blob: f708936cdd23c415fab8cc577a4b0226194e84ae [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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070038
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010039/* A. Checksumming of received packets by device.
40 *
41 * CHECKSUM_NONE:
42 *
43 * Device failed to checksum this packet e.g. due to lack of capabilities.
44 * The packet contains full (though not verified) checksum in packet but
45 * not in skb->csum. Thus, skb->csum is undefined in this case.
46 *
47 * CHECKSUM_UNNECESSARY:
48 *
49 * The hardware you're dealing with doesn't calculate the full checksum
50 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
Tom Herbert77cffe22014-08-27 21:26:46 -070051 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
52 * if their checksums are okay. skb->csum is still undefined in this case
53 * though. It is a bad option, but, unfortunately, nowadays most vendors do
54 * this. Apparently with the secret goal to sell you new devices, when you
55 * will add new protocol to your host, f.e. IPv6 8)
56 *
57 * CHECKSUM_UNNECESSARY is applicable to following protocols:
58 * TCP: IPv6 and IPv4.
59 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
60 * zero UDP checksum for either IPv4 or IPv6, the networking stack
61 * may perform further validation in this case.
62 * GRE: only if the checksum is present in the header.
63 * SCTP: indicates the CRC in SCTP header has been validated.
64 *
65 * skb->csum_level indicates the number of consecutive checksums found in
66 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
67 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
68 * and a device is able to verify the checksums for UDP (possibly zero),
69 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
70 * two. If the device were only able to verify the UDP checksum and not
71 * GRE, either because it doesn't support GRE checksum of because GRE
72 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
73 * not considered in this case).
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010074 *
75 * CHECKSUM_COMPLETE:
76 *
77 * This is the most generic way. The device supplied checksum of the _whole_
78 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
79 * hardware doesn't need to parse L3/L4 headers to implement this.
80 *
81 * Note: Even if device supports only some protocols, but is able to produce
82 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
83 *
84 * CHECKSUM_PARTIAL:
85 *
Tom Herbert6edec0e2015-02-10 16:30:28 -080086 * A checksum is set up to be offloaded to a device as described in the
87 * output description for CHECKSUM_PARTIAL. This may occur on a packet
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010088 * received directly from another Linux OS, e.g., a virtualized Linux kernel
Tom Herbert6edec0e2015-02-10 16:30:28 -080089 * on the same host, or it may be set in the input path in GRO or remote
90 * checksum offload. For the purposes of checksum verification, the checksum
91 * referred to by skb->csum_start + skb->csum_offset and any preceding
92 * checksums in the packet are considered verified. Any checksums in the
93 * packet that are after the checksum being offloaded are not considered to
94 * be verified.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010095 *
96 * B. Checksumming on output.
97 *
98 * CHECKSUM_NONE:
99 *
100 * The skb was already checksummed by the protocol, or a checksum is not
101 * required.
102 *
103 * CHECKSUM_PARTIAL:
104 *
105 * The device is required to checksum the packet as seen by hard_start_xmit()
106 * from skb->csum_start up to the end, and to record/write the checksum at
107 * offset skb->csum_start + skb->csum_offset.
108 *
109 * The device must show its capabilities in dev->features, set up at device
110 * setup time, e.g. netdev_features.h:
111 *
112 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
113 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
114 * IPv4. Sigh. Vendors like this way for an unknown reason.
115 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
116 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
117 * NETIF_F_... - Well, you get the picture.
118 *
119 * CHECKSUM_UNNECESSARY:
120 *
121 * Normally, the device will do per protocol specific checksumming. Protocol
122 * implementations that do not want the NIC to perform the checksum
123 * calculation should use this flag in their outgoing skbs.
124 *
125 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
126 * offload. Correspondingly, the FCoE protocol driver
127 * stack should use CHECKSUM_UNNECESSARY.
128 *
129 * Any questions? No questions, good. --ANK
130 */
131
Herbert Xu60476372007-04-09 11:59:39 -0700132/* Don't change this without changing skb_csum_unnecessary! */
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100133#define CHECKSUM_NONE 0
134#define CHECKSUM_UNNECESSARY 1
135#define CHECKSUM_COMPLETE 2
136#define CHECKSUM_PARTIAL 3
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137
Tom Herbert77cffe22014-08-27 21:26:46 -0700138/* Maximum value in skb->csum_level */
139#define SKB_MAX_CSUM_LEVEL 3
140
Tobias Klauser0bec8c82014-07-22 12:06:23 +0200141#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
David S. Millerfc910a22007-03-25 20:27:59 -0700142#define SKB_WITH_OVERHEAD(X) \
Herbert Xudeea84b2007-10-21 16:27:46 -0700143 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
David S. Millerfc910a22007-03-25 20:27:59 -0700144#define SKB_MAX_ORDER(X, ORDER) \
145 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
147#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
148
Eric Dumazet87fb4b72011-10-13 07:28:54 +0000149/* return minimum truesize of one skb containing X bytes of data */
150#define SKB_TRUESIZE(X) ((X) + \
151 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
152 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
153
Linus Torvalds1da177e2005-04-16 15:20:36 -0700154struct net_device;
David Howells716ea3a2007-04-02 20:19:53 -0700155struct scatterlist;
Jens Axboe9c55e012007-11-06 23:30:13 -0800156struct pipe_inode_info;
Herbert Xua8f820aa2014-11-07 21:22:22 +0800157struct iov_iter;
Alexander Duyckfd11a832014-12-09 19:40:49 -0800158struct napi_struct;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700159
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700160#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700161struct nf_conntrack {
162 atomic_t use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700163};
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700164#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +0200166#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700167struct nf_bridge_info {
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000168 atomic_t use;
Florian Westphal3eaf4022015-04-02 14:31:44 +0200169 enum {
170 BRNF_PROTO_UNCHANGED,
171 BRNF_PROTO_8021Q,
172 BRNF_PROTO_PPPOE
Florian Westphal7fb48c52015-05-03 22:05:28 +0200173 } orig_proto:8;
Florian Westphala1e67952015-04-02 14:31:45 +0200174 bool pkt_otherhost;
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000175 unsigned int mask;
176 struct net_device *physindev;
Florian Westphal7fb48c52015-05-03 22:05:28 +0200177 union {
178 struct net_device *physoutdev;
179 char neigh_header[8];
180 };
Florian Westphalfaecbb42015-05-20 13:42:25 +0200181 __be32 ipv4_daddr;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700182};
183#endif
184
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185struct sk_buff_head {
186 /* These two members must be first. */
187 struct sk_buff *next;
188 struct sk_buff *prev;
189
190 __u32 qlen;
191 spinlock_t lock;
192};
193
194struct sk_buff;
195
Ian Campbell9d4dde52011-12-22 23:39:14 +0000196/* To allow 64K frame to be packed as single skb without frag_list we
197 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
198 * buffers which do not start on a page boundary.
199 *
200 * Since GRO uses frags we allocate at least 16 regardless of page
201 * size.
Anton Blancharda715dea2011-03-27 14:57:26 +0000202 */
Ian Campbell9d4dde52011-12-22 23:39:14 +0000203#if (65536/PAGE_SIZE + 1) < 16
David S. Millereec00952011-03-29 23:34:08 -0700204#define MAX_SKB_FRAGS 16UL
Anton Blancharda715dea2011-03-27 14:57:26 +0000205#else
Ian Campbell9d4dde52011-12-22 23:39:14 +0000206#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
Anton Blancharda715dea2011-03-27 14:57:26 +0000207#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700208
209typedef struct skb_frag_struct skb_frag_t;
210
211struct skb_frag_struct {
Ian Campbella8605c62011-10-19 23:01:49 +0000212 struct {
213 struct page *p;
214 } page;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000215#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
David S. Millera309bb02007-07-30 18:47:03 -0700216 __u32 page_offset;
217 __u32 size;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000218#else
219 __u16 page_offset;
220 __u16 size;
221#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222};
223
Eric Dumazet9e903e02011-10-18 21:00:24 +0000224static inline unsigned int skb_frag_size(const skb_frag_t *frag)
225{
226 return frag->size;
227}
228
229static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
230{
231 frag->size = size;
232}
233
234static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
235{
236 frag->size += delta;
237}
238
239static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
240{
241 frag->size -= delta;
242}
243
Patrick Ohlyac45f602009-02-12 05:03:37 +0000244#define HAVE_HW_TIME_STAMP
245
246/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800247 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000248 * @hwtstamp: hardware time stamp transformed into duration
249 * since arbitrary point in time
Patrick Ohlyac45f602009-02-12 05:03:37 +0000250 *
251 * Software time stamps generated by ktime_get_real() are stored in
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400252 * skb->tstamp.
Patrick Ohlyac45f602009-02-12 05:03:37 +0000253 *
254 * hwtstamps can only be compared against other hwtstamps from
255 * the same device.
256 *
257 * This structure is attached to packets as part of the
258 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
259 */
260struct skb_shared_hwtstamps {
261 ktime_t hwtstamp;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000262};
263
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000264/* Definitions for tx_flags in struct skb_shared_info */
265enum {
266 /* generate hardware time stamp */
267 SKBTX_HW_TSTAMP = 1 << 0,
268
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400269 /* generate software time stamp when queueing packet to NIC */
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000270 SKBTX_SW_TSTAMP = 1 << 1,
271
272 /* device driver is going to provide hardware time stamp */
273 SKBTX_IN_PROGRESS = 1 << 2,
274
Shirley Maa6686f22011-07-06 12:22:12 +0000275 /* device driver supports TX zero-copy buffers */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000276 SKBTX_DEV_ZEROCOPY = 1 << 3,
Johannes Berg6e3e9392011-11-09 10:15:42 +0100277
278 /* generate wifi status information (where possible) */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000279 SKBTX_WIFI_STATUS = 1 << 4,
Pravin B Shelarc9af6db2013-02-11 09:27:41 +0000280
281 /* This indicates at least one fragment might be overwritten
282 * (as in vmsplice(), sendfile() ...)
283 * If we need to compute a TX checksum, we'll need to copy
284 * all frags to avoid possible bad checksum
285 */
286 SKBTX_SHARED_FRAG = 1 << 5,
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400287
288 /* generate software time stamp when entering packet scheduling */
289 SKBTX_SCHED_TSTAMP = 1 << 6,
Willem de Bruijne1c8a602014-08-04 22:11:50 -0400290
291 /* generate software timestamp on peer data acknowledgment */
292 SKBTX_ACK_TSTAMP = 1 << 7,
Shirley Maa6686f22011-07-06 12:22:12 +0000293};
294
Willem de Bruijne1c8a602014-08-04 22:11:50 -0400295#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
296 SKBTX_SCHED_TSTAMP | \
297 SKBTX_ACK_TSTAMP)
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400298#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
299
Shirley Maa6686f22011-07-06 12:22:12 +0000300/*
301 * The callback notifies userspace to release buffers when skb DMA is done in
302 * lower device, the skb last reference should be 0 when calling this.
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000303 * The zerocopy_success argument is true if zero copy transmit occurred,
304 * false on data copy or out of memory error caused by data copy attempt.
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000305 * The ctx field is used to track device context.
306 * The desc field is used to track userspace buffer index.
Shirley Maa6686f22011-07-06 12:22:12 +0000307 */
308struct ubuf_info {
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000309 void (*callback)(struct ubuf_info *, bool zerocopy_success);
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000310 void *ctx;
Shirley Maa6686f22011-07-06 12:22:12 +0000311 unsigned long desc;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000312};
313
Linus Torvalds1da177e2005-04-16 15:20:36 -0700314/* This data is invariant across clones and lives at
315 * the end of the header data, ie. at skb->end.
316 */
317struct skb_shared_info {
Ian Campbell9f42f122012-01-05 07:13:39 +0000318 unsigned char nr_frags;
319 __u8 tx_flags;
Herbert Xu79671682006-06-22 02:40:14 -0700320 unsigned short gso_size;
321 /* Warning: this field is not always filled in (UFO)! */
322 unsigned short gso_segs;
323 unsigned short gso_type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000325 struct skb_shared_hwtstamps hwtstamps;
Willem de Bruijn09c2d252014-08-04 22:11:47 -0400326 u32 tskey;
Ian Campbell9f42f122012-01-05 07:13:39 +0000327 __be32 ip6_frag_id;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700328
329 /*
330 * Warning : all fields before dataref are cleared in __alloc_skb()
331 */
332 atomic_t dataref;
333
Johann Baudy69e3c752009-05-18 22:11:22 -0700334 /* Intermediate layers must ensure that destructor_arg
335 * remains valid until skb destructor */
336 void * destructor_arg;
Shirley Maa6686f22011-07-06 12:22:12 +0000337
Eric Dumazetfed66382010-07-22 19:09:08 +0000338 /* must be last field, see pskb_expand_head() */
339 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340};
341
342/* We divide dataref into two halves. The higher 16 bits hold references
343 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700344 * the entire skb->data. A clone of a headerless skb holds the length of
345 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346 *
347 * All users must obey the rule that the skb->data reference count must be
348 * greater than or equal to the payload reference count.
349 *
350 * Holding a reference to the payload part means that the user does not
351 * care about modifications to the header part of skb->data.
352 */
353#define SKB_DATAREF_SHIFT 16
354#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
355
David S. Millerd179cd12005-08-17 14:57:30 -0700356
357enum {
Vijay Subramanianc8753d52014-10-02 10:00:43 -0700358 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
359 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
360 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
David S. Millerd179cd12005-08-17 14:57:30 -0700361};
362
Herbert Xu79671682006-06-22 02:40:14 -0700363enum {
364 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700365 SKB_GSO_UDP = 1 << 1,
Herbert Xu576a30e2006-06-27 13:22:38 -0700366
367 /* This indicates the skb is from an untrusted source. */
368 SKB_GSO_DODGY = 1 << 2,
Michael Chanb0da85372006-06-29 12:30:00 -0700369
370 /* This indicates the tcp segment has CWR set. */
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700371 SKB_GSO_TCP_ECN = 1 << 3,
372
373 SKB_GSO_TCPV6 = 1 << 4,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800374
375 SKB_GSO_FCOE = 1 << 5,
Pravin B Shelar68c33162013-02-14 14:02:41 +0000376
377 SKB_GSO_GRE = 1 << 6,
Pravin B Shelar73136262013-03-07 13:21:51 +0000378
Tom Herbert4b282522014-06-14 23:23:52 -0700379 SKB_GSO_GRE_CSUM = 1 << 7,
Simon Horman0d89d202013-05-23 21:02:52 +0000380
Tom Herbert4b282522014-06-14 23:23:52 -0700381 SKB_GSO_IPIP = 1 << 8,
Eric Dumazetcb32f512013-10-19 11:42:57 -0700382
Tom Herbert4b282522014-06-14 23:23:52 -0700383 SKB_GSO_SIT = 1 << 9,
Eric Dumazet61c1db72013-10-20 20:47:30 -0700384
Tom Herbert4b282522014-06-14 23:23:52 -0700385 SKB_GSO_UDP_TUNNEL = 1 << 10,
Tom Herbert0f4f4ff2014-06-04 17:20:16 -0700386
387 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
Tom Herbert4749c092014-06-04 17:20:23 -0700388
Pravin B Shelar59b93b42014-11-05 15:27:48 -0800389 SKB_GSO_TUNNEL_REMCSUM = 1 << 12,
Herbert Xu79671682006-06-22 02:40:14 -0700390};
391
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700392#if BITS_PER_LONG > 32
393#define NET_SKBUFF_DATA_USES_OFFSET 1
394#endif
395
396#ifdef NET_SKBUFF_DATA_USES_OFFSET
397typedef unsigned int sk_buff_data_t;
398#else
399typedef unsigned char *sk_buff_data_t;
400#endif
401
Eric Dumazet363ec39232014-02-26 14:02:11 -0800402/**
403 * struct skb_mstamp - multi resolution time stamps
404 * @stamp_us: timestamp in us resolution
405 * @stamp_jiffies: timestamp in jiffies
406 */
407struct skb_mstamp {
408 union {
409 u64 v64;
410 struct {
411 u32 stamp_us;
412 u32 stamp_jiffies;
413 };
414 };
415};
416
417/**
418 * skb_mstamp_get - get current timestamp
419 * @cl: place to store timestamps
420 */
421static inline void skb_mstamp_get(struct skb_mstamp *cl)
422{
423 u64 val = local_clock();
424
425 do_div(val, NSEC_PER_USEC);
426 cl->stamp_us = (u32)val;
427 cl->stamp_jiffies = (u32)jiffies;
428}
429
430/**
431 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
432 * @t1: pointer to newest sample
433 * @t0: pointer to oldest sample
434 */
435static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
436 const struct skb_mstamp *t0)
437{
438 s32 delta_us = t1->stamp_us - t0->stamp_us;
439 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
440
441 /* If delta_us is negative, this might be because interval is too big,
442 * or local_clock() drift is too big : fallback using jiffies.
443 */
444 if (delta_us <= 0 ||
445 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
446
447 delta_us = jiffies_to_usecs(delta_jiffies);
448
449 return delta_us;
450}
451
452
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453/**
454 * struct sk_buff - socket buffer
455 * @next: Next buffer in list
456 * @prev: Previous buffer in list
Eric Dumazet363ec39232014-02-26 14:02:11 -0800457 * @tstamp: Time we arrived/left
Eric Dumazet56b17422014-11-03 08:19:53 -0800458 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700459 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700461 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000462 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700463 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700464 * @len: Length of actual data
465 * @data_len: Data length
466 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700467 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700468 * @csum: Checksum (must include start/offset pair)
469 * @csum_start: Offset from skb->head where checksumming should start
470 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700471 * @priority: Packet queueing priority
WANG Cong60ff7462014-05-04 16:39:18 -0700472 * @ignore_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700473 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700474 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700475 * @nohdr: Payload reference only, must not modify header
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700476 * @nfctinfo: Relationship of this skb to the connection
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700478 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700479 * @ipvs_property: skbuff is owned by ipvs
Randy Dunlap31729362008-02-18 20:52:13 -0800480 * @peeked: this packet has been seen already, so stats have been
481 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700482 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700483 * @protocol: Packet protocol from driver
484 * @destructor: Destruct function
485 * @nfct: Associated connection, if any
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800487 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700488 * @tc_index: Traffic control index
489 * @tc_verd: traffic control verdict
Tom Herbert61b905d2014-03-24 15:34:47 -0700490 * @hash: the packet hash
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700491 * @queue_mapping: Queue mapping for multiqueue devices
David S. Miller0b725a22014-08-25 15:51:53 -0700492 * @xmit_more: More SKBs are pending for this queue
Randy Dunlap553a5672008-04-20 10:51:01 -0700493 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700494 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Tom Herbert61b905d2014-03-24 15:34:47 -0700495 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
Changli Gao4ca24622011-08-19 07:26:44 -0700496 * ports.
Tom Herberta3b18dd2014-07-01 21:33:17 -0700497 * @sw_hash: indicates hash was computed in software stack
Johannes Berg6e3e9392011-11-09 10:15:42 +0100498 * @wifi_acked_valid: wifi_acked was set
499 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000500 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Eliezer Tamir06021292013-06-10 11:39:50 +0300501 * @napi_id: id of the NAPI struct this skb came from
James Morris984bc162006-06-09 00:29:17 -0700502 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700503 * @mark: Generic packet mark
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000504 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b2008-07-14 22:49:06 -0700505 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000506 * @inner_protocol: Protocol (encapsulation)
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000507 * @inner_transport_header: Inner transport layer header (encapsulation)
508 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000509 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700510 * @transport_header: Transport layer header
511 * @network_header: Network layer header
512 * @mac_header: Link layer header
513 * @tail: Tail pointer
514 * @end: End pointer
515 * @head: Head of buffer
516 * @data: Data head pointer
517 * @truesize: Buffer size
518 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 */
520
521struct sk_buff {
Eric Dumazet363ec39232014-02-26 14:02:11 -0800522 union {
Eric Dumazet56b17422014-11-03 08:19:53 -0800523 struct {
524 /* These two members must be first. */
525 struct sk_buff *next;
526 struct sk_buff *prev;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000527
Eric Dumazet56b17422014-11-03 08:19:53 -0800528 union {
529 ktime_t tstamp;
530 struct skb_mstamp skb_mstamp;
531 };
532 };
533 struct rb_node rbnode; /* used in netem & tcp stack */
534 };
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000535 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700536 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700537
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538 /*
539 * This is the control buffer. It is free to use for every
540 * layer. Please put your private variables there. If you
541 * want to keep them across layers you have to do a skb_clone()
542 * first. This is owned by whoever has the skb queued ATM.
543 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000544 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545
Eric Dumazet7fee2262010-05-11 23:19:48 +0000546 unsigned long _skb_refdst;
Eric Dumazetb1937222014-09-28 22:18:47 -0700547 void (*destructor)(struct sk_buff *skb);
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000548#ifdef CONFIG_XFRM
549 struct sec_path *sp;
550#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700551#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
552 struct nf_conntrack *nfct;
553#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +0200554#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Eric Dumazetb1937222014-09-28 22:18:47 -0700555 struct nf_bridge_info *nf_bridge;
556#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700558 data_len;
559 __u16 mac_len,
560 hdr_len;
Eric Dumazetb1937222014-09-28 22:18:47 -0700561
562 /* Following fields are _not_ copied in __copy_skb_header()
563 * Note that queue_mapping is here mostly to fill a hole.
564 */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200565 kmemcheck_bitfield_begin(flags1);
Eric Dumazetb1937222014-09-28 22:18:47 -0700566 __u16 queue_mapping;
567 __u8 cloned:1,
Harald Welte6869c4d2005-08-09 19:24:19 -0700568 nohdr:1,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800569 fclone:2,
Herbert Xua59322b2007-12-05 01:53:40 -0800570 peeked:1,
Eric Dumazetb1937222014-09-28 22:18:47 -0700571 head_frag:1,
572 xmit_more:1;
573 /* one bit hole */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200574 kmemcheck_bitfield_end(flags1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700575
Eric Dumazetb1937222014-09-28 22:18:47 -0700576 /* fields enclosed in headers_start/headers_end are copied
577 * using a single memcpy() in __copy_skb_header()
578 */
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700579 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700580 __u32 headers_start[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700581 /* public: */
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200582
583/* if you move pkt_type around you also must adapt those constants */
584#ifdef __BIG_ENDIAN_BITFIELD
585#define PKT_TYPE_MAX (7 << 5)
586#else
587#define PKT_TYPE_MAX 7
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100588#endif
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200589#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
590
591 __u8 __pkt_type_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700592 __u8 pkt_type:3;
593 __u8 pfmemalloc:1;
594 __u8 ignore_df:1;
595 __u8 nfctinfo:3;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596
Eric Dumazetb1937222014-09-28 22:18:47 -0700597 __u8 nf_trace:1;
598 __u8 ip_summed:2;
599 __u8 ooo_okay:1;
600 __u8 l4_hash:1;
601 __u8 sw_hash:1;
602 __u8 wifi_acked_valid:1;
603 __u8 wifi_acked:1;
604
605 __u8 no_fcs:1;
606 /* Indicates the inner headers are valid in the skbuff. */
607 __u8 encapsulation:1;
608 __u8 encap_hdr_csum:1;
609 __u8 csum_valid:1;
610 __u8 csum_complete_sw:1;
611 __u8 csum_level:2;
612 __u8 csum_bad:1;
613
614#ifdef CONFIG_IPV6_NDISC_NODETYPE
615 __u8 ndisc_nodetype:2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700617 __u8 ipvs_property:1;
Tom Herbert8bce6d72014-09-29 20:22:29 -0700618 __u8 inner_protocol_type:1;
Tom Herberte585f232014-11-04 09:06:54 -0800619 __u8 remcsum_offload:1;
620 /* 3 or 5 bit hole */
Alexander Duyck4031ae62012-01-27 06:22:53 +0000621
Linus Torvalds1da177e2005-04-16 15:20:36 -0700622#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700623 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700625 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200628
Eric Dumazetb1937222014-09-28 22:18:47 -0700629 union {
630 __wsum csum;
631 struct {
632 __u16 csum_start;
633 __u16 csum_offset;
634 };
635 };
636 __u32 priority;
637 int skb_iif;
638 __u32 hash;
639 __be16 vlan_proto;
640 __u16 vlan_tci;
Eric Dumazet2bd82482015-02-03 23:48:24 -0800641#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
642 union {
643 unsigned int napi_id;
644 unsigned int sender_cpu;
645 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700646#endif
James Morris984bc162006-06-09 00:29:17 -0700647#ifdef CONFIG_NETWORK_SECMARK
648 __u32 secmark;
649#endif
Neil Horman3b885782009-10-12 13:26:31 -0700650 union {
651 __u32 mark;
Eric Dumazet16fad692013-03-14 05:40:32 +0000652 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700653 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654
Tom Herbert8bce6d72014-09-29 20:22:29 -0700655 union {
656 __be16 inner_protocol;
657 __u8 inner_ipproto;
658 };
659
Simon Horman1a37e412013-05-23 21:02:51 +0000660 __u16 inner_transport_header;
661 __u16 inner_network_header;
662 __u16 inner_mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700663
664 __be16 protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000665 __u16 transport_header;
666 __u16 network_header;
667 __u16 mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700668
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700669 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700670 __u32 headers_end[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700671 /* public: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700672
Linus Torvalds1da177e2005-04-16 15:20:36 -0700673 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700674 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700675 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700676 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700677 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700678 unsigned int truesize;
679 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700680};
681
682#ifdef __KERNEL__
683/*
684 * Handling routines are only of interest to the kernel
685 */
686#include <linux/slab.h>
687
Linus Torvalds1da177e2005-04-16 15:20:36 -0700688
Mel Gormanc93bdd02012-07-31 16:44:19 -0700689#define SKB_ALLOC_FCLONE 0x01
690#define SKB_ALLOC_RX 0x02
Alexander Duyckfd11a832014-12-09 19:40:49 -0800691#define SKB_ALLOC_NAPI 0x04
Mel Gormanc93bdd02012-07-31 16:44:19 -0700692
693/* Returns true if the skb was allocated from PFMEMALLOC reserves */
694static inline bool skb_pfmemalloc(const struct sk_buff *skb)
695{
696 return unlikely(skb->pfmemalloc);
697}
698
Eric Dumazet7fee2262010-05-11 23:19:48 +0000699/*
700 * skb might have a dst pointer attached, refcounted or not.
701 * _skb_refdst low order bit is set if refcount was _not_ taken
702 */
703#define SKB_DST_NOREF 1UL
704#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
705
706/**
707 * skb_dst - returns skb dst_entry
708 * @skb: buffer
709 *
710 * Returns skb dst_entry, regardless of reference taken or not.
711 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000712static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
713{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000714 /* If refdst was not refcounted, check we still are in a
715 * rcu_read_lock section
716 */
717 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
718 !rcu_read_lock_held() &&
719 !rcu_read_lock_bh_held());
720 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000721}
722
Eric Dumazet7fee2262010-05-11 23:19:48 +0000723/**
724 * skb_dst_set - sets skb dst
725 * @skb: buffer
726 * @dst: dst entry
727 *
728 * Sets skb dst, assuming a reference was taken on dst and should
729 * be released by skb_dst_drop()
730 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000731static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
732{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000733 skb->_skb_refdst = (unsigned long)dst;
734}
735
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200736/**
737 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
738 * @skb: buffer
739 * @dst: dst entry
740 *
741 * Sets skb dst, assuming a reference was not taken on dst.
742 * If dst entry is cached, we do not take reference and dst_release
743 * will be avoided by refdst_drop. If dst entry is not cached, we take
744 * reference, so that last dst_release can destroy the dst immediately.
745 */
746static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
747{
Hannes Frederic Sowadbfc4fb2014-12-06 19:19:42 +0100748 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
749 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200750}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000751
752/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300753 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000754 * @skb: buffer
755 */
756static inline bool skb_dst_is_noref(const struct sk_buff *skb)
757{
758 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000759}
760
Eric Dumazet511c3f92009-06-02 05:14:27 +0000761static inline struct rtable *skb_rtable(const struct sk_buff *skb)
762{
Eric Dumazetadf30902009-06-02 05:19:30 +0000763 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000764}
765
Joe Perches7965bd42013-09-26 14:48:15 -0700766void kfree_skb(struct sk_buff *skb);
767void kfree_skb_list(struct sk_buff *segs);
768void skb_tx_error(struct sk_buff *skb);
769void consume_skb(struct sk_buff *skb);
770void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000771extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000772
Joe Perches7965bd42013-09-26 14:48:15 -0700773void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
774bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
775 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000776
Joe Perches7965bd42013-09-26 14:48:15 -0700777struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
778 int node);
Eric Dumazet2ea2f622015-04-24 16:05:01 -0700779struct sk_buff *__build_skb(void *data, unsigned int frag_size);
Joe Perches7965bd42013-09-26 14:48:15 -0700780struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700781static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100782 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700783{
Eric Dumazet564824b2010-10-11 19:05:25 +0000784 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700785}
786
Eric Dumazet2e4e4412014-09-17 04:49:49 -0700787struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
788 unsigned long data_len,
789 int max_page_order,
790 int *errcode,
791 gfp_t gfp_mask);
792
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700793/* Layout of fast clones : [skb1][skb2][fclone_ref] */
794struct sk_buff_fclones {
795 struct sk_buff skb1;
796
797 struct sk_buff skb2;
798
799 atomic_t fclone_ref;
800};
801
802/**
803 * skb_fclone_busy - check if fclone is busy
804 * @skb: buffer
805 *
806 * Returns true is skb is a fast clone, and its clone is not freed.
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700807 * Some drivers call skb_orphan() in their ndo_start_xmit(),
808 * so we also check that this didnt happen.
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700809 */
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700810static inline bool skb_fclone_busy(const struct sock *sk,
811 const struct sk_buff *skb)
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700812{
813 const struct sk_buff_fclones *fclones;
814
815 fclones = container_of(skb, struct sk_buff_fclones, skb1);
816
817 return skb->fclone == SKB_FCLONE_ORIG &&
Eric Dumazet6ffe75eb2014-12-03 17:04:39 -0800818 atomic_read(&fclones->fclone_ref) > 1 &&
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700819 fclones->skb2.sk == sk;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700820}
821
David S. Millerd179cd12005-08-17 14:57:30 -0700822static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100823 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700824{
Mel Gormanc93bdd02012-07-31 16:44:19 -0700825 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700826}
827
Joe Perches7965bd42013-09-26 14:48:15 -0700828struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
Patrick McHardy0ebd0ac2013-04-17 06:46:58 +0000829static inline struct sk_buff *alloc_skb_head(gfp_t priority)
830{
831 return __alloc_skb_head(priority, -1);
832}
833
Joe Perches7965bd42013-09-26 14:48:15 -0700834struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
835int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
836struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
837struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
Octavian Purdilabad93e92014-06-12 01:36:26 +0300838struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
839 gfp_t gfp_mask, bool fclone);
840static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
841 gfp_t gfp_mask)
842{
843 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
844}
Eric Dumazet117632e2011-12-03 21:39:53 +0000845
Joe Perches7965bd42013-09-26 14:48:15 -0700846int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
847struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
848 unsigned int headroom);
849struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
850 int newtailroom, gfp_t priority);
Fan Du25a91d82014-01-18 09:54:23 +0800851int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
852 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -0700853int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
854 int len);
855int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
856int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000857#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858
Joe Perches7965bd42013-09-26 14:48:15 -0700859int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
860 int getfrag(void *from, char *to, int offset,
861 int len, int odd, struct sk_buff *skb),
862 void *from, int length);
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700863
Hannes Frederic Sowabe12a1f2015-05-21 16:59:58 +0200864int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
865 int offset, size_t size);
866
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800867struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700868 __u32 lower_offset;
869 __u32 upper_offset;
870 __u32 frag_idx;
871 __u32 stepped_offset;
872 struct sk_buff *root_skb;
873 struct sk_buff *cur_skb;
874 __u8 *frag_data;
875};
876
Joe Perches7965bd42013-09-26 14:48:15 -0700877void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
878 unsigned int to, struct skb_seq_state *st);
879unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
880 struct skb_seq_state *st);
881void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -0700882
Joe Perches7965bd42013-09-26 14:48:15 -0700883unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
Bojan Prtvar059a2442015-02-22 11:46:35 +0100884 unsigned int to, struct ts_config *config);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700885
Tom Herbert09323cc2013-12-15 22:16:19 -0800886/*
887 * Packet hash types specify the type of hash in skb_set_hash.
888 *
889 * Hash types refer to the protocol layer addresses which are used to
890 * construct a packet's hash. The hashes are used to differentiate or identify
891 * flows of the protocol layer for the hash type. Hash types are either
892 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
893 *
894 * Properties of hashes:
895 *
896 * 1) Two packets in different flows have different hash values
897 * 2) Two packets in the same flow should have the same hash value
898 *
899 * A hash at a higher layer is considered to be more specific. A driver should
900 * set the most specific hash possible.
901 *
902 * A driver cannot indicate a more specific hash than the layer at which a hash
903 * was computed. For instance an L3 hash cannot be set as an L4 hash.
904 *
905 * A driver may indicate a hash level which is less specific than the
906 * actual layer the hash was computed on. For instance, a hash computed
907 * at L4 may be considered an L3 hash. This should only be done if the
908 * driver can't unambiguously determine that the HW computed the hash at
909 * the higher layer. Note that the "should" in the second property above
910 * permits this.
911 */
912enum pkt_hash_types {
913 PKT_HASH_TYPE_NONE, /* Undefined type */
914 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
915 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
916 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
917};
918
919static inline void
920skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
921{
Tom Herbert61b905d2014-03-24 15:34:47 -0700922 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
Tom Herberta3b18dd2014-07-01 21:33:17 -0700923 skb->sw_hash = 0;
Tom Herbert61b905d2014-03-24 15:34:47 -0700924 skb->hash = hash;
Tom Herbert09323cc2013-12-15 22:16:19 -0800925}
926
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800927static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000928{
Tom Herberta3b18dd2014-07-01 21:33:17 -0700929 if (!skb->l4_hash && !skb->sw_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800930 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000931
Tom Herbert61b905d2014-03-24 15:34:47 -0700932 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000933}
934
Tom Herbert50fb7992015-05-01 11:30:12 -0700935__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
936
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800937static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
938{
Tom Herbert61b905d2014-03-24 15:34:47 -0700939 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800940}
941
Tom Herbert7539fad2013-12-15 22:12:18 -0800942static inline void skb_clear_hash(struct sk_buff *skb)
943{
Tom Herbert61b905d2014-03-24 15:34:47 -0700944 skb->hash = 0;
Tom Herberta3b18dd2014-07-01 21:33:17 -0700945 skb->sw_hash = 0;
Tom Herbert61b905d2014-03-24 15:34:47 -0700946 skb->l4_hash = 0;
Tom Herbert7539fad2013-12-15 22:12:18 -0800947}
948
949static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
950{
Tom Herbert61b905d2014-03-24 15:34:47 -0700951 if (!skb->l4_hash)
Tom Herbert7539fad2013-12-15 22:12:18 -0800952 skb_clear_hash(skb);
953}
954
Tom Herbert3df7a742013-12-15 22:16:29 -0800955static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
956{
Tom Herbert61b905d2014-03-24 15:34:47 -0700957 to->hash = from->hash;
Tom Herberta3b18dd2014-07-01 21:33:17 -0700958 to->sw_hash = from->sw_hash;
Tom Herbert61b905d2014-03-24 15:34:47 -0700959 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -0800960};
961
Eric Dumazetc29390c2015-03-11 18:42:02 -0700962static inline void skb_sender_cpu_clear(struct sk_buff *skb)
963{
964#ifdef CONFIG_XPS
965 skb->sender_cpu = 0;
966#endif
967}
968
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700969#ifdef NET_SKBUFF_DATA_USES_OFFSET
970static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
971{
972 return skb->head + skb->end;
973}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000974
975static inline unsigned int skb_end_offset(const struct sk_buff *skb)
976{
977 return skb->end;
978}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700979#else
980static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
981{
982 return skb->end;
983}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000984
985static inline unsigned int skb_end_offset(const struct sk_buff *skb)
986{
987 return skb->end - skb->head;
988}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700989#endif
990
Linus Torvalds1da177e2005-04-16 15:20:36 -0700991/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700992#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700993
Patrick Ohlyac45f602009-02-12 05:03:37 +0000994static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
995{
996 return &skb_shinfo(skb)->hwtstamps;
997}
998
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999/**
1000 * skb_queue_empty - check if a queue is empty
1001 * @list: queue head
1002 *
1003 * Returns true if the queue is empty, false otherwise.
1004 */
1005static inline int skb_queue_empty(const struct sk_buff_head *list)
1006{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001007 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001008}
1009
1010/**
David S. Millerfc7ebb22008-09-23 00:34:07 -07001011 * skb_queue_is_last - check if skb is the last entry in the queue
1012 * @list: queue head
1013 * @skb: buffer
1014 *
1015 * Returns true if @skb is the last buffer on the list.
1016 */
1017static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1018 const struct sk_buff *skb)
1019{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001020 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -07001021}
1022
1023/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001024 * skb_queue_is_first - check if skb is the first entry in the queue
1025 * @list: queue head
1026 * @skb: buffer
1027 *
1028 * Returns true if @skb is the first buffer on the list.
1029 */
1030static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1031 const struct sk_buff *skb)
1032{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001033 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001034}
1035
1036/**
David S. Miller249c8b42008-09-23 00:44:42 -07001037 * skb_queue_next - return the next packet in the queue
1038 * @list: queue head
1039 * @skb: current buffer
1040 *
1041 * Return the next packet in @list after @skb. It is only valid to
1042 * call this if skb_queue_is_last() evaluates to false.
1043 */
1044static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1045 const struct sk_buff *skb)
1046{
1047 /* This BUG_ON may seem severe, but if we just return then we
1048 * are going to dereference garbage.
1049 */
1050 BUG_ON(skb_queue_is_last(list, skb));
1051 return skb->next;
1052}
1053
1054/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001055 * skb_queue_prev - return the prev packet in the queue
1056 * @list: queue head
1057 * @skb: current buffer
1058 *
1059 * Return the prev packet in @list before @skb. It is only valid to
1060 * call this if skb_queue_is_first() evaluates to false.
1061 */
1062static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1063 const struct sk_buff *skb)
1064{
1065 /* This BUG_ON may seem severe, but if we just return then we
1066 * are going to dereference garbage.
1067 */
1068 BUG_ON(skb_queue_is_first(list, skb));
1069 return skb->prev;
1070}
1071
1072/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001073 * skb_get - reference buffer
1074 * @skb: buffer to reference
1075 *
1076 * Makes another reference to a socket buffer and returns a pointer
1077 * to the buffer.
1078 */
1079static inline struct sk_buff *skb_get(struct sk_buff *skb)
1080{
1081 atomic_inc(&skb->users);
1082 return skb;
1083}
1084
1085/*
1086 * If users == 1, we are the only owner and are can avoid redundant
1087 * atomic change.
1088 */
1089
1090/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001091 * skb_cloned - is the buffer a clone
1092 * @skb: buffer to check
1093 *
1094 * Returns true if the buffer was generated with skb_clone() and is
1095 * one of multiple shared copies of the buffer. Cloned buffers are
1096 * shared data so must not be written to under normal circumstances.
1097 */
1098static inline int skb_cloned(const struct sk_buff *skb)
1099{
1100 return skb->cloned &&
1101 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1102}
1103
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001104static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1105{
1106 might_sleep_if(pri & __GFP_WAIT);
1107
1108 if (skb_cloned(skb))
1109 return pskb_expand_head(skb, 0, 0, pri);
1110
1111 return 0;
1112}
1113
Linus Torvalds1da177e2005-04-16 15:20:36 -07001114/**
1115 * skb_header_cloned - is the header a clone
1116 * @skb: buffer to check
1117 *
1118 * Returns true if modifying the header part of the buffer requires
1119 * the data to be copied.
1120 */
1121static inline int skb_header_cloned(const struct sk_buff *skb)
1122{
1123 int dataref;
1124
1125 if (!skb->cloned)
1126 return 0;
1127
1128 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1129 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1130 return dataref != 1;
1131}
1132
1133/**
1134 * skb_header_release - release reference to header
1135 * @skb: buffer to operate on
1136 *
1137 * Drop a reference to the header part of the buffer. This is done
1138 * by acquiring a payload reference. You must not read from the header
1139 * part of skb->data after this.
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001140 * Note : Check if you can use __skb_header_release() instead.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001141 */
1142static inline void skb_header_release(struct sk_buff *skb)
1143{
1144 BUG_ON(skb->nohdr);
1145 skb->nohdr = 1;
1146 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1147}
1148
1149/**
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001150 * __skb_header_release - release reference to header
1151 * @skb: buffer to operate on
1152 *
1153 * Variant of skb_header_release() assuming skb is private to caller.
1154 * We can avoid one atomic operation.
1155 */
1156static inline void __skb_header_release(struct sk_buff *skb)
1157{
1158 skb->nohdr = 1;
1159 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1160}
1161
1162
1163/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001164 * skb_shared - is the buffer shared
1165 * @skb: buffer to check
1166 *
1167 * Returns true if more than one person has a reference to this
1168 * buffer.
1169 */
1170static inline int skb_shared(const struct sk_buff *skb)
1171{
1172 return atomic_read(&skb->users) != 1;
1173}
1174
1175/**
1176 * skb_share_check - check if buffer is shared and if so clone it
1177 * @skb: buffer to check
1178 * @pri: priority for memory allocation
1179 *
1180 * If the buffer is shared the buffer is cloned and the old copy
1181 * drops a reference. A new clone with a single reference is returned.
1182 * If the buffer is not shared the original buffer is returned. When
1183 * being called from interrupt status or with spinlocks held pri must
1184 * be GFP_ATOMIC.
1185 *
1186 * NULL is returned on a memory allocation failure.
1187 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001188static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001189{
1190 might_sleep_if(pri & __GFP_WAIT);
1191 if (skb_shared(skb)) {
1192 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001193
1194 if (likely(nskb))
1195 consume_skb(skb);
1196 else
1197 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001198 skb = nskb;
1199 }
1200 return skb;
1201}
1202
1203/*
1204 * Copy shared buffers into a new sk_buff. We effectively do COW on
1205 * packets to handle cases where we have a local reader and forward
1206 * and a couple of other messy ones. The normal one is tcpdumping
1207 * a packet thats being forwarded.
1208 */
1209
1210/**
1211 * skb_unshare - make a copy of a shared buffer
1212 * @skb: buffer to check
1213 * @pri: priority for memory allocation
1214 *
1215 * If the socket buffer is a clone then this function creates a new
1216 * copy of the data, drops a reference count on the old copy and returns
1217 * the new copy with the reference count at 1. If the buffer is not a clone
1218 * the original buffer is returned. When called with a spinlock held or
1219 * from interrupt state @pri must be %GFP_ATOMIC
1220 *
1221 * %NULL is returned on a memory allocation failure.
1222 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001223static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001224 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001225{
1226 might_sleep_if(pri & __GFP_WAIT);
1227 if (skb_cloned(skb)) {
1228 struct sk_buff *nskb = skb_copy(skb, pri);
Alexander Aring31eff812014-10-10 23:10:47 +02001229
1230 /* Free our shared copy */
1231 if (likely(nskb))
1232 consume_skb(skb);
1233 else
1234 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001235 skb = nskb;
1236 }
1237 return skb;
1238}
1239
1240/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001241 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001242 * @list_: list to peek at
1243 *
1244 * Peek an &sk_buff. Unlike most other operations you _MUST_
1245 * be careful with this one. A peek leaves the buffer on the
1246 * list and someone else may run off with it. You must hold
1247 * the appropriate locks or have a private queue to do this.
1248 *
1249 * Returns %NULL for an empty list or a pointer to the head element.
1250 * The reference count is not incremented and the reference is therefore
1251 * volatile. Use with caution.
1252 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001253static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001254{
Eric Dumazet18d07002012-04-30 16:31:46 +00001255 struct sk_buff *skb = list_->next;
1256
1257 if (skb == (struct sk_buff *)list_)
1258 skb = NULL;
1259 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001260}
1261
1262/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001263 * skb_peek_next - peek skb following the given one from a queue
1264 * @skb: skb to start from
1265 * @list_: list to peek at
1266 *
1267 * Returns %NULL when the end of the list is met or a pointer to the
1268 * next element. The reference count is not incremented and the
1269 * reference is therefore volatile. Use with caution.
1270 */
1271static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1272 const struct sk_buff_head *list_)
1273{
1274 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001275
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001276 if (next == (struct sk_buff *)list_)
1277 next = NULL;
1278 return next;
1279}
1280
1281/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001282 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001283 * @list_: list to peek at
1284 *
1285 * Peek an &sk_buff. Unlike most other operations you _MUST_
1286 * be careful with this one. A peek leaves the buffer on the
1287 * list and someone else may run off with it. You must hold
1288 * the appropriate locks or have a private queue to do this.
1289 *
1290 * Returns %NULL for an empty list or a pointer to the tail element.
1291 * The reference count is not incremented and the reference is therefore
1292 * volatile. Use with caution.
1293 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001294static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001295{
Eric Dumazet18d07002012-04-30 16:31:46 +00001296 struct sk_buff *skb = list_->prev;
1297
1298 if (skb == (struct sk_buff *)list_)
1299 skb = NULL;
1300 return skb;
1301
Linus Torvalds1da177e2005-04-16 15:20:36 -07001302}
1303
1304/**
1305 * skb_queue_len - get queue length
1306 * @list_: list to measure
1307 *
1308 * Return the length of an &sk_buff queue.
1309 */
1310static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1311{
1312 return list_->qlen;
1313}
1314
David S. Miller67fed452008-09-21 22:36:24 -07001315/**
1316 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1317 * @list: queue to initialize
1318 *
1319 * This initializes only the list and queue length aspects of
1320 * an sk_buff_head object. This allows to initialize the list
1321 * aspects of an sk_buff_head without reinitializing things like
1322 * the spinlock. It can also be used for on-stack sk_buff_head
1323 * objects where the spinlock is known to not be used.
1324 */
1325static inline void __skb_queue_head_init(struct sk_buff_head *list)
1326{
1327 list->prev = list->next = (struct sk_buff *)list;
1328 list->qlen = 0;
1329}
1330
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001331/*
1332 * This function creates a split out lock class for each invocation;
1333 * this is needed for now since a whole lot of users of the skb-queue
1334 * infrastructure in drivers have different locking usage (in hardirq)
1335 * than the networking core (in softirq only). In the long run either the
1336 * network layer or drivers should need annotation to consolidate the
1337 * main types of usage into 3 classes.
1338 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001339static inline void skb_queue_head_init(struct sk_buff_head *list)
1340{
1341 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001342 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001343}
1344
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001345static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1346 struct lock_class_key *class)
1347{
1348 skb_queue_head_init(list);
1349 lockdep_set_class(&list->lock, class);
1350}
1351
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001353 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001354 *
1355 * The "__skb_xxxx()" functions are the non-atomic ones that
1356 * can only be called with interrupts disabled.
1357 */
Joe Perches7965bd42013-09-26 14:48:15 -07001358void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1359 struct sk_buff_head *list);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001360static inline void __skb_insert(struct sk_buff *newsk,
1361 struct sk_buff *prev, struct sk_buff *next,
1362 struct sk_buff_head *list)
1363{
1364 newsk->next = next;
1365 newsk->prev = prev;
1366 next->prev = prev->next = newsk;
1367 list->qlen++;
1368}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001369
David S. Miller67fed452008-09-21 22:36:24 -07001370static inline void __skb_queue_splice(const struct sk_buff_head *list,
1371 struct sk_buff *prev,
1372 struct sk_buff *next)
1373{
1374 struct sk_buff *first = list->next;
1375 struct sk_buff *last = list->prev;
1376
1377 first->prev = prev;
1378 prev->next = first;
1379
1380 last->next = next;
1381 next->prev = last;
1382}
1383
1384/**
1385 * skb_queue_splice - join two skb lists, this is designed for stacks
1386 * @list: the new list to add
1387 * @head: the place to add it in the first list
1388 */
1389static inline void skb_queue_splice(const struct sk_buff_head *list,
1390 struct sk_buff_head *head)
1391{
1392 if (!skb_queue_empty(list)) {
1393 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001394 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001395 }
1396}
1397
1398/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001399 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001400 * @list: the new list to add
1401 * @head: the place to add it in the first list
1402 *
1403 * The list at @list is reinitialised
1404 */
1405static inline void skb_queue_splice_init(struct sk_buff_head *list,
1406 struct sk_buff_head *head)
1407{
1408 if (!skb_queue_empty(list)) {
1409 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001410 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001411 __skb_queue_head_init(list);
1412 }
1413}
1414
1415/**
1416 * skb_queue_splice_tail - join two skb lists, each list being a queue
1417 * @list: the new list to add
1418 * @head: the place to add it in the first list
1419 */
1420static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1421 struct sk_buff_head *head)
1422{
1423 if (!skb_queue_empty(list)) {
1424 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001425 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001426 }
1427}
1428
1429/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001430 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001431 * @list: the new list to add
1432 * @head: the place to add it in the first list
1433 *
1434 * Each of the lists is a queue.
1435 * The list at @list is reinitialised
1436 */
1437static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1438 struct sk_buff_head *head)
1439{
1440 if (!skb_queue_empty(list)) {
1441 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001442 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001443 __skb_queue_head_init(list);
1444 }
1445}
1446
Linus Torvalds1da177e2005-04-16 15:20:36 -07001447/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001448 * __skb_queue_after - queue a buffer at the list head
1449 * @list: list to use
1450 * @prev: place after this buffer
1451 * @newsk: buffer to queue
1452 *
1453 * Queue a buffer int the middle of a list. This function takes no locks
1454 * and you must therefore hold required locks before calling it.
1455 *
1456 * A buffer cannot be placed on two lists at the same time.
1457 */
1458static inline void __skb_queue_after(struct sk_buff_head *list,
1459 struct sk_buff *prev,
1460 struct sk_buff *newsk)
1461{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001462 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001463}
1464
Joe Perches7965bd42013-09-26 14:48:15 -07001465void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1466 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001467
Gerrit Renkerf5572852008-04-14 00:05:28 -07001468static inline void __skb_queue_before(struct sk_buff_head *list,
1469 struct sk_buff *next,
1470 struct sk_buff *newsk)
1471{
1472 __skb_insert(newsk, next->prev, next, list);
1473}
1474
Stephen Hemminger300ce172005-10-30 13:47:34 -08001475/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476 * __skb_queue_head - queue a buffer at the list head
1477 * @list: list to use
1478 * @newsk: buffer to queue
1479 *
1480 * Queue a buffer at the start of a list. This function takes no locks
1481 * and you must therefore hold required locks before calling it.
1482 *
1483 * A buffer cannot be placed on two lists at the same time.
1484 */
Joe Perches7965bd42013-09-26 14:48:15 -07001485void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001486static inline void __skb_queue_head(struct sk_buff_head *list,
1487 struct sk_buff *newsk)
1488{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001489 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001490}
1491
1492/**
1493 * __skb_queue_tail - queue a buffer at the list tail
1494 * @list: list to use
1495 * @newsk: buffer to queue
1496 *
1497 * Queue a buffer at the end of a list. This function takes no locks
1498 * and you must therefore hold required locks before calling it.
1499 *
1500 * A buffer cannot be placed on two lists at the same time.
1501 */
Joe Perches7965bd42013-09-26 14:48:15 -07001502void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001503static inline void __skb_queue_tail(struct sk_buff_head *list,
1504 struct sk_buff *newsk)
1505{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001506 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001507}
1508
Linus Torvalds1da177e2005-04-16 15:20:36 -07001509/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001510 * remove sk_buff from list. _Must_ be called atomically, and with
1511 * the list known..
1512 */
Joe Perches7965bd42013-09-26 14:48:15 -07001513void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001514static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1515{
1516 struct sk_buff *next, *prev;
1517
1518 list->qlen--;
1519 next = skb->next;
1520 prev = skb->prev;
1521 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001522 next->prev = prev;
1523 prev->next = next;
1524}
1525
Gerrit Renkerf525c062008-04-14 00:04:12 -07001526/**
1527 * __skb_dequeue - remove from the head of the queue
1528 * @list: list to dequeue from
1529 *
1530 * Remove the head of the list. This function does not take any locks
1531 * so must be used with appropriate locks held only. The head item is
1532 * returned or %NULL if the list is empty.
1533 */
Joe Perches7965bd42013-09-26 14:48:15 -07001534struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Gerrit Renkerf525c062008-04-14 00:04:12 -07001535static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1536{
1537 struct sk_buff *skb = skb_peek(list);
1538 if (skb)
1539 __skb_unlink(skb, list);
1540 return skb;
1541}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001542
1543/**
1544 * __skb_dequeue_tail - remove from the tail of the queue
1545 * @list: list to dequeue from
1546 *
1547 * Remove the tail of the list. This function does not take any locks
1548 * so must be used with appropriate locks held only. The tail item is
1549 * returned or %NULL if the list is empty.
1550 */
Joe Perches7965bd42013-09-26 14:48:15 -07001551struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001552static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1553{
1554 struct sk_buff *skb = skb_peek_tail(list);
1555 if (skb)
1556 __skb_unlink(skb, list);
1557 return skb;
1558}
1559
1560
David S. Millerbdcc0922012-03-07 20:53:36 -05001561static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001562{
1563 return skb->data_len;
1564}
1565
1566static inline unsigned int skb_headlen(const struct sk_buff *skb)
1567{
1568 return skb->len - skb->data_len;
1569}
1570
1571static inline int skb_pagelen(const struct sk_buff *skb)
1572{
1573 int i, len = 0;
1574
1575 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001576 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001577 return len + skb_headlen(skb);
1578}
1579
Ian Campbell131ea662011-08-19 06:25:00 +00001580/**
1581 * __skb_fill_page_desc - initialise a paged fragment in an skb
1582 * @skb: buffer containing fragment to be initialised
1583 * @i: paged fragment index to initialise
1584 * @page: the page to use for this fragment
1585 * @off: the offset to the data with @page
1586 * @size: the length of the data
1587 *
1588 * Initialises the @i'th fragment of @skb to point to &size bytes at
1589 * offset @off within @page.
1590 *
1591 * Does not take any additional reference on the fragment.
1592 */
1593static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1594 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001595{
1596 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1597
Mel Gormanc48a11c2012-07-31 16:44:23 -07001598 /*
1599 * Propagate page->pfmemalloc to the skb if we can. The problem is
1600 * that not all callers have unique ownership of the page. If
1601 * pfmemalloc is set, we check the mapping as a mapping implies
1602 * page->index is set (index and pfmemalloc share space).
1603 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1604 * do not lose pfmemalloc information as the pages would not be
1605 * allocated using __GFP_MEMALLOC.
1606 */
Ian Campbella8605c62011-10-19 23:01:49 +00001607 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001608 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001609 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001610
1611 page = compound_head(page);
1612 if (page->pfmemalloc && !page->mapping)
1613 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00001614}
1615
1616/**
1617 * skb_fill_page_desc - initialise a paged fragment in an skb
1618 * @skb: buffer containing fragment to be initialised
1619 * @i: paged fragment index to initialise
1620 * @page: the page to use for this fragment
1621 * @off: the offset to the data with @page
1622 * @size: the length of the data
1623 *
1624 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01001625 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00001626 * addition updates @skb such that @i is the last fragment.
1627 *
1628 * Does not take any additional reference on the fragment.
1629 */
1630static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1631 struct page *page, int off, int size)
1632{
1633 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001634 skb_shinfo(skb)->nr_frags = i + 1;
1635}
1636
Joe Perches7965bd42013-09-26 14:48:15 -07001637void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1638 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001639
Jason Wangf8e617e2013-11-01 14:07:47 +08001640void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1641 unsigned int truesize);
1642
Linus Torvalds1da177e2005-04-16 15:20:36 -07001643#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001644#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001645#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1646
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001647#ifdef NET_SKBUFF_DATA_USES_OFFSET
1648static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1649{
1650 return skb->head + skb->tail;
1651}
1652
1653static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1654{
1655 skb->tail = skb->data - skb->head;
1656}
1657
1658static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1659{
1660 skb_reset_tail_pointer(skb);
1661 skb->tail += offset;
1662}
Simon Horman7cc46192013-05-28 20:34:29 +00001663
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001664#else /* NET_SKBUFF_DATA_USES_OFFSET */
1665static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1666{
1667 return skb->tail;
1668}
1669
1670static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1671{
1672 skb->tail = skb->data;
1673}
1674
1675static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1676{
1677 skb->tail = skb->data + offset;
1678}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001679
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001680#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1681
Linus Torvalds1da177e2005-04-16 15:20:36 -07001682/*
1683 * Add data to an sk_buff
1684 */
Mathias Krause0c7ddf32013-11-07 14:18:24 +01001685unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001686unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1688{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001689 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001690 SKB_LINEAR_ASSERT(skb);
1691 skb->tail += len;
1692 skb->len += len;
1693 return tmp;
1694}
1695
Joe Perches7965bd42013-09-26 14:48:15 -07001696unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001697static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1698{
1699 skb->data -= len;
1700 skb->len += len;
1701 return skb->data;
1702}
1703
Joe Perches7965bd42013-09-26 14:48:15 -07001704unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001705static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1706{
1707 skb->len -= len;
1708 BUG_ON(skb->len < skb->data_len);
1709 return skb->data += len;
1710}
1711
David S. Miller47d29642010-05-02 02:21:44 -07001712static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1713{
1714 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1715}
1716
Joe Perches7965bd42013-09-26 14:48:15 -07001717unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001718
1719static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1720{
1721 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001722 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001723 return NULL;
1724 skb->len -= len;
1725 return skb->data += len;
1726}
1727
1728static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1729{
1730 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1731}
1732
1733static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1734{
1735 if (likely(len <= skb_headlen(skb)))
1736 return 1;
1737 if (unlikely(len > skb->len))
1738 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001739 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001740}
1741
1742/**
1743 * skb_headroom - bytes at buffer head
1744 * @skb: buffer to check
1745 *
1746 * Return the number of bytes of free space at the head of an &sk_buff.
1747 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001748static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001749{
1750 return skb->data - skb->head;
1751}
1752
1753/**
1754 * skb_tailroom - bytes at buffer end
1755 * @skb: buffer to check
1756 *
1757 * Return the number of bytes of free space at the tail of an sk_buff
1758 */
1759static inline int skb_tailroom(const struct sk_buff *skb)
1760{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001761 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001762}
1763
1764/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00001765 * skb_availroom - bytes at buffer end
1766 * @skb: buffer to check
1767 *
1768 * Return the number of bytes of free space at the tail of an sk_buff
1769 * allocated by sk_stream_alloc()
1770 */
1771static inline int skb_availroom(const struct sk_buff *skb)
1772{
Eric Dumazet16fad692013-03-14 05:40:32 +00001773 if (skb_is_nonlinear(skb))
1774 return 0;
1775
1776 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00001777}
1778
1779/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001780 * skb_reserve - adjust headroom
1781 * @skb: buffer to alter
1782 * @len: bytes to move
1783 *
1784 * Increase the headroom of an empty &sk_buff by reducing the tail
1785 * room. This is only allowed for an empty buffer.
1786 */
David S. Miller8243126c2006-01-17 02:54:21 -08001787static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001788{
1789 skb->data += len;
1790 skb->tail += len;
1791}
1792
Tom Herbert8bce6d72014-09-29 20:22:29 -07001793#define ENCAP_TYPE_ETHER 0
1794#define ENCAP_TYPE_IPPROTO 1
1795
1796static inline void skb_set_inner_protocol(struct sk_buff *skb,
1797 __be16 protocol)
1798{
1799 skb->inner_protocol = protocol;
1800 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
1801}
1802
1803static inline void skb_set_inner_ipproto(struct sk_buff *skb,
1804 __u8 ipproto)
1805{
1806 skb->inner_ipproto = ipproto;
1807 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
1808}
1809
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001810static inline void skb_reset_inner_headers(struct sk_buff *skb)
1811{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001812 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001813 skb->inner_network_header = skb->network_header;
1814 skb->inner_transport_header = skb->transport_header;
1815}
1816
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00001817static inline void skb_reset_mac_len(struct sk_buff *skb)
1818{
1819 skb->mac_len = skb->network_header - skb->mac_header;
1820}
1821
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001822static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1823 *skb)
1824{
1825 return skb->head + skb->inner_transport_header;
1826}
1827
1828static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1829{
1830 skb->inner_transport_header = skb->data - skb->head;
1831}
1832
1833static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1834 const int offset)
1835{
1836 skb_reset_inner_transport_header(skb);
1837 skb->inner_transport_header += offset;
1838}
1839
1840static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1841{
1842 return skb->head + skb->inner_network_header;
1843}
1844
1845static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1846{
1847 skb->inner_network_header = skb->data - skb->head;
1848}
1849
1850static inline void skb_set_inner_network_header(struct sk_buff *skb,
1851 const int offset)
1852{
1853 skb_reset_inner_network_header(skb);
1854 skb->inner_network_header += offset;
1855}
1856
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001857static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1858{
1859 return skb->head + skb->inner_mac_header;
1860}
1861
1862static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1863{
1864 skb->inner_mac_header = skb->data - skb->head;
1865}
1866
1867static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1868 const int offset)
1869{
1870 skb_reset_inner_mac_header(skb);
1871 skb->inner_mac_header += offset;
1872}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001873static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1874{
Cong Wang35d04612013-05-29 15:16:05 +08001875 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001876}
1877
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001878static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1879{
1880 return skb->head + skb->transport_header;
1881}
1882
1883static inline void skb_reset_transport_header(struct sk_buff *skb)
1884{
1885 skb->transport_header = skb->data - skb->head;
1886}
1887
1888static inline void skb_set_transport_header(struct sk_buff *skb,
1889 const int offset)
1890{
1891 skb_reset_transport_header(skb);
1892 skb->transport_header += offset;
1893}
1894
1895static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1896{
1897 return skb->head + skb->network_header;
1898}
1899
1900static inline void skb_reset_network_header(struct sk_buff *skb)
1901{
1902 skb->network_header = skb->data - skb->head;
1903}
1904
1905static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1906{
1907 skb_reset_network_header(skb);
1908 skb->network_header += offset;
1909}
1910
1911static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1912{
1913 return skb->head + skb->mac_header;
1914}
1915
1916static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1917{
Cong Wang35d04612013-05-29 15:16:05 +08001918 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001919}
1920
1921static inline void skb_reset_mac_header(struct sk_buff *skb)
1922{
1923 skb->mac_header = skb->data - skb->head;
1924}
1925
1926static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1927{
1928 skb_reset_mac_header(skb);
1929 skb->mac_header += offset;
1930}
1931
Timo Teräs0e3da5b2013-12-16 11:02:09 +02001932static inline void skb_pop_mac_header(struct sk_buff *skb)
1933{
1934 skb->mac_header = skb->network_header;
1935}
1936
Ying Xuefbbdb8f2013-03-27 16:46:06 +00001937static inline void skb_probe_transport_header(struct sk_buff *skb,
1938 const int offset_hint)
1939{
1940 struct flow_keys keys;
1941
1942 if (skb_transport_header_was_set(skb))
1943 return;
Jiri Pirko06635a32015-05-12 14:56:16 +02001944 else if (skb_flow_dissect_flow_keys(skb, &keys))
1945 skb_set_transport_header(skb, keys.basic.thoff);
Ying Xuefbbdb8f2013-03-27 16:46:06 +00001946 else
1947 skb_set_transport_header(skb, offset_hint);
1948}
1949
Eric Dumazet03606892012-02-23 10:55:02 +00001950static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1951{
1952 if (skb_mac_header_was_set(skb)) {
1953 const unsigned char *old_mac = skb_mac_header(skb);
1954
1955 skb_set_mac_header(skb, -skb->mac_len);
1956 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1957 }
1958}
1959
Michał Mirosław04fb4512010-12-14 15:24:08 +00001960static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1961{
1962 return skb->csum_start - skb_headroom(skb);
1963}
1964
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001965static inline int skb_transport_offset(const struct sk_buff *skb)
1966{
1967 return skb_transport_header(skb) - skb->data;
1968}
1969
1970static inline u32 skb_network_header_len(const struct sk_buff *skb)
1971{
1972 return skb->transport_header - skb->network_header;
1973}
1974
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001975static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1976{
1977 return skb->inner_transport_header - skb->inner_network_header;
1978}
1979
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001980static inline int skb_network_offset(const struct sk_buff *skb)
1981{
1982 return skb_network_header(skb) - skb->data;
1983}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001984
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001985static inline int skb_inner_network_offset(const struct sk_buff *skb)
1986{
1987 return skb_inner_network_header(skb) - skb->data;
1988}
1989
Changli Gaof9599ce2010-08-04 04:43:44 +00001990static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1991{
1992 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1993}
1994
Linus Torvalds1da177e2005-04-16 15:20:36 -07001995/*
1996 * CPUs often take a performance hit when accessing unaligned memory
1997 * locations. The actual performance hit varies, it can be small if the
1998 * hardware handles it or large if we have to take an exception and fix it
1999 * in software.
2000 *
2001 * Since an ethernet header is 14 bytes network drivers often end up with
2002 * the IP header at an unaligned offset. The IP header can be aligned by
2003 * shifting the start of the packet by 2 bytes. Drivers should do this
2004 * with:
2005 *
Tobias Klauser8660c122009-07-13 22:48:16 +00002006 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002007 *
2008 * The downside to this alignment of the IP header is that the DMA is now
2009 * unaligned. On some architectures the cost of an unaligned DMA is high
2010 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00002011 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002012 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2013 * to be overridden.
2014 */
2015#ifndef NET_IP_ALIGN
2016#define NET_IP_ALIGN 2
2017#endif
2018
Anton Blanchard025be812006-03-31 02:27:06 -08002019/*
2020 * The networking layer reserves some headroom in skb data (via
2021 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2022 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08002023 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08002024 *
2025 * Unfortunately this headroom changes the DMA alignment of the resulting
2026 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2027 * on some architectures. An architecture can override this value,
2028 * perhaps setting it to a cacheline in size (since that will maintain
2029 * cacheline alignment of the DMA). It must be a power of 2.
2030 *
David S. Millerd6301d32009-02-08 19:24:13 -08002031 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08002032 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07002033 *
2034 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2035 * to reduce average number of cache lines per packet.
2036 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07002037 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08002038 */
2039#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07002040#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08002041#endif
2042
Joe Perches7965bd42013-09-26 14:48:15 -07002043int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002044
2045static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2046{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00002047 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07002048 WARN_ON(1);
2049 return;
2050 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002051 skb->len = len;
2052 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002053}
2054
Joe Perches7965bd42013-09-26 14:48:15 -07002055void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002056
2057static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2058{
Herbert Xu3cc0e872006-06-09 16:13:38 -07002059 if (skb->data_len)
2060 return ___pskb_trim(skb, len);
2061 __skb_trim(skb, len);
2062 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002063}
2064
2065static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2066{
2067 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2068}
2069
2070/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07002071 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2072 * @skb: buffer to alter
2073 * @len: new length
2074 *
2075 * This is identical to pskb_trim except that the caller knows that
2076 * the skb is not cloned so we should never get an error due to out-
2077 * of-memory.
2078 */
2079static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2080{
2081 int err = pskb_trim(skb, len);
2082 BUG_ON(err);
2083}
2084
2085/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002086 * skb_orphan - orphan a buffer
2087 * @skb: buffer to orphan
2088 *
2089 * If a buffer currently has an owner then we call the owner's
2090 * destructor function and make the @skb unowned. The buffer continues
2091 * to exist but is no longer charged to its former owner.
2092 */
2093static inline void skb_orphan(struct sk_buff *skb)
2094{
Eric Dumazetc34a7612013-07-30 16:11:15 -07002095 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002096 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002097 skb->destructor = NULL;
2098 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07002099 } else {
2100 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002101 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002102}
2103
2104/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002105 * skb_orphan_frags - orphan the frags contained in a buffer
2106 * @skb: buffer to orphan frags from
2107 * @gfp_mask: allocation mask for replacement pages
2108 *
2109 * For each frag in the SKB which needs a destructor (i.e. has an
2110 * owner) create a copy of that frag and release the original
2111 * page by calling the destructor.
2112 */
2113static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2114{
2115 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2116 return 0;
2117 return skb_copy_ubufs(skb, gfp_mask);
2118}
2119
2120/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002121 * __skb_queue_purge - empty a list
2122 * @list: list to empty
2123 *
2124 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2125 * the list and one reference dropped. This function does not take the
2126 * list lock and the caller must hold the relevant locks to use it.
2127 */
Joe Perches7965bd42013-09-26 14:48:15 -07002128void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002129static inline void __skb_queue_purge(struct sk_buff_head *list)
2130{
2131 struct sk_buff *skb;
2132 while ((skb = __skb_dequeue(list)) != NULL)
2133 kfree_skb(skb);
2134}
2135
Joe Perches7965bd42013-09-26 14:48:15 -07002136void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002137
Joe Perches7965bd42013-09-26 14:48:15 -07002138struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2139 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07002140
2141/**
2142 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2143 * @dev: network device to receive on
2144 * @length: length to allocate
2145 *
2146 * Allocate a new &sk_buff and assign it a usage count of one. The
2147 * buffer has unspecified headroom built in. Users should allocate
2148 * the headroom they think they need without accounting for the
2149 * built in space. The built in space is used for optimisations.
2150 *
2151 * %NULL is returned if there is no free memory. Although this function
2152 * allocates memory it can be called from an interrupt.
2153 */
2154static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002155 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002156{
2157 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2158}
2159
Eric Dumazet6f532612012-05-18 05:12:12 +00002160/* legacy helper around __netdev_alloc_skb() */
2161static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2162 gfp_t gfp_mask)
2163{
2164 return __netdev_alloc_skb(NULL, length, gfp_mask);
2165}
2166
2167/* legacy helper around netdev_alloc_skb() */
2168static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2169{
2170 return netdev_alloc_skb(NULL, length);
2171}
2172
2173
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002174static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2175 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002176{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002177 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002178
2179 if (NET_IP_ALIGN && skb)
2180 skb_reserve(skb, NET_IP_ALIGN);
2181 return skb;
2182}
2183
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002184static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2185 unsigned int length)
2186{
2187 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2188}
2189
Alexander Duyck181edb22015-05-06 21:12:03 -07002190static inline void skb_free_frag(void *addr)
2191{
2192 __free_page_frag(addr);
2193}
2194
Alexander Duyckffde7322014-12-09 19:40:42 -08002195void *napi_alloc_frag(unsigned int fragsz);
Alexander Duyckfd11a832014-12-09 19:40:49 -08002196struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2197 unsigned int length, gfp_t gfp_mask);
2198static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2199 unsigned int length)
2200{
2201 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2202}
Alexander Duyckffde7322014-12-09 19:40:42 -08002203
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002204/**
Alexander Duyck71dfda52014-11-11 09:26:34 -08002205 * __dev_alloc_pages - allocate page for network Rx
2206 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2207 * @order: size of the allocation
2208 *
2209 * Allocate a new page.
2210 *
2211 * %NULL is returned if there is no free memory.
2212*/
2213static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2214 unsigned int order)
2215{
2216 /* This piece of code contains several assumptions.
2217 * 1. This is for device Rx, therefor a cold page is preferred.
2218 * 2. The expectation is the user wants a compound page.
2219 * 3. If requesting a order 0 page it will not be compound
2220 * due to the check to see if order has a value in prep_new_page
2221 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2222 * code in gfp_to_alloc_flags that should be enforcing this.
2223 */
2224 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2225
2226 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2227}
2228
2229static inline struct page *dev_alloc_pages(unsigned int order)
2230{
2231 return __dev_alloc_pages(GFP_ATOMIC, order);
2232}
2233
2234/**
2235 * __dev_alloc_page - allocate a page for network Rx
2236 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2237 *
2238 * Allocate a new page.
2239 *
2240 * %NULL is returned if there is no free memory.
2241 */
2242static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2243{
2244 return __dev_alloc_pages(gfp_mask, 0);
2245}
2246
2247static inline struct page *dev_alloc_page(void)
2248{
2249 return __dev_alloc_page(GFP_ATOMIC);
2250}
2251
2252/**
Mel Gorman06140022012-07-31 16:44:24 -07002253 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2254 * @page: The page that was allocated from skb_alloc_page
2255 * @skb: The skb that may need pfmemalloc set
2256 */
2257static inline void skb_propagate_pfmemalloc(struct page *page,
2258 struct sk_buff *skb)
2259{
2260 if (page && page->pfmemalloc)
2261 skb->pfmemalloc = true;
2262}
2263
Eric Dumazet564824b2010-10-11 19:05:25 +00002264/**
Masanari Iidae2278672014-02-18 22:54:36 +09002265 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002266 * @frag: the paged fragment
2267 *
2268 * Returns the &struct page associated with @frag.
2269 */
2270static inline struct page *skb_frag_page(const skb_frag_t *frag)
2271{
Ian Campbella8605c62011-10-19 23:01:49 +00002272 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00002273}
2274
2275/**
2276 * __skb_frag_ref - take an addition reference on a paged fragment.
2277 * @frag: the paged fragment
2278 *
2279 * Takes an additional reference on the paged fragment @frag.
2280 */
2281static inline void __skb_frag_ref(skb_frag_t *frag)
2282{
2283 get_page(skb_frag_page(frag));
2284}
2285
2286/**
2287 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2288 * @skb: the buffer
2289 * @f: the fragment offset.
2290 *
2291 * Takes an additional reference on the @f'th paged fragment of @skb.
2292 */
2293static inline void skb_frag_ref(struct sk_buff *skb, int f)
2294{
2295 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2296}
2297
2298/**
2299 * __skb_frag_unref - release a reference on a paged fragment.
2300 * @frag: the paged fragment
2301 *
2302 * Releases a reference on the paged fragment @frag.
2303 */
2304static inline void __skb_frag_unref(skb_frag_t *frag)
2305{
2306 put_page(skb_frag_page(frag));
2307}
2308
2309/**
2310 * skb_frag_unref - release a reference on a paged fragment of an skb.
2311 * @skb: the buffer
2312 * @f: the fragment offset
2313 *
2314 * Releases a reference on the @f'th paged fragment of @skb.
2315 */
2316static inline void skb_frag_unref(struct sk_buff *skb, int f)
2317{
2318 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2319}
2320
2321/**
2322 * skb_frag_address - gets the address of the data contained in a paged fragment
2323 * @frag: the paged fragment buffer
2324 *
2325 * Returns the address of the data within @frag. The page must already
2326 * be mapped.
2327 */
2328static inline void *skb_frag_address(const skb_frag_t *frag)
2329{
2330 return page_address(skb_frag_page(frag)) + frag->page_offset;
2331}
2332
2333/**
2334 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2335 * @frag: the paged fragment buffer
2336 *
2337 * Returns the address of the data within @frag. Checks that the page
2338 * is mapped and returns %NULL otherwise.
2339 */
2340static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2341{
2342 void *ptr = page_address(skb_frag_page(frag));
2343 if (unlikely(!ptr))
2344 return NULL;
2345
2346 return ptr + frag->page_offset;
2347}
2348
2349/**
2350 * __skb_frag_set_page - sets the page contained in a paged fragment
2351 * @frag: the paged fragment
2352 * @page: the page to set
2353 *
2354 * Sets the fragment @frag to contain @page.
2355 */
2356static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2357{
Ian Campbella8605c62011-10-19 23:01:49 +00002358 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002359}
2360
2361/**
2362 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2363 * @skb: the buffer
2364 * @f: the fragment offset
2365 * @page: the page to set
2366 *
2367 * Sets the @f'th fragment of @skb to contain @page.
2368 */
2369static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2370 struct page *page)
2371{
2372 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2373}
2374
Eric Dumazet400dfd32013-10-17 16:27:07 -07002375bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2376
Ian Campbell131ea662011-08-19 06:25:00 +00002377/**
2378 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002379 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002380 * @frag: the paged fragment to map
2381 * @offset: the offset within the fragment (starting at the
2382 * fragment's own offset)
2383 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002384 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00002385 *
2386 * Maps the page associated with @frag to @device.
2387 */
2388static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2389 const skb_frag_t *frag,
2390 size_t offset, size_t size,
2391 enum dma_data_direction dir)
2392{
2393 return dma_map_page(dev, skb_frag_page(frag),
2394 frag->page_offset + offset, size, dir);
2395}
2396
Eric Dumazet117632e2011-12-03 21:39:53 +00002397static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2398 gfp_t gfp_mask)
2399{
2400 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2401}
2402
Octavian Purdilabad93e92014-06-12 01:36:26 +03002403
2404static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2405 gfp_t gfp_mask)
2406{
2407 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2408}
2409
2410
Ian Campbell131ea662011-08-19 06:25:00 +00002411/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002412 * skb_clone_writable - is the header of a clone writable
2413 * @skb: buffer to check
2414 * @len: length up to which to write
2415 *
2416 * Returns true if modifying the header part of the cloned buffer
2417 * does not requires the data to be copied.
2418 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002419static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002420{
2421 return !skb_header_cloned(skb) &&
2422 skb_headroom(skb) + len <= skb->hdr_len;
2423}
2424
Herbert Xud9cc2042007-09-16 16:21:16 -07002425static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2426 int cloned)
2427{
2428 int delta = 0;
2429
Herbert Xud9cc2042007-09-16 16:21:16 -07002430 if (headroom > skb_headroom(skb))
2431 delta = headroom - skb_headroom(skb);
2432
2433 if (delta || cloned)
2434 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2435 GFP_ATOMIC);
2436 return 0;
2437}
2438
Patrick McHardy334a8132007-06-25 04:35:20 -07002439/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002440 * skb_cow - copy header of skb when it is required
2441 * @skb: buffer to cow
2442 * @headroom: needed headroom
2443 *
2444 * If the skb passed lacks sufficient headroom or its data part
2445 * is shared, data is reallocated. If reallocation fails, an error
2446 * is returned and original skb is not changed.
2447 *
2448 * The result is skb with writable area skb->head...skb->tail
2449 * and at least @headroom of space at head.
2450 */
2451static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2452{
Herbert Xud9cc2042007-09-16 16:21:16 -07002453 return __skb_cow(skb, headroom, skb_cloned(skb));
2454}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002455
Herbert Xud9cc2042007-09-16 16:21:16 -07002456/**
2457 * skb_cow_head - skb_cow but only making the head writable
2458 * @skb: buffer to cow
2459 * @headroom: needed headroom
2460 *
2461 * This function is identical to skb_cow except that we replace the
2462 * skb_cloned check by skb_header_cloned. It should be used when
2463 * you only need to push on some header and do not need to modify
2464 * the data.
2465 */
2466static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2467{
2468 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002469}
2470
2471/**
2472 * skb_padto - pad an skbuff up to a minimal size
2473 * @skb: buffer to pad
2474 * @len: minimal length
2475 *
2476 * Pads up a buffer to ensure the trailing bytes exist and are
2477 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002478 * is untouched. Otherwise it is extended. Returns zero on
2479 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002480 */
Herbert Xu5b057c62006-06-23 02:06:41 -07002481static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002482{
2483 unsigned int size = skb->len;
2484 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002485 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002486 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002487}
2488
Alexander Duyck9c0c1122014-12-03 08:17:33 -08002489/**
2490 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2491 * @skb: buffer to pad
2492 * @len: minimal length
2493 *
2494 * Pads up a buffer to ensure the trailing bytes exist and are
2495 * blanked. If the buffer already contains sufficient data it
2496 * is untouched. Otherwise it is extended. Returns zero on
2497 * success. The skb is freed on error.
2498 */
2499static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2500{
2501 unsigned int size = skb->len;
2502
2503 if (unlikely(size < len)) {
2504 len -= size;
2505 if (skb_pad(skb, len))
2506 return -ENOMEM;
2507 __skb_put(skb, len);
2508 }
2509 return 0;
2510}
2511
Linus Torvalds1da177e2005-04-16 15:20:36 -07002512static inline int skb_add_data(struct sk_buff *skb,
Al Viroaf2b0402014-11-27 21:44:24 -05002513 struct iov_iter *from, int copy)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002514{
2515 const int off = skb->len;
2516
2517 if (skb->ip_summed == CHECKSUM_NONE) {
Al Viroaf2b0402014-11-27 21:44:24 -05002518 __wsum csum = 0;
2519 if (csum_and_copy_from_iter(skb_put(skb, copy), copy,
2520 &csum, from) == copy) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002521 skb->csum = csum_block_add(skb->csum, csum, off);
2522 return 0;
2523 }
Al Viroaf2b0402014-11-27 21:44:24 -05002524 } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002525 return 0;
2526
2527 __skb_trim(skb, off);
2528 return -EFAULT;
2529}
2530
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002531static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2532 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002533{
2534 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002535 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002536
Ian Campbellea2ab692011-08-22 23:44:58 +00002537 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002538 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002539 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002540 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002541}
2542
Herbert Xu364c6ba2006-06-09 16:10:40 -07002543static inline int __skb_linearize(struct sk_buff *skb)
2544{
2545 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2546}
2547
Linus Torvalds1da177e2005-04-16 15:20:36 -07002548/**
2549 * skb_linearize - convert paged skb to linear one
2550 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002551 *
2552 * If there is no free memory -ENOMEM is returned, otherwise zero
2553 * is returned and the old skb data released.
2554 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002555static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002556{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002557 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2558}
2559
2560/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00002561 * skb_has_shared_frag - can any frag be overwritten
2562 * @skb: buffer to test
2563 *
2564 * Return true if the skb has at least one frag that might be modified
2565 * by an external entity (as in vmsplice()/sendfile())
2566 */
2567static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2568{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00002569 return skb_is_nonlinear(skb) &&
2570 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00002571}
2572
2573/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07002574 * skb_linearize_cow - make sure skb is linear and writable
2575 * @skb: buffer to process
2576 *
2577 * If there is no free memory -ENOMEM is returned, otherwise zero
2578 * is returned and the old skb data released.
2579 */
2580static inline int skb_linearize_cow(struct sk_buff *skb)
2581{
2582 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2583 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002584}
2585
2586/**
2587 * skb_postpull_rcsum - update checksum for received skb after pull
2588 * @skb: buffer to update
2589 * @start: start of data before pull
2590 * @len: length of data pulled
2591 *
2592 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002593 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2594 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002595 */
2596
2597static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002598 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002599{
Patrick McHardy84fa7932006-08-29 16:44:56 -07002600 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002601 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2602}
2603
Herbert Xucbb042f2006-03-20 22:43:56 -08002604unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2605
David S. Miller7ce5a272013-12-02 17:26:05 -05002606/**
2607 * pskb_trim_rcsum - trim received skb and update checksum
2608 * @skb: buffer to trim
2609 * @len: new length
2610 *
2611 * This is exactly the same as pskb_trim except that it ensures the
2612 * checksum of received packets are still valid after the operation.
2613 */
2614
2615static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2616{
2617 if (likely(len >= skb->len))
2618 return 0;
2619 if (skb->ip_summed == CHECKSUM_COMPLETE)
2620 skb->ip_summed = CHECKSUM_NONE;
2621 return __pskb_trim(skb, len);
2622}
2623
Linus Torvalds1da177e2005-04-16 15:20:36 -07002624#define skb_queue_walk(queue, skb) \
2625 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002626 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002627 skb = skb->next)
2628
James Chapman46f89142007-04-30 00:07:31 -07002629#define skb_queue_walk_safe(queue, skb, tmp) \
2630 for (skb = (queue)->next, tmp = skb->next; \
2631 skb != (struct sk_buff *)(queue); \
2632 skb = tmp, tmp = skb->next)
2633
David S. Miller1164f522008-09-23 00:49:44 -07002634#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002635 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002636 skb = skb->next)
2637
2638#define skb_queue_walk_from_safe(queue, skb, tmp) \
2639 for (tmp = skb->next; \
2640 skb != (struct sk_buff *)(queue); \
2641 skb = tmp, tmp = skb->next)
2642
Stephen Hemminger300ce172005-10-30 13:47:34 -08002643#define skb_queue_reverse_walk(queue, skb) \
2644 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002645 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08002646 skb = skb->prev)
2647
David S. Miller686a2952011-01-20 22:47:32 -08002648#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2649 for (skb = (queue)->prev, tmp = skb->prev; \
2650 skb != (struct sk_buff *)(queue); \
2651 skb = tmp, tmp = skb->prev)
2652
2653#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2654 for (tmp = skb->prev; \
2655 skb != (struct sk_buff *)(queue); \
2656 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002657
David S. Miller21dc3302010-08-23 00:13:46 -07002658static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07002659{
2660 return skb_shinfo(skb)->frag_list != NULL;
2661}
2662
2663static inline void skb_frag_list_init(struct sk_buff *skb)
2664{
2665 skb_shinfo(skb)->frag_list = NULL;
2666}
2667
2668static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2669{
2670 frag->next = skb_shinfo(skb)->frag_list;
2671 skb_shinfo(skb)->frag_list = frag;
2672}
2673
2674#define skb_walk_frags(skb, iter) \
2675 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2676
Joe Perches7965bd42013-09-26 14:48:15 -07002677struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2678 int *peeked, int *off, int *err);
2679struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2680 int *err);
2681unsigned int datagram_poll(struct file *file, struct socket *sock,
2682 struct poll_table_struct *wait);
Al Viroc0371da2014-11-24 10:42:55 -05002683int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
2684 struct iov_iter *to, int size);
David S. Miller51f3d022014-11-05 16:46:40 -05002685static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
2686 struct msghdr *msg, int size)
2687{
Al Viroe5a4b0b2014-11-24 18:17:55 -05002688 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
David S. Miller51f3d022014-11-05 16:46:40 -05002689}
Al Viroe5a4b0b2014-11-24 18:17:55 -05002690int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
2691 struct msghdr *msg);
Al Viro3a654f92014-06-19 14:15:22 -04002692int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
2693 struct iov_iter *from, int len);
Al Viro3a654f92014-06-19 14:15:22 -04002694int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
Joe Perches7965bd42013-09-26 14:48:15 -07002695void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2696void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2697int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07002698int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2699int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2700__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2701 int len, __wsum csum);
2702int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2703 struct pipe_inode_info *pipe, unsigned int len,
2704 unsigned int flags);
2705void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01002706unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00002707int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2708 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07002709void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2710int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2711void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Florian Westphalde960aa2014-01-26 10:58:16 +01002712unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -07002713struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Vlad Yasevich0d5501c2014-08-08 14:42:13 -04002714struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
Jiri Pirkoe2195122014-11-19 14:05:01 +01002715int skb_ensure_writable(struct sk_buff *skb, int write_len);
Jiri Pirko93515d52014-11-19 14:05:02 +01002716int skb_vlan_pop(struct sk_buff *skb);
2717int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002718
Al Viro6ce8e9c2014-04-06 21:25:44 -04002719static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
2720{
Al Viro21226ab2014-11-28 15:48:29 -05002721 return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro6ce8e9c2014-04-06 21:25:44 -04002722}
2723
Al Viro7eab8d92014-04-06 21:51:23 -04002724static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
2725{
Al Viroe5a4b0b2014-11-24 18:17:55 -05002726 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro7eab8d92014-04-06 21:51:23 -04002727}
2728
Daniel Borkmann2817a332013-10-30 11:50:51 +01002729struct skb_checksum_ops {
2730 __wsum (*update)(const void *mem, int len, __wsum wsum);
2731 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2732};
2733
2734__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2735 __wsum csum, const struct skb_checksum_ops *ops);
2736__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2737 __wsum csum);
2738
David S. Miller690e36e2014-08-23 12:13:41 -07002739static inline void *__skb_header_pointer(const struct sk_buff *skb, int offset,
2740 int len, void *data, int hlen, void *buffer)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002741{
Patrick McHardy55820ee2005-07-05 14:08:10 -07002742 if (hlen - offset >= len)
David S. Miller690e36e2014-08-23 12:13:41 -07002743 return data + offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002744
David S. Miller690e36e2014-08-23 12:13:41 -07002745 if (!skb ||
2746 skb_copy_bits(skb, offset, buffer, len) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002747 return NULL;
2748
2749 return buffer;
2750}
2751
David S. Miller690e36e2014-08-23 12:13:41 -07002752static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2753 int len, void *buffer)
2754{
2755 return __skb_header_pointer(skb, offset, len, skb->data,
2756 skb_headlen(skb), buffer);
2757}
2758
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01002759/**
2760 * skb_needs_linearize - check if we need to linearize a given skb
2761 * depending on the given device features.
2762 * @skb: socket buffer to check
2763 * @features: net device features
2764 *
2765 * Returns true if either:
2766 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2767 * 2. skb is fragmented and the device does not support SG.
2768 */
2769static inline bool skb_needs_linearize(struct sk_buff *skb,
2770 netdev_features_t features)
2771{
2772 return skb_is_nonlinear(skb) &&
2773 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2774 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2775}
2776
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03002777static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2778 void *to,
2779 const unsigned int len)
2780{
2781 memcpy(to, skb->data, len);
2782}
2783
2784static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2785 const int offset, void *to,
2786 const unsigned int len)
2787{
2788 memcpy(to, skb->data + offset, len);
2789}
2790
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03002791static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2792 const void *from,
2793 const unsigned int len)
2794{
2795 memcpy(skb->data, from, len);
2796}
2797
2798static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2799 const int offset,
2800 const void *from,
2801 const unsigned int len)
2802{
2803 memcpy(skb->data + offset, from, len);
2804}
2805
Joe Perches7965bd42013-09-26 14:48:15 -07002806void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002807
Patrick Ohlyac45f602009-02-12 05:03:37 +00002808static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2809{
2810 return skb->tstamp;
2811}
2812
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002813/**
2814 * skb_get_timestamp - get timestamp from a skb
2815 * @skb: skb to get stamp from
2816 * @stamp: pointer to struct timeval to store stamp in
2817 *
2818 * Timestamps are stored in the skb as offsets to a base timestamp.
2819 * This function converts the offset back to a struct timeval and stores
2820 * it in stamp.
2821 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00002822static inline void skb_get_timestamp(const struct sk_buff *skb,
2823 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002824{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002825 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002826}
2827
Patrick Ohlyac45f602009-02-12 05:03:37 +00002828static inline void skb_get_timestampns(const struct sk_buff *skb,
2829 struct timespec *stamp)
2830{
2831 *stamp = ktime_to_timespec(skb->tstamp);
2832}
2833
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002834static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002835{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002836 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002837}
2838
Stephen Hemminger164891a2007-04-23 22:26:16 -07002839static inline ktime_t net_timedelta(ktime_t t)
2840{
2841 return ktime_sub(ktime_get_real(), t);
2842}
2843
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07002844static inline ktime_t net_invalid_timestamp(void)
2845{
2846 return ktime_set(0, 0);
2847}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002848
Alexander Duyck62bccb82014-09-04 13:31:35 -04002849struct sk_buff *skb_clone_sk(struct sk_buff *skb);
2850
Richard Cochranc1f19b52010-07-17 08:49:36 +00002851#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2852
Joe Perches7965bd42013-09-26 14:48:15 -07002853void skb_clone_tx_timestamp(struct sk_buff *skb);
2854bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00002855
2856#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2857
2858static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2859{
2860}
2861
2862static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2863{
2864 return false;
2865}
2866
2867#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2868
2869/**
2870 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2871 *
Richard Cochranda92b192011-10-21 00:49:15 +00002872 * PHY drivers may accept clones of transmitted packets for
2873 * timestamping via their phy_driver.txtstamp method. These drivers
2874 * must call this function to return the skb back to the stack, with
2875 * or without a timestamp.
2876 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00002877 * @skb: clone of the the original outgoing packet
Richard Cochranda92b192011-10-21 00:49:15 +00002878 * @hwtstamps: hardware time stamps, may be NULL if not available
Richard Cochranc1f19b52010-07-17 08:49:36 +00002879 *
2880 */
2881void skb_complete_tx_timestamp(struct sk_buff *skb,
2882 struct skb_shared_hwtstamps *hwtstamps);
2883
Willem de Bruijne7fd2882014-08-04 22:11:48 -04002884void __skb_tstamp_tx(struct sk_buff *orig_skb,
2885 struct skb_shared_hwtstamps *hwtstamps,
2886 struct sock *sk, int tstype);
2887
Patrick Ohlyac45f602009-02-12 05:03:37 +00002888/**
2889 * skb_tstamp_tx - queue clone of skb with send time stamps
2890 * @orig_skb: the original outgoing packet
2891 * @hwtstamps: hardware time stamps, may be NULL if not available
2892 *
2893 * If the skb has a socket associated, then this function clones the
2894 * skb (thus sharing the actual data and optional structures), stores
2895 * the optional hardware time stamping information (if non NULL) or
2896 * generates a software time stamp (otherwise), then queues the clone
2897 * to the error queue of the socket. Errors are silently ignored.
2898 */
Joe Perches7965bd42013-09-26 14:48:15 -07002899void skb_tstamp_tx(struct sk_buff *orig_skb,
2900 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00002901
Richard Cochran4507a712010-07-17 08:48:28 +00002902static inline void sw_tx_timestamp(struct sk_buff *skb)
2903{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002904 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2905 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00002906 skb_tstamp_tx(skb, NULL);
2907}
2908
2909/**
2910 * skb_tx_timestamp() - Driver hook for transmit timestamping
2911 *
2912 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00002913 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00002914 *
David S. Miller73409f32013-12-27 13:04:33 -05002915 * Specifically, one should make absolutely sure that this function is
2916 * called before TX completion of this packet can trigger. Otherwise
2917 * the packet could potentially already be freed.
2918 *
Richard Cochran4507a712010-07-17 08:48:28 +00002919 * @skb: A socket buffer.
2920 */
2921static inline void skb_tx_timestamp(struct sk_buff *skb)
2922{
Richard Cochranc1f19b52010-07-17 08:49:36 +00002923 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002924 sw_tx_timestamp(skb);
2925}
2926
Johannes Berg6e3e9392011-11-09 10:15:42 +01002927/**
2928 * skb_complete_wifi_ack - deliver skb with wifi status
2929 *
2930 * @skb: the original outgoing packet
2931 * @acked: ack status
2932 *
2933 */
2934void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2935
Joe Perches7965bd42013-09-26 14:48:15 -07002936__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2937__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002938
Herbert Xu60476372007-04-09 11:59:39 -07002939static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2940{
Tom Herbert6edec0e2015-02-10 16:30:28 -08002941 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
2942 skb->csum_valid ||
2943 (skb->ip_summed == CHECKSUM_PARTIAL &&
2944 skb_checksum_start_offset(skb) >= 0));
Herbert Xu60476372007-04-09 11:59:39 -07002945}
2946
Herbert Xufb286bb2005-11-10 13:01:24 -08002947/**
2948 * skb_checksum_complete - Calculate checksum of an entire packet
2949 * @skb: packet to process
2950 *
2951 * This function calculates the checksum over the entire packet plus
2952 * the value of skb->csum. The latter can be used to supply the
2953 * checksum of a pseudo header as used by TCP/UDP. It returns the
2954 * checksum.
2955 *
2956 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2957 * this function can be used to verify that checksum on received
2958 * packets. In that case the function should return zero if the
2959 * checksum is correct. In particular, this function will return zero
2960 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2961 * hardware has already verified the correctness of the checksum.
2962 */
Al Viro4381ca32007-07-15 21:00:11 +01002963static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002964{
Herbert Xu60476372007-04-09 11:59:39 -07002965 return skb_csum_unnecessary(skb) ?
2966 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002967}
2968
Tom Herbert77cffe22014-08-27 21:26:46 -07002969static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
2970{
2971 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2972 if (skb->csum_level == 0)
2973 skb->ip_summed = CHECKSUM_NONE;
2974 else
2975 skb->csum_level--;
2976 }
2977}
2978
2979static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
2980{
2981 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
2982 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
2983 skb->csum_level++;
2984 } else if (skb->ip_summed == CHECKSUM_NONE) {
2985 skb->ip_summed = CHECKSUM_UNNECESSARY;
2986 skb->csum_level = 0;
2987 }
2988}
2989
Tom Herbert5a212322014-08-31 15:12:41 -07002990static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
2991{
2992 /* Mark current checksum as bad (typically called from GRO
2993 * path). In the case that ip_summed is CHECKSUM_NONE
2994 * this must be the first checksum encountered in the packet.
2995 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
2996 * checksum after the last one validated. For UDP, a zero
2997 * checksum can not be marked as bad.
2998 */
2999
3000 if (skb->ip_summed == CHECKSUM_NONE ||
3001 skb->ip_summed == CHECKSUM_UNNECESSARY)
3002 skb->csum_bad = 1;
3003}
3004
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003005/* Check if we need to perform checksum complete validation.
3006 *
3007 * Returns true if checksum complete is needed, false otherwise
3008 * (either checksum is unnecessary or zero checksum is allowed).
3009 */
3010static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3011 bool zero_okay,
3012 __sum16 check)
3013{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003014 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3015 skb->csum_valid = 1;
Tom Herbert77cffe22014-08-27 21:26:46 -07003016 __skb_decr_checksum_unnecessary(skb);
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003017 return false;
3018 }
3019
3020 return true;
3021}
3022
3023/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3024 * in checksum_init.
3025 */
3026#define CHECKSUM_BREAK 76
3027
Tom Herbert4e18b9a2015-04-20 14:10:04 -07003028/* Unset checksum-complete
3029 *
3030 * Unset checksum complete can be done when packet is being modified
3031 * (uncompressed for instance) and checksum-complete value is
3032 * invalidated.
3033 */
3034static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3035{
3036 if (skb->ip_summed == CHECKSUM_COMPLETE)
3037 skb->ip_summed = CHECKSUM_NONE;
3038}
3039
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003040/* Validate (init) checksum based on checksum complete.
3041 *
3042 * Return values:
3043 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3044 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3045 * checksum is stored in skb->csum for use in __skb_checksum_complete
3046 * non-zero: value of invalid checksum
3047 *
3048 */
3049static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3050 bool complete,
3051 __wsum psum)
3052{
3053 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3054 if (!csum_fold(csum_add(psum, skb->csum))) {
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003055 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003056 return 0;
3057 }
Tom Herbert5a212322014-08-31 15:12:41 -07003058 } else if (skb->csum_bad) {
3059 /* ip_summed == CHECKSUM_NONE in this case */
Eric Dumazetc91d4602015-05-15 05:48:07 -07003060 return (__force __sum16)1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003061 }
3062
3063 skb->csum = psum;
3064
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003065 if (complete || skb->len <= CHECKSUM_BREAK) {
3066 __sum16 csum;
3067
3068 csum = __skb_checksum_complete(skb);
3069 skb->csum_valid = !csum;
3070 return csum;
3071 }
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003072
3073 return 0;
3074}
3075
3076static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3077{
3078 return 0;
3079}
3080
3081/* Perform checksum validate (init). Note that this is a macro since we only
3082 * want to calculate the pseudo header which is an input function if necessary.
3083 * First we try to validate without any computation (checksum unnecessary) and
3084 * then calculate based on checksum complete calling the function to compute
3085 * pseudo header.
3086 *
3087 * Return values:
3088 * 0: checksum is validated or try to in skb_checksum_complete
3089 * non-zero: value of invalid checksum
3090 */
3091#define __skb_checksum_validate(skb, proto, complete, \
3092 zero_okay, check, compute_pseudo) \
3093({ \
3094 __sum16 __ret = 0; \
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003095 skb->csum_valid = 0; \
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003096 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3097 __ret = __skb_checksum_validate_complete(skb, \
3098 complete, compute_pseudo(skb, proto)); \
3099 __ret; \
3100})
3101
3102#define skb_checksum_init(skb, proto, compute_pseudo) \
3103 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3104
3105#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3106 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3107
3108#define skb_checksum_validate(skb, proto, compute_pseudo) \
3109 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3110
3111#define skb_checksum_validate_zero_check(skb, proto, check, \
3112 compute_pseudo) \
Sabrina Dubroca096a4cf2015-02-06 18:54:19 +01003113 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003114
3115#define skb_checksum_simple_validate(skb) \
3116 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3117
Tom Herbertd96535a2014-08-31 15:12:42 -07003118static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3119{
3120 return (skb->ip_summed == CHECKSUM_NONE &&
3121 skb->csum_valid && !skb->csum_bad);
3122}
3123
3124static inline void __skb_checksum_convert(struct sk_buff *skb,
3125 __sum16 check, __wsum pseudo)
3126{
3127 skb->csum = ~pseudo;
3128 skb->ip_summed = CHECKSUM_COMPLETE;
3129}
3130
3131#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3132do { \
3133 if (__skb_checksum_convert_check(skb)) \
3134 __skb_checksum_convert(skb, check, \
3135 compute_pseudo(skb, proto)); \
3136} while (0)
3137
Tom Herbert15e23962015-02-10 16:30:31 -08003138static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3139 u16 start, u16 offset)
3140{
3141 skb->ip_summed = CHECKSUM_PARTIAL;
3142 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3143 skb->csum_offset = offset - start;
3144}
3145
Tom Herbertdcdc8992015-02-02 16:07:34 -08003146/* Update skbuf and packet to reflect the remote checksum offload operation.
3147 * When called, ptr indicates the starting point for skb->csum when
3148 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3149 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3150 */
3151static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
Tom Herbert15e23962015-02-10 16:30:31 -08003152 int start, int offset, bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08003153{
3154 __wsum delta;
3155
Tom Herbert15e23962015-02-10 16:30:31 -08003156 if (!nopartial) {
3157 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3158 return;
3159 }
3160
Tom Herbertdcdc8992015-02-02 16:07:34 -08003161 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3162 __skb_checksum_complete(skb);
3163 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3164 }
3165
3166 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3167
3168 /* Adjust skb->csum since we changed the packet */
3169 skb->csum = csum_add(skb->csum, delta);
3170}
3171
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003172#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07003173void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003174static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3175{
3176 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07003177 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003178}
3179static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3180{
3181 if (nfct)
3182 atomic_inc(&nfct->use);
3183}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003184#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003185#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003186static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3187{
3188 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3189 kfree(nf_bridge);
3190}
3191static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3192{
3193 if (nf_bridge)
3194 atomic_inc(&nf_bridge->use);
3195}
3196#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08003197static inline void nf_reset(struct sk_buff *skb)
3198{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003199#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08003200 nf_conntrack_put(skb->nfct);
3201 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003202#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003203#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Patrick McHardya193a4a2006-03-20 19:23:05 -08003204 nf_bridge_put(skb->nf_bridge);
3205 skb->nf_bridge = NULL;
3206#endif
3207}
3208
Patrick McHardy124dff02013-04-05 20:42:05 +02003209static inline void nf_reset_trace(struct sk_buff *skb)
3210{
Florian Westphal478b3602014-02-15 23:48:45 +01003211#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00003212 skb->nf_trace = 0;
3213#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003214}
3215
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003216/* Note: This doesn't put any conntrack and bridge info in dst. */
Eric Dumazetb1937222014-09-28 22:18:47 -07003217static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3218 bool copy)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003219{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003220#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003221 dst->nfct = src->nfct;
3222 nf_conntrack_get(src->nfct);
Eric Dumazetb1937222014-09-28 22:18:47 -07003223 if (copy)
3224 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003225#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003226#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003227 dst->nf_bridge = src->nf_bridge;
3228 nf_bridge_get(src->nf_bridge);
3229#endif
Florian Westphal478b3602014-02-15 23:48:45 +01003230#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Eric Dumazetb1937222014-09-28 22:18:47 -07003231 if (copy)
3232 dst->nf_trace = src->nf_trace;
Florian Westphal478b3602014-02-15 23:48:45 +01003233#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003234}
3235
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003236static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3237{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003238#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003239 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003240#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003241#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003242 nf_bridge_put(dst->nf_bridge);
3243#endif
Eric Dumazetb1937222014-09-28 22:18:47 -07003244 __nf_copy(dst, src, true);
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003245}
3246
James Morris984bc162006-06-09 00:29:17 -07003247#ifdef CONFIG_NETWORK_SECMARK
3248static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3249{
3250 to->secmark = from->secmark;
3251}
3252
3253static inline void skb_init_secmark(struct sk_buff *skb)
3254{
3255 skb->secmark = 0;
3256}
3257#else
3258static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3259{ }
3260
3261static inline void skb_init_secmark(struct sk_buff *skb)
3262{ }
3263#endif
3264
Eric W. Biederman574f7192014-04-01 12:20:24 -07003265static inline bool skb_irq_freeable(const struct sk_buff *skb)
3266{
3267 return !skb->destructor &&
3268#if IS_ENABLED(CONFIG_XFRM)
3269 !skb->sp &&
3270#endif
3271#if IS_ENABLED(CONFIG_NF_CONNTRACK)
3272 !skb->nfct &&
3273#endif
3274 !skb->_skb_refdst &&
3275 !skb_has_frag_list(skb);
3276}
3277
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003278static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3279{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003280 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003281}
3282
Stephen Hemminger92477442009-03-21 13:39:26 -07003283static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003284{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003285 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003286}
3287
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003288static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3289{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003290 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003291}
3292
David S. Millerd5a9e242009-01-27 16:22:11 -08003293static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3294{
3295 skb->queue_mapping = rx_queue + 1;
3296}
3297
Stephen Hemminger92477442009-03-21 13:39:26 -07003298static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003299{
3300 return skb->queue_mapping - 1;
3301}
3302
Stephen Hemminger92477442009-03-21 13:39:26 -07003303static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003304{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003305 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08003306}
3307
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003308static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3309{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003310#ifdef CONFIG_XFRM
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003311 return skb->sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003312#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003313 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003314#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003315}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003316
Pravin B Shelar68c33162013-02-14 14:02:41 +00003317/* Keeps track of mac header offset relative to skb->head.
3318 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3319 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07003320 * tunnel skb it points to outer mac header.
3321 * Keeps track of level of encapsulation of network headers.
3322 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00003323struct skb_gso_cb {
Eric Dumazet3347c962013-10-19 11:42:56 -07003324 int mac_offset;
3325 int encap_level;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003326 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00003327};
3328#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3329
3330static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3331{
3332 return (skb_mac_header(inner_skb) - inner_skb->head) -
3333 SKB_GSO_CB(inner_skb)->mac_offset;
3334}
3335
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00003336static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3337{
3338 int new_headroom, headroom;
3339 int ret;
3340
3341 headroom = skb_headroom(skb);
3342 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3343 if (ret)
3344 return ret;
3345
3346 new_headroom = skb_headroom(skb);
3347 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3348 return 0;
3349}
3350
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003351/* Compute the checksum for a gso segment. First compute the checksum value
3352 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3353 * then add in skb->csum (checksum from csum_start to end of packet).
3354 * skb->csum and csum_start are then updated to reflect the checksum of the
3355 * resultant packet starting from the transport header-- the resultant checksum
3356 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3357 * header.
3358 */
3359static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3360{
3361 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
Eric Dumazetc91d4602015-05-15 05:48:07 -07003362 skb_transport_offset(skb);
3363 __wsum partial;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003364
Eric Dumazetc91d4602015-05-15 05:48:07 -07003365 partial = csum_partial(skb_transport_header(skb), plen, skb->csum);
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003366 skb->csum = res;
3367 SKB_GSO_CB(skb)->csum_start -= plen;
3368
Eric Dumazetc91d4602015-05-15 05:48:07 -07003369 return csum_fold(partial);
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003370}
3371
David S. Millerbdcc0922012-03-07 20:53:36 -05003372static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07003373{
3374 return skb_shinfo(skb)->gso_size;
3375}
3376
Eric Dumazet36a8f392013-09-29 01:21:32 -07003377/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05003378static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02003379{
3380 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3381}
3382
Joe Perches7965bd42013-09-26 14:48:15 -07003383void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07003384
3385static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3386{
3387 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3388 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003389 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3390
Alexander Duyckb78462e2010-06-02 12:24:37 +00003391 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3392 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07003393 __skb_warn_lro_forwarding(skb);
3394 return true;
3395 }
3396 return false;
3397}
3398
Herbert Xu35fc92a2007-03-26 23:22:20 -07003399static inline void skb_forward_csum(struct sk_buff *skb)
3400{
3401 /* Unfortunately we don't support this one. Any brave souls? */
3402 if (skb->ip_summed == CHECKSUM_COMPLETE)
3403 skb->ip_summed = CHECKSUM_NONE;
3404}
3405
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003406/**
3407 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3408 * @skb: skb to check
3409 *
3410 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3411 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3412 * use this helper, to document places where we make this assertion.
3413 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003414static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003415{
3416#ifdef DEBUG
3417 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3418#endif
3419}
3420
Rusty Russellf35d9d82008-02-04 23:49:54 -05003421bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00003422
Paul Durranted1f50c2014-01-09 10:02:46 +00003423int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
Linus Lüssing9afd85c2015-05-02 14:01:07 +02003424struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
3425 unsigned int transport_len,
3426 __sum16(*skb_chkf)(struct sk_buff *skb));
Paul Durranted1f50c2014-01-09 10:02:46 +00003427
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00003428/**
3429 * skb_head_is_locked - Determine if the skb->head is locked down
3430 * @skb: skb to check
3431 *
3432 * The head on skbs build around a head frag can be removed if they are
3433 * not cloned. This function returns true if the skb head is locked down
3434 * due to either being allocated via kmalloc, or by being a clone with
3435 * multiple references to the head.
3436 */
3437static inline bool skb_head_is_locked(const struct sk_buff *skb)
3438{
3439 return !skb->head_frag || skb_cloned(skb);
3440}
Florian Westphalfe6cc552014-02-13 23:09:12 +01003441
3442/**
3443 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3444 *
3445 * @skb: GSO skb
3446 *
3447 * skb_gso_network_seglen is used to determine the real size of the
3448 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3449 *
3450 * The MAC/L2 header is not accounted for.
3451 */
3452static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3453{
3454 unsigned int hdr_len = skb_transport_header(skb) -
3455 skb_network_header(skb);
3456 return hdr_len + skb_gso_transport_seglen(skb);
3457}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003458#endif /* __KERNEL__ */
3459#endif /* _LINUX_SKBUFF_H */