
2627 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002628 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002629 skb = skb->next)
2630
James Chapman46f89142007-04-30 00:07:31 -07002631#define skb_queue_walk_safe(queue, skb, tmp) \
2632 for (skb = (queue)->next, tmp = skb->next; \
2633 skb != (struct sk_buff *)(queue); \
2634 skb = tmp, tmp = skb->next)
2635
David S. Miller1164f522008-09-23 00:49:44 -07002636#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002637 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002638 skb = skb->next)
2639
2640#define skb_queue_walk_from_safe(queue, skb, tmp) \
2641 for (tmp = skb->next; \
2642 skb != (struct sk_buff *)(queue); \
2643 skb = tmp, tmp = skb->next)
2644
Stephen Hemminger300ce172005-10-30 13:47:34 -08002645#define skb_queue_reverse_walk(queue, skb) \
2646 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002647 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08002648 skb = skb->prev)
2649
David S. Miller686a2952011-01-20 22:47:32 -08002650#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2651 for (skb = (queue)->prev, tmp = skb->prev; \
2652 skb != (struct sk_buff *)(queue); \
2653 skb = tmp, tmp = skb->prev)
2654
2655#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2656 for (tmp = skb->prev; \
2657 skb != (struct sk_buff *)(queue); \
2658 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002659
David S. Miller21dc3302010-08-23 00:13:46 -07002660static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07002661{
2662 return skb_shinfo(skb)->frag_list != NULL;
2663}
2664
2665static inline void skb_frag_list_init(struct sk_buff *skb)
2666{
2667 skb_shinfo(skb)->frag_list = NULL;
2668}
2669
2670static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2671{
2672 frag->next = skb_shinfo(skb)->frag_list;
2673 skb_shinfo(skb)->frag_list = frag;
2674}
2675
2676#define skb_walk_frags(skb, iter) \
2677 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2678
Joe Perches7965bd42013-09-26 14:48:15 -07002679struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2680 int *peeked, int *off, int *err);
2681struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2682 int *err);
2683unsigned int datagram_poll(struct file *file, struct socket *sock,
2684 struct poll_table_struct *wait);
Al Viroc0371da2014-11-24 10:42:55 -05002685int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
2686 struct iov_iter *to, int size);
David S. Miller51f3d022014-11-05 16:46:40 -05002687static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
2688 struct msghdr *msg, int size)
2689{
Al Viroe5a4b0b2014-11-24 18:17:55 -05002690 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
David S. Miller51f3d022014-11-05 16:46:40 -05002691}
Al Viroe5a4b0b2014-11-24 18:17:55 -05002692int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
2693 struct msghdr *msg);
Al Viro3a654f92014-06-19 14:15:22 -04002694int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
2695 struct iov_iter *from, int len);
Al Viro3a654f92014-06-19 14:15:22 -04002696int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
Joe Perches7965bd42013-09-26 14:48:15 -07002697void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2698void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2699int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07002700int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2701int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2702__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2703 int len, __wsum csum);
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +02002704ssize_t skb_socket_splice(struct sock *sk,
2705 struct pipe_inode_info *pipe,
2706 struct splice_pipe_desc *spd);
2707int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
Joe Perches7965bd42013-09-26 14:48:15 -07002708 struct pipe_inode_info *pipe, unsigned int len,
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +02002709 unsigned int flags,
2710 ssize_t (*splice_cb)(struct sock *,
2711 struct pipe_inode_info *,
2712 struct splice_pipe_desc *));
Joe Perches7965bd42013-09-26 14:48:15 -07002713void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01002714unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00002715int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2716 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07002717void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2718int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2719void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Florian Westphalde960aa2014-01-26 10:58:16 +01002720unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -07002721struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Vlad Yasevich0d5501c2014-08-08 14:42:13 -04002722struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
Jiri Pirkoe2195122014-11-19 14:05:01 +01002723int skb_ensure_writable(struct sk_buff *skb, int write_len);
Jiri Pirko93515d52014-11-19 14:05:02 +01002724int skb_vlan_pop(struct sk_buff *skb);
2725int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002726
Al Viro6ce8e9c2014-04-06 21:25:44 -04002727static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
2728{
Al Viro21226ab2014-11-28 15:48:29 -05002729 return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro6ce8e9c2014-04-06 21:25:44 -04002730}
2731
Al Viro7eab8d92014-04-06 21:51:23 -04002732static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
2733{
Al Viroe5a4b0b2014-11-24 18:17:55 -05002734 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro7eab8d92014-04-06 21:51:23 -04002735}
2736
Daniel Borkmann2817a332013-10-30 11:50:51 +01002737struct skb_checksum_ops {
2738 __wsum (*update)(const void *mem, int len, __wsum wsum);
2739 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2740};
2741
2742__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2743 __wsum csum, const struct skb_checksum_ops *ops);
2744__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2745 __wsum csum);
2746
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07002747static inline void * __must_check
2748__skb_header_pointer(const struct sk_buff *skb, int offset,
2749 int len, void *data, int hlen, void *buffer)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002750{
Patrick McHardy55820ee2005-07-05 14:08:10 -07002751 if (hlen - offset >= len)
David S. Miller690e36e2014-08-23 12:13:41 -07002752 return data + offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002753
David S. Miller690e36e2014-08-23 12:13:41 -07002754 if (!skb ||
2755 skb_copy_bits(skb, offset, buffer, len) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002756 return NULL;
2757
2758 return buffer;
2759}
2760
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07002761static inline void * __must_check
2762skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
David S. Miller690e36e2014-08-23 12:13:41 -07002763{
2764 return __skb_header_pointer(skb, offset, len, skb->data,
2765 skb_headlen(skb), buffer);
2766}
2767
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01002768/**
2769 * skb_needs_linearize - check if we need to linearize a given skb
2770 * depending on the given device features.
2771 * @skb: socket buffer to check
2772 * @features: net device features
2773 *
2774 * Returns true if either:
2775 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2776 * 2. skb is fragmented and the device does not support SG.
2777 */
2778static inline bool skb_needs_linearize(struct sk_buff *skb,
2779 netdev_features_t features)
2780{
2781 return skb_is_nonlinear(skb) &&
2782 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2783 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2784}
2785
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03002786static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2787 void *to,
2788 const unsigned int len)
2789{
2790 memcpy(to, skb->data, len);
2791}
2792
2793static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2794 const int offset, void *to,
2795 const unsigned int len)
2796{
2797 memcpy(to, skb->data + offset, len);
2798}
2799
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03002800static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2801 const void *from,
2802 const unsigned int len)
2803{
2804 memcpy(skb->data, from, len);
2805}
2806
2807static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2808 const int offset,
2809 const void *from,
2810 const unsigned int len)
2811{
2812 memcpy(skb->data + offset, from, len);
2813}
2814
Joe Perches7965bd42013-09-26 14:48:15 -07002815void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002816
Patrick Ohlyac45f602009-02-12 05:03:37 +00002817static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2818{
2819 return skb->tstamp;
2820}
2821
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002822/**
2823 * skb_get_timestamp - get timestamp from a skb
2824 * @skb: skb to get stamp from
2825 * @stamp: pointer to struct timeval to store stamp in
2826 *
2827 * Timestamps are stored in the skb as offsets to a base timestamp.
2828 * This function converts the offset back to a struct timeval and stores
2829 * it in stamp.
2830 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00002831static inline void skb_get_timestamp(const struct sk_buff *skb,
2832 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002833{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002834 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002835}
2836
Patrick Ohlyac45f602009-02-12 05:03:37 +00002837static inline void skb_get_timestampns(const struct sk_buff *skb,
2838 struct timespec *stamp)
2839{
2840 *stamp = ktime_to_timespec(skb->tstamp);
2841}
2842
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002843static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002844{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002845 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002846}
2847
Stephen Hemminger164891a2007-04-23 22:26:16 -07002848static inline ktime_t net_timedelta(ktime_t t)
2849{
2850 return ktime_sub(ktime_get_real(), t);
2851}
2852
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07002853static inline ktime_t net_invalid_timestamp(void)
2854{
2855 return ktime_set(0, 0);
2856}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002857
Alexander Duyck62bccb82014-09-04 13:31:35 -04002858struct sk_buff *skb_clone_sk(struct sk_buff *skb);
2859
Richard Cochranc1f19b52010-07-17 08:49:36 +00002860#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2861
Joe Perches7965bd42013-09-26 14:48:15 -07002862void skb_clone_tx_timestamp(struct sk_buff *skb);
2863bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00002864
2865#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2866
2867static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2868{
2869}
2870
2871static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2872{
2873 return false;
2874}
2875
2876#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2877
2878/**
2879 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2880 *
Richard Cochranda92b192011-10-21 00:49:15 +00002881 * PHY drivers may accept clones of transmitted packets for
2882 * timestamping via their phy_driver.txtstamp method. These drivers
Benjamin Poirier7a76a022015-08-07 09:32:21 -07002883 * must call this function to return the skb back to the stack with a
2884 * timestamp.
Richard Cochranda92b192011-10-21 00:49:15 +00002885 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00002886 * @skb: clone of the the original outgoing packet
Benjamin Poirier7a76a022015-08-07 09:32:21 -07002887 * @hwtstamps: hardware time stamps
Richard Cochranc1f19b52010-07-17 08:49:36 +00002888 *
2889 */
2890void skb_complete_tx_timestamp(struct sk_buff *skb,
2891 struct skb_shared_hwtstamps *hwtstamps);
2892
Willem de Bruijne7fd2882014-08-04 22:11:48 -04002893void __skb_tstamp_tx(struct sk_buff *orig_skb,
2894 struct skb_shared_hwtstamps *hwtstamps,
2895 struct sock *sk, int tstype);
2896
Patrick Ohlyac45f602009-02-12 05:03:37 +00002897/**
2898 * skb_tstamp_tx - queue clone of skb with send time stamps
2899 * @orig_skb: the original outgoing packet
2900 * @hwtstamps: hardware time stamps, may be NULL if not available
2901 *
2902 * If the skb has a socket associated, then this function clones the
2903 * skb (thus sharing the actual data and optional structures), stores
2904 * the optional hardware time stamping information (if non NULL) or
2905 * generates a software time stamp (otherwise), then queues the clone
2906 * to the error queue of the socket. Errors are silently ignored.
2907 */
Joe Perches7965bd42013-09-26 14:48:15 -07002908void skb_tstamp_tx(struct sk_buff *orig_skb,
2909 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00002910
Richard Cochran4507a712010-07-17 08:48:28 +00002911static inline void sw_tx_timestamp(struct sk_buff *skb)
2912{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002913 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2914 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00002915 skb_tstamp_tx(skb, NULL);
2916}
2917
2918/**
2919 * skb_tx_timestamp() - Driver hook for transmit timestamping
2920 *
2921 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00002922 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00002923 *
David S. Miller73409f32013-12-27 13:04:33 -05002924 * Specifically, one should make absolutely sure that this function is
2925 * called before TX completion of this packet can trigger. Otherwise
2926 * the packet could potentially already be freed.
2927 *
Richard Cochran4507a712010-07-17 08:48:28 +00002928 * @skb: A socket buffer.
2929 */
2930static inline void skb_tx_timestamp(struct sk_buff *skb)
2931{
Richard Cochranc1f19b52010-07-17 08:49:36 +00002932 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002933 sw_tx_timestamp(skb);
2934}
2935
Johannes Berg6e3e9392011-11-09 10:15:42 +01002936/**
2937 * skb_complete_wifi_ack - deliver skb with wifi status
2938 *
2939 * @skb: the original outgoing packet
2940 * @acked: ack status
2941 *
2942 */
2943void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2944
Joe Perches7965bd42013-09-26 14:48:15 -07002945__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2946__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002947
Herbert Xu60476372007-04-09 11:59:39 -07002948static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2949{
Tom Herbert6edec0e2015-02-10 16:30:28 -08002950 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
2951 skb->csum_valid ||
2952 (skb->ip_summed == CHECKSUM_PARTIAL &&
2953 skb_checksum_start_offset(skb) >= 0));
Herbert Xu60476372007-04-09 11:59:39 -07002954}
2955
Herbert Xufb286bb2005-11-10 13:01:24 -08002956/**
2957 * skb_checksum_complete - Calculate checksum of an entire packet
2958 * @skb: packet to process
2959 *
2960 * This function calculates the checksum over the entire packet plus
2961 * the value of skb->csum. The latter can be used to supply the
2962 * checksum of a pseudo header as used by TCP/UDP. It returns the
2963 * checksum.
2964 *
2965 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2966 * this function can be used to verify that checksum on received
2967 * packets. In that case the function should return zero if the
2968 * checksum is correct. In particular, this function will return zero
2969 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2970 * hardware has already verified the correctness of the checksum.
2971 */
Al Viro4381ca32007-07-15 21:00:11 +01002972static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002973{
Herbert Xu60476372007-04-09 11:59:39 -07002974 return skb_csum_unnecessary(skb) ?
2975 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002976}
2977
Tom Herbert77cffe22014-08-27 21:26:46 -07002978static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
2979{
2980 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2981 if (skb->csum_level == 0)
2982 skb->ip_summed = CHECKSUM_NONE;
2983 else
2984 skb->csum_level--;
2985 }
2986}
2987
2988static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
2989{
2990 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2991 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
2992 skb->csum_level++;
2993 } else if (skb->ip_summed == CHECKSUM_NONE) {
2994 skb->ip_summed = CHECKSUM_UNNECESSARY;
2995 skb->csum_level = 0;
2996 }
2997}
2998
Tom Herbert5a212322014-08-31 15:12:41 -07002999static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
3000{
3001 /* Mark current checksum as bad (typically called from GRO
3002 * path). In the case that ip_summed is CHECKSUM_NONE
3003 * this must be the first checksum encountered in the packet.
3004 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
3005 * checksum after the last one validated. For UDP, a zero
3006 * checksum can not be marked as bad.
3007 */
3008
3009 if (skb->ip_summed == CHECKSUM_NONE ||
3010 skb->ip_summed == CHECKSUM_UNNECESSARY)
3011 skb->csum_bad = 1;
3012}
3013
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003014/* Check if we need to perform checksum complete validation.
3015 *
3016 * Returns true if checksum complete is needed, false otherwise
3017 * (either checksum is unnecessary or zero checksum is allowed).
3018 */
3019static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3020 bool zero_okay,
3021 __sum16 check)
3022{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003023 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3024 skb->csum_valid = 1;
Tom Herbert77cffe22014-08-27 21:26:46 -07003025 __skb_decr_checksum_unnecessary(skb);
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003026 return false;
3027 }
3028
3029 return true;
3030}
3031
3032/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3033 * in checksum_init.
3034 */
3035#define CHECKSUM_BREAK 76
3036
Tom Herbert4e18b9a2015-04-20 14:10:04 -07003037/* Unset checksum-complete
3038 *
3039 * Unset checksum complete can be done when packet is being modified
3040 * (uncompressed for instance) and checksum-complete value is
3041 * invalidated.
3042 */
3043static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3044{
3045 if (skb->ip_summed == CHECKSUM_COMPLETE)
3046 skb->ip_summed = CHECKSUM_NONE;
3047}
3048
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003049/* Validate (init) checksum based on checksum complete.
3050 *
3051 * Return values:
3052 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3053 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3054 * checksum is stored in skb->csum for use in __skb_checksum_complete
3055 * non-zero: value of invalid checksum
3056 *
3057 */
3058static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3059 bool complete,
3060 __wsum psum)
3061{
3062 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3063 if (!csum_fold(csum_add(psum, skb->csum))) {
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003064 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003065 return 0;
3066 }
Tom Herbert5a212322014-08-31 15:12:41 -07003067 } else if (skb->csum_bad) {
3068 /* ip_summed == CHECKSUM_NONE in this case */
Eric Dumazetc91d4602015-05-15 05:48:07 -07003069 return (__force __sum16)1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003070 }
3071
3072 skb->csum = psum;
3073
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003074 if (complete || skb->len <= CHECKSUM_BREAK) {
3075 __sum16 csum;
3076
3077 csum = __skb_checksum_complete(skb);
3078 skb->csum_valid = !csum;
3079 return csum;
3080 }
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003081
3082 return 0;
3083}
3084
3085static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3086{
3087 return 0;
3088}
3089
3090/* Perform checksum validate (init). Note that this is a macro since we only
3091 * want to calculate the pseudo header which is an input function if necessary.
3092 * First we try to validate without any computation (checksum unnecessary) and
3093 * then calculate based on checksum complete calling the function to compute
3094 * pseudo header.
3095 *
3096 * Return values:
3097 * 0: checksum is validated or try to in skb_checksum_complete
3098 * non-zero: value of invalid checksum
3099 */
3100#define __skb_checksum_validate(skb, proto, complete, \
3101 zero_okay, check, compute_pseudo) \
3102({ \
3103 __sum16 __ret = 0; \
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003104 skb->csum_valid = 0; \
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003105 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3106 __ret = __skb_checksum_validate_complete(skb, \
3107 complete, compute_pseudo(skb, proto)); \
3108 __ret; \
3109})
3110
3111#define skb_checksum_init(skb, proto, compute_pseudo) \
3112 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3113
3114#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3115 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3116
3117#define skb_checksum_validate(skb, proto, compute_pseudo) \
3118 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3119
3120#define skb_checksum_validate_zero_check(skb, proto, check, \
3121 compute_pseudo) \
Sabrina Dubroca096a4cf2015-02-06 18:54:19 +01003122 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003123
3124#define skb_checksum_simple_validate(skb) \
3125 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3126
Tom Herbertd96535a2014-08-31 15:12:42 -07003127static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3128{
3129 return (skb->ip_summed == CHECKSUM_NONE &&
3130 skb->csum_valid && !skb->csum_bad);
3131}
3132
3133static inline void __skb_checksum_convert(struct sk_buff *skb,
3134 __sum16 check, __wsum pseudo)
3135{
3136 skb->csum = ~pseudo;
3137 skb->ip_summed = CHECKSUM_COMPLETE;
3138}
3139
3140#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3141do { \
3142 if (__skb_checksum_convert_check(skb)) \
3143 __skb_checksum_convert(skb, check, \
3144 compute_pseudo(skb, proto)); \
3145} while (0)
3146
Tom Herbert15e23962015-02-10 16:30:31 -08003147static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3148 u16 start, u16 offset)
3149{
3150 skb->ip_summed = CHECKSUM_PARTIAL;
3151 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3152 skb->csum_offset = offset - start;
3153}
3154
Tom Herbertdcdc8992015-02-02 16:07:34 -08003155/* Update skbuf and packet to reflect the remote checksum offload operation.
3156 * When called, ptr indicates the starting point for skb->csum when
3157 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3158 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3159 */
3160static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
Tom Herbert15e23962015-02-10 16:30:31 -08003161 int start, int offset, bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08003162{
3163 __wsum delta;
3164
Tom Herbert15e23962015-02-10 16:30:31 -08003165 if (!nopartial) {
3166 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3167 return;
3168 }
3169
Tom Herbertdcdc8992015-02-02 16:07:34 -08003170 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3171 __skb_checksum_complete(skb);
3172 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3173 }
3174
3175 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3176
3177 /* Adjust skb->csum since we changed the packet */
3178 skb->csum = csum_add(skb->csum, delta);
3179}
3180
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003181#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07003182void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003183static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3184{
3185 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07003186 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003187}
3188static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3189{
3190 if (nfct)
3191 atomic_inc(&nfct->use);
3192}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003193#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003194#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003195static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3196{
3197 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3198 kfree(nf_bridge);
3199}
3200static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3201{
3202 if (nf_bridge)
3203 atomic_inc(&nf_bridge->use);
3204}
3205#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08003206static inline void nf_reset(struct sk_buff *skb)
3207{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003208#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08003209 nf_conntrack_put(skb->nfct);
3210 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003211#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003212#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Patrick McHardya193a4a2006-03-20 19:23:05 -08003213 nf_bridge_put(skb->nf_bridge);
3214 skb->nf_bridge = NULL;
3215#endif
3216}
3217
Patrick McHardy124dff02013-04-05 20:42:05 +02003218static inline void nf_reset_trace(struct sk_buff *skb)
3219{
Florian Westphal478b3602014-02-15 23:48:45 +01003220#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00003221 skb->nf_trace = 0;
3222#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003223}
3224
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003225/* Note: This doesn't put any conntrack and bridge info in dst. */
Eric Dumazetb1937222014-09-28 22:18:47 -07003226static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3227 bool copy)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003228{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003229#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003230 dst->nfct = src->nfct;
3231 nf_conntrack_get(src->nfct);
Eric Dumazetb1937222014-09-28 22:18:47 -07003232 if (copy)
3233 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003234#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003235#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003236 dst->nf_bridge = src->nf_bridge;
3237 nf_bridge_get(src->nf_bridge);
3238#endif
Florian Westphal478b3602014-02-15 23:48:45 +01003239#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Eric Dumazetb1937222014-09-28 22:18:47 -07003240 if (copy)
3241 dst->nf_trace = src->nf_trace;
Florian Westphal478b3602014-02-15 23:48:45 +01003242#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003243}
3244
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003245static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3246{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003247#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003248 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003249#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003250#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003251 nf_bridge_put(dst->nf_bridge);
3252#endif
Eric Dumazetb1937222014-09-28 22:18:47 -07003253 __nf_copy(dst, src, true);
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003254}
3255
James Morris984bc162006-06-09 00:29:17 -07003256#ifdef CONFIG_NETWORK_SECMARK
3257static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3258{
3259 to->secmark = from->secmark;
3260}
3261
3262static inline void skb_init_secmark(struct sk_buff *skb)
3263{
3264 skb->secmark = 0;
3265}
3266#else
3267static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3268{ }
3269
3270static inline void skb_init_secmark(struct sk_buff *skb)
3271{ }
3272#endif
3273
Eric W. Biederman574f7192014-04-01 12:20:24 -07003274static inline bool skb_irq_freeable(const struct sk_buff *skb)
3275{
3276 return !skb->destructor &&
3277#if IS_ENABLED(CONFIG_XFRM)
3278 !skb->sp &&
3279#endif
3280#if IS_ENABLED(CONFIG_NF_CONNTRACK)
3281 !skb->nfct &&
3282#endif
3283 !skb->_skb_refdst &&
3284 !skb_has_frag_list(skb);
3285}
3286
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003287static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3288{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003289 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003290}
3291
Stephen Hemminger92477442009-03-21 13:39:26 -07003292static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003293{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003294 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003295}
3296
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003297static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3298{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003299 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003300}
3301
David S. Millerd5a9e242009-01-27 16:22:11 -08003302static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3303{
3304 skb->queue_mapping = rx_queue + 1;
3305}
3306
Stephen Hemminger92477442009-03-21 13:39:26 -07003307static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003308{
3309 return skb->queue_mapping - 1;
3310}
3311
Stephen Hemminger92477442009-03-21 13:39:26 -07003312static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003313{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003314 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08003315}
3316
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003317static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3318{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003319#ifdef CONFIG_XFRM
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003320 return skb->sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003321#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003322 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003323#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003324}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003325
Pravin B Shelar68c33162013-02-14 14:02:41 +00003326/* Keeps track of mac header offset relative to skb->head.
3327 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3328 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07003329 * tunnel skb it points to outer mac header.
3330 * Keeps track of level of encapsulation of network headers.
3331 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00003332struct skb_gso_cb {
Eric Dumazet3347c962013-10-19 11:42:56 -07003333 int mac_offset;
3334 int encap_level;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003335 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00003336};
3337#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3338
3339static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3340{
3341 return (skb_mac_header(inner_skb) - inner_skb->head) -
3342 SKB_GSO_CB(inner_skb)->mac_offset;
3343}
3344
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00003345static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3346{
3347 int new_headroom, headroom;
3348 int ret;
3349
3350 headroom = skb_headroom(skb);
3351 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3352 if (ret)
3353 return ret;
3354
3355 new_headroom = skb_headroom(skb);
3356 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3357 return 0;
3358}
3359
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003360/* Compute the checksum for a gso segment. First compute the checksum value
3361 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3362 * then add in skb->csum (checksum from csum_start to end of packet).
3363 * skb->csum and csum_start are then updated to reflect the checksum of the
3364 * resultant packet starting from the transport header-- the resultant checksum
3365 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3366 * header.
3367 */
3368static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3369{
3370 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
Eric Dumazetc91d4602015-05-15 05:48:07 -07003371 skb_transport_offset(skb);
3372 __wsum partial;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003373
Eric Dumazetc91d4602015-05-15 05:48:07 -07003374 partial = csum_partial(skb_transport_header(skb), plen, skb->csum);
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003375 skb->csum = res;
3376 SKB_GSO_CB(skb)->csum_start -= plen;
3377
Eric Dumazetc91d4602015-05-15 05:48:07 -07003378 return csum_fold(partial);
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003379}
3380
David S. Millerbdcc0922012-03-07 20:53:36 -05003381static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07003382{
3383 return skb_shinfo(skb)->gso_size;
3384}
3385
Eric Dumazet36a8f392013-09-29 01:21:32 -07003386/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05003387static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02003388{
3389 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3390}
3391
Joe Perches7965bd42013-09-26 14:48:15 -07003392void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07003393
3394static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3395{
3396 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3397 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003398 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3399
Alexander Duyckb78462e2010-06-02 12:24:37 +00003400 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3401 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07003402 __skb_warn_lro_forwarding(skb);
3403 return true;
3404 }
3405 return false;
3406}
3407
Herbert Xu35fc92a2007-03-26 23:22:20 -07003408static inline void skb_forward_csum(struct sk_buff *skb)
3409{
3410 /* Unfortunately we don't support this one. Any brave souls? */
3411 if (skb->ip_summed == CHECKSUM_COMPLETE)
3412 skb->ip_summed = CHECKSUM_NONE;
3413}
3414
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003415/**
3416 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3417 * @skb: skb to check
3418 *
3419 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3420 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3421 * use this helper, to document places where we make this assertion.
3422 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003423static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003424{
3425#ifdef DEBUG
3426 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3427#endif
3428}
3429
Rusty Russellf35d9d82008-02-04 23:49:54 -05003430bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00003431
Paul Durranted1f50c2014-01-09 10:02:46 +00003432int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
Linus Lüssing9afd85c2015-05-02 14:01:07 +02003433struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
3434 unsigned int transport_len,
3435 __sum16(*skb_chkf)(struct sk_buff *skb));
Paul Durranted1f50c2014-01-09 10:02:46 +00003436
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00003437/**
3438 * skb_head_is_locked - Determine if the skb->head is locked down
3439 * @skb: skb to check
3440 *
3441 * The head on skbs build around a head frag can be removed if they are
3442 * not cloned. This function returns true if the skb head is locked down
3443 * due to either being allocated via kmalloc, or by being a clone with
3444 * multiple references to the head.
3445 */
3446static inline bool skb_head_is_locked(const struct sk_buff *skb)
3447{
3448 return !skb->head_frag || skb_cloned(skb);
3449}
Florian Westphalfe6cc552014-02-13 23:09:12 +01003450
3451/**
3452 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3453 *
3454 * @skb: GSO skb
3455 *
3456 * skb_gso_network_seglen is used to determine the real size of the
3457 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3458 *
3459 * The MAC/L2 header is not accounted for.
3460 */
3461static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3462{
3463 unsigned int hdr_len = skb_transport_header(skb) -
3464 skb_network_header(skb);
3465 return hdr_len + skb_gso_transport_seglen(skb);
3466}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003467#endif /* __KERNEL__ */
3468#endif /* _LINUX_SKBUFF_H */