blob: b7c1286e247d9c50951ad8d9d1256ccc9d691481 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * Definitions for the 'struct sk_buff' memory handlers.
3 *
4 * Authors:
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14#ifndef _LINUX_SKBUFF_H
15#define _LINUX_SKBUFF_H
16
Linus Torvalds1da177e2005-04-16 15:20:36 -070017#include <linux/kernel.h>
Vegard Nossumfe55f6d2008-08-30 12:16:35 +020018#include <linux/kmemcheck.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070019#include <linux/compiler.h>
20#include <linux/time.h>
Paul Gortmaker187f1882011-11-23 20:12:59 -050021#include <linux/bug.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070022#include <linux/cache.h>
Eric Dumazet56b17422014-11-03 08:19:53 -080023#include <linux/rbtree.h>
David S. Miller51f3d022014-11-05 16:46:40 -050024#include <linux/socket.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025
Arun Sharma600634972011-07-26 16:09:06 -070026#include <linux/atomic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <asm/types.h>
28#include <linux/spinlock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070029#include <linux/net.h>
Thomas Graf3fc7e8a2005-06-23 21:00:17 -070030#include <linux/textsearch.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070031#include <net/checksum.h>
Al Viroa80958f2006-12-04 20:41:19 +000032#include <linux/rcupdate.h>
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -070033#include <linux/hrtimer.h>
Ian Campbell131ea662011-08-19 06:25:00 +000034#include <linux/dma-mapping.h>
Michał Mirosławc8f44af2011-11-15 15:29:55 +000035#include <linux/netdev_features.h>
Eric Dumazet363ec39232014-02-26 14:02:11 -080036#include <linux/sched.h>
Jiri Pirko1bd758e2015-05-12 14:56:07 +020037#include <net/flow_dissector.h>
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +020038#include <linux/splice.h>
Bernhard Thaler72b31f72015-05-30 15:27:40 +020039#include <linux/in6.h>
Tom Herbertf70ea012015-07-31 16:52:10 -070040#include <net/flow.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070041
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010042/* A. Checksumming of received packets by device.
43 *
44 * CHECKSUM_NONE:
45 *
46 * Device failed to checksum this packet e.g. due to lack of capabilities.
47 * The packet contains full (though not verified) checksum in packet but
48 * not in skb->csum. Thus, skb->csum is undefined in this case.
49 *
50 * CHECKSUM_UNNECESSARY:
51 *
52 * The hardware you're dealing with doesn't calculate the full checksum
53 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
Tom Herbert77cffe22014-08-27 21:26:46 -070054 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
55 * if their checksums are okay. skb->csum is still undefined in this case
56 * though. It is a bad option, but, unfortunately, nowadays most vendors do
57 * this. Apparently with the secret goal to sell you new devices, when you
58 * will add new protocol to your host, f.e. IPv6 8)
59 *
60 * CHECKSUM_UNNECESSARY is applicable to following protocols:
61 * TCP: IPv6 and IPv4.
62 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
63 * zero UDP checksum for either IPv4 or IPv6, the networking stack
64 * may perform further validation in this case.
65 * GRE: only if the checksum is present in the header.
66 * SCTP: indicates the CRC in SCTP header has been validated.
67 *
68 * skb->csum_level indicates the number of consecutive checksums found in
69 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
70 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
71 * and a device is able to verify the checksums for UDP (possibly zero),
72 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
73 * two. If the device were only able to verify the UDP checksum and not
74 * GRE, either because it doesn't support GRE checksum of because GRE
75 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
76 * not considered in this case).
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010077 *
78 * CHECKSUM_COMPLETE:
79 *
80 * This is the most generic way. The device supplied checksum of the _whole_
81 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
82 * hardware doesn't need to parse L3/L4 headers to implement this.
83 *
84 * Note: Even if device supports only some protocols, but is able to produce
85 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
86 *
87 * CHECKSUM_PARTIAL:
88 *
Tom Herbert6edec0e2015-02-10 16:30:28 -080089 * A checksum is set up to be offloaded to a device as described in the
90 * output description for CHECKSUM_PARTIAL. This may occur on a packet
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010091 * received directly from another Linux OS, e.g., a virtualized Linux kernel
Tom Herbert6edec0e2015-02-10 16:30:28 -080092 * on the same host, or it may be set in the input path in GRO or remote
93 * checksum offload. For the purposes of checksum verification, the checksum
94 * referred to by skb->csum_start + skb->csum_offset and any preceding
95 * checksums in the packet are considered verified. Any checksums in the
96 * packet that are after the checksum being offloaded are not considered to
97 * be verified.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010098 *
99 * B. Checksumming on output.
100 *
101 * CHECKSUM_NONE:
102 *
103 * The skb was already checksummed by the protocol, or a checksum is not
104 * required.
105 *
106 * CHECKSUM_PARTIAL:
107 *
108 * The device is required to checksum the packet as seen by hard_start_xmit()
109 * from skb->csum_start up to the end, and to record/write the checksum at
110 * offset skb->csum_start + skb->csum_offset.
111 *
112 * The device must show its capabilities in dev->features, set up at device
113 * setup time, e.g. netdev_features.h:
114 *
115 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
116 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
117 * IPv4. Sigh. Vendors like this way for an unknown reason.
118 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
119 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
120 * NETIF_F_... - Well, you get the picture.
121 *
122 * CHECKSUM_UNNECESSARY:
123 *
124 * Normally, the device will do per protocol specific checksumming. Protocol
125 * implementations that do not want the NIC to perform the checksum
126 * calculation should use this flag in their outgoing skbs.
127 *
128 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
129 * offload. Correspondingly, the FCoE protocol driver
130 * stack should use CHECKSUM_UNNECESSARY.
131 *
132 * Any questions? No questions, good. --ANK
133 */
134
Herbert Xu60476372007-04-09 11:59:39 -0700135/* Don't change this without changing skb_csum_unnecessary! */
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100136#define CHECKSUM_NONE 0
137#define CHECKSUM_UNNECESSARY 1
138#define CHECKSUM_COMPLETE 2
139#define CHECKSUM_PARTIAL 3
Linus Torvalds1da177e2005-04-16 15:20:36 -0700140
Tom Herbert77cffe22014-08-27 21:26:46 -0700141/* Maximum value in skb->csum_level */
142#define SKB_MAX_CSUM_LEVEL 3
143
Tobias Klauser0bec8c82014-07-22 12:06:23 +0200144#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
David S. Millerfc910a22007-03-25 20:27:59 -0700145#define SKB_WITH_OVERHEAD(X) \
Herbert Xudeea84b2007-10-21 16:27:46 -0700146 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
David S. Millerfc910a22007-03-25 20:27:59 -0700147#define SKB_MAX_ORDER(X, ORDER) \
148 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
150#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
151
Eric Dumazet87fb4b72011-10-13 07:28:54 +0000152/* return minimum truesize of one skb containing X bytes of data */
153#define SKB_TRUESIZE(X) ((X) + \
154 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
155 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
156
Linus Torvalds1da177e2005-04-16 15:20:36 -0700157struct net_device;
David Howells716ea3a2007-04-02 20:19:53 -0700158struct scatterlist;
Jens Axboe9c55e012007-11-06 23:30:13 -0800159struct pipe_inode_info;
Herbert Xua8f820aa2014-11-07 21:22:22 +0800160struct iov_iter;
Alexander Duyckfd11a832014-12-09 19:40:49 -0800161struct napi_struct;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700162
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700163#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700164struct nf_conntrack {
165 atomic_t use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700166};
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700167#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700168
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +0200169#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700170struct nf_bridge_info {
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000171 atomic_t use;
Florian Westphal3eaf4022015-04-02 14:31:44 +0200172 enum {
173 BRNF_PROTO_UNCHANGED,
174 BRNF_PROTO_8021Q,
175 BRNF_PROTO_PPPOE
Florian Westphal7fb48c52015-05-03 22:05:28 +0200176 } orig_proto:8;
Florian Westphala1e67952015-04-02 14:31:45 +0200177 bool pkt_otherhost;
Bernhard Thaler411ffb42015-05-30 15:28:28 +0200178 __u16 frag_max_size;
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000179 unsigned int mask;
180 struct net_device *physindev;
Florian Westphal7fb48c52015-05-03 22:05:28 +0200181 union {
182 struct net_device *physoutdev;
183 char neigh_header[8];
184 };
Bernhard Thaler72b31f72015-05-30 15:27:40 +0200185 union {
186 __be32 ipv4_daddr;
187 struct in6_addr ipv6_daddr;
188 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700189};
190#endif
191
Linus Torvalds1da177e2005-04-16 15:20:36 -0700192struct sk_buff_head {
193 /* These two members must be first. */
194 struct sk_buff *next;
195 struct sk_buff *prev;
196
197 __u32 qlen;
198 spinlock_t lock;
199};
200
201struct sk_buff;
202
Ian Campbell9d4dde52011-12-22 23:39:14 +0000203/* To allow 64K frame to be packed as single skb without frag_list we
204 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
205 * buffers which do not start on a page boundary.
206 *
207 * Since GRO uses frags we allocate at least 16 regardless of page
208 * size.
Anton Blancharda715dea2011-03-27 14:57:26 +0000209 */
Ian Campbell9d4dde52011-12-22 23:39:14 +0000210#if (65536/PAGE_SIZE + 1) < 16
David S. Millereec00952011-03-29 23:34:08 -0700211#define MAX_SKB_FRAGS 16UL
Anton Blancharda715dea2011-03-27 14:57:26 +0000212#else
Ian Campbell9d4dde52011-12-22 23:39:14 +0000213#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
Anton Blancharda715dea2011-03-27 14:57:26 +0000214#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700215
216typedef struct skb_frag_struct skb_frag_t;
217
218struct skb_frag_struct {
Ian Campbella8605c62011-10-19 23:01:49 +0000219 struct {
220 struct page *p;
221 } page;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000222#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
David S. Millera309bb02007-07-30 18:47:03 -0700223 __u32 page_offset;
224 __u32 size;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000225#else
226 __u16 page_offset;
227 __u16 size;
228#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229};
230
Eric Dumazet9e903e02011-10-18 21:00:24 +0000231static inline unsigned int skb_frag_size(const skb_frag_t *frag)
232{
233 return frag->size;
234}
235
236static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
237{
238 frag->size = size;
239}
240
241static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
242{
243 frag->size += delta;
244}
245
246static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
247{
248 frag->size -= delta;
249}
250
Patrick Ohlyac45f602009-02-12 05:03:37 +0000251#define HAVE_HW_TIME_STAMP
252
253/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800254 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000255 * @hwtstamp: hardware time stamp transformed into duration
256 * since arbitrary point in time
Patrick Ohlyac45f602009-02-12 05:03:37 +0000257 *
258 * Software time stamps generated by ktime_get_real() are stored in
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400259 * skb->tstamp.
Patrick Ohlyac45f602009-02-12 05:03:37 +0000260 *
261 * hwtstamps can only be compared against other hwtstamps from
262 * the same device.
263 *
264 * This structure is attached to packets as part of the
265 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
266 */
267struct skb_shared_hwtstamps {
268 ktime_t hwtstamp;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000269};
270
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000271/* Definitions for tx_flags in struct skb_shared_info */
272enum {
273 /* generate hardware time stamp */
274 SKBTX_HW_TSTAMP = 1 << 0,
275
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400276 /* generate software time stamp when queueing packet to NIC */
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000277 SKBTX_SW_TSTAMP = 1 << 1,
278
279 /* device driver is going to provide hardware time stamp */
280 SKBTX_IN_PROGRESS = 1 << 2,
281
Shirley Maa6686f22011-07-06 12:22:12 +0000282 /* device driver supports TX zero-copy buffers */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000283 SKBTX_DEV_ZEROCOPY = 1 << 3,
Johannes Berg6e3e9392011-11-09 10:15:42 +0100284
285 /* generate wifi status information (where possible) */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000286 SKBTX_WIFI_STATUS = 1 << 4,
Pravin B Shelarc9af6db2013-02-11 09:27:41 +0000287
288 /* This indicates at least one fragment might be overwritten
289 * (as in vmsplice(), sendfile() ...)
290 * If we need to compute a TX checksum, we'll need to copy
291 * all frags to avoid possible bad checksum
292 */
293 SKBTX_SHARED_FRAG = 1 << 5,
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400294
295 /* generate software time stamp when entering packet scheduling */
296 SKBTX_SCHED_TSTAMP = 1 << 6,
Willem de Bruijne1c8a602014-08-04 22:11:50 -0400297
298 /* generate software timestamp on peer data acknowledgment */
299 SKBTX_ACK_TSTAMP = 1 << 7,
Shirley Maa6686f22011-07-06 12:22:12 +0000300};
301
Willem de Bruijne1c8a602014-08-04 22:11:50 -0400302#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
303 SKBTX_SCHED_TSTAMP | \
304 SKBTX_ACK_TSTAMP)
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400305#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
306
Shirley Maa6686f22011-07-06 12:22:12 +0000307/*
308 * The callback notifies userspace to release buffers when skb DMA is done in
309 * lower device, the skb last reference should be 0 when calling this.
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000310 * The zerocopy_success argument is true if zero copy transmit occurred,
311 * false on data copy or out of memory error caused by data copy attempt.
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000312 * The ctx field is used to track device context.
313 * The desc field is used to track userspace buffer index.
Shirley Maa6686f22011-07-06 12:22:12 +0000314 */
315struct ubuf_info {
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000316 void (*callback)(struct ubuf_info *, bool zerocopy_success);
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000317 void *ctx;
Shirley Maa6686f22011-07-06 12:22:12 +0000318 unsigned long desc;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000319};
320
Linus Torvalds1da177e2005-04-16 15:20:36 -0700321/* This data is invariant across clones and lives at
322 * the end of the header data, ie. at skb->end.
323 */
324struct skb_shared_info {
Ian Campbell9f42f122012-01-05 07:13:39 +0000325 unsigned char nr_frags;
326 __u8 tx_flags;
Herbert Xu79671682006-06-22 02:40:14 -0700327 unsigned short gso_size;
328 /* Warning: this field is not always filled in (UFO)! */
329 unsigned short gso_segs;
330 unsigned short gso_type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000332 struct skb_shared_hwtstamps hwtstamps;
Willem de Bruijn09c2d252014-08-04 22:11:47 -0400333 u32 tskey;
Ian Campbell9f42f122012-01-05 07:13:39 +0000334 __be32 ip6_frag_id;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700335
336 /*
337 * Warning : all fields before dataref are cleared in __alloc_skb()
338 */
339 atomic_t dataref;
340
Johann Baudy69e3c752009-05-18 22:11:22 -0700341 /* Intermediate layers must ensure that destructor_arg
342 * remains valid until skb destructor */
343 void * destructor_arg;
Shirley Maa6686f22011-07-06 12:22:12 +0000344
Eric Dumazetfed66382010-07-22 19:09:08 +0000345 /* must be last field, see pskb_expand_head() */
346 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700347};
348
349/* We divide dataref into two halves. The higher 16 bits hold references
350 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700351 * the entire skb->data. A clone of a headerless skb holds the length of
352 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700353 *
354 * All users must obey the rule that the skb->data reference count must be
355 * greater than or equal to the payload reference count.
356 *
357 * Holding a reference to the payload part means that the user does not
358 * care about modifications to the header part of skb->data.
359 */
360#define SKB_DATAREF_SHIFT 16
361#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
362
David S. Millerd179cd12005-08-17 14:57:30 -0700363
364enum {
Vijay Subramanianc8753d52014-10-02 10:00:43 -0700365 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
366 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
367 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
David S. Millerd179cd12005-08-17 14:57:30 -0700368};
369
Herbert Xu79671682006-06-22 02:40:14 -0700370enum {
371 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700372 SKB_GSO_UDP = 1 << 1,
Herbert Xu576a30e2006-06-27 13:22:38 -0700373
374 /* This indicates the skb is from an untrusted source. */
375 SKB_GSO_DODGY = 1 << 2,
Michael Chanb0da85372006-06-29 12:30:00 -0700376
377 /* This indicates the tcp segment has CWR set. */
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700378 SKB_GSO_TCP_ECN = 1 << 3,
379
380 SKB_GSO_TCPV6 = 1 << 4,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800381
382 SKB_GSO_FCOE = 1 << 5,
Pravin B Shelar68c33162013-02-14 14:02:41 +0000383
384 SKB_GSO_GRE = 1 << 6,
Pravin B Shelar73136262013-03-07 13:21:51 +0000385
Tom Herbert4b282522014-06-14 23:23:52 -0700386 SKB_GSO_GRE_CSUM = 1 << 7,
Simon Horman0d89d202013-05-23 21:02:52 +0000387
Tom Herbert4b282522014-06-14 23:23:52 -0700388 SKB_GSO_IPIP = 1 << 8,
Eric Dumazetcb32f512013-10-19 11:42:57 -0700389
Tom Herbert4b282522014-06-14 23:23:52 -0700390 SKB_GSO_SIT = 1 << 9,
Eric Dumazet61c1db72013-10-20 20:47:30 -0700391
Tom Herbert4b282522014-06-14 23:23:52 -0700392 SKB_GSO_UDP_TUNNEL = 1 << 10,
Tom Herbert0f4f4ff2014-06-04 17:20:16 -0700393
394 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
Tom Herbert4749c092014-06-04 17:20:23 -0700395
Pravin B Shelar59b93b42014-11-05 15:27:48 -0800396 SKB_GSO_TUNNEL_REMCSUM = 1 << 12,
Herbert Xu79671682006-06-22 02:40:14 -0700397};
398
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700399#if BITS_PER_LONG > 32
400#define NET_SKBUFF_DATA_USES_OFFSET 1
401#endif
402
403#ifdef NET_SKBUFF_DATA_USES_OFFSET
404typedef unsigned int sk_buff_data_t;
405#else
406typedef unsigned char *sk_buff_data_t;
407#endif
408
Eric Dumazet363ec39232014-02-26 14:02:11 -0800409/**
410 * struct skb_mstamp - multi resolution time stamps
411 * @stamp_us: timestamp in us resolution
412 * @stamp_jiffies: timestamp in jiffies
413 */
414struct skb_mstamp {
415 union {
416 u64 v64;
417 struct {
418 u32 stamp_us;
419 u32 stamp_jiffies;
420 };
421 };
422};
423
424/**
425 * skb_mstamp_get - get current timestamp
426 * @cl: place to store timestamps
427 */
428static inline void skb_mstamp_get(struct skb_mstamp *cl)
429{
430 u64 val = local_clock();
431
432 do_div(val, NSEC_PER_USEC);
433 cl->stamp_us = (u32)val;
434 cl->stamp_jiffies = (u32)jiffies;
435}
436
437/**
438 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
439 * @t1: pointer to newest sample
440 * @t0: pointer to oldest sample
441 */
442static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
443 const struct skb_mstamp *t0)
444{
445 s32 delta_us = t1->stamp_us - t0->stamp_us;
446 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
447
448 /* If delta_us is negative, this might be because interval is too big,
449 * or local_clock() drift is too big : fallback using jiffies.
450 */
451 if (delta_us <= 0 ||
452 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
453
454 delta_us = jiffies_to_usecs(delta_jiffies);
455
456 return delta_us;
457}
458
459
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460/**
461 * struct sk_buff - socket buffer
462 * @next: Next buffer in list
463 * @prev: Previous buffer in list
Eric Dumazet363ec39232014-02-26 14:02:11 -0800464 * @tstamp: Time we arrived/left
Eric Dumazet56b17422014-11-03 08:19:53 -0800465 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700466 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700467 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700468 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000469 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700470 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700471 * @len: Length of actual data
472 * @data_len: Data length
473 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700474 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700475 * @csum: Checksum (must include start/offset pair)
476 * @csum_start: Offset from skb->head where checksumming should start
477 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700478 * @priority: Packet queueing priority
WANG Cong60ff7462014-05-04 16:39:18 -0700479 * @ignore_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700480 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700481 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482 * @nohdr: Payload reference only, must not modify header
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700483 * @nfctinfo: Relationship of this skb to the connection
Linus Torvalds1da177e2005-04-16 15:20:36 -0700484 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700485 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700486 * @ipvs_property: skbuff is owned by ipvs
Randy Dunlap31729362008-02-18 20:52:13 -0800487 * @peeked: this packet has been seen already, so stats have been
488 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700489 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700490 * @protocol: Packet protocol from driver
491 * @destructor: Destruct function
492 * @nfct: Associated connection, if any
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800494 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495 * @tc_index: Traffic control index
496 * @tc_verd: traffic control verdict
Tom Herbert61b905d2014-03-24 15:34:47 -0700497 * @hash: the packet hash
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700498 * @queue_mapping: Queue mapping for multiqueue devices
David S. Miller0b725a22014-08-25 15:51:53 -0700499 * @xmit_more: More SKBs are pending for this queue
Randy Dunlap553a5672008-04-20 10:51:01 -0700500 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700501 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Tom Herbert61b905d2014-03-24 15:34:47 -0700502 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
Changli Gao4ca24622011-08-19 07:26:44 -0700503 * ports.
Tom Herberta3b18dd2014-07-01 21:33:17 -0700504 * @sw_hash: indicates hash was computed in software stack
Johannes Berg6e3e9392011-11-09 10:15:42 +0100505 * @wifi_acked_valid: wifi_acked was set
506 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000507 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Eliezer Tamir06021292013-06-10 11:39:50 +0300508 * @napi_id: id of the NAPI struct this skb came from
James Morris984bc162006-06-09 00:29:17 -0700509 * @secmark: security marking
Scott Feldman0c4f6912015-07-18 18:24:48 -0700510 * @offload_fwd_mark: fwding offload mark
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700511 * @mark: Generic packet mark
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000512 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b2008-07-14 22:49:06 -0700513 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000514 * @inner_protocol: Protocol (encapsulation)
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000515 * @inner_transport_header: Inner transport layer header (encapsulation)
516 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000517 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700518 * @transport_header: Transport layer header
519 * @network_header: Network layer header
520 * @mac_header: Link layer header
521 * @tail: Tail pointer
522 * @end: End pointer
523 * @head: Head of buffer
524 * @data: Data head pointer
525 * @truesize: Buffer size
526 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 */
528
529struct sk_buff {
Eric Dumazet363ec39232014-02-26 14:02:11 -0800530 union {
Eric Dumazet56b17422014-11-03 08:19:53 -0800531 struct {
532 /* These two members must be first. */
533 struct sk_buff *next;
534 struct sk_buff *prev;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000535
Eric Dumazet56b17422014-11-03 08:19:53 -0800536 union {
537 ktime_t tstamp;
538 struct skb_mstamp skb_mstamp;
539 };
540 };
541 struct rb_node rbnode; /* used in netem & tcp stack */
542 };
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000543 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700544 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546 /*
547 * This is the control buffer. It is free to use for every
548 * layer. Please put your private variables there. If you
549 * want to keep them across layers you have to do a skb_clone()
550 * first. This is owned by whoever has the skb queued ATM.
551 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000552 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553
Eric Dumazet7fee2262010-05-11 23:19:48 +0000554 unsigned long _skb_refdst;
Eric Dumazetb1937222014-09-28 22:18:47 -0700555 void (*destructor)(struct sk_buff *skb);
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000556#ifdef CONFIG_XFRM
557 struct sec_path *sp;
558#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700559#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
560 struct nf_conntrack *nfct;
561#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +0200562#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Eric Dumazetb1937222014-09-28 22:18:47 -0700563 struct nf_bridge_info *nf_bridge;
564#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700566 data_len;
567 __u16 mac_len,
568 hdr_len;
Eric Dumazetb1937222014-09-28 22:18:47 -0700569
570 /* Following fields are _not_ copied in __copy_skb_header()
571 * Note that queue_mapping is here mostly to fill a hole.
572 */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200573 kmemcheck_bitfield_begin(flags1);
Eric Dumazetb1937222014-09-28 22:18:47 -0700574 __u16 queue_mapping;
575 __u8 cloned:1,
Harald Welte6869c4d2005-08-09 19:24:19 -0700576 nohdr:1,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800577 fclone:2,
Herbert Xua59322b2007-12-05 01:53:40 -0800578 peeked:1,
Eric Dumazetb1937222014-09-28 22:18:47 -0700579 head_frag:1,
580 xmit_more:1;
581 /* one bit hole */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200582 kmemcheck_bitfield_end(flags1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700583
Eric Dumazetb1937222014-09-28 22:18:47 -0700584 /* fields enclosed in headers_start/headers_end are copied
585 * using a single memcpy() in __copy_skb_header()
586 */
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700587 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700588 __u32 headers_start[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700589 /* public: */
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200590
591/* if you move pkt_type around you also must adapt those constants */
592#ifdef __BIG_ENDIAN_BITFIELD
593#define PKT_TYPE_MAX (7 << 5)
594#else
595#define PKT_TYPE_MAX 7
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100596#endif
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200597#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
598
599 __u8 __pkt_type_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700600 __u8 pkt_type:3;
601 __u8 pfmemalloc:1;
602 __u8 ignore_df:1;
603 __u8 nfctinfo:3;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604
Eric Dumazetb1937222014-09-28 22:18:47 -0700605 __u8 nf_trace:1;
606 __u8 ip_summed:2;
607 __u8 ooo_okay:1;
608 __u8 l4_hash:1;
609 __u8 sw_hash:1;
610 __u8 wifi_acked_valid:1;
611 __u8 wifi_acked:1;
612
613 __u8 no_fcs:1;
614 /* Indicates the inner headers are valid in the skbuff. */
615 __u8 encapsulation:1;
616 __u8 encap_hdr_csum:1;
617 __u8 csum_valid:1;
618 __u8 csum_complete_sw:1;
619 __u8 csum_level:2;
620 __u8 csum_bad:1;
621
622#ifdef CONFIG_IPV6_NDISC_NODETYPE
623 __u8 ndisc_nodetype:2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700624#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700625 __u8 ipvs_property:1;
Tom Herbert8bce6d72014-09-29 20:22:29 -0700626 __u8 inner_protocol_type:1;
Tom Herberte585f232014-11-04 09:06:54 -0800627 __u8 remcsum_offload:1;
628 /* 3 or 5 bit hole */
Alexander Duyck4031ae62012-01-27 06:22:53 +0000629
Linus Torvalds1da177e2005-04-16 15:20:36 -0700630#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700631 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700633 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200636
Eric Dumazetb1937222014-09-28 22:18:47 -0700637 union {
638 __wsum csum;
639 struct {
640 __u16 csum_start;
641 __u16 csum_offset;
642 };
643 };
644 __u32 priority;
645 int skb_iif;
646 __u32 hash;
647 __be16 vlan_proto;
648 __u16 vlan_tci;
Eric Dumazet2bd82482015-02-03 23:48:24 -0800649#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
650 union {
651 unsigned int napi_id;
652 unsigned int sender_cpu;
653 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700654#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -0700655 union {
James Morris984bc162006-06-09 00:29:17 -0700656#ifdef CONFIG_NETWORK_SECMARK
Scott Feldman0c4f6912015-07-18 18:24:48 -0700657 __u32 secmark;
James Morris984bc162006-06-09 00:29:17 -0700658#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -0700659#ifdef CONFIG_NET_SWITCHDEV
660 __u32 offload_fwd_mark;
661#endif
662 };
663
Neil Horman3b885782009-10-12 13:26:31 -0700664 union {
665 __u32 mark;
Eric Dumazet16fad692013-03-14 05:40:32 +0000666 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700667 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700668
Tom Herbert8bce6d72014-09-29 20:22:29 -0700669 union {
670 __be16 inner_protocol;
671 __u8 inner_ipproto;
672 };
673
Simon Horman1a37e412013-05-23 21:02:51 +0000674 __u16 inner_transport_header;
675 __u16 inner_network_header;
676 __u16 inner_mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700677
678 __be16 protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000679 __u16 transport_header;
680 __u16 network_header;
681 __u16 mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700682
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700683 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700684 __u32 headers_end[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700685 /* public: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700686
Linus Torvalds1da177e2005-04-16 15:20:36 -0700687 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700688 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700689 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700690 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700691 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700692 unsigned int truesize;
693 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700694};
695
696#ifdef __KERNEL__
697/*
698 * Handling routines are only of interest to the kernel
699 */
700#include <linux/slab.h>
701
Linus Torvalds1da177e2005-04-16 15:20:36 -0700702
Mel Gormanc93bdd02012-07-31 16:44:19 -0700703#define SKB_ALLOC_FCLONE 0x01
704#define SKB_ALLOC_RX 0x02
Alexander Duyckfd11a832014-12-09 19:40:49 -0800705#define SKB_ALLOC_NAPI 0x04
Mel Gormanc93bdd02012-07-31 16:44:19 -0700706
707/* Returns true if the skb was allocated from PFMEMALLOC reserves */
708static inline bool skb_pfmemalloc(const struct sk_buff *skb)
709{
710 return unlikely(skb->pfmemalloc);
711}
712
Eric Dumazet7fee2262010-05-11 23:19:48 +0000713/*
714 * skb might have a dst pointer attached, refcounted or not.
715 * _skb_refdst low order bit is set if refcount was _not_ taken
716 */
717#define SKB_DST_NOREF 1UL
718#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
719
720/**
721 * skb_dst - returns skb dst_entry
722 * @skb: buffer
723 *
724 * Returns skb dst_entry, regardless of reference taken or not.
725 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000726static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
727{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000728 /* If refdst was not refcounted, check we still are in a
729 * rcu_read_lock section
730 */
731 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
732 !rcu_read_lock_held() &&
733 !rcu_read_lock_bh_held());
734 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000735}
736
Eric Dumazet7fee2262010-05-11 23:19:48 +0000737/**
738 * skb_dst_set - sets skb dst
739 * @skb: buffer
740 * @dst: dst entry
741 *
742 * Sets skb dst, assuming a reference was taken on dst and should
743 * be released by skb_dst_drop()
744 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000745static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
746{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000747 skb->_skb_refdst = (unsigned long)dst;
748}
749
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200750/**
751 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
752 * @skb: buffer
753 * @dst: dst entry
754 *
755 * Sets skb dst, assuming a reference was not taken on dst.
756 * If dst entry is cached, we do not take reference and dst_release
757 * will be avoided by refdst_drop. If dst entry is not cached, we take
758 * reference, so that last dst_release can destroy the dst immediately.
759 */
760static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
761{
Hannes Frederic Sowadbfc4fb2014-12-06 19:19:42 +0100762 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
763 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200764}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000765
766/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300767 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000768 * @skb: buffer
769 */
770static inline bool skb_dst_is_noref(const struct sk_buff *skb)
771{
772 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000773}
774
Eric Dumazet511c3f92009-06-02 05:14:27 +0000775static inline struct rtable *skb_rtable(const struct sk_buff *skb)
776{
Eric Dumazetadf30902009-06-02 05:19:30 +0000777 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000778}
779
Joe Perches7965bd42013-09-26 14:48:15 -0700780void kfree_skb(struct sk_buff *skb);
781void kfree_skb_list(struct sk_buff *segs);
782void skb_tx_error(struct sk_buff *skb);
783void consume_skb(struct sk_buff *skb);
784void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000785extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000786
Joe Perches7965bd42013-09-26 14:48:15 -0700787void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
788bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
789 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000790
Joe Perches7965bd42013-09-26 14:48:15 -0700791struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
792 int node);
Eric Dumazet2ea2f622015-04-24 16:05:01 -0700793struct sk_buff *__build_skb(void *data, unsigned int frag_size);
Joe Perches7965bd42013-09-26 14:48:15 -0700794struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700795static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100796 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700797{
Eric Dumazet564824b2010-10-11 19:05:25 +0000798 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700799}
800
Eric Dumazet2e4e4412014-09-17 04:49:49 -0700801struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
802 unsigned long data_len,
803 int max_page_order,
804 int *errcode,
805 gfp_t gfp_mask);
806
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700807/* Layout of fast clones : [skb1][skb2][fclone_ref] */
808struct sk_buff_fclones {
809 struct sk_buff skb1;
810
811 struct sk_buff skb2;
812
813 atomic_t fclone_ref;
814};
815
816/**
817 * skb_fclone_busy - check if fclone is busy
818 * @skb: buffer
819 *
820 * Returns true is skb is a fast clone, and its clone is not freed.
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700821 * Some drivers call skb_orphan() in their ndo_start_xmit(),
822 * so we also check that this didnt happen.
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700823 */
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700824static inline bool skb_fclone_busy(const struct sock *sk,
825 const struct sk_buff *skb)
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700826{
827 const struct sk_buff_fclones *fclones;
828
829 fclones = container_of(skb, struct sk_buff_fclones, skb1);
830
831 return skb->fclone == SKB_FCLONE_ORIG &&
Eric Dumazet6ffe75eb2014-12-03 17:04:39 -0800832 atomic_read(&fclones->fclone_ref) > 1 &&
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700833 fclones->skb2.sk == sk;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700834}
835
David S. Millerd179cd12005-08-17 14:57:30 -0700836static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100837 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700838{
Mel Gormanc93bdd02012-07-31 16:44:19 -0700839 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700840}
841
Joe Perches7965bd42013-09-26 14:48:15 -0700842struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
Patrick McHardy0ebd0ac2013-04-17 06:46:58 +0000843static inline struct sk_buff *alloc_skb_head(gfp_t priority)
844{
845 return __alloc_skb_head(priority, -1);
846}
847
Joe Perches7965bd42013-09-26 14:48:15 -0700848struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
849int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
850struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
851struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
Octavian Purdilabad93e92014-06-12 01:36:26 +0300852struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
853 gfp_t gfp_mask, bool fclone);
854static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
855 gfp_t gfp_mask)
856{
857 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
858}
Eric Dumazet117632e2011-12-03 21:39:53 +0000859
Joe Perches7965bd42013-09-26 14:48:15 -0700860int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
861struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
862 unsigned int headroom);
863struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
864 int newtailroom, gfp_t priority);
Fan Du25a91d82014-01-18 09:54:23 +0800865int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
866 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -0700867int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
868 int len);
869int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
870int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000871#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700872
Joe Perches7965bd42013-09-26 14:48:15 -0700873int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
874 int getfrag(void *from, char *to, int offset,
875 int len, int odd, struct sk_buff *skb),
876 void *from, int length);
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700877
Hannes Frederic Sowabe12a1f2015-05-21 16:59:58 +0200878int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
879 int offset, size_t size);
880
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800881struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700882 __u32 lower_offset;
883 __u32 upper_offset;
884 __u32 frag_idx;
885 __u32 stepped_offset;
886 struct sk_buff *root_skb;
887 struct sk_buff *cur_skb;
888 __u8 *frag_data;
889};
890
Joe Perches7965bd42013-09-26 14:48:15 -0700891void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
892 unsigned int to, struct skb_seq_state *st);
893unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
894 struct skb_seq_state *st);
895void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -0700896
Joe Perches7965bd42013-09-26 14:48:15 -0700897unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
Bojan Prtvar059a2442015-02-22 11:46:35 +0100898 unsigned int to, struct ts_config *config);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700899
Tom Herbert09323cc2013-12-15 22:16:19 -0800900/*
901 * Packet hash types specify the type of hash in skb_set_hash.
902 *
903 * Hash types refer to the protocol layer addresses which are used to
904 * construct a packet's hash. The hashes are used to differentiate or identify
905 * flows of the protocol layer for the hash type. Hash types are either
906 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
907 *
908 * Properties of hashes:
909 *
910 * 1) Two packets in different flows have different hash values
911 * 2) Two packets in the same flow should have the same hash value
912 *
913 * A hash at a higher layer is considered to be more specific. A driver should
914 * set the most specific hash possible.
915 *
916 * A driver cannot indicate a more specific hash than the layer at which a hash
917 * was computed. For instance an L3 hash cannot be set as an L4 hash.
918 *
919 * A driver may indicate a hash level which is less specific than the
920 * actual layer the hash was computed on. For instance, a hash computed
921 * at L4 may be considered an L3 hash. This should only be done if the
922 * driver can't unambiguously determine that the HW computed the hash at
923 * the higher layer. Note that the "should" in the second property above
924 * permits this.
925 */
926enum pkt_hash_types {
927 PKT_HASH_TYPE_NONE, /* Undefined type */
928 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
929 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
930 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
931};
932
933static inline void
934skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
935{
Tom Herbert61b905d2014-03-24 15:34:47 -0700936 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
Tom Herberta3b18dd2014-07-01 21:33:17 -0700937 skb->sw_hash = 0;
Tom Herbert61b905d2014-03-24 15:34:47 -0700938 skb->hash = hash;
Tom Herbert09323cc2013-12-15 22:16:19 -0800939}
940
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800941static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000942{
Tom Herberta3b18dd2014-07-01 21:33:17 -0700943 if (!skb->l4_hash && !skb->sw_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800944 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000945
Tom Herbert61b905d2014-03-24 15:34:47 -0700946 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000947}
948
Tom Herbertf70ea012015-07-31 16:52:10 -0700949__u32 __skb_get_hash_flowi6(struct sk_buff *skb, struct flowi6 *fl6);
950
951static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, struct flowi6 *fl6)
952{
953 if (!skb->l4_hash && !skb->sw_hash)
954 __skb_get_hash_flowi6(skb, fl6);
955
956 return skb->hash;
957}
958
959__u32 __skb_get_hash_flowi4(struct sk_buff *skb, struct flowi4 *fl);
960
961static inline __u32 skb_get_hash_flowi4(struct sk_buff *skb, struct flowi4 *fl4)
962{
963 if (!skb->l4_hash && !skb->sw_hash)
964 __skb_get_hash_flowi4(skb, fl4);
965
966 return skb->hash;
967}
968
Tom Herbert50fb7992015-05-01 11:30:12 -0700969__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
970
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800971static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
972{
Tom Herbert61b905d2014-03-24 15:34:47 -0700973 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800974}
975
Tom Herbert7539fad2013-12-15 22:12:18 -0800976static inline void skb_clear_hash(struct sk_buff *skb)
977{
Tom Herbert61b905d2014-03-24 15:34:47 -0700978 skb->hash = 0;
Tom Herberta3b18dd2014-07-01 21:33:17 -0700979 skb->sw_hash = 0;
Tom Herbert61b905d2014-03-24 15:34:47 -0700980 skb->l4_hash = 0;
Tom Herbert7539fad2013-12-15 22:12:18 -0800981}
982
983static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
984{
Tom Herbert61b905d2014-03-24 15:34:47 -0700985 if (!skb->l4_hash)
Tom Herbert7539fad2013-12-15 22:12:18 -0800986 skb_clear_hash(skb);
987}
988
Tom Herbert3df7a742013-12-15 22:16:29 -0800989static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
990{
Tom Herbert61b905d2014-03-24 15:34:47 -0700991 to->hash = from->hash;
Tom Herberta3b18dd2014-07-01 21:33:17 -0700992 to->sw_hash = from->sw_hash;
Tom Herbert61b905d2014-03-24 15:34:47 -0700993 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -0800994};
995
Eric Dumazetc29390c2015-03-11 18:42:02 -0700996static inline void skb_sender_cpu_clear(struct sk_buff *skb)
997{
998#ifdef CONFIG_XPS
999 skb->sender_cpu = 0;
1000#endif
1001}
1002
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001003#ifdef NET_SKBUFF_DATA_USES_OFFSET
1004static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1005{
1006 return skb->head + skb->end;
1007}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001008
1009static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1010{
1011 return skb->end;
1012}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001013#else
1014static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1015{
1016 return skb->end;
1017}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001018
1019static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1020{
1021 return skb->end - skb->head;
1022}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001023#endif
1024
Linus Torvalds1da177e2005-04-16 15:20:36 -07001025/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001026#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001027
Patrick Ohlyac45f602009-02-12 05:03:37 +00001028static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1029{
1030 return &skb_shinfo(skb)->hwtstamps;
1031}
1032
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033/**
1034 * skb_queue_empty - check if a queue is empty
1035 * @list: queue head
1036 *
1037 * Returns true if the queue is empty, false otherwise.
1038 */
1039static inline int skb_queue_empty(const struct sk_buff_head *list)
1040{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001041 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001042}
1043
1044/**
David S. Millerfc7ebb22008-09-23 00:34:07 -07001045 * skb_queue_is_last - check if skb is the last entry in the queue
1046 * @list: queue head
1047 * @skb: buffer
1048 *
1049 * Returns true if @skb is the last buffer on the list.
1050 */
1051static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1052 const struct sk_buff *skb)
1053{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001054 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -07001055}
1056
1057/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001058 * skb_queue_is_first - check if skb is the first entry in the queue
1059 * @list: queue head
1060 * @skb: buffer
1061 *
1062 * Returns true if @skb is the first buffer on the list.
1063 */
1064static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1065 const struct sk_buff *skb)
1066{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001067 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001068}
1069
1070/**
David S. Miller249c8b42008-09-23 00:44:42 -07001071 * skb_queue_next - return the next packet in the queue
1072 * @list: queue head
1073 * @skb: current buffer
1074 *
1075 * Return the next packet in @list after @skb. It is only valid to
1076 * call this if skb_queue_is_last() evaluates to false.
1077 */
1078static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1079 const struct sk_buff *skb)
1080{
1081 /* This BUG_ON may seem severe, but if we just return then we
1082 * are going to dereference garbage.
1083 */
1084 BUG_ON(skb_queue_is_last(list, skb));
1085 return skb->next;
1086}
1087
1088/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001089 * skb_queue_prev - return the prev packet in the queue
1090 * @list: queue head
1091 * @skb: current buffer
1092 *
1093 * Return the prev packet in @list before @skb. It is only valid to
1094 * call this if skb_queue_is_first() evaluates to false.
1095 */
1096static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1097 const struct sk_buff *skb)
1098{
1099 /* This BUG_ON may seem severe, but if we just return then we
1100 * are going to dereference garbage.
1101 */
1102 BUG_ON(skb_queue_is_first(list, skb));
1103 return skb->prev;
1104}
1105
1106/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107 * skb_get - reference buffer
1108 * @skb: buffer to reference
1109 *
1110 * Makes another reference to a socket buffer and returns a pointer
1111 * to the buffer.
1112 */
1113static inline struct sk_buff *skb_get(struct sk_buff *skb)
1114{
1115 atomic_inc(&skb->users);
1116 return skb;
1117}
1118
1119/*
1120 * If users == 1, we are the only owner and are can avoid redundant
1121 * atomic change.
1122 */
1123
1124/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001125 * skb_cloned - is the buffer a clone
1126 * @skb: buffer to check
1127 *
1128 * Returns true if the buffer was generated with skb_clone() and is
1129 * one of multiple shared copies of the buffer. Cloned buffers are
1130 * shared data so must not be written to under normal circumstances.
1131 */
1132static inline int skb_cloned(const struct sk_buff *skb)
1133{
1134 return skb->cloned &&
1135 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1136}
1137
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001138static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1139{
1140 might_sleep_if(pri & __GFP_WAIT);
1141
1142 if (skb_cloned(skb))
1143 return pskb_expand_head(skb, 0, 0, pri);
1144
1145 return 0;
1146}
1147
Linus Torvalds1da177e2005-04-16 15:20:36 -07001148/**
1149 * skb_header_cloned - is the header a clone
1150 * @skb: buffer to check
1151 *
1152 * Returns true if modifying the header part of the buffer requires
1153 * the data to be copied.
1154 */
1155static inline int skb_header_cloned(const struct sk_buff *skb)
1156{
1157 int dataref;
1158
1159 if (!skb->cloned)
1160 return 0;
1161
1162 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1163 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1164 return dataref != 1;
1165}
1166
1167/**
1168 * skb_header_release - release reference to header
1169 * @skb: buffer to operate on
1170 *
1171 * Drop a reference to the header part of the buffer. This is done
1172 * by acquiring a payload reference. You must not read from the header
1173 * part of skb->data after this.
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001174 * Note : Check if you can use __skb_header_release() instead.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001175 */
1176static inline void skb_header_release(struct sk_buff *skb)
1177{
1178 BUG_ON(skb->nohdr);
1179 skb->nohdr = 1;
1180 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1181}
1182
1183/**
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001184 * __skb_header_release - release reference to header
1185 * @skb: buffer to operate on
1186 *
1187 * Variant of skb_header_release() assuming skb is private to caller.
1188 * We can avoid one atomic operation.
1189 */
1190static inline void __skb_header_release(struct sk_buff *skb)
1191{
1192 skb->nohdr = 1;
1193 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1194}
1195
1196
1197/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001198 * skb_shared - is the buffer shared
1199 * @skb: buffer to check
1200 *
1201 * Returns true if more than one person has a reference to this
1202 * buffer.
1203 */
1204static inline int skb_shared(const struct sk_buff *skb)
1205{
1206 return atomic_read(&skb->users) != 1;
1207}
1208
1209/**
1210 * skb_share_check - check if buffer is shared and if so clone it
1211 * @skb: buffer to check
1212 * @pri: priority for memory allocation
1213 *
1214 * If the buffer is shared the buffer is cloned and the old copy
1215 * drops a reference. A new clone with a single reference is returned.
1216 * If the buffer is not shared the original buffer is returned. When
1217 * being called from interrupt status or with spinlocks held pri must
1218 * be GFP_ATOMIC.
1219 *
1220 * NULL is returned on a memory allocation failure.
1221 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001222static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001223{
1224 might_sleep_if(pri & __GFP_WAIT);
1225 if (skb_shared(skb)) {
1226 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001227
1228 if (likely(nskb))
1229 consume_skb(skb);
1230 else
1231 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001232 skb = nskb;
1233 }
1234 return skb;
1235}
1236
1237/*
1238 * Copy shared buffers into a new sk_buff. We effectively do COW on
1239 * packets to handle cases where we have a local reader and forward
1240 * and a couple of other messy ones. The normal one is tcpdumping
1241 * a packet thats being forwarded.
1242 */
1243
1244/**
1245 * skb_unshare - make a copy of a shared buffer
1246 * @skb: buffer to check
1247 * @pri: priority for memory allocation
1248 *
1249 * If the socket buffer is a clone then this function creates a new
1250 * copy of the data, drops a reference count on the old copy and returns
1251 * the new copy with the reference count at 1. If the buffer is not a clone
1252 * the original buffer is returned. When called with a spinlock held or
1253 * from interrupt state @pri must be %GFP_ATOMIC
1254 *
1255 * %NULL is returned on a memory allocation failure.
1256 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001257static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001258 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001259{
1260 might_sleep_if(pri & __GFP_WAIT);
1261 if (skb_cloned(skb)) {
1262 struct sk_buff *nskb = skb_copy(skb, pri);
Alexander Aring31eff812014-10-10 23:10:47 +02001263
1264 /* Free our shared copy */
1265 if (likely(nskb))
1266 consume_skb(skb);
1267 else
1268 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001269 skb = nskb;
1270 }
1271 return skb;
1272}
1273
1274/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001275 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001276 * @list_: list to peek at
1277 *
1278 * Peek an &sk_buff. Unlike most other operations you _MUST_
1279 * be careful with this one. A peek leaves the buffer on the
1280 * list and someone else may run off with it. You must hold
1281 * the appropriate locks or have a private queue to do this.
1282 *
1283 * Returns %NULL for an empty list or a pointer to the head element.
1284 * The reference count is not incremented and the reference is therefore
1285 * volatile. Use with caution.
1286 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001287static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001288{
Eric Dumazet18d07002012-04-30 16:31:46 +00001289 struct sk_buff *skb = list_->next;
1290
1291 if (skb == (struct sk_buff *)list_)
1292 skb = NULL;
1293 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001294}
1295
1296/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001297 * skb_peek_next - peek skb following the given one from a queue
1298 * @skb: skb to start from
1299 * @list_: list to peek at
1300 *
1301 * Returns %NULL when the end of the list is met or a pointer to the
1302 * next element. The reference count is not incremented and the
1303 * reference is therefore volatile. Use with caution.
1304 */
1305static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1306 const struct sk_buff_head *list_)
1307{
1308 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001309
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001310 if (next == (struct sk_buff *)list_)
1311 next = NULL;
1312 return next;
1313}
1314
1315/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001316 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001317 * @list_: list to peek at
1318 *
1319 * Peek an &sk_buff. Unlike most other operations you _MUST_
1320 * be careful with this one. A peek leaves the buffer on the
1321 * list and someone else may run off with it. You must hold
1322 * the appropriate locks or have a private queue to do this.
1323 *
1324 * Returns %NULL for an empty list or a pointer to the tail element.
1325 * The reference count is not incremented and the reference is therefore
1326 * volatile. Use with caution.
1327 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001328static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001329{
Eric Dumazet18d07002012-04-30 16:31:46 +00001330 struct sk_buff *skb = list_->prev;
1331
1332 if (skb == (struct sk_buff *)list_)
1333 skb = NULL;
1334 return skb;
1335
Linus Torvalds1da177e2005-04-16 15:20:36 -07001336}
1337
1338/**
1339 * skb_queue_len - get queue length
1340 * @list_: list to measure
1341 *
1342 * Return the length of an &sk_buff queue.
1343 */
1344static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1345{
1346 return list_->qlen;
1347}
1348
David S. Miller67fed452008-09-21 22:36:24 -07001349/**
1350 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1351 * @list: queue to initialize
1352 *
1353 * This initializes only the list and queue length aspects of
1354 * an sk_buff_head object. This allows to initialize the list
1355 * aspects of an sk_buff_head without reinitializing things like
1356 * the spinlock. It can also be used for on-stack sk_buff_head
1357 * objects where the spinlock is known to not be used.
1358 */
1359static inline void __skb_queue_head_init(struct sk_buff_head *list)
1360{
1361 list->prev = list->next = (struct sk_buff *)list;
1362 list->qlen = 0;
1363}
1364
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001365/*
1366 * This function creates a split out lock class for each invocation;
1367 * this is needed for now since a whole lot of users of the skb-queue
1368 * infrastructure in drivers have different locking usage (in hardirq)
1369 * than the networking core (in softirq only). In the long run either the
1370 * network layer or drivers should need annotation to consolidate the
1371 * main types of usage into 3 classes.
1372 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001373static inline void skb_queue_head_init(struct sk_buff_head *list)
1374{
1375 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001376 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001377}
1378
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001379static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1380 struct lock_class_key *class)
1381{
1382 skb_queue_head_init(list);
1383 lockdep_set_class(&list->lock, class);
1384}
1385
Linus Torvalds1da177e2005-04-16 15:20:36 -07001386/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001387 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001388 *
1389 * The "__skb_xxxx()" functions are the non-atomic ones that
1390 * can only be called with interrupts disabled.
1391 */
Joe Perches7965bd42013-09-26 14:48:15 -07001392void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1393 struct sk_buff_head *list);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001394static inline void __skb_insert(struct sk_buff *newsk,
1395 struct sk_buff *prev, struct sk_buff *next,
1396 struct sk_buff_head *list)
1397{
1398 newsk->next = next;
1399 newsk->prev = prev;
1400 next->prev = prev->next = newsk;
1401 list->qlen++;
1402}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001403
David S. Miller67fed452008-09-21 22:36:24 -07001404static inline void __skb_queue_splice(const struct sk_buff_head *list,
1405 struct sk_buff *prev,
1406 struct sk_buff *next)
1407{
1408 struct sk_buff *first = list->next;
1409 struct sk_buff *last = list->prev;
1410
1411 first->prev = prev;
1412 prev->next = first;
1413
1414 last->next = next;
1415 next->prev = last;
1416}
1417
1418/**
1419 * skb_queue_splice - join two skb lists, this is designed for stacks
1420 * @list: the new list to add
1421 * @head: the place to add it in the first list
1422 */
1423static inline void skb_queue_splice(const struct sk_buff_head *list,
1424 struct sk_buff_head *head)
1425{
1426 if (!skb_queue_empty(list)) {
1427 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001428 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001429 }
1430}
1431
1432/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001433 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001434 * @list: the new list to add
1435 * @head: the place to add it in the first list
1436 *
1437 * The list at @list is reinitialised
1438 */
1439static inline void skb_queue_splice_init(struct sk_buff_head *list,
1440 struct sk_buff_head *head)
1441{
1442 if (!skb_queue_empty(list)) {
1443 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001444 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001445 __skb_queue_head_init(list);
1446 }
1447}
1448
1449/**
1450 * skb_queue_splice_tail - join two skb lists, each list being a queue
1451 * @list: the new list to add
1452 * @head: the place to add it in the first list
1453 */
1454static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1455 struct sk_buff_head *head)
1456{
1457 if (!skb_queue_empty(list)) {
1458 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001459 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001460 }
1461}
1462
1463/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001464 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001465 * @list: the new list to add
1466 * @head: the place to add it in the first list
1467 *
1468 * Each of the lists is a queue.
1469 * The list at @list is reinitialised
1470 */
1471static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1472 struct sk_buff_head *head)
1473{
1474 if (!skb_queue_empty(list)) {
1475 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001476 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001477 __skb_queue_head_init(list);
1478 }
1479}
1480
Linus Torvalds1da177e2005-04-16 15:20:36 -07001481/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001482 * __skb_queue_after - queue a buffer at the list head
1483 * @list: list to use
1484 * @prev: place after this buffer
1485 * @newsk: buffer to queue
1486 *
1487 * Queue a buffer int the middle of a list. This function takes no locks
1488 * and you must therefore hold required locks before calling it.
1489 *
1490 * A buffer cannot be placed on two lists at the same time.
1491 */
1492static inline void __skb_queue_after(struct sk_buff_head *list,
1493 struct sk_buff *prev,
1494 struct sk_buff *newsk)
1495{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001496 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001497}
1498
Joe Perches7965bd42013-09-26 14:48:15 -07001499void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1500 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001501
Gerrit Renkerf5572852008-04-14 00:05:28 -07001502static inline void __skb_queue_before(struct sk_buff_head *list,
1503 struct sk_buff *next,
1504 struct sk_buff *newsk)
1505{
1506 __skb_insert(newsk, next->prev, next, list);
1507}
1508
Stephen Hemminger300ce172005-10-30 13:47:34 -08001509/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001510 * __skb_queue_head - queue a buffer at the list head
1511 * @list: list to use
1512 * @newsk: buffer to queue
1513 *
1514 * Queue a buffer at the start of a list. This function takes no locks
1515 * and you must therefore hold required locks before calling it.
1516 *
1517 * A buffer cannot be placed on two lists at the same time.
1518 */
Joe Perches7965bd42013-09-26 14:48:15 -07001519void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001520static inline void __skb_queue_head(struct sk_buff_head *list,
1521 struct sk_buff *newsk)
1522{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001523 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524}
1525
1526/**
1527 * __skb_queue_tail - queue a buffer at the list tail
1528 * @list: list to use
1529 * @newsk: buffer to queue
1530 *
1531 * Queue a buffer at the end of a list. This function takes no locks
1532 * and you must therefore hold required locks before calling it.
1533 *
1534 * A buffer cannot be placed on two lists at the same time.
1535 */
Joe Perches7965bd42013-09-26 14:48:15 -07001536void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001537static inline void __skb_queue_tail(struct sk_buff_head *list,
1538 struct sk_buff *newsk)
1539{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001540 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001541}
1542
Linus Torvalds1da177e2005-04-16 15:20:36 -07001543/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001544 * remove sk_buff from list. _Must_ be called atomically, and with
1545 * the list known..
1546 */
Joe Perches7965bd42013-09-26 14:48:15 -07001547void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001548static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1549{
1550 struct sk_buff *next, *prev;
1551
1552 list->qlen--;
1553 next = skb->next;
1554 prev = skb->prev;
1555 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001556 next->prev = prev;
1557 prev->next = next;
1558}
1559
Gerrit Renkerf525c062008-04-14 00:04:12 -07001560/**
1561 * __skb_dequeue - remove from the head of the queue
1562 * @list: list to dequeue from
1563 *
1564 * Remove the head of the list. This function does not take any locks
1565 * so must be used with appropriate locks held only. The head item is
1566 * returned or %NULL if the list is empty.
1567 */
Joe Perches7965bd42013-09-26 14:48:15 -07001568struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Gerrit Renkerf525c062008-04-14 00:04:12 -07001569static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1570{
1571 struct sk_buff *skb = skb_peek(list);
1572 if (skb)
1573 __skb_unlink(skb, list);
1574 return skb;
1575}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001576
1577/**
1578 * __skb_dequeue_tail - remove from the tail of the queue
1579 * @list: list to dequeue from
1580 *
1581 * Remove the tail of the list. This function does not take any locks
1582 * so must be used with appropriate locks held only. The tail item is
1583 * returned or %NULL if the list is empty.
1584 */
Joe Perches7965bd42013-09-26 14:48:15 -07001585struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001586static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1587{
1588 struct sk_buff *skb = skb_peek_tail(list);
1589 if (skb)
1590 __skb_unlink(skb, list);
1591 return skb;
1592}
1593
1594
David S. Millerbdcc0922012-03-07 20:53:36 -05001595static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001596{
1597 return skb->data_len;
1598}
1599
1600static inline unsigned int skb_headlen(const struct sk_buff *skb)
1601{
1602 return skb->len - skb->data_len;
1603}
1604
1605static inline int skb_pagelen(const struct sk_buff *skb)
1606{
1607 int i, len = 0;
1608
1609 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001610 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001611 return len + skb_headlen(skb);
1612}
1613
Ian Campbell131ea662011-08-19 06:25:00 +00001614/**
1615 * __skb_fill_page_desc - initialise a paged fragment in an skb
1616 * @skb: buffer containing fragment to be initialised
1617 * @i: paged fragment index to initialise
1618 * @page: the page to use for this fragment
1619 * @off: the offset to the data with @page
1620 * @size: the length of the data
1621 *
1622 * Initialises the @i'th fragment of @skb to point to &size bytes at
1623 * offset @off within @page.
1624 *
1625 * Does not take any additional reference on the fragment.
1626 */
1627static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1628 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001629{
1630 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1631
Mel Gormanc48a11c2012-07-31 16:44:23 -07001632 /*
1633 * Propagate page->pfmemalloc to the skb if we can. The problem is
1634 * that not all callers have unique ownership of the page. If
1635 * pfmemalloc is set, we check the mapping as a mapping implies
1636 * page->index is set (index and pfmemalloc share space).
1637 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1638 * do not lose pfmemalloc information as the pages would not be
1639 * allocated using __GFP_MEMALLOC.
1640 */
Ian Campbella8605c62011-10-19 23:01:49 +00001641 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001642 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001643 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001644
1645 page = compound_head(page);
1646 if (page->pfmemalloc && !page->mapping)
1647 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00001648}
1649
1650/**
1651 * skb_fill_page_desc - initialise a paged fragment in an skb
1652 * @skb: buffer containing fragment to be initialised
1653 * @i: paged fragment index to initialise
1654 * @page: the page to use for this fragment
1655 * @off: the offset to the data with @page
1656 * @size: the length of the data
1657 *
1658 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01001659 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00001660 * addition updates @skb such that @i is the last fragment.
1661 *
1662 * Does not take any additional reference on the fragment.
1663 */
1664static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1665 struct page *page, int off, int size)
1666{
1667 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001668 skb_shinfo(skb)->nr_frags = i + 1;
1669}
1670
Joe Perches7965bd42013-09-26 14:48:15 -07001671void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1672 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001673
Jason Wangf8e617e2013-11-01 14:07:47 +08001674void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1675 unsigned int truesize);
1676
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001678#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001679#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1680
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001681#ifdef NET_SKBUFF_DATA_USES_OFFSET
1682static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1683{
1684 return skb->head + skb->tail;
1685}
1686
1687static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1688{
1689 skb->tail = skb->data - skb->head;
1690}
1691
1692static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1693{
1694 skb_reset_tail_pointer(skb);
1695 skb->tail += offset;
1696}
Simon Horman7cc46192013-05-28 20:34:29 +00001697
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001698#else /* NET_SKBUFF_DATA_USES_OFFSET */
1699static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1700{
1701 return skb->tail;
1702}
1703
1704static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1705{
1706 skb->tail = skb->data;
1707}
1708
1709static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1710{
1711 skb->tail = skb->data + offset;
1712}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001713
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001714#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1715
Linus Torvalds1da177e2005-04-16 15:20:36 -07001716/*
1717 * Add data to an sk_buff
1718 */
Mathias Krause0c7ddf32013-11-07 14:18:24 +01001719unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001720unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001721static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1722{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001723 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724 SKB_LINEAR_ASSERT(skb);
1725 skb->tail += len;
1726 skb->len += len;
1727 return tmp;
1728}
1729
Joe Perches7965bd42013-09-26 14:48:15 -07001730unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1732{
1733 skb->data -= len;
1734 skb->len += len;
1735 return skb->data;
1736}
1737
Joe Perches7965bd42013-09-26 14:48:15 -07001738unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001739static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1740{
1741 skb->len -= len;
1742 BUG_ON(skb->len < skb->data_len);
1743 return skb->data += len;
1744}
1745
David S. Miller47d29642010-05-02 02:21:44 -07001746static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1747{
1748 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1749}
1750
Joe Perches7965bd42013-09-26 14:48:15 -07001751unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001752
1753static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1754{
1755 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001756 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001757 return NULL;
1758 skb->len -= len;
1759 return skb->data += len;
1760}
1761
1762static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1763{
1764 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1765}
1766
1767static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1768{
1769 if (likely(len <= skb_headlen(skb)))
1770 return 1;
1771 if (unlikely(len > skb->len))
1772 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001773 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001774}
1775
1776/**
1777 * skb_headroom - bytes at buffer head
1778 * @skb: buffer to check
1779 *
1780 * Return the number of bytes of free space at the head of an &sk_buff.
1781 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001782static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001783{
1784 return skb->data - skb->head;
1785}
1786
1787/**
1788 * skb_tailroom - bytes at buffer end
1789 * @skb: buffer to check
1790 *
1791 * Return the number of bytes of free space at the tail of an sk_buff
1792 */
1793static inline int skb_tailroom(const struct sk_buff *skb)
1794{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001795 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001796}
1797
1798/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00001799 * skb_availroom - bytes at buffer end
1800 * @skb: buffer to check
1801 *
1802 * Return the number of bytes of free space at the tail of an sk_buff
1803 * allocated by sk_stream_alloc()
1804 */
1805static inline int skb_availroom(const struct sk_buff *skb)
1806{
Eric Dumazet16fad692013-03-14 05:40:32 +00001807 if (skb_is_nonlinear(skb))
1808 return 0;
1809
1810 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00001811}
1812
1813/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001814 * skb_reserve - adjust headroom
1815 * @skb: buffer to alter
1816 * @len: bytes to move
1817 *
1818 * Increase the headroom of an empty &sk_buff by reducing the tail
1819 * room. This is only allowed for an empty buffer.
1820 */
David S. Miller8243126c2006-01-17 02:54:21 -08001821static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001822{
1823 skb->data += len;
1824 skb->tail += len;
1825}
1826
Tom Herbert8bce6d72014-09-29 20:22:29 -07001827#define ENCAP_TYPE_ETHER 0
1828#define ENCAP_TYPE_IPPROTO 1
1829
1830static inline void skb_set_inner_protocol(struct sk_buff *skb,
1831 __be16 protocol)
1832{
1833 skb->inner_protocol = protocol;
1834 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
1835}
1836
1837static inline void skb_set_inner_ipproto(struct sk_buff *skb,
1838 __u8 ipproto)
1839{
1840 skb->inner_ipproto = ipproto;
1841 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
1842}
1843
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001844static inline void skb_reset_inner_headers(struct sk_buff *skb)
1845{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001846 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001847 skb->inner_network_header = skb->network_header;
1848 skb->inner_transport_header = skb->transport_header;
1849}
1850
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00001851static inline void skb_reset_mac_len(struct sk_buff *skb)
1852{
1853 skb->mac_len = skb->network_header - skb->mac_header;
1854}
1855
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001856static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1857 *skb)
1858{
1859 return skb->head + skb->inner_transport_header;
1860}
1861
1862static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1863{
1864 skb->inner_transport_header = skb->data - skb->head;
1865}
1866
1867static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1868 const int offset)
1869{
1870 skb_reset_inner_transport_header(skb);
1871 skb->inner_transport_header += offset;
1872}
1873
1874static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1875{
1876 return skb->head + skb->inner_network_header;
1877}
1878
1879static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1880{
1881 skb->inner_network_header = skb->data - skb->head;
1882}
1883
1884static inline void skb_set_inner_network_header(struct sk_buff *skb,
1885 const int offset)
1886{
1887 skb_reset_inner_network_header(skb);
1888 skb->inner_network_header += offset;
1889}
1890
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001891static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1892{
1893 return skb->head + skb->inner_mac_header;
1894}
1895
1896static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1897{
1898 skb->inner_mac_header = skb->data - skb->head;
1899}
1900
1901static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1902 const int offset)
1903{
1904 skb_reset_inner_mac_header(skb);
1905 skb->inner_mac_header += offset;
1906}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001907static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1908{
Cong Wang35d04612013-05-29 15:16:05 +08001909 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001910}
1911
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001912static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1913{
1914 return skb->head + skb->transport_header;
1915}
1916
1917static inline void skb_reset_transport_header(struct sk_buff *skb)
1918{
1919 skb->transport_header = skb->data - skb->head;
1920}
1921
1922static inline void skb_set_transport_header(struct sk_buff *skb,
1923 const int offset)
1924{
1925 skb_reset_transport_header(skb);
1926 skb->transport_header += offset;
1927}
1928
1929static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1930{
1931 return skb->head + skb->network_header;
1932}
1933
1934static inline void skb_reset_network_header(struct sk_buff *skb)
1935{
1936 skb->network_header = skb->data - skb->head;
1937}
1938
1939static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1940{
1941 skb_reset_network_header(skb);
1942 skb->network_header += offset;
1943}
1944
1945static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1946{
1947 return skb->head + skb->mac_header;
1948}
1949
1950static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1951{
Cong Wang35d04612013-05-29 15:16:05 +08001952 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001953}
1954
1955static inline void skb_reset_mac_header(struct sk_buff *skb)
1956{
1957 skb->mac_header = skb->data - skb->head;
1958}
1959
1960static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1961{
1962 skb_reset_mac_header(skb);
1963 skb->mac_header += offset;
1964}
1965
Timo Teräs0e3da5b2013-12-16 11:02:09 +02001966static inline void skb_pop_mac_header(struct sk_buff *skb)
1967{
1968 skb->mac_header = skb->network_header;
1969}
1970
Ying Xuefbbdb8f2013-03-27 16:46:06 +00001971static inline void skb_probe_transport_header(struct sk_buff *skb,
1972 const int offset_hint)
1973{
1974 struct flow_keys keys;
1975
1976 if (skb_transport_header_was_set(skb))
1977 return;
Jiri Pirko06635a32015-05-12 14:56:16 +02001978 else if (skb_flow_dissect_flow_keys(skb, &keys))
Tom Herbert42aecaa2015-06-04 09:16:39 -07001979 skb_set_transport_header(skb, keys.control.thoff);
Ying Xuefbbdb8f2013-03-27 16:46:06 +00001980 else
1981 skb_set_transport_header(skb, offset_hint);
1982}
1983
Eric Dumazet03606892012-02-23 10:55:02 +00001984static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1985{
1986 if (skb_mac_header_was_set(skb)) {
1987 const unsigned char *old_mac = skb_mac_header(skb);
1988
1989 skb_set_mac_header(skb, -skb->mac_len);
1990 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1991 }
1992}
1993
Michał Mirosław04fb4512010-12-14 15:24:08 +00001994static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1995{
1996 return skb->csum_start - skb_headroom(skb);
1997}
1998
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001999static inline int skb_transport_offset(const struct sk_buff *skb)
2000{
2001 return skb_transport_header(skb) - skb->data;
2002}
2003
2004static inline u32 skb_network_header_len(const struct sk_buff *skb)
2005{
2006 return skb->transport_header - skb->network_header;
2007}
2008
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002009static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2010{
2011 return skb->inner_transport_header - skb->inner_network_header;
2012}
2013
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002014static inline int skb_network_offset(const struct sk_buff *skb)
2015{
2016 return skb_network_header(skb) - skb->data;
2017}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03002018
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002019static inline int skb_inner_network_offset(const struct sk_buff *skb)
2020{
2021 return skb_inner_network_header(skb) - skb->data;
2022}
2023
Changli Gaof9599ce2010-08-04 04:43:44 +00002024static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2025{
2026 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2027}
2028
Linus Torvalds1da177e2005-04-16 15:20:36 -07002029/*
2030 * CPUs often take a performance hit when accessing unaligned memory
2031 * locations. The actual performance hit varies, it can be small if the
2032 * hardware handles it or large if we have to take an exception and fix it
2033 * in software.
2034 *
2035 * Since an ethernet header is 14 bytes network drivers often end up with
2036 * the IP header at an unaligned offset. The IP header can be aligned by
2037 * shifting the start of the packet by 2 bytes. Drivers should do this
2038 * with:
2039 *
Tobias Klauser8660c122009-07-13 22:48:16 +00002040 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002041 *
2042 * The downside to this alignment of the IP header is that the DMA is now
2043 * unaligned. On some architectures the cost of an unaligned DMA is high
2044 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00002045 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002046 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2047 * to be overridden.
2048 */
2049#ifndef NET_IP_ALIGN
2050#define NET_IP_ALIGN 2
2051#endif
2052
Anton Blanchard025be812006-03-31 02:27:06 -08002053/*
2054 * The networking layer reserves some headroom in skb data (via
2055 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2056 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08002057 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08002058 *
2059 * Unfortunately this headroom changes the DMA alignment of the resulting
2060 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2061 * on some architectures. An architecture can override this value,
2062 * perhaps setting it to a cacheline in size (since that will maintain
2063 * cacheline alignment of the DMA). It must be a power of 2.
2064 *
David S. Millerd6301d32009-02-08 19:24:13 -08002065 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08002066 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07002067 *
2068 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2069 * to reduce average number of cache lines per packet.
2070 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07002071 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08002072 */
2073#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07002074#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08002075#endif
2076
Joe Perches7965bd42013-09-26 14:48:15 -07002077int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002078
2079static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2080{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00002081 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07002082 WARN_ON(1);
2083 return;
2084 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002085 skb->len = len;
2086 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002087}
2088
Joe Perches7965bd42013-09-26 14:48:15 -07002089void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002090
2091static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2092{
Herbert Xu3cc0e872006-06-09 16:13:38 -07002093 if (skb->data_len)
2094 return ___pskb_trim(skb, len);
2095 __skb_trim(skb, len);
2096 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002097}
2098
2099static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2100{
2101 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2102}
2103
2104/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07002105 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2106 * @skb: buffer to alter
2107 * @len: new length
2108 *
2109 * This is identical to pskb_trim except that the caller knows that
2110 * the skb is not cloned so we should never get an error due to out-
2111 * of-memory.
2112 */
2113static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2114{
2115 int err = pskb_trim(skb, len);
2116 BUG_ON(err);
2117}
2118
2119/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002120 * skb_orphan - orphan a buffer
2121 * @skb: buffer to orphan
2122 *
2123 * If a buffer currently has an owner then we call the owner's
2124 * destructor function and make the @skb unowned. The buffer continues
2125 * to exist but is no longer charged to its former owner.
2126 */
2127static inline void skb_orphan(struct sk_buff *skb)
2128{
Eric Dumazetc34a7612013-07-30 16:11:15 -07002129 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002130 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002131 skb->destructor = NULL;
2132 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07002133 } else {
2134 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002135 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002136}
2137
2138/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002139 * skb_orphan_frags - orphan the frags contained in a buffer
2140 * @skb: buffer to orphan frags from
2141 * @gfp_mask: allocation mask for replacement pages
2142 *
2143 * For each frag in the SKB which needs a destructor (i.e. has an
2144 * owner) create a copy of that frag and release the original
2145 * page by calling the destructor.
2146 */
2147static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2148{
2149 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2150 return 0;
2151 return skb_copy_ubufs(skb, gfp_mask);
2152}
2153
2154/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002155 * __skb_queue_purge - empty a list
2156 * @list: list to empty
2157 *
2158 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2159 * the list and one reference dropped. This function does not take the
2160 * list lock and the caller must hold the relevant locks to use it.
2161 */
Joe Perches7965bd42013-09-26 14:48:15 -07002162void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002163static inline void __skb_queue_purge(struct sk_buff_head *list)
2164{
2165 struct sk_buff *skb;
2166 while ((skb = __skb_dequeue(list)) != NULL)
2167 kfree_skb(skb);
2168}
2169
Joe Perches7965bd42013-09-26 14:48:15 -07002170void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002171
Joe Perches7965bd42013-09-26 14:48:15 -07002172struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2173 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07002174
2175/**
2176 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2177 * @dev: network device to receive on
2178 * @length: length to allocate
2179 *
2180 * Allocate a new &sk_buff and assign it a usage count of one. The
2181 * buffer has unspecified headroom built in. Users should allocate
2182 * the headroom they think they need without accounting for the
2183 * built in space. The built in space is used for optimisations.
2184 *
2185 * %NULL is returned if there is no free memory. Although this function
2186 * allocates memory it can be called from an interrupt.
2187 */
2188static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002189 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002190{
2191 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2192}
2193
Eric Dumazet6f532612012-05-18 05:12:12 +00002194/* legacy helper around __netdev_alloc_skb() */
2195static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2196 gfp_t gfp_mask)
2197{
2198 return __netdev_alloc_skb(NULL, length, gfp_mask);
2199}
2200
2201/* legacy helper around netdev_alloc_skb() */
2202static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2203{
2204 return netdev_alloc_skb(NULL, length);
2205}
2206
2207
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002208static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2209 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002210{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002211 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002212
2213 if (NET_IP_ALIGN && skb)
2214 skb_reserve(skb, NET_IP_ALIGN);
2215 return skb;
2216}
2217
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002218static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2219 unsigned int length)
2220{
2221 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2222}
2223
Alexander Duyck181edb22015-05-06 21:12:03 -07002224static inline void skb_free_frag(void *addr)
2225{
2226 __free_page_frag(addr);
2227}
2228
Alexander Duyckffde7322014-12-09 19:40:42 -08002229void *napi_alloc_frag(unsigned int fragsz);
Alexander Duyckfd11a832014-12-09 19:40:49 -08002230struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2231 unsigned int length, gfp_t gfp_mask);
2232static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2233 unsigned int length)
2234{
2235 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2236}
Alexander Duyckffde7322014-12-09 19:40:42 -08002237
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002238/**
Alexander Duyck71dfda52014-11-11 09:26:34 -08002239 * __dev_alloc_pages - allocate page for network Rx
2240 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2241 * @order: size of the allocation
2242 *
2243 * Allocate a new page.
2244 *
2245 * %NULL is returned if there is no free memory.
2246*/
2247static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2248 unsigned int order)
2249{
2250 /* This piece of code contains several assumptions.
2251 * 1. This is for device Rx, therefor a cold page is preferred.
2252 * 2. The expectation is the user wants a compound page.
2253 * 3. If requesting a order 0 page it will not be compound
2254 * due to the check to see if order has a value in prep_new_page
2255 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2256 * code in gfp_to_alloc_flags that should be enforcing this.
2257 */
2258 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2259
2260 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2261}
2262
2263static inline struct page *dev_alloc_pages(unsigned int order)
2264{
2265 return __dev_alloc_pages(GFP_ATOMIC, order);
2266}
2267
2268/**
2269 * __dev_alloc_page - allocate a page for network Rx
2270 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2271 *
2272 * Allocate a new page.
2273 *
2274 * %NULL is returned if there is no free memory.
2275 */
2276static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2277{
2278 return __dev_alloc_pages(gfp_mask, 0);
2279}
2280
2281static inline struct page *dev_alloc_page(void)
2282{
2283 return __dev_alloc_page(GFP_ATOMIC);
2284}
2285
2286/**
Mel Gorman06140022012-07-31 16:44:24 -07002287 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2288 * @page: The page that was allocated from skb_alloc_page
2289 * @skb: The skb that may need pfmemalloc set
2290 */
2291static inline void skb_propagate_pfmemalloc(struct page *page,
2292 struct sk_buff *skb)
2293{
2294 if (page && page->pfmemalloc)
2295 skb->pfmemalloc = true;
2296}
2297
Eric Dumazet564824b2010-10-11 19:05:25 +00002298/**
Masanari Iidae2278672014-02-18 22:54:36 +09002299 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002300 * @frag: the paged fragment
2301 *
2302 * Returns the &struct page associated with @frag.
2303 */
2304static inline struct page *skb_frag_page(const skb_frag_t *frag)
2305{
Ian Campbella8605c62011-10-19 23:01:49 +00002306 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00002307}
2308
2309/**
2310 * __skb_frag_ref - take an addition reference on a paged fragment.
2311 * @frag: the paged fragment
2312 *
2313 * Takes an additional reference on the paged fragment @frag.
2314 */
2315static inline void __skb_frag_ref(skb_frag_t *frag)
2316{
2317 get_page(skb_frag_page(frag));
2318}
2319
2320/**
2321 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2322 * @skb: the buffer
2323 * @f: the fragment offset.
2324 *
2325 * Takes an additional reference on the @f'th paged fragment of @skb.
2326 */
2327static inline void skb_frag_ref(struct sk_buff *skb, int f)
2328{
2329 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2330}
2331
2332/**
2333 * __skb_frag_unref - release a reference on a paged fragment.
2334 * @frag: the paged fragment
2335 *
2336 * Releases a reference on the paged fragment @frag.
2337 */
2338static inline void __skb_frag_unref(skb_frag_t *frag)
2339{
2340 put_page(skb_frag_page(frag));
2341}
2342
2343/**
2344 * skb_frag_unref - release a reference on a paged fragment of an skb.
2345 * @skb: the buffer
2346 * @f: the fragment offset
2347 *
2348 * Releases a reference on the @f'th paged fragment of @skb.
2349 */
2350static inline void skb_frag_unref(struct sk_buff *skb, int f)
2351{
2352 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2353}
2354
2355/**
2356 * skb_frag_address - gets the address of the data contained in a paged fragment
2357 * @frag: the paged fragment buffer
2358 *
2359 * Returns the address of the data within @frag. The page must already
2360 * be mapped.
2361 */
2362static inline void *skb_frag_address(const skb_frag_t *frag)
2363{
2364 return page_address(skb_frag_page(frag)) + frag->page_offset;
2365}
2366
2367/**
2368 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2369 * @frag: the paged fragment buffer
2370 *
2371 * Returns the address of the data within @frag. Checks that the page
2372 * is mapped and returns %NULL otherwise.
2373 */
2374static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2375{
2376 void *ptr = page_address(skb_frag_page(frag));
2377 if (unlikely(!ptr))
2378 return NULL;
2379
2380 return ptr + frag->page_offset;
2381}
2382
2383/**
2384 * __skb_frag_set_page - sets the page contained in a paged fragment
2385 * @frag: the paged fragment
2386 * @page: the page to set
2387 *
2388 * Sets the fragment @frag to contain @page.
2389 */
2390static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2391{
Ian Campbella8605c62011-10-19 23:01:49 +00002392 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002393}
2394
2395/**
2396 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2397 * @skb: the buffer
2398 * @f: the fragment offset
2399 * @page: the page to set
2400 *
2401 * Sets the @f'th fragment of @skb to contain @page.
2402 */
2403static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2404 struct page *page)
2405{
2406 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2407}
2408
Eric Dumazet400dfd32013-10-17 16:27:07 -07002409bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2410
Ian Campbell131ea662011-08-19 06:25:00 +00002411/**
2412 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002413 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002414 * @frag: the paged fragment to map
2415 * @offset: the offset within the fragment (starting at the
2416 * fragment's own offset)
2417 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002418 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00002419 *
2420 * Maps the page associated with @frag to @device.
2421 */
2422static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2423 const skb_frag_t *frag,
2424 size_t offset, size_t size,
2425 enum dma_data_direction dir)
2426{
2427 return dma_map_page(dev, skb_frag_page(frag),
2428 frag->page_offset + offset, size, dir);
2429}
2430
Eric Dumazet117632e2011-12-03 21:39:53 +00002431static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2432 gfp_t gfp_mask)
2433{
2434 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2435}
2436
Octavian Purdilabad93e92014-06-12 01:36:26 +03002437
2438static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2439 gfp_t gfp_mask)
2440{
2441 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2442}
2443
2444
Ian Campbell131ea662011-08-19 06:25:00 +00002445/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002446 * skb_clone_writable - is the header of a clone writable
2447 * @skb: buffer to check
2448 * @len: length up to which to write
2449 *
2450 * Returns true if modifying the header part of the cloned buffer
2451 * does not requires the data to be copied.
2452 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002453static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002454{
2455 return !skb_header_cloned(skb) &&
2456 skb_headroom(skb) + len <= skb->hdr_len;
2457}
2458
Herbert Xud9cc2042007-09-16 16:21:16 -07002459static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2460 int cloned)
2461{
2462 int delta = 0;
2463
Herbert Xud9cc2042007-09-16 16:21:16 -07002464 if (headroom > skb_headroom(skb))
2465 delta = headroom - skb_headroom(skb);
2466
2467 if (delta || cloned)
2468 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2469 GFP_ATOMIC);
2470 return 0;
2471}
2472
Patrick McHardy334a8132007-06-25 04:35:20 -07002473/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002474 * skb_cow - copy header of skb when it is required
2475 * @skb: buffer to cow
2476 * @headroom: needed headroom
2477 *
2478 * If the skb passed lacks sufficient headroom or its data part
2479 * is shared, data is reallocated. If reallocation fails, an error
2480 * is returned and original skb is not changed.
2481 *
2482 * The result is skb with writable area skb->head...skb->tail
2483 * and at least @headroom of space at head.
2484 */
2485static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2486{
Herbert Xud9cc2042007-09-16 16:21:16 -07002487 return __skb_cow(skb, headroom, skb_cloned(skb));
2488}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002489
Herbert Xud9cc2042007-09-16 16:21:16 -07002490/**
2491 * skb_cow_head - skb_cow but only making the head writable
2492 * @skb: buffer to cow
2493 * @headroom: needed headroom
2494 *
2495 * This function is identical to skb_cow except that we replace the
2496 * skb_cloned check by skb_header_cloned. It should be used when
2497 * you only need to push on some header and do not need to modify
2498 * the data.
2499 */
2500static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2501{
2502 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002503}
2504
2505/**
2506 * skb_padto - pad an skbuff up to a minimal size
2507 * @skb: buffer to pad
2508 * @len: minimal length
2509 *
2510 * Pads up a buffer to ensure the trailing bytes exist and are
2511 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002512 * is untouched. Otherwise it is extended. Returns zero on
2513 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002514 */
Herbert Xu5b057c62006-06-23 02:06:41 -07002515static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002516{
2517 unsigned int size = skb->len;
2518 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002519 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002520 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002521}
2522
Alexander Duyck9c0c1122014-12-03 08:17:33 -08002523/**
2524 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2525 * @skb: buffer to pad
2526 * @len: minimal length
2527 *
2528 * Pads up a buffer to ensure the trailing bytes exist and are
2529 * blanked. If the buffer already contains sufficient data it
2530 * is untouched. Otherwise it is extended. Returns zero on
2531 * success. The skb is freed on error.
2532 */
2533static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2534{
2535 unsigned int size = skb->len;
2536
2537 if (unlikely(size < len)) {
2538 len -= size;
2539 if (skb_pad(skb, len))
2540 return -ENOMEM;
2541 __skb_put(skb, len);
2542 }
2543 return 0;
2544}
2545
Linus Torvalds1da177e2005-04-16 15:20:36 -07002546static inline int skb_add_data(struct sk_buff *skb,
Al Viroaf2b0402014-11-27 21:44:24 -05002547 struct iov_iter *from, int copy)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002548{
2549 const int off = skb->len;
2550
2551 if (skb->ip_summed == CHECKSUM_NONE) {
Al Viroaf2b0402014-11-27 21:44:24 -05002552 __wsum csum = 0;
2553 if (csum_and_copy_from_iter(skb_put(skb, copy), copy,
2554 &csum, from) == copy) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002555 skb->csum = csum_block_add(skb->csum, csum, off);
2556 return 0;
2557 }
Al Viroaf2b0402014-11-27 21:44:24 -05002558 } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002559 return 0;
2560
2561 __skb_trim(skb, off);
2562 return -EFAULT;
2563}
2564
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002565static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2566 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002567{
2568 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002569 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002570
Ian Campbellea2ab692011-08-22 23:44:58 +00002571 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002572 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002573 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002574 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002575}
2576
Herbert Xu364c6ba2006-06-09 16:10:40 -07002577static inline int __skb_linearize(struct sk_buff *skb)
2578{
2579 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2580}
2581
Linus Torvalds1da177e2005-04-16 15:20:36 -07002582/**
2583 * skb_linearize - convert paged skb to linear one
2584 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002585 *
2586 * If there is no free memory -ENOMEM is returned, otherwise zero
2587 * is returned and the old skb data released.
2588 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002589static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002590{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002591 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2592}
2593
2594/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00002595 * skb_has_shared_frag - can any frag be overwritten
2596 * @skb: buffer to test
2597 *
2598 * Return true if the skb has at least one frag that might be modified
2599 * by an external entity (as in vmsplice()/sendfile())
2600 */
2601static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2602{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00002603 return skb_is_nonlinear(skb) &&
2604 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00002605}
2606
2607/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07002608 * skb_linearize_cow - make sure skb is linear and writable
2609 * @skb: buffer to process
2610 *
2611 * If there is no free memory -ENOMEM is returned, otherwise zero
2612 * is returned and the old skb data released.
2613 */
2614static inline int skb_linearize_cow(struct sk_buff *skb)
2615{
2616 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2617 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002618}
2619
2620/**
2621 * skb_postpull_rcsum - update checksum for received skb after pull
2622 * @skb: buffer to update
2623 * @start: start of data before pull
2624 * @len: length of data pulled
2625 *
2626 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002627 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2628 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002629 */
2630
2631static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f92006-03-20 22:43:56 -08002632 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002633{
Patrick McHardy84fa7932006-08-29 16:44:56 -07002634 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002635 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2636}
2637
Herbert Xucbb042f92006-03-20 22:43:56 -08002638unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2639
David S. Miller7ce5a272013-12-02 17:26:05 -05002640/**
2641 * pskb_trim_rcsum - trim received skb and update checksum
2642 * @skb: buffer to trim
2643 * @len: new length
2644 *
2645 * This is exactly the same as pskb_trim except that it ensures the
2646 * checksum of received packets are still valid after the operation.
2647 */
2648
2649static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2650{
2651 if (likely(len >= skb->len))
2652 return 0;
2653 if (skb->ip_summed == CHECKSUM_COMPLETE)
2654 skb->ip_summed = CHECKSUM_NONE;
2655 return __pskb_trim(skb, len);
2656}
2657
Linus Torvalds1da177e2005-04-16 15:20:36 -07002658#define skb_queue_walk(queue, skb) \
2659 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002660 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002661 skb = skb->next)
2662
James Chapman46f89142007-04-30 00:07:31 -07002663#define skb_queue_walk_safe(queue, skb, tmp) \
2664 for (skb = (queue)->next, tmp = skb->next; \
2665 skb != (struct sk_buff *)(queue); \
2666 skb = tmp, tmp = skb->next)
2667
David S. Miller1164f522008-09-23 00:49:44 -07002668#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002669 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002670 skb = skb->next)
2671
2672#define skb_queue_walk_from_safe(queue, skb, tmp) \
2673 for (tmp = skb->next; \
2674 skb != (struct sk_buff *)(queue); \
2675 skb = tmp, tmp = skb->next)
2676
Stephen Hemminger300ce172005-10-30 13:47:34 -08002677#define skb_queue_reverse_walk(queue, skb) \
2678 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002679 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08002680 skb = skb->prev)
2681
David S. Miller686a2952011-01-20 22:47:32 -08002682#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2683 for (skb = (queue)->prev, tmp = skb->prev; \
2684 skb != (struct sk_buff *)(queue); \
2685 skb = tmp, tmp = skb->prev)
2686
2687#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2688 for (tmp = skb->prev; \
2689 skb != (struct sk_buff *)(queue); \
2690 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002691
David S. Miller21dc3302010-08-23 00:13:46 -07002692static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07002693{
2694 return skb_shinfo(skb)->frag_list != NULL;
2695}
2696
2697static inline void skb_frag_list_init(struct sk_buff *skb)
2698{
2699 skb_shinfo(skb)->frag_list = NULL;
2700}
2701
David S. Milleree039872009-06-09 00:17:13 -07002702#define skb_walk_frags(skb, iter) \
2703 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2704
Joe Perches7965bd42013-09-26 14:48:15 -07002705struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2706 int *peeked, int *off, int *err);
2707struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2708 int *err);
2709unsigned int datagram_poll(struct file *file, struct socket *sock,
2710 struct poll_table_struct *wait);
Al Viroc0371da2014-11-24 10:42:55 -05002711int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
2712 struct iov_iter *to, int size);
David S. Miller51f3d022014-11-05 16:46:40 -05002713static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
2714 struct msghdr *msg, int size)
2715{
Al Viroe5a4b0b2014-11-24 18:17:55 -05002716 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
David S. Miller51f3d022014-11-05 16:46:40 -05002717}
Al Viroe5a4b0b2014-11-24 18:17:55 -05002718int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
2719 struct msghdr *msg);
Al Viro3a654f92014-06-19 14:15:22 -04002720int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
2721 struct iov_iter *from, int len);
Al Viro3a654f92014-06-19 14:15:22 -04002722int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
Joe Perches7965bd42013-09-26 14:48:15 -07002723void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2724void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2725int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07002726int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2727int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2728__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2729 int len, __wsum csum);
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +02002730ssize_t skb_socket_splice(struct sock *sk,
2731 struct pipe_inode_info *pipe,
2732 struct splice_pipe_desc *spd);
2733int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
Joe Perches7965bd42013-09-26 14:48:15 -07002734 struct pipe_inode_info *pipe, unsigned int len,
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +02002735 unsigned int flags,
2736 ssize_t (*splice_cb)(struct sock *,
2737 struct pipe_inode_info *,
2738 struct splice_pipe_desc *));
Joe Perches7965bd42013-09-26 14:48:15 -07002739void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01002740unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00002741int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2742 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07002743void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2744int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2745void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Florian Westphalde960aa2014-01-26 10:58:16 +01002746unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -07002747struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Vlad Yasevich0d5501c2014-08-08 14:42:13 -04002748struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
Jiri Pirkoe2195122014-11-19 14:05:01 +01002749int skb_ensure_writable(struct sk_buff *skb, int write_len);
Jiri Pirko93515d52014-11-19 14:05:02 +01002750int skb_vlan_pop(struct sk_buff *skb);
2751int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002752
Al Viro6ce8e9c2014-04-06 21:25:44 -04002753static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
2754{
Al Viro21226ab2014-11-28 15:48:29 -05002755 return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro6ce8e9c2014-04-06 21:25:44 -04002756}
2757
Al Viro7eab8d92014-04-06 21:51:23 -04002758static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
2759{
Al Viroe5a4b0b2014-11-24 18:17:55 -05002760 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro7eab8d92014-04-06 21:51:23 -04002761}
2762
Daniel Borkmann2817a332013-10-30 11:50:51 +01002763struct skb_checksum_ops {
2764 __wsum (*update)(const void *mem, int len, __wsum wsum);
2765 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2766};
2767
2768__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2769 __wsum csum, const struct skb_checksum_ops *ops);
2770__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2771 __wsum csum);
2772
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07002773static inline void * __must_check
2774__skb_header_pointer(const struct sk_buff *skb, int offset,
2775 int len, void *data, int hlen, void *buffer)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002776{
Patrick McHardy55820ee2005-07-05 14:08:10 -07002777 if (hlen - offset >= len)
David S. Miller690e36e2014-08-23 12:13:41 -07002778 return data + offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002779
David S. Miller690e36e2014-08-23 12:13:41 -07002780 if (!skb ||
2781 skb_copy_bits(skb, offset, buffer, len) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002782 return NULL;
2783
2784 return buffer;
2785}
2786
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07002787static inline void * __must_check
2788skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
David S. Miller690e36e2014-08-23 12:13:41 -07002789{
2790 return __skb_header_pointer(skb, offset, len, skb->data,
2791 skb_headlen(skb), buffer);
2792}
2793
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01002794/**
2795 * skb_needs_linearize - check if we need to linearize a given skb
2796 * depending on the given device features.
2797 * @skb: socket buffer to check
2798 * @features: net device features
2799 *
2800 * Returns true if either:
2801 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2802 * 2. skb is fragmented and the device does not support SG.
2803 */
2804static inline bool skb_needs_linearize(struct sk_buff *skb,
2805 netdev_features_t features)
2806{
2807 return skb_is_nonlinear(skb) &&
2808 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2809 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2810}
2811
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03002812static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2813 void *to,
2814 const unsigned int len)
2815{
2816 memcpy(to, skb->data, len);
2817}
2818
2819static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2820 const int offset, void *to,
2821 const unsigned int len)
2822{
2823 memcpy(to, skb->data + offset, len);
2824}
2825
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03002826static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2827 const void *from,
2828 const unsigned int len)
2829{
2830 memcpy(skb->data, from, len);
2831}
2832
2833static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2834 const int offset,
2835 const void *from,
2836 const unsigned int len)
2837{
2838 memcpy(skb->data + offset, from, len);
2839}
2840
Joe Perches7965bd42013-09-26 14:48:15 -07002841void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002842
Patrick Ohlyac45f602009-02-12 05:03:37 +00002843static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2844{
2845 return skb->tstamp;
2846}
2847
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002848/**
2849 * skb_get_timestamp - get timestamp from a skb
2850 * @skb: skb to get stamp from
2851 * @stamp: pointer to struct timeval to store stamp in
2852 *
2853 * Timestamps are stored in the skb as offsets to a base timestamp.
2854 * This function converts the offset back to a struct timeval and stores
2855 * it in stamp.
2856 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00002857static inline void skb_get_timestamp(const struct sk_buff *skb,
2858 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002859{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002860 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002861}
2862
Patrick Ohlyac45f602009-02-12 05:03:37 +00002863static inline void skb_get_timestampns(const struct sk_buff *skb,
2864 struct timespec *stamp)
2865{
2866 *stamp = ktime_to_timespec(skb->tstamp);
2867}
2868
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002869static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002870{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002871 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002872}
2873
Stephen Hemminger164891a2007-04-23 22:26:16 -07002874static inline ktime_t net_timedelta(ktime_t t)
2875{
2876 return ktime_sub(ktime_get_real(), t);
2877}
2878
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07002879static inline ktime_t net_invalid_timestamp(void)
2880{
2881 return ktime_set(0, 0);
2882}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002883
Alexander Duyck62bccb82014-09-04 13:31:35 -04002884struct sk_buff *skb_clone_sk(struct sk_buff *skb);
2885
Richard Cochranc1f19b52010-07-17 08:49:36 +00002886#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2887
Joe Perches7965bd42013-09-26 14:48:15 -07002888void skb_clone_tx_timestamp(struct sk_buff *skb);
2889bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00002890
2891#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2892
2893static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2894{
2895}
2896
2897static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2898{
2899 return false;
2900}
2901
2902#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2903
2904/**
2905 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2906 *
Richard Cochranda92b192011-10-21 00:49:15 +00002907 * PHY drivers may accept clones of transmitted packets for
2908 * timestamping via their phy_driver.txtstamp method. These drivers
2909 * must call this function to return the skb back to the stack, with
2910 * or without a timestamp.
2911 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00002912 * @skb: clone of the the original outgoing packet
Richard Cochranda92b192011-10-21 00:49:15 +00002913 * @hwtstamps: hardware time stamps, may be NULL if not available
Richard Cochranc1f19b52010-07-17 08:49:36 +00002914 *
2915 */
2916void skb_complete_tx_timestamp(struct sk_buff *skb,
2917 struct skb_shared_hwtstamps *hwtstamps);
2918
Willem de Bruijne7fd2882014-08-04 22:11:48 -04002919void __skb_tstamp_tx(struct sk_buff *orig_skb,
2920 struct skb_shared_hwtstamps *hwtstamps,
2921 struct sock *sk, int tstype);
2922
Patrick Ohlyac45f602009-02-12 05:03:37 +00002923/**
2924 * skb_tstamp_tx - queue clone of skb with send time stamps
2925 * @orig_skb: the original outgoing packet
2926 * @hwtstamps: hardware time stamps, may be NULL if not available
2927 *
2928 * If the skb has a socket associated, then this function clones the
2929 * skb (thus sharing the actual data and optional structures), stores
2930 * the optional hardware time stamping information (if non NULL) or
2931 * generates a software time stamp (otherwise), then queues the clone
2932 * to the error queue of the socket. Errors are silently ignored.
2933 */
Joe Perches7965bd42013-09-26 14:48:15 -07002934void skb_tstamp_tx(struct sk_buff *orig_skb,
2935 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00002936
Richard Cochran4507a712010-07-17 08:48:28 +00002937static inline void sw_tx_timestamp(struct sk_buff *skb)
2938{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002939 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2940 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00002941 skb_tstamp_tx(skb, NULL);
2942}
2943
2944/**
2945 * skb_tx_timestamp() - Driver hook for transmit timestamping
2946 *
2947 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00002948 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00002949 *
David S. Miller73409f32013-12-27 13:04:33 -05002950 * Specifically, one should make absolutely sure that this function is
2951 * called before TX completion of this packet can trigger. Otherwise
2952 * the packet could potentially already be freed.
2953 *
Richard Cochran4507a712010-07-17 08:48:28 +00002954 * @skb: A socket buffer.
2955 */
2956static inline void skb_tx_timestamp(struct sk_buff *skb)
2957{
Richard Cochranc1f19b52010-07-17 08:49:36 +00002958 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002959 sw_tx_timestamp(skb);
2960}
2961
Johannes Berg6e3e9392011-11-09 10:15:42 +01002962/**
2963 * skb_complete_wifi_ack - deliver skb with wifi status
2964 *
2965 * @skb: the original outgoing packet
2966 * @acked: ack status
2967 *
2968 */
2969void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2970
Joe Perches7965bd42013-09-26 14:48:15 -07002971__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2972__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002973
Herbert Xu60476372007-04-09 11:59:39 -07002974static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2975{
Tom Herbert6edec0e2015-02-10 16:30:28 -08002976 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
2977 skb->csum_valid ||
2978 (skb->ip_summed == CHECKSUM_PARTIAL &&
2979 skb_checksum_start_offset(skb) >= 0));
Herbert Xu60476372007-04-09 11:59:39 -07002980}
2981
Herbert Xufb286bb2005-11-10 13:01:24 -08002982/**
2983 * skb_checksum_complete - Calculate checksum of an entire packet
2984 * @skb: packet to process
2985 *
2986 * This function calculates the checksum over the entire packet plus
2987 * the value of skb->csum. The latter can be used to supply the
2988 * checksum of a pseudo header as used by TCP/UDP. It returns the
2989 * checksum.
2990 *
2991 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2992 * this function can be used to verify that checksum on received
2993 * packets. In that case the function should return zero if the
2994 * checksum is correct. In particular, this function will return zero
2995 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2996 * hardware has already verified the correctness of the checksum.
2997 */
Al Viro4381ca32007-07-15 21:00:11 +01002998static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002999{
Herbert Xu60476372007-04-09 11:59:39 -07003000 return skb_csum_unnecessary(skb) ?
3001 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003002}
3003
Tom Herbert77cffe22014-08-27 21:26:46 -07003004static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3005{
3006 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3007 if (skb->csum_level == 0)
3008 skb->ip_summed = CHECKSUM_NONE;
3009 else
3010 skb->csum_level--;
3011 }
3012}
3013
3014static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3015{
3016 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3017 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3018 skb->csum_level++;
3019 } else if (skb->ip_summed == CHECKSUM_NONE) {
3020 skb->ip_summed = CHECKSUM_UNNECESSARY;
3021 skb->csum_level = 0;
3022 }
3023}
3024
Tom Herbert5a212322014-08-31 15:12:41 -07003025static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
3026{
3027 /* Mark current checksum as bad (typically called from GRO
3028 * path). In the case that ip_summed is CHECKSUM_NONE
3029 * this must be the first checksum encountered in the packet.
3030 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
3031 * checksum after the last one validated. For UDP, a zero
3032 * checksum can not be marked as bad.
3033 */
3034
3035 if (skb->ip_summed == CHECKSUM_NONE ||
3036 skb->ip_summed == CHECKSUM_UNNECESSARY)
3037 skb->csum_bad = 1;
3038}
3039
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003040/* Check if we need to perform checksum complete validation.
3041 *
3042 * Returns true if checksum complete is needed, false otherwise
3043 * (either checksum is unnecessary or zero checksum is allowed).
3044 */
3045static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3046 bool zero_okay,
3047 __sum16 check)
3048{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003049 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3050 skb->csum_valid = 1;
Tom Herbert77cffe22014-08-27 21:26:46 -07003051 __skb_decr_checksum_unnecessary(skb);
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003052 return false;
3053 }
3054
3055 return true;
3056}
3057
3058/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3059 * in checksum_init.
3060 */
3061#define CHECKSUM_BREAK 76
3062
Tom Herbert4e18b9a2015-04-20 14:10:04 -07003063/* Unset checksum-complete
3064 *
3065 * Unset checksum complete can be done when packet is being modified
3066 * (uncompressed for instance) and checksum-complete value is
3067 * invalidated.
3068 */
3069static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3070{
3071 if (skb->ip_summed == CHECKSUM_COMPLETE)
3072 skb->ip_summed = CHECKSUM_NONE;
3073}
3074
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003075/* Validate (init) checksum based on checksum complete.
3076 *
3077 * Return values:
3078 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3079 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3080 * checksum is stored in skb->csum for use in __skb_checksum_complete
3081 * non-zero: value of invalid checksum
3082 *
3083 */
3084static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3085 bool complete,
3086 __wsum psum)
3087{
3088 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3089 if (!csum_fold(csum_add(psum, skb->csum))) {
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003090 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003091 return 0;
3092 }
Tom Herbert5a212322014-08-31 15:12:41 -07003093 } else if (skb->csum_bad) {
3094 /* ip_summed == CHECKSUM_NONE in this case */
Eric Dumazetc91d4602015-05-15 05:48:07 -07003095 return (__force __sum16)1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003096 }
3097
3098 skb->csum = psum;
3099
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003100 if (complete || skb->len <= CHECKSUM_BREAK) {
3101 __sum16 csum;
3102
3103 csum = __skb_checksum_complete(skb);
3104 skb->csum_valid = !csum;
3105 return csum;
3106 }
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003107
3108 return 0;
3109}
3110
3111static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3112{
3113 return 0;
3114}
3115
3116/* Perform checksum validate (init). Note that this is a macro since we only
3117 * want to calculate the pseudo header which is an input function if necessary.
3118 * First we try to validate without any computation (checksum unnecessary) and
3119 * then calculate based on checksum complete calling the function to compute
3120 * pseudo header.
3121 *
3122 * Return values:
3123 * 0: checksum is validated or try to in skb_checksum_complete
3124 * non-zero: value of invalid checksum
3125 */
3126#define __skb_checksum_validate(skb, proto, complete, \
3127 zero_okay, check, compute_pseudo) \
3128({ \
3129 __sum16 __ret = 0; \
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003130 skb->csum_valid = 0; \
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003131 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3132 __ret = __skb_checksum_validate_complete(skb, \
3133 complete, compute_pseudo(skb, proto)); \
3134 __ret; \
3135})
3136
3137#define skb_checksum_init(skb, proto, compute_pseudo) \
3138 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3139
3140#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3141 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3142
3143#define skb_checksum_validate(skb, proto, compute_pseudo) \
3144 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3145
3146#define skb_checksum_validate_zero_check(skb, proto, check, \
3147 compute_pseudo) \
Sabrina Dubroca096a4cf2015-02-06 18:54:19 +01003148 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003149
3150#define skb_checksum_simple_validate(skb) \
3151 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3152
Tom Herbertd96535a2014-08-31 15:12:42 -07003153static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3154{
3155 return (skb->ip_summed == CHECKSUM_NONE &&
3156 skb->csum_valid && !skb->csum_bad);
3157}
3158
3159static inline void __skb_checksum_convert(struct sk_buff *skb,
3160 __sum16 check, __wsum pseudo)
3161{
3162 skb->csum = ~pseudo;
3163 skb->ip_summed = CHECKSUM_COMPLETE;
3164}
3165
3166#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3167do { \
3168 if (__skb_checksum_convert_check(skb)) \
3169 __skb_checksum_convert(skb, check, \
3170 compute_pseudo(skb, proto)); \
3171} while (0)
3172
Tom Herbert15e23962015-02-10 16:30:31 -08003173static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3174 u16 start, u16 offset)
3175{
3176 skb->ip_summed = CHECKSUM_PARTIAL;
3177 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3178 skb->csum_offset = offset - start;
3179}
3180
Tom Herbertdcdc8992015-02-02 16:07:34 -08003181/* Update skbuf and packet to reflect the remote checksum offload operation.
3182 * When called, ptr indicates the starting point for skb->csum when
3183 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3184 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3185 */
3186static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
Tom Herbert15e23962015-02-10 16:30:31 -08003187 int start, int offset, bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08003188{
3189 __wsum delta;
3190
Tom Herbert15e23962015-02-10 16:30:31 -08003191 if (!nopartial) {
3192 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3193 return;
3194 }
3195
Tom Herbertdcdc8992015-02-02 16:07:34 -08003196 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3197 __skb_checksum_complete(skb);
3198 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3199 }
3200
3201 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3202
3203 /* Adjust skb->csum since we changed the packet */
3204 skb->csum = csum_add(skb->csum, delta);
3205}
3206
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003207#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07003208void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003209static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3210{
3211 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07003212 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003213}
3214static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3215{
3216 if (nfct)
3217 atomic_inc(&nfct->use);
3218}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003219#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003220#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003221static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3222{
3223 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3224 kfree(nf_bridge);
3225}
3226static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3227{
3228 if (nf_bridge)
3229 atomic_inc(&nf_bridge->use);
3230}
3231#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08003232static inline void nf_reset(struct sk_buff *skb)
3233{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003234#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08003235 nf_conntrack_put(skb->nfct);
3236 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003237#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003238#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Patrick McHardya193a4a2006-03-20 19:23:05 -08003239 nf_bridge_put(skb->nf_bridge);
3240 skb->nf_bridge = NULL;
3241#endif
3242}
3243
Patrick McHardy124dff02013-04-05 20:42:05 +02003244static inline void nf_reset_trace(struct sk_buff *skb)
3245{
Florian Westphal478b3602014-02-15 23:48:45 +01003246#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00003247 skb->nf_trace = 0;
3248#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003249}
3250
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003251/* Note: This doesn't put any conntrack and bridge info in dst. */
Eric Dumazetb1937222014-09-28 22:18:47 -07003252static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3253 bool copy)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003254{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003255#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003256 dst->nfct = src->nfct;
3257 nf_conntrack_get(src->nfct);
Eric Dumazetb1937222014-09-28 22:18:47 -07003258 if (copy)
3259 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003260#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003261#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003262 dst->nf_bridge = src->nf_bridge;
3263 nf_bridge_get(src->nf_bridge);
3264#endif
Florian Westphal478b3602014-02-15 23:48:45 +01003265#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Eric Dumazetb1937222014-09-28 22:18:47 -07003266 if (copy)
3267 dst->nf_trace = src->nf_trace;
Florian Westphal478b3602014-02-15 23:48:45 +01003268#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003269}
3270
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003271static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3272{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003273#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003274 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003275#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003276#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003277 nf_bridge_put(dst->nf_bridge);
3278#endif
Eric Dumazetb1937222014-09-28 22:18:47 -07003279 __nf_copy(dst, src, true);
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003280}
3281
James Morris984bc162006-06-09 00:29:17 -07003282#ifdef CONFIG_NETWORK_SECMARK
3283static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3284{
3285 to->secmark = from->secmark;
3286}
3287
3288static inline void skb_init_secmark(struct sk_buff *skb)
3289{
3290 skb->secmark = 0;
3291}
3292#else
3293static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3294{ }
3295
3296static inline void skb_init_secmark(struct sk_buff *skb)
3297{ }
3298#endif
3299
Eric W. Biederman574f7192014-04-01 12:20:24 -07003300static inline bool skb_irq_freeable(const struct sk_buff *skb)
3301{
3302 return !skb->destructor &&
3303#if IS_ENABLED(CONFIG_XFRM)
3304 !skb->sp &&
3305#endif
3306#if IS_ENABLED(CONFIG_NF_CONNTRACK)
3307 !skb->nfct &&
3308#endif
3309 !skb->_skb_refdst &&
3310 !skb_has_frag_list(skb);
3311}
3312
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003313static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3314{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003315 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003316}
3317
Stephen Hemminger92477442009-03-21 13:39:26 -07003318static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003319{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003320 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003321}
3322
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003323static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3324{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003325 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003326}
3327
David S. Millerd5a9e242009-01-27 16:22:11 -08003328static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3329{
3330 skb->queue_mapping = rx_queue + 1;
3331}
3332
Stephen Hemminger92477442009-03-21 13:39:26 -07003333static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003334{
3335 return skb->queue_mapping - 1;
3336}
3337
Stephen Hemminger92477442009-03-21 13:39:26 -07003338static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003339{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003340 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08003341}
3342
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003343static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3344{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003345#ifdef CONFIG_XFRM
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003346 return skb->sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003347#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003348 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003349#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003350}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003351
Pravin B Shelar68c33162013-02-14 14:02:41 +00003352/* Keeps track of mac header offset relative to skb->head.
3353 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3354 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07003355 * tunnel skb it points to outer mac header.
3356 * Keeps track of level of encapsulation of network headers.
3357 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00003358struct skb_gso_cb {
Eric Dumazet3347c962013-10-19 11:42:56 -07003359 int mac_offset;
3360 int encap_level;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003361 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00003362};
3363#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3364
3365static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3366{
3367 return (skb_mac_header(inner_skb) - inner_skb->head) -
3368 SKB_GSO_CB(inner_skb)->mac_offset;
3369}
3370
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00003371static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3372{
3373 int new_headroom, headroom;
3374 int ret;
3375
3376 headroom = skb_headroom(skb);
3377 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3378 if (ret)
3379 return ret;
3380
3381 new_headroom = skb_headroom(skb);
3382 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3383 return 0;
3384}
3385
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003386/* Compute the checksum for a gso segment. First compute the checksum value
3387 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3388 * then add in skb->csum (checksum from csum_start to end of packet).
3389 * skb->csum and csum_start are then updated to reflect the checksum of the
3390 * resultant packet starting from the transport header-- the resultant checksum
3391 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3392 * header.
3393 */
3394static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3395{
3396 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
Eric Dumazetc91d4602015-05-15 05:48:07 -07003397 skb_transport_offset(skb);
3398 __wsum partial;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003399
Eric Dumazetc91d4602015-05-15 05:48:07 -07003400 partial = csum_partial(skb_transport_header(skb), plen, skb->csum);
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003401 skb->csum = res;
3402 SKB_GSO_CB(skb)->csum_start -= plen;
3403
Eric Dumazetc91d4602015-05-15 05:48:07 -07003404 return csum_fold(partial);
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003405}
3406
David S. Millerbdcc0922012-03-07 20:53:36 -05003407static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07003408{
3409 return skb_shinfo(skb)->gso_size;
3410}
3411
Eric Dumazet36a8f392013-09-29 01:21:32 -07003412/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05003413static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02003414{
3415 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3416}
3417
Joe Perches7965bd42013-09-26 14:48:15 -07003418void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07003419
3420static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3421{
3422 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3423 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003424 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3425
Alexander Duyckb78462e2010-06-02 12:24:37 +00003426 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3427 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07003428 __skb_warn_lro_forwarding(skb);
3429 return true;
3430 }
3431 return false;
3432}
3433
Herbert Xu35fc92a2007-03-26 23:22:20 -07003434static inline void skb_forward_csum(struct sk_buff *skb)
3435{
3436 /* Unfortunately we don't support this one. Any brave souls? */
3437 if (skb->ip_summed == CHECKSUM_COMPLETE)
3438 skb->ip_summed = CHECKSUM_NONE;
3439}
3440
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003441/**
3442 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3443 * @skb: skb to check
3444 *
3445 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3446 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3447 * use this helper, to document places where we make this assertion.
3448 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003449static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003450{
3451#ifdef DEBUG
3452 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3453#endif
3454}
3455
Rusty Russellf35d9d82008-02-04 23:49:54 -05003456bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00003457
Paul Durranted1f50c2014-01-09 10:02:46 +00003458int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
Linus Lüssing9afd85c2015-05-02 14:01:07 +02003459struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
3460 unsigned int transport_len,
3461 __sum16(*skb_chkf)(struct sk_buff *skb));
Paul Durranted1f50c2014-01-09 10:02:46 +00003462
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00003463/**
3464 * skb_head_is_locked - Determine if the skb->head is locked down
3465 * @skb: skb to check
3466 *
3467 * The head on skbs build around a head frag can be removed if they are
3468 * not cloned. This function returns true if the skb head is locked down
3469 * due to either being allocated via kmalloc, or by being a clone with
3470 * multiple references to the head.
3471 */
3472static inline bool skb_head_is_locked(const struct sk_buff *skb)
3473{
3474 return !skb->head_frag || skb_cloned(skb);
3475}
Florian Westphalfe6cc552014-02-13 23:09:12 +01003476
3477/**
3478 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3479 *
3480 * @skb: GSO skb
3481 *
3482 * skb_gso_network_seglen is used to determine the real size of the
3483 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3484 *
3485 * The MAC/L2 header is not accounted for.
3486 */
3487static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3488{
3489 unsigned int hdr_len = skb_transport_header(skb) -
3490 skb_network_header(skb);
3491 return hdr_len + skb_gso_transport_seglen(skb);
3492}
Thomas Grafee122c72015-07-21 10:43:58 +02003493
Linus Torvalds1da177e2005-04-16 15:20:36 -07003494#endif /* __KERNEL__ */
3495#endif /* _LINUX_SKBUFF_H */