blob: d97f2d07d02b71a967488bd445d2e91dd0b51c58 [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>
23
Arun Sharma600634972011-07-26 16:09:06 -070024#include <linux/atomic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <asm/types.h>
26#include <linux/spinlock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <linux/net.h>
Thomas Graf3fc7e8a2005-06-23 21:00:17 -070028#include <linux/textsearch.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070029#include <net/checksum.h>
Al Viroa80958f2006-12-04 20:41:19 +000030#include <linux/rcupdate.h>
Chris Leech97fc2f02006-05-23 17:55:33 -070031#include <linux/dmaengine.h>
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -070032#include <linux/hrtimer.h>
Ian Campbell131ea662011-08-19 06:25:00 +000033#include <linux/dma-mapping.h>
Michał Mirosławc8f44af2011-11-15 15:29:55 +000034#include <linux/netdev_features.h>
Jason Wang5203cd22013-03-26 23:11:21 +000035#include <net/flow_keys.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070036
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010037/* A. Checksumming of received packets by device.
38 *
39 * CHECKSUM_NONE:
40 *
41 * Device failed to checksum this packet e.g. due to lack of capabilities.
42 * The packet contains full (though not verified) checksum in packet but
43 * not in skb->csum. Thus, skb->csum is undefined in this case.
44 *
45 * CHECKSUM_UNNECESSARY:
46 *
47 * The hardware you're dealing with doesn't calculate the full checksum
48 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
49 * for specific protocols e.g. TCP/UDP/SCTP, then, for such packets it will
50 * set CHECKSUM_UNNECESSARY if their checksums are okay. skb->csum is still
51 * undefined in this case though. It is a bad option, but, unfortunately,
52 * nowadays most vendors do this. Apparently with the secret goal to sell
53 * you new devices, when you will add new protocol to your host, f.e. IPv6 8)
54 *
55 * CHECKSUM_COMPLETE:
56 *
57 * This is the most generic way. The device supplied checksum of the _whole_
58 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
59 * hardware doesn't need to parse L3/L4 headers to implement this.
60 *
61 * Note: Even if device supports only some protocols, but is able to produce
62 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
63 *
64 * CHECKSUM_PARTIAL:
65 *
66 * This is identical to the case for output below. This may occur on a packet
67 * received directly from another Linux OS, e.g., a virtualized Linux kernel
68 * on the same host. The packet can be treated in the same way as
69 * CHECKSUM_UNNECESSARY, except that on output (i.e., forwarding) the
70 * checksum must be filled in by the OS or the hardware.
71 *
72 * B. Checksumming on output.
73 *
74 * CHECKSUM_NONE:
75 *
76 * The skb was already checksummed by the protocol, or a checksum is not
77 * required.
78 *
79 * CHECKSUM_PARTIAL:
80 *
81 * The device is required to checksum the packet as seen by hard_start_xmit()
82 * from skb->csum_start up to the end, and to record/write the checksum at
83 * offset skb->csum_start + skb->csum_offset.
84 *
85 * The device must show its capabilities in dev->features, set up at device
86 * setup time, e.g. netdev_features.h:
87 *
88 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
89 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
90 * IPv4. Sigh. Vendors like this way for an unknown reason.
91 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
92 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
93 * NETIF_F_... - Well, you get the picture.
94 *
95 * CHECKSUM_UNNECESSARY:
96 *
97 * Normally, the device will do per protocol specific checksumming. Protocol
98 * implementations that do not want the NIC to perform the checksum
99 * calculation should use this flag in their outgoing skbs.
100 *
101 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
102 * offload. Correspondingly, the FCoE protocol driver
103 * stack should use CHECKSUM_UNNECESSARY.
104 *
105 * Any questions? No questions, good. --ANK
106 */
107
Herbert Xu60476372007-04-09 11:59:39 -0700108/* Don't change this without changing skb_csum_unnecessary! */
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100109#define CHECKSUM_NONE 0
110#define CHECKSUM_UNNECESSARY 1
111#define CHECKSUM_COMPLETE 2
112#define CHECKSUM_PARTIAL 3
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113
114#define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
115 ~(SMP_CACHE_BYTES - 1))
David S. Millerfc910a22007-03-25 20:27:59 -0700116#define SKB_WITH_OVERHEAD(X) \
Herbert Xudeea84b2007-10-21 16:27:46 -0700117 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
David S. Millerfc910a22007-03-25 20:27:59 -0700118#define SKB_MAX_ORDER(X, ORDER) \
119 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700120#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
121#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
122
Eric Dumazet87fb4b72011-10-13 07:28:54 +0000123/* return minimum truesize of one skb containing X bytes of data */
124#define SKB_TRUESIZE(X) ((X) + \
125 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
126 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
127
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128struct net_device;
David Howells716ea3a2007-04-02 20:19:53 -0700129struct scatterlist;
Jens Axboe9c55e012007-11-06 23:30:13 -0800130struct pipe_inode_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700131
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700132#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700133struct nf_conntrack {
134 atomic_t use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700135};
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700136#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700137
138#ifdef CONFIG_BRIDGE_NETFILTER
139struct nf_bridge_info {
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000140 atomic_t use;
141 unsigned int mask;
142 struct net_device *physindev;
143 struct net_device *physoutdev;
144 unsigned long data[32 / sizeof(unsigned long)];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700145};
146#endif
147
Linus Torvalds1da177e2005-04-16 15:20:36 -0700148struct sk_buff_head {
149 /* These two members must be first. */
150 struct sk_buff *next;
151 struct sk_buff *prev;
152
153 __u32 qlen;
154 spinlock_t lock;
155};
156
157struct sk_buff;
158
Ian Campbell9d4dde52011-12-22 23:39:14 +0000159/* To allow 64K frame to be packed as single skb without frag_list we
160 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
161 * buffers which do not start on a page boundary.
162 *
163 * Since GRO uses frags we allocate at least 16 regardless of page
164 * size.
Anton Blancharda715dea2011-03-27 14:57:26 +0000165 */
Ian Campbell9d4dde52011-12-22 23:39:14 +0000166#if (65536/PAGE_SIZE + 1) < 16
David S. Millereec00952011-03-29 23:34:08 -0700167#define MAX_SKB_FRAGS 16UL
Anton Blancharda715dea2011-03-27 14:57:26 +0000168#else
Ian Campbell9d4dde52011-12-22 23:39:14 +0000169#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
Anton Blancharda715dea2011-03-27 14:57:26 +0000170#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700171
172typedef struct skb_frag_struct skb_frag_t;
173
174struct skb_frag_struct {
Ian Campbella8605c62011-10-19 23:01:49 +0000175 struct {
176 struct page *p;
177 } page;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000178#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
David S. Millera309bb02007-07-30 18:47:03 -0700179 __u32 page_offset;
180 __u32 size;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000181#else
182 __u16 page_offset;
183 __u16 size;
184#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700185};
186
Eric Dumazet9e903e02011-10-18 21:00:24 +0000187static inline unsigned int skb_frag_size(const skb_frag_t *frag)
188{
189 return frag->size;
190}
191
192static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
193{
194 frag->size = size;
195}
196
197static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
198{
199 frag->size += delta;
200}
201
202static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
203{
204 frag->size -= delta;
205}
206
Patrick Ohlyac45f602009-02-12 05:03:37 +0000207#define HAVE_HW_TIME_STAMP
208
209/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800210 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000211 * @hwtstamp: hardware time stamp transformed into duration
212 * since arbitrary point in time
213 * @syststamp: hwtstamp transformed to system time base
214 *
215 * Software time stamps generated by ktime_get_real() are stored in
216 * skb->tstamp. The relation between the different kinds of time
217 * stamps is as follows:
218 *
219 * syststamp and tstamp can be compared against each other in
220 * arbitrary combinations. The accuracy of a
221 * syststamp/tstamp/"syststamp from other device" comparison is
222 * limited by the accuracy of the transformation into system time
223 * base. This depends on the device driver and its underlying
224 * hardware.
225 *
226 * hwtstamps can only be compared against other hwtstamps from
227 * the same device.
228 *
229 * This structure is attached to packets as part of the
230 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
231 */
232struct skb_shared_hwtstamps {
233 ktime_t hwtstamp;
234 ktime_t syststamp;
235};
236
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000237/* Definitions for tx_flags in struct skb_shared_info */
238enum {
239 /* generate hardware time stamp */
240 SKBTX_HW_TSTAMP = 1 << 0,
241
242 /* generate software time stamp */
243 SKBTX_SW_TSTAMP = 1 << 1,
244
245 /* device driver is going to provide hardware time stamp */
246 SKBTX_IN_PROGRESS = 1 << 2,
247
Shirley Maa6686f22011-07-06 12:22:12 +0000248 /* device driver supports TX zero-copy buffers */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000249 SKBTX_DEV_ZEROCOPY = 1 << 3,
Johannes Berg6e3e9392011-11-09 10:15:42 +0100250
251 /* generate wifi status information (where possible) */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000252 SKBTX_WIFI_STATUS = 1 << 4,
Pravin B Shelarc9af6db2013-02-11 09:27:41 +0000253
254 /* This indicates at least one fragment might be overwritten
255 * (as in vmsplice(), sendfile() ...)
256 * If we need to compute a TX checksum, we'll need to copy
257 * all frags to avoid possible bad checksum
258 */
259 SKBTX_SHARED_FRAG = 1 << 5,
Shirley Maa6686f22011-07-06 12:22:12 +0000260};
261
262/*
263 * The callback notifies userspace to release buffers when skb DMA is done in
264 * lower device, the skb last reference should be 0 when calling this.
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000265 * The zerocopy_success argument is true if zero copy transmit occurred,
266 * false on data copy or out of memory error caused by data copy attempt.
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000267 * The ctx field is used to track device context.
268 * The desc field is used to track userspace buffer index.
Shirley Maa6686f22011-07-06 12:22:12 +0000269 */
270struct ubuf_info {
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000271 void (*callback)(struct ubuf_info *, bool zerocopy_success);
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000272 void *ctx;
Shirley Maa6686f22011-07-06 12:22:12 +0000273 unsigned long desc;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000274};
275
Linus Torvalds1da177e2005-04-16 15:20:36 -0700276/* This data is invariant across clones and lives at
277 * the end of the header data, ie. at skb->end.
278 */
279struct skb_shared_info {
Ian Campbell9f42f122012-01-05 07:13:39 +0000280 unsigned char nr_frags;
281 __u8 tx_flags;
Herbert Xu79671682006-06-22 02:40:14 -0700282 unsigned short gso_size;
283 /* Warning: this field is not always filled in (UFO)! */
284 unsigned short gso_segs;
285 unsigned short gso_type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000287 struct skb_shared_hwtstamps hwtstamps;
Ian Campbell9f42f122012-01-05 07:13:39 +0000288 __be32 ip6_frag_id;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700289
290 /*
291 * Warning : all fields before dataref are cleared in __alloc_skb()
292 */
293 atomic_t dataref;
294
Johann Baudy69e3c752009-05-18 22:11:22 -0700295 /* Intermediate layers must ensure that destructor_arg
296 * remains valid until skb destructor */
297 void * destructor_arg;
Shirley Maa6686f22011-07-06 12:22:12 +0000298
Eric Dumazetfed66382010-07-22 19:09:08 +0000299 /* must be last field, see pskb_expand_head() */
300 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700301};
302
303/* We divide dataref into two halves. The higher 16 bits hold references
304 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700305 * the entire skb->data. A clone of a headerless skb holds the length of
306 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307 *
308 * All users must obey the rule that the skb->data reference count must be
309 * greater than or equal to the payload reference count.
310 *
311 * Holding a reference to the payload part means that the user does not
312 * care about modifications to the header part of skb->data.
313 */
314#define SKB_DATAREF_SHIFT 16
315#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
316
David S. Millerd179cd12005-08-17 14:57:30 -0700317
318enum {
319 SKB_FCLONE_UNAVAILABLE,
320 SKB_FCLONE_ORIG,
321 SKB_FCLONE_CLONE,
322};
323
Herbert Xu79671682006-06-22 02:40:14 -0700324enum {
325 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700326 SKB_GSO_UDP = 1 << 1,
Herbert Xu576a30e2006-06-27 13:22:38 -0700327
328 /* This indicates the skb is from an untrusted source. */
329 SKB_GSO_DODGY = 1 << 2,
Michael Chanb0da85372006-06-29 12:30:00 -0700330
331 /* This indicates the tcp segment has CWR set. */
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700332 SKB_GSO_TCP_ECN = 1 << 3,
333
334 SKB_GSO_TCPV6 = 1 << 4,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800335
336 SKB_GSO_FCOE = 1 << 5,
Pravin B Shelar68c33162013-02-14 14:02:41 +0000337
338 SKB_GSO_GRE = 1 << 6,
Pravin B Shelar73136262013-03-07 13:21:51 +0000339
Eric Dumazetcb32f512013-10-19 11:42:57 -0700340 SKB_GSO_IPIP = 1 << 7,
Simon Horman0d89d202013-05-23 21:02:52 +0000341
Eric Dumazet61c1db72013-10-20 20:47:30 -0700342 SKB_GSO_SIT = 1 << 8,
Eric Dumazetcb32f512013-10-19 11:42:57 -0700343
Eric Dumazet61c1db72013-10-20 20:47:30 -0700344 SKB_GSO_UDP_TUNNEL = 1 << 9,
345
346 SKB_GSO_MPLS = 1 << 10,
Herbert Xu79671682006-06-22 02:40:14 -0700347};
348
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700349#if BITS_PER_LONG > 32
350#define NET_SKBUFF_DATA_USES_OFFSET 1
351#endif
352
353#ifdef NET_SKBUFF_DATA_USES_OFFSET
354typedef unsigned int sk_buff_data_t;
355#else
356typedef unsigned char *sk_buff_data_t;
357#endif
358
Linus Torvalds1da177e2005-04-16 15:20:36 -0700359/**
360 * struct sk_buff - socket buffer
361 * @next: Next buffer in list
362 * @prev: Previous buffer in list
Herbert Xu325ed822005-10-03 13:57:23 -0700363 * @tstamp: Time we arrived
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700364 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700365 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700366 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000367 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700368 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700369 * @len: Length of actual data
370 * @data_len: Data length
371 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700372 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700373 * @csum: Checksum (must include start/offset pair)
374 * @csum_start: Offset from skb->head where checksumming should start
375 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700376 * @priority: Packet queueing priority
Martin Waitz67be2dd2005-05-01 08:59:26 -0700377 * @local_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700378 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700379 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700380 * @nohdr: Payload reference only, must not modify header
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700381 * @nfctinfo: Relationship of this skb to the connection
Linus Torvalds1da177e2005-04-16 15:20:36 -0700382 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700383 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700384 * @ipvs_property: skbuff is owned by ipvs
Randy Dunlap31729362008-02-18 20:52:13 -0800385 * @peeked: this packet has been seen already, so stats have been
386 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700387 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700388 * @protocol: Packet protocol from driver
389 * @destructor: Destruct function
390 * @nfct: Associated connection, if any
Linus Torvalds1da177e2005-04-16 15:20:36 -0700391 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800392 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700393 * @tc_index: Traffic control index
394 * @tc_verd: traffic control verdict
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700395 * @rxhash: the packet hash computed on receive
396 * @queue_mapping: Queue mapping for multiqueue devices
Randy Dunlap553a5672008-04-20 10:51:01 -0700397 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700398 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Changli Gao4ca24622011-08-19 07:26:44 -0700399 * @l4_rxhash: indicate rxhash is a canonical 4-tuple hash over transport
400 * ports.
Johannes Berg6e3e9392011-11-09 10:15:42 +0100401 * @wifi_acked_valid: wifi_acked was set
402 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000403 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Randy Dunlapf4b8ea72006-06-22 16:00:11 -0700404 * @dma_cookie: a cookie to one of several possible DMA operations
405 * done by skb DMA functions
Eliezer Tamir06021292013-06-10 11:39:50 +0300406 * @napi_id: id of the NAPI struct this skb came from
James Morris984bc162006-06-09 00:29:17 -0700407 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700408 * @mark: Generic packet mark
409 * @dropcount: total number of sk_receive_queue overflows
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000410 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b02008-07-14 22:49:06 -0700411 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000412 * @inner_protocol: Protocol (encapsulation)
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000413 * @inner_transport_header: Inner transport layer header (encapsulation)
414 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000415 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700416 * @transport_header: Transport layer header
417 * @network_header: Network layer header
418 * @mac_header: Link layer header
419 * @tail: Tail pointer
420 * @end: End pointer
421 * @head: Head of buffer
422 * @data: Data head pointer
423 * @truesize: Buffer size
424 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425 */
426
427struct sk_buff {
428 /* These two members must be first. */
429 struct sk_buff *next;
430 struct sk_buff *prev;
431
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700432 ktime_t tstamp;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000433
434 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437 /*
438 * This is the control buffer. It is free to use for every
439 * layer. Please put your private variables there. If you
440 * want to keep them across layers you have to do a skb_clone()
441 * first. This is owned by whoever has the skb queued ATM.
442 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000443 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700444
Eric Dumazet7fee2262010-05-11 23:19:48 +0000445 unsigned long _skb_refdst;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000446#ifdef CONFIG_XFRM
447 struct sec_path *sp;
448#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700450 data_len;
451 __u16 mac_len,
452 hdr_len;
Al Viroff1dcad2006-11-20 18:07:29 -0800453 union {
454 __wsum csum;
Herbert Xu663ead32007-04-09 11:59:07 -0700455 struct {
456 __u16 csum_start;
457 __u16 csum_offset;
458 };
Al Viroff1dcad2006-11-20 18:07:29 -0800459 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460 __u32 priority;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200461 kmemcheck_bitfield_begin(flags1);
Thomas Graf1cbb3382005-07-05 14:13:41 -0700462 __u8 local_df:1,
463 cloned:1,
464 ip_summed:2,
Harald Welte6869c4d2005-08-09 19:24:19 -0700465 nohdr:1,
466 nfctinfo:3;
David S. Millerd179cd12005-08-17 14:57:30 -0700467 __u8 pkt_type:3,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800468 fclone:2,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700469 ipvs_property:1,
Herbert Xua59322b2007-12-05 01:53:40 -0800470 peeked:1,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700471 nf_trace:1;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200472 kmemcheck_bitfield_end(flags1);
Eric Dumazet4ab408d2010-03-01 03:09:26 +0000473 __be16 protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700474
475 void (*destructor)(struct sk_buff *skb);
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800476#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700477 struct nf_conntrack *nfct;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100478#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700479#ifdef CONFIG_BRIDGE_NETFILTER
480 struct nf_bridge_info *nf_bridge;
481#endif
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700482
Eric Dumazet8964be42009-11-20 15:35:04 -0800483 int skb_iif;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000484
485 __u32 rxhash;
486
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000487 __be16 vlan_proto;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000488 __u16 vlan_tci;
489
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700491 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700493 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200496
Eric Dumazet0a148422011-04-20 09:27:32 +0000497 __u16 queue_mapping;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200498 kmemcheck_bitfield_begin(flags2);
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400499#ifdef CONFIG_IPV6_NDISC_NODETYPE
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000500 __u8 ndisc_nodetype:2;
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400501#endif
Mel Gormanc93bdd02012-07-31 16:44:19 -0700502 __u8 pfmemalloc:1;
Tom Herbert3853b582010-11-21 13:17:29 +0000503 __u8 ooo_okay:1;
Tom Herbertbdeab992011-08-14 19:45:55 +0000504 __u8 l4_rxhash:1;
Johannes Berg6e3e9392011-11-09 10:15:42 +0100505 __u8 wifi_acked_valid:1;
506 __u8 wifi_acked:1;
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000507 __u8 no_fcs:1;
Eric Dumazetd3836f22012-04-27 00:33:38 +0000508 __u8 head_frag:1;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000509 /* Encapsulation protocol and NIC drivers should use
510 * this flag to indicate to each other if the skb contains
511 * encapsulated packet or not and maybe use the inner packet
512 * headers if needed
513 */
514 __u8 encapsulation:1;
Nicolas Dichtel45906722013-09-30 14:16:41 +0200515 /* 6/8 bit hole (depending on ndisc_nodetype presence) */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200516 kmemcheck_bitfield_end(flags2);
517
Cong Wange0d10952013-08-01 11:10:25 +0800518#if defined CONFIG_NET_DMA || defined CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir06021292013-06-10 11:39:50 +0300519 union {
520 unsigned int napi_id;
521 dma_cookie_t dma_cookie;
522 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700523#endif
James Morris984bc162006-06-09 00:29:17 -0700524#ifdef CONFIG_NETWORK_SECMARK
525 __u32 secmark;
526#endif
Neil Horman3b885782009-10-12 13:26:31 -0700527 union {
528 __u32 mark;
529 __u32 dropcount;
Eric Dumazet16fad692013-03-14 05:40:32 +0000530 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700531 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700532
Simon Horman0d89d202013-05-23 21:02:52 +0000533 __be16 inner_protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000534 __u16 inner_transport_header;
535 __u16 inner_network_header;
536 __u16 inner_mac_header;
537 __u16 transport_header;
538 __u16 network_header;
539 __u16 mac_header;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700541 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700542 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700544 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700545 unsigned int truesize;
546 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547};
548
549#ifdef __KERNEL__
550/*
551 * Handling routines are only of interest to the kernel
552 */
553#include <linux/slab.h>
554
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555
Mel Gormanc93bdd02012-07-31 16:44:19 -0700556#define SKB_ALLOC_FCLONE 0x01
557#define SKB_ALLOC_RX 0x02
558
559/* Returns true if the skb was allocated from PFMEMALLOC reserves */
560static inline bool skb_pfmemalloc(const struct sk_buff *skb)
561{
562 return unlikely(skb->pfmemalloc);
563}
564
Eric Dumazet7fee2262010-05-11 23:19:48 +0000565/*
566 * skb might have a dst pointer attached, refcounted or not.
567 * _skb_refdst low order bit is set if refcount was _not_ taken
568 */
569#define SKB_DST_NOREF 1UL
570#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
571
572/**
573 * skb_dst - returns skb dst_entry
574 * @skb: buffer
575 *
576 * Returns skb dst_entry, regardless of reference taken or not.
577 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000578static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
579{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000580 /* If refdst was not refcounted, check we still are in a
581 * rcu_read_lock section
582 */
583 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
584 !rcu_read_lock_held() &&
585 !rcu_read_lock_bh_held());
586 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000587}
588
Eric Dumazet7fee2262010-05-11 23:19:48 +0000589/**
590 * skb_dst_set - sets skb dst
591 * @skb: buffer
592 * @dst: dst entry
593 *
594 * Sets skb dst, assuming a reference was taken on dst and should
595 * be released by skb_dst_drop()
596 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000597static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
598{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000599 skb->_skb_refdst = (unsigned long)dst;
600}
601
Joe Perches7965bd42013-09-26 14:48:15 -0700602void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
603 bool force);
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200604
605/**
606 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
607 * @skb: buffer
608 * @dst: dst entry
609 *
610 * Sets skb dst, assuming a reference was not taken on dst.
611 * If dst entry is cached, we do not take reference and dst_release
612 * will be avoided by refdst_drop. If dst entry is not cached, we take
613 * reference, so that last dst_release can destroy the dst immediately.
614 */
615static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
616{
617 __skb_dst_set_noref(skb, dst, false);
618}
619
620/**
621 * skb_dst_set_noref_force - sets skb dst, without taking reference
622 * @skb: buffer
623 * @dst: dst entry
624 *
625 * Sets skb dst, assuming a reference was not taken on dst.
626 * No reference is taken and no dst_release will be called. While for
627 * cached dsts deferred reclaim is a basic feature, for entries that are
628 * not cached it is caller's job to guarantee that last dst_release for
629 * provided dst happens when nobody uses it, eg. after a RCU grace period.
630 */
631static inline void skb_dst_set_noref_force(struct sk_buff *skb,
632 struct dst_entry *dst)
633{
634 __skb_dst_set_noref(skb, dst, true);
635}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000636
637/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300638 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000639 * @skb: buffer
640 */
641static inline bool skb_dst_is_noref(const struct sk_buff *skb)
642{
643 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000644}
645
Eric Dumazet511c3f92009-06-02 05:14:27 +0000646static inline struct rtable *skb_rtable(const struct sk_buff *skb)
647{
Eric Dumazetadf30902009-06-02 05:19:30 +0000648 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000649}
650
Joe Perches7965bd42013-09-26 14:48:15 -0700651void kfree_skb(struct sk_buff *skb);
652void kfree_skb_list(struct sk_buff *segs);
653void skb_tx_error(struct sk_buff *skb);
654void consume_skb(struct sk_buff *skb);
655void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000656extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000657
Joe Perches7965bd42013-09-26 14:48:15 -0700658void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
659bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
660 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000661
Joe Perches7965bd42013-09-26 14:48:15 -0700662struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
663 int node);
664struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700665static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100666 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700667{
Eric Dumazet564824b2010-10-11 19:05:25 +0000668 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700669}
670
671static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100672 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700673{
Mel Gormanc93bdd02012-07-31 16:44:19 -0700674 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700675}
676
Joe Perches7965bd42013-09-26 14:48:15 -0700677struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
Patrick McHardy0ebd0ac2013-04-17 06:46:58 +0000678static inline struct sk_buff *alloc_skb_head(gfp_t priority)
679{
680 return __alloc_skb_head(priority, -1);
681}
682
Joe Perches7965bd42013-09-26 14:48:15 -0700683struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
684int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
685struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
686struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
687struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom, gfp_t gfp_mask);
Eric Dumazet117632e2011-12-03 21:39:53 +0000688
Joe Perches7965bd42013-09-26 14:48:15 -0700689int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
690struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
691 unsigned int headroom);
692struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
693 int newtailroom, gfp_t priority);
694int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
695 int len);
696int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
697int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000698#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700699
Joe Perches7965bd42013-09-26 14:48:15 -0700700int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
701 int getfrag(void *from, char *to, int offset,
702 int len, int odd, struct sk_buff *skb),
703 void *from, int length);
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700704
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800705struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700706 __u32 lower_offset;
707 __u32 upper_offset;
708 __u32 frag_idx;
709 __u32 stepped_offset;
710 struct sk_buff *root_skb;
711 struct sk_buff *cur_skb;
712 __u8 *frag_data;
713};
714
Joe Perches7965bd42013-09-26 14:48:15 -0700715void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
716 unsigned int to, struct skb_seq_state *st);
717unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
718 struct skb_seq_state *st);
719void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -0700720
Joe Perches7965bd42013-09-26 14:48:15 -0700721unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
722 unsigned int to, struct ts_config *config,
723 struct ts_state *state);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700724
Tom Herbert09323cc2013-12-15 22:16:19 -0800725/*
726 * Packet hash types specify the type of hash in skb_set_hash.
727 *
728 * Hash types refer to the protocol layer addresses which are used to
729 * construct a packet's hash. The hashes are used to differentiate or identify
730 * flows of the protocol layer for the hash type. Hash types are either
731 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
732 *
733 * Properties of hashes:
734 *
735 * 1) Two packets in different flows have different hash values
736 * 2) Two packets in the same flow should have the same hash value
737 *
738 * A hash at a higher layer is considered to be more specific. A driver should
739 * set the most specific hash possible.
740 *
741 * A driver cannot indicate a more specific hash than the layer at which a hash
742 * was computed. For instance an L3 hash cannot be set as an L4 hash.
743 *
744 * A driver may indicate a hash level which is less specific than the
745 * actual layer the hash was computed on. For instance, a hash computed
746 * at L4 may be considered an L3 hash. This should only be done if the
747 * driver can't unambiguously determine that the HW computed the hash at
748 * the higher layer. Note that the "should" in the second property above
749 * permits this.
750 */
751enum pkt_hash_types {
752 PKT_HASH_TYPE_NONE, /* Undefined type */
753 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
754 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
755 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
756};
757
758static inline void
759skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
760{
761 skb->l4_rxhash = (type == PKT_HASH_TYPE_L4);
762 skb->rxhash = hash;
763}
764
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800765void __skb_get_hash(struct sk_buff *skb);
766static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000767{
Willem de Bruijnecd5cf52012-10-26 11:52:08 +0000768 if (!skb->l4_rxhash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800769 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000770
771 return skb->rxhash;
772}
773
Tom Herbert7539fad2013-12-15 22:12:18 -0800774static inline void skb_clear_hash(struct sk_buff *skb)
775{
776 skb->rxhash = 0;
777 skb->l4_rxhash = 0;
778}
779
780static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
781{
782 if (!skb->l4_rxhash)
783 skb_clear_hash(skb);
784}
785
Tom Herbert3df7a742013-12-15 22:16:29 -0800786static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
787{
788 to->rxhash = from->rxhash;
789 to->l4_rxhash = from->l4_rxhash;
790};
791
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700792#ifdef NET_SKBUFF_DATA_USES_OFFSET
793static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
794{
795 return skb->head + skb->end;
796}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000797
798static inline unsigned int skb_end_offset(const struct sk_buff *skb)
799{
800 return skb->end;
801}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700802#else
803static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
804{
805 return skb->end;
806}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000807
808static inline unsigned int skb_end_offset(const struct sk_buff *skb)
809{
810 return skb->end - skb->head;
811}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700812#endif
813
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700815#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816
Patrick Ohlyac45f602009-02-12 05:03:37 +0000817static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
818{
819 return &skb_shinfo(skb)->hwtstamps;
820}
821
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822/**
823 * skb_queue_empty - check if a queue is empty
824 * @list: queue head
825 *
826 * Returns true if the queue is empty, false otherwise.
827 */
828static inline int skb_queue_empty(const struct sk_buff_head *list)
829{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100830 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700831}
832
833/**
David S. Millerfc7ebb22008-09-23 00:34:07 -0700834 * skb_queue_is_last - check if skb is the last entry in the queue
835 * @list: queue head
836 * @skb: buffer
837 *
838 * Returns true if @skb is the last buffer on the list.
839 */
840static inline bool skb_queue_is_last(const struct sk_buff_head *list,
841 const struct sk_buff *skb)
842{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100843 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -0700844}
845
846/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800847 * skb_queue_is_first - check if skb is the first entry in the queue
848 * @list: queue head
849 * @skb: buffer
850 *
851 * Returns true if @skb is the first buffer on the list.
852 */
853static inline bool skb_queue_is_first(const struct sk_buff_head *list,
854 const struct sk_buff *skb)
855{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100856 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800857}
858
859/**
David S. Miller249c8b42008-09-23 00:44:42 -0700860 * skb_queue_next - return the next packet in the queue
861 * @list: queue head
862 * @skb: current buffer
863 *
864 * Return the next packet in @list after @skb. It is only valid to
865 * call this if skb_queue_is_last() evaluates to false.
866 */
867static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
868 const struct sk_buff *skb)
869{
870 /* This BUG_ON may seem severe, but if we just return then we
871 * are going to dereference garbage.
872 */
873 BUG_ON(skb_queue_is_last(list, skb));
874 return skb->next;
875}
876
877/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800878 * skb_queue_prev - return the prev packet in the queue
879 * @list: queue head
880 * @skb: current buffer
881 *
882 * Return the prev packet in @list before @skb. It is only valid to
883 * call this if skb_queue_is_first() evaluates to false.
884 */
885static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
886 const struct sk_buff *skb)
887{
888 /* This BUG_ON may seem severe, but if we just return then we
889 * are going to dereference garbage.
890 */
891 BUG_ON(skb_queue_is_first(list, skb));
892 return skb->prev;
893}
894
895/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896 * skb_get - reference buffer
897 * @skb: buffer to reference
898 *
899 * Makes another reference to a socket buffer and returns a pointer
900 * to the buffer.
901 */
902static inline struct sk_buff *skb_get(struct sk_buff *skb)
903{
904 atomic_inc(&skb->users);
905 return skb;
906}
907
908/*
909 * If users == 1, we are the only owner and are can avoid redundant
910 * atomic change.
911 */
912
913/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700914 * skb_cloned - is the buffer a clone
915 * @skb: buffer to check
916 *
917 * Returns true if the buffer was generated with skb_clone() and is
918 * one of multiple shared copies of the buffer. Cloned buffers are
919 * shared data so must not be written to under normal circumstances.
920 */
921static inline int skb_cloned(const struct sk_buff *skb)
922{
923 return skb->cloned &&
924 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
925}
926
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +0000927static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
928{
929 might_sleep_if(pri & __GFP_WAIT);
930
931 if (skb_cloned(skb))
932 return pskb_expand_head(skb, 0, 0, pri);
933
934 return 0;
935}
936
Linus Torvalds1da177e2005-04-16 15:20:36 -0700937/**
938 * skb_header_cloned - is the header a clone
939 * @skb: buffer to check
940 *
941 * Returns true if modifying the header part of the buffer requires
942 * the data to be copied.
943 */
944static inline int skb_header_cloned(const struct sk_buff *skb)
945{
946 int dataref;
947
948 if (!skb->cloned)
949 return 0;
950
951 dataref = atomic_read(&skb_shinfo(skb)->dataref);
952 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
953 return dataref != 1;
954}
955
956/**
957 * skb_header_release - release reference to header
958 * @skb: buffer to operate on
959 *
960 * Drop a reference to the header part of the buffer. This is done
961 * by acquiring a payload reference. You must not read from the header
962 * part of skb->data after this.
963 */
964static inline void skb_header_release(struct sk_buff *skb)
965{
966 BUG_ON(skb->nohdr);
967 skb->nohdr = 1;
968 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
969}
970
971/**
972 * skb_shared - is the buffer shared
973 * @skb: buffer to check
974 *
975 * Returns true if more than one person has a reference to this
976 * buffer.
977 */
978static inline int skb_shared(const struct sk_buff *skb)
979{
980 return atomic_read(&skb->users) != 1;
981}
982
983/**
984 * skb_share_check - check if buffer is shared and if so clone it
985 * @skb: buffer to check
986 * @pri: priority for memory allocation
987 *
988 * If the buffer is shared the buffer is cloned and the old copy
989 * drops a reference. A new clone with a single reference is returned.
990 * If the buffer is not shared the original buffer is returned. When
991 * being called from interrupt status or with spinlocks held pri must
992 * be GFP_ATOMIC.
993 *
994 * NULL is returned on a memory allocation failure.
995 */
Eric Dumazet47061bc2012-08-03 20:54:15 +0000996static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700997{
998 might_sleep_if(pri & __GFP_WAIT);
999 if (skb_shared(skb)) {
1000 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001001
1002 if (likely(nskb))
1003 consume_skb(skb);
1004 else
1005 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006 skb = nskb;
1007 }
1008 return skb;
1009}
1010
1011/*
1012 * Copy shared buffers into a new sk_buff. We effectively do COW on
1013 * packets to handle cases where we have a local reader and forward
1014 * and a couple of other messy ones. The normal one is tcpdumping
1015 * a packet thats being forwarded.
1016 */
1017
1018/**
1019 * skb_unshare - make a copy of a shared buffer
1020 * @skb: buffer to check
1021 * @pri: priority for memory allocation
1022 *
1023 * If the socket buffer is a clone then this function creates a new
1024 * copy of the data, drops a reference count on the old copy and returns
1025 * the new copy with the reference count at 1. If the buffer is not a clone
1026 * the original buffer is returned. When called with a spinlock held or
1027 * from interrupt state @pri must be %GFP_ATOMIC
1028 *
1029 * %NULL is returned on a memory allocation failure.
1030 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001031static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001032 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001033{
1034 might_sleep_if(pri & __GFP_WAIT);
1035 if (skb_cloned(skb)) {
1036 struct sk_buff *nskb = skb_copy(skb, pri);
1037 kfree_skb(skb); /* Free our shared copy */
1038 skb = nskb;
1039 }
1040 return skb;
1041}
1042
1043/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001044 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001045 * @list_: list to peek at
1046 *
1047 * Peek an &sk_buff. Unlike most other operations you _MUST_
1048 * be careful with this one. A peek leaves the buffer on the
1049 * list and someone else may run off with it. You must hold
1050 * the appropriate locks or have a private queue to do this.
1051 *
1052 * Returns %NULL for an empty list or a pointer to the head element.
1053 * The reference count is not incremented and the reference is therefore
1054 * volatile. Use with caution.
1055 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001056static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001057{
Eric Dumazet18d07002012-04-30 16:31:46 +00001058 struct sk_buff *skb = list_->next;
1059
1060 if (skb == (struct sk_buff *)list_)
1061 skb = NULL;
1062 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001063}
1064
1065/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001066 * skb_peek_next - peek skb following the given one from a queue
1067 * @skb: skb to start from
1068 * @list_: list to peek at
1069 *
1070 * Returns %NULL when the end of the list is met or a pointer to the
1071 * next element. The reference count is not incremented and the
1072 * reference is therefore volatile. Use with caution.
1073 */
1074static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1075 const struct sk_buff_head *list_)
1076{
1077 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001078
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001079 if (next == (struct sk_buff *)list_)
1080 next = NULL;
1081 return next;
1082}
1083
1084/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001085 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001086 * @list_: list to peek at
1087 *
1088 * Peek an &sk_buff. Unlike most other operations you _MUST_
1089 * be careful with this one. A peek leaves the buffer on the
1090 * list and someone else may run off with it. You must hold
1091 * the appropriate locks or have a private queue to do this.
1092 *
1093 * Returns %NULL for an empty list or a pointer to the tail element.
1094 * The reference count is not incremented and the reference is therefore
1095 * volatile. Use with caution.
1096 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001097static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001098{
Eric Dumazet18d07002012-04-30 16:31:46 +00001099 struct sk_buff *skb = list_->prev;
1100
1101 if (skb == (struct sk_buff *)list_)
1102 skb = NULL;
1103 return skb;
1104
Linus Torvalds1da177e2005-04-16 15:20:36 -07001105}
1106
1107/**
1108 * skb_queue_len - get queue length
1109 * @list_: list to measure
1110 *
1111 * Return the length of an &sk_buff queue.
1112 */
1113static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1114{
1115 return list_->qlen;
1116}
1117
David S. Miller67fed452008-09-21 22:36:24 -07001118/**
1119 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1120 * @list: queue to initialize
1121 *
1122 * This initializes only the list and queue length aspects of
1123 * an sk_buff_head object. This allows to initialize the list
1124 * aspects of an sk_buff_head without reinitializing things like
1125 * the spinlock. It can also be used for on-stack sk_buff_head
1126 * objects where the spinlock is known to not be used.
1127 */
1128static inline void __skb_queue_head_init(struct sk_buff_head *list)
1129{
1130 list->prev = list->next = (struct sk_buff *)list;
1131 list->qlen = 0;
1132}
1133
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001134/*
1135 * This function creates a split out lock class for each invocation;
1136 * this is needed for now since a whole lot of users of the skb-queue
1137 * infrastructure in drivers have different locking usage (in hardirq)
1138 * than the networking core (in softirq only). In the long run either the
1139 * network layer or drivers should need annotation to consolidate the
1140 * main types of usage into 3 classes.
1141 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142static inline void skb_queue_head_init(struct sk_buff_head *list)
1143{
1144 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001145 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146}
1147
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001148static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1149 struct lock_class_key *class)
1150{
1151 skb_queue_head_init(list);
1152 lockdep_set_class(&list->lock, class);
1153}
1154
Linus Torvalds1da177e2005-04-16 15:20:36 -07001155/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001156 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001157 *
1158 * The "__skb_xxxx()" functions are the non-atomic ones that
1159 * can only be called with interrupts disabled.
1160 */
Joe Perches7965bd42013-09-26 14:48:15 -07001161void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1162 struct sk_buff_head *list);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001163static inline void __skb_insert(struct sk_buff *newsk,
1164 struct sk_buff *prev, struct sk_buff *next,
1165 struct sk_buff_head *list)
1166{
1167 newsk->next = next;
1168 newsk->prev = prev;
1169 next->prev = prev->next = newsk;
1170 list->qlen++;
1171}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172
David S. Miller67fed452008-09-21 22:36:24 -07001173static inline void __skb_queue_splice(const struct sk_buff_head *list,
1174 struct sk_buff *prev,
1175 struct sk_buff *next)
1176{
1177 struct sk_buff *first = list->next;
1178 struct sk_buff *last = list->prev;
1179
1180 first->prev = prev;
1181 prev->next = first;
1182
1183 last->next = next;
1184 next->prev = last;
1185}
1186
1187/**
1188 * skb_queue_splice - join two skb lists, this is designed for stacks
1189 * @list: the new list to add
1190 * @head: the place to add it in the first list
1191 */
1192static inline void skb_queue_splice(const struct sk_buff_head *list,
1193 struct sk_buff_head *head)
1194{
1195 if (!skb_queue_empty(list)) {
1196 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001197 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001198 }
1199}
1200
1201/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001202 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001203 * @list: the new list to add
1204 * @head: the place to add it in the first list
1205 *
1206 * The list at @list is reinitialised
1207 */
1208static inline void skb_queue_splice_init(struct sk_buff_head *list,
1209 struct sk_buff_head *head)
1210{
1211 if (!skb_queue_empty(list)) {
1212 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001213 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001214 __skb_queue_head_init(list);
1215 }
1216}
1217
1218/**
1219 * skb_queue_splice_tail - join two skb lists, each list being a queue
1220 * @list: the new list to add
1221 * @head: the place to add it in the first list
1222 */
1223static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1224 struct sk_buff_head *head)
1225{
1226 if (!skb_queue_empty(list)) {
1227 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001228 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001229 }
1230}
1231
1232/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001233 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001234 * @list: the new list to add
1235 * @head: the place to add it in the first list
1236 *
1237 * Each of the lists is a queue.
1238 * The list at @list is reinitialised
1239 */
1240static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1241 struct sk_buff_head *head)
1242{
1243 if (!skb_queue_empty(list)) {
1244 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001245 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001246 __skb_queue_head_init(list);
1247 }
1248}
1249
Linus Torvalds1da177e2005-04-16 15:20:36 -07001250/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001251 * __skb_queue_after - queue a buffer at the list head
1252 * @list: list to use
1253 * @prev: place after this buffer
1254 * @newsk: buffer to queue
1255 *
1256 * Queue a buffer int the middle of a list. This function takes no locks
1257 * and you must therefore hold required locks before calling it.
1258 *
1259 * A buffer cannot be placed on two lists at the same time.
1260 */
1261static inline void __skb_queue_after(struct sk_buff_head *list,
1262 struct sk_buff *prev,
1263 struct sk_buff *newsk)
1264{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001265 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001266}
1267
Joe Perches7965bd42013-09-26 14:48:15 -07001268void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1269 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001270
Gerrit Renkerf5572852008-04-14 00:05:28 -07001271static inline void __skb_queue_before(struct sk_buff_head *list,
1272 struct sk_buff *next,
1273 struct sk_buff *newsk)
1274{
1275 __skb_insert(newsk, next->prev, next, list);
1276}
1277
Stephen Hemminger300ce172005-10-30 13:47:34 -08001278/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001279 * __skb_queue_head - queue a buffer at the list head
1280 * @list: list to use
1281 * @newsk: buffer to queue
1282 *
1283 * Queue a buffer at the start of a list. This function takes no locks
1284 * and you must therefore hold required locks before calling it.
1285 *
1286 * A buffer cannot be placed on two lists at the same time.
1287 */
Joe Perches7965bd42013-09-26 14:48:15 -07001288void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001289static inline void __skb_queue_head(struct sk_buff_head *list,
1290 struct sk_buff *newsk)
1291{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001292 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001293}
1294
1295/**
1296 * __skb_queue_tail - queue a buffer at the list tail
1297 * @list: list to use
1298 * @newsk: buffer to queue
1299 *
1300 * Queue a buffer at the end of a list. This function takes no locks
1301 * and you must therefore hold required locks before calling it.
1302 *
1303 * A buffer cannot be placed on two lists at the same time.
1304 */
Joe Perches7965bd42013-09-26 14:48:15 -07001305void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001306static inline void __skb_queue_tail(struct sk_buff_head *list,
1307 struct sk_buff *newsk)
1308{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001309 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001310}
1311
Linus Torvalds1da177e2005-04-16 15:20:36 -07001312/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001313 * remove sk_buff from list. _Must_ be called atomically, and with
1314 * the list known..
1315 */
Joe Perches7965bd42013-09-26 14:48:15 -07001316void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001317static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1318{
1319 struct sk_buff *next, *prev;
1320
1321 list->qlen--;
1322 next = skb->next;
1323 prev = skb->prev;
1324 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325 next->prev = prev;
1326 prev->next = next;
1327}
1328
Gerrit Renkerf525c062008-04-14 00:04:12 -07001329/**
1330 * __skb_dequeue - remove from the head of the queue
1331 * @list: list to dequeue from
1332 *
1333 * Remove the head of the list. This function does not take any locks
1334 * so must be used with appropriate locks held only. The head item is
1335 * returned or %NULL if the list is empty.
1336 */
Joe Perches7965bd42013-09-26 14:48:15 -07001337struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Gerrit Renkerf525c062008-04-14 00:04:12 -07001338static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1339{
1340 struct sk_buff *skb = skb_peek(list);
1341 if (skb)
1342 __skb_unlink(skb, list);
1343 return skb;
1344}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001345
1346/**
1347 * __skb_dequeue_tail - remove from the tail of the queue
1348 * @list: list to dequeue from
1349 *
1350 * Remove the tail of the list. This function does not take any locks
1351 * so must be used with appropriate locks held only. The tail item is
1352 * returned or %NULL if the list is empty.
1353 */
Joe Perches7965bd42013-09-26 14:48:15 -07001354struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001355static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1356{
1357 struct sk_buff *skb = skb_peek_tail(list);
1358 if (skb)
1359 __skb_unlink(skb, list);
1360 return skb;
1361}
1362
1363
David S. Millerbdcc0922012-03-07 20:53:36 -05001364static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365{
1366 return skb->data_len;
1367}
1368
1369static inline unsigned int skb_headlen(const struct sk_buff *skb)
1370{
1371 return skb->len - skb->data_len;
1372}
1373
1374static inline int skb_pagelen(const struct sk_buff *skb)
1375{
1376 int i, len = 0;
1377
1378 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001379 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001380 return len + skb_headlen(skb);
1381}
1382
Ian Campbell131ea662011-08-19 06:25:00 +00001383/**
1384 * __skb_fill_page_desc - initialise a paged fragment in an skb
1385 * @skb: buffer containing fragment to be initialised
1386 * @i: paged fragment index to initialise
1387 * @page: the page to use for this fragment
1388 * @off: the offset to the data with @page
1389 * @size: the length of the data
1390 *
1391 * Initialises the @i'th fragment of @skb to point to &size bytes at
1392 * offset @off within @page.
1393 *
1394 * Does not take any additional reference on the fragment.
1395 */
1396static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1397 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398{
1399 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1400
Mel Gormanc48a11c2012-07-31 16:44:23 -07001401 /*
1402 * Propagate page->pfmemalloc to the skb if we can. The problem is
1403 * that not all callers have unique ownership of the page. If
1404 * pfmemalloc is set, we check the mapping as a mapping implies
1405 * page->index is set (index and pfmemalloc share space).
1406 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1407 * do not lose pfmemalloc information as the pages would not be
1408 * allocated using __GFP_MEMALLOC.
1409 */
Ian Campbella8605c62011-10-19 23:01:49 +00001410 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001411 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001412 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001413
1414 page = compound_head(page);
1415 if (page->pfmemalloc && !page->mapping)
1416 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00001417}
1418
1419/**
1420 * skb_fill_page_desc - initialise a paged fragment in an skb
1421 * @skb: buffer containing fragment to be initialised
1422 * @i: paged fragment index to initialise
1423 * @page: the page to use for this fragment
1424 * @off: the offset to the data with @page
1425 * @size: the length of the data
1426 *
1427 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01001428 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00001429 * addition updates @skb such that @i is the last fragment.
1430 *
1431 * Does not take any additional reference on the fragment.
1432 */
1433static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1434 struct page *page, int off, int size)
1435{
1436 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437 skb_shinfo(skb)->nr_frags = i + 1;
1438}
1439
Joe Perches7965bd42013-09-26 14:48:15 -07001440void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1441 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001442
Jason Wangf8e617e2013-11-01 14:07:47 +08001443void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1444 unsigned int truesize);
1445
Linus Torvalds1da177e2005-04-16 15:20:36 -07001446#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001447#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001448#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1449
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001450#ifdef NET_SKBUFF_DATA_USES_OFFSET
1451static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1452{
1453 return skb->head + skb->tail;
1454}
1455
1456static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1457{
1458 skb->tail = skb->data - skb->head;
1459}
1460
1461static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1462{
1463 skb_reset_tail_pointer(skb);
1464 skb->tail += offset;
1465}
Simon Horman7cc46192013-05-28 20:34:29 +00001466
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001467#else /* NET_SKBUFF_DATA_USES_OFFSET */
1468static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1469{
1470 return skb->tail;
1471}
1472
1473static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1474{
1475 skb->tail = skb->data;
1476}
1477
1478static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1479{
1480 skb->tail = skb->data + offset;
1481}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001482
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001483#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1484
Linus Torvalds1da177e2005-04-16 15:20:36 -07001485/*
1486 * Add data to an sk_buff
1487 */
Mathias Krause0c7ddf32013-11-07 14:18:24 +01001488unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001489unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001490static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1491{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001492 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001493 SKB_LINEAR_ASSERT(skb);
1494 skb->tail += len;
1495 skb->len += len;
1496 return tmp;
1497}
1498
Joe Perches7965bd42013-09-26 14:48:15 -07001499unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001500static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1501{
1502 skb->data -= len;
1503 skb->len += len;
1504 return skb->data;
1505}
1506
Joe Perches7965bd42013-09-26 14:48:15 -07001507unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001508static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1509{
1510 skb->len -= len;
1511 BUG_ON(skb->len < skb->data_len);
1512 return skb->data += len;
1513}
1514
David S. Miller47d29642010-05-02 02:21:44 -07001515static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1516{
1517 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1518}
1519
Joe Perches7965bd42013-09-26 14:48:15 -07001520unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001521
1522static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1523{
1524 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001525 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001526 return NULL;
1527 skb->len -= len;
1528 return skb->data += len;
1529}
1530
1531static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1532{
1533 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1534}
1535
1536static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1537{
1538 if (likely(len <= skb_headlen(skb)))
1539 return 1;
1540 if (unlikely(len > skb->len))
1541 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001542 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001543}
1544
1545/**
1546 * skb_headroom - bytes at buffer head
1547 * @skb: buffer to check
1548 *
1549 * Return the number of bytes of free space at the head of an &sk_buff.
1550 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001551static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001552{
1553 return skb->data - skb->head;
1554}
1555
1556/**
1557 * skb_tailroom - bytes at buffer end
1558 * @skb: buffer to check
1559 *
1560 * Return the number of bytes of free space at the tail of an sk_buff
1561 */
1562static inline int skb_tailroom(const struct sk_buff *skb)
1563{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001564 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565}
1566
1567/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00001568 * skb_availroom - bytes at buffer end
1569 * @skb: buffer to check
1570 *
1571 * Return the number of bytes of free space at the tail of an sk_buff
1572 * allocated by sk_stream_alloc()
1573 */
1574static inline int skb_availroom(const struct sk_buff *skb)
1575{
Eric Dumazet16fad692013-03-14 05:40:32 +00001576 if (skb_is_nonlinear(skb))
1577 return 0;
1578
1579 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00001580}
1581
1582/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583 * skb_reserve - adjust headroom
1584 * @skb: buffer to alter
1585 * @len: bytes to move
1586 *
1587 * Increase the headroom of an empty &sk_buff by reducing the tail
1588 * room. This is only allowed for an empty buffer.
1589 */
David S. Miller8243126c2006-01-17 02:54:21 -08001590static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001591{
1592 skb->data += len;
1593 skb->tail += len;
1594}
1595
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001596static inline void skb_reset_inner_headers(struct sk_buff *skb)
1597{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001598 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001599 skb->inner_network_header = skb->network_header;
1600 skb->inner_transport_header = skb->transport_header;
1601}
1602
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00001603static inline void skb_reset_mac_len(struct sk_buff *skb)
1604{
1605 skb->mac_len = skb->network_header - skb->mac_header;
1606}
1607
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001608static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1609 *skb)
1610{
1611 return skb->head + skb->inner_transport_header;
1612}
1613
1614static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1615{
1616 skb->inner_transport_header = skb->data - skb->head;
1617}
1618
1619static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1620 const int offset)
1621{
1622 skb_reset_inner_transport_header(skb);
1623 skb->inner_transport_header += offset;
1624}
1625
1626static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1627{
1628 return skb->head + skb->inner_network_header;
1629}
1630
1631static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1632{
1633 skb->inner_network_header = skb->data - skb->head;
1634}
1635
1636static inline void skb_set_inner_network_header(struct sk_buff *skb,
1637 const int offset)
1638{
1639 skb_reset_inner_network_header(skb);
1640 skb->inner_network_header += offset;
1641}
1642
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001643static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1644{
1645 return skb->head + skb->inner_mac_header;
1646}
1647
1648static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1649{
1650 skb->inner_mac_header = skb->data - skb->head;
1651}
1652
1653static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1654 const int offset)
1655{
1656 skb_reset_inner_mac_header(skb);
1657 skb->inner_mac_header += offset;
1658}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001659static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1660{
Cong Wang35d04612013-05-29 15:16:05 +08001661 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001662}
1663
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001664static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1665{
1666 return skb->head + skb->transport_header;
1667}
1668
1669static inline void skb_reset_transport_header(struct sk_buff *skb)
1670{
1671 skb->transport_header = skb->data - skb->head;
1672}
1673
1674static inline void skb_set_transport_header(struct sk_buff *skb,
1675 const int offset)
1676{
1677 skb_reset_transport_header(skb);
1678 skb->transport_header += offset;
1679}
1680
1681static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1682{
1683 return skb->head + skb->network_header;
1684}
1685
1686static inline void skb_reset_network_header(struct sk_buff *skb)
1687{
1688 skb->network_header = skb->data - skb->head;
1689}
1690
1691static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1692{
1693 skb_reset_network_header(skb);
1694 skb->network_header += offset;
1695}
1696
1697static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1698{
1699 return skb->head + skb->mac_header;
1700}
1701
1702static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1703{
Cong Wang35d04612013-05-29 15:16:05 +08001704 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001705}
1706
1707static inline void skb_reset_mac_header(struct sk_buff *skb)
1708{
1709 skb->mac_header = skb->data - skb->head;
1710}
1711
1712static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1713{
1714 skb_reset_mac_header(skb);
1715 skb->mac_header += offset;
1716}
1717
Timo Teräs0e3da5b2013-12-16 11:02:09 +02001718static inline void skb_pop_mac_header(struct sk_buff *skb)
1719{
1720 skb->mac_header = skb->network_header;
1721}
1722
Ying Xuefbbdb8f2013-03-27 16:46:06 +00001723static inline void skb_probe_transport_header(struct sk_buff *skb,
1724 const int offset_hint)
1725{
1726 struct flow_keys keys;
1727
1728 if (skb_transport_header_was_set(skb))
1729 return;
1730 else if (skb_flow_dissect(skb, &keys))
1731 skb_set_transport_header(skb, keys.thoff);
1732 else
1733 skb_set_transport_header(skb, offset_hint);
1734}
1735
Eric Dumazet03606892012-02-23 10:55:02 +00001736static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1737{
1738 if (skb_mac_header_was_set(skb)) {
1739 const unsigned char *old_mac = skb_mac_header(skb);
1740
1741 skb_set_mac_header(skb, -skb->mac_len);
1742 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1743 }
1744}
1745
Michał Mirosław04fb4512010-12-14 15:24:08 +00001746static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1747{
1748 return skb->csum_start - skb_headroom(skb);
1749}
1750
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001751static inline int skb_transport_offset(const struct sk_buff *skb)
1752{
1753 return skb_transport_header(skb) - skb->data;
1754}
1755
1756static inline u32 skb_network_header_len(const struct sk_buff *skb)
1757{
1758 return skb->transport_header - skb->network_header;
1759}
1760
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001761static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1762{
1763 return skb->inner_transport_header - skb->inner_network_header;
1764}
1765
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001766static inline int skb_network_offset(const struct sk_buff *skb)
1767{
1768 return skb_network_header(skb) - skb->data;
1769}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001770
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001771static inline int skb_inner_network_offset(const struct sk_buff *skb)
1772{
1773 return skb_inner_network_header(skb) - skb->data;
1774}
1775
Changli Gaof9599ce2010-08-04 04:43:44 +00001776static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1777{
1778 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1779}
1780
Linus Torvalds1da177e2005-04-16 15:20:36 -07001781/*
1782 * CPUs often take a performance hit when accessing unaligned memory
1783 * locations. The actual performance hit varies, it can be small if the
1784 * hardware handles it or large if we have to take an exception and fix it
1785 * in software.
1786 *
1787 * Since an ethernet header is 14 bytes network drivers often end up with
1788 * the IP header at an unaligned offset. The IP header can be aligned by
1789 * shifting the start of the packet by 2 bytes. Drivers should do this
1790 * with:
1791 *
Tobias Klauser8660c122009-07-13 22:48:16 +00001792 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001793 *
1794 * The downside to this alignment of the IP header is that the DMA is now
1795 * unaligned. On some architectures the cost of an unaligned DMA is high
1796 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00001797 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001798 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1799 * to be overridden.
1800 */
1801#ifndef NET_IP_ALIGN
1802#define NET_IP_ALIGN 2
1803#endif
1804
Anton Blanchard025be812006-03-31 02:27:06 -08001805/*
1806 * The networking layer reserves some headroom in skb data (via
1807 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1808 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08001809 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08001810 *
1811 * Unfortunately this headroom changes the DMA alignment of the resulting
1812 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1813 * on some architectures. An architecture can override this value,
1814 * perhaps setting it to a cacheline in size (since that will maintain
1815 * cacheline alignment of the DMA). It must be a power of 2.
1816 *
David S. Millerd6301d32009-02-08 19:24:13 -08001817 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08001818 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07001819 *
1820 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1821 * to reduce average number of cache lines per packet.
1822 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07001823 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08001824 */
1825#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07001826#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08001827#endif
1828
Joe Perches7965bd42013-09-26 14:48:15 -07001829int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001830
1831static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1832{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00001833 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07001834 WARN_ON(1);
1835 return;
1836 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001837 skb->len = len;
1838 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001839}
1840
Joe Perches7965bd42013-09-26 14:48:15 -07001841void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001842
1843static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1844{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001845 if (skb->data_len)
1846 return ___pskb_trim(skb, len);
1847 __skb_trim(skb, len);
1848 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001849}
1850
1851static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1852{
1853 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1854}
1855
1856/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07001857 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1858 * @skb: buffer to alter
1859 * @len: new length
1860 *
1861 * This is identical to pskb_trim except that the caller knows that
1862 * the skb is not cloned so we should never get an error due to out-
1863 * of-memory.
1864 */
1865static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1866{
1867 int err = pskb_trim(skb, len);
1868 BUG_ON(err);
1869}
1870
1871/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001872 * skb_orphan - orphan a buffer
1873 * @skb: buffer to orphan
1874 *
1875 * If a buffer currently has an owner then we call the owner's
1876 * destructor function and make the @skb unowned. The buffer continues
1877 * to exist but is no longer charged to its former owner.
1878 */
1879static inline void skb_orphan(struct sk_buff *skb)
1880{
Eric Dumazetc34a7612013-07-30 16:11:15 -07001881 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001882 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07001883 skb->destructor = NULL;
1884 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07001885 } else {
1886 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07001887 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001888}
1889
1890/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00001891 * skb_orphan_frags - orphan the frags contained in a buffer
1892 * @skb: buffer to orphan frags from
1893 * @gfp_mask: allocation mask for replacement pages
1894 *
1895 * For each frag in the SKB which needs a destructor (i.e. has an
1896 * owner) create a copy of that frag and release the original
1897 * page by calling the destructor.
1898 */
1899static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1900{
1901 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1902 return 0;
1903 return skb_copy_ubufs(skb, gfp_mask);
1904}
1905
1906/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001907 * __skb_queue_purge - empty a list
1908 * @list: list to empty
1909 *
1910 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1911 * the list and one reference dropped. This function does not take the
1912 * list lock and the caller must hold the relevant locks to use it.
1913 */
Joe Perches7965bd42013-09-26 14:48:15 -07001914void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001915static inline void __skb_queue_purge(struct sk_buff_head *list)
1916{
1917 struct sk_buff *skb;
1918 while ((skb = __skb_dequeue(list)) != NULL)
1919 kfree_skb(skb);
1920}
1921
Alexander Duycke5e67302013-02-08 10:17:15 +00001922#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
1923#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
1924#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
1925
Joe Perches7965bd42013-09-26 14:48:15 -07001926void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001927
Joe Perches7965bd42013-09-26 14:48:15 -07001928struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
1929 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07001930
1931/**
1932 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1933 * @dev: network device to receive on
1934 * @length: length to allocate
1935 *
1936 * Allocate a new &sk_buff and assign it a usage count of one. The
1937 * buffer has unspecified headroom built in. Users should allocate
1938 * the headroom they think they need without accounting for the
1939 * built in space. The built in space is used for optimisations.
1940 *
1941 * %NULL is returned if there is no free memory. Although this function
1942 * allocates memory it can be called from an interrupt.
1943 */
1944static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00001945 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07001946{
1947 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1948}
1949
Eric Dumazet6f532612012-05-18 05:12:12 +00001950/* legacy helper around __netdev_alloc_skb() */
1951static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1952 gfp_t gfp_mask)
1953{
1954 return __netdev_alloc_skb(NULL, length, gfp_mask);
1955}
1956
1957/* legacy helper around netdev_alloc_skb() */
1958static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1959{
1960 return netdev_alloc_skb(NULL, length);
1961}
1962
1963
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001964static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
1965 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00001966{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001967 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00001968
1969 if (NET_IP_ALIGN && skb)
1970 skb_reserve(skb, NET_IP_ALIGN);
1971 return skb;
1972}
1973
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001974static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1975 unsigned int length)
1976{
1977 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
1978}
1979
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01001980/**
1981 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
Mel Gorman06140022012-07-31 16:44:24 -07001982 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1983 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1984 * @order: size of the allocation
1985 *
1986 * Allocate a new page.
1987 *
1988 * %NULL is returned if there is no free memory.
1989*/
1990static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
1991 struct sk_buff *skb,
1992 unsigned int order)
1993{
1994 struct page *page;
1995
1996 gfp_mask |= __GFP_COLD;
1997
1998 if (!(gfp_mask & __GFP_NOMEMALLOC))
1999 gfp_mask |= __GFP_MEMALLOC;
2000
2001 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2002 if (skb && page && page->pfmemalloc)
2003 skb->pfmemalloc = true;
2004
2005 return page;
2006}
2007
2008/**
2009 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2010 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2011 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2012 *
2013 * Allocate a new page.
2014 *
2015 * %NULL is returned if there is no free memory.
2016 */
2017static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
2018 struct sk_buff *skb)
2019{
2020 return __skb_alloc_pages(gfp_mask, skb, 0);
2021}
2022
2023/**
2024 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2025 * @page: The page that was allocated from skb_alloc_page
2026 * @skb: The skb that may need pfmemalloc set
2027 */
2028static inline void skb_propagate_pfmemalloc(struct page *page,
2029 struct sk_buff *skb)
2030{
2031 if (page && page->pfmemalloc)
2032 skb->pfmemalloc = true;
2033}
2034
Eric Dumazet564824b2010-10-11 19:05:25 +00002035/**
Ian Campbell131ea662011-08-19 06:25:00 +00002036 * skb_frag_page - retrieve the page refered to by a paged fragment
2037 * @frag: the paged fragment
2038 *
2039 * Returns the &struct page associated with @frag.
2040 */
2041static inline struct page *skb_frag_page(const skb_frag_t *frag)
2042{
Ian Campbella8605c62011-10-19 23:01:49 +00002043 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00002044}
2045
2046/**
2047 * __skb_frag_ref - take an addition reference on a paged fragment.
2048 * @frag: the paged fragment
2049 *
2050 * Takes an additional reference on the paged fragment @frag.
2051 */
2052static inline void __skb_frag_ref(skb_frag_t *frag)
2053{
2054 get_page(skb_frag_page(frag));
2055}
2056
2057/**
2058 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2059 * @skb: the buffer
2060 * @f: the fragment offset.
2061 *
2062 * Takes an additional reference on the @f'th paged fragment of @skb.
2063 */
2064static inline void skb_frag_ref(struct sk_buff *skb, int f)
2065{
2066 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2067}
2068
2069/**
2070 * __skb_frag_unref - release a reference on a paged fragment.
2071 * @frag: the paged fragment
2072 *
2073 * Releases a reference on the paged fragment @frag.
2074 */
2075static inline void __skb_frag_unref(skb_frag_t *frag)
2076{
2077 put_page(skb_frag_page(frag));
2078}
2079
2080/**
2081 * skb_frag_unref - release a reference on a paged fragment of an skb.
2082 * @skb: the buffer
2083 * @f: the fragment offset
2084 *
2085 * Releases a reference on the @f'th paged fragment of @skb.
2086 */
2087static inline void skb_frag_unref(struct sk_buff *skb, int f)
2088{
2089 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2090}
2091
2092/**
2093 * skb_frag_address - gets the address of the data contained in a paged fragment
2094 * @frag: the paged fragment buffer
2095 *
2096 * Returns the address of the data within @frag. The page must already
2097 * be mapped.
2098 */
2099static inline void *skb_frag_address(const skb_frag_t *frag)
2100{
2101 return page_address(skb_frag_page(frag)) + frag->page_offset;
2102}
2103
2104/**
2105 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2106 * @frag: the paged fragment buffer
2107 *
2108 * Returns the address of the data within @frag. Checks that the page
2109 * is mapped and returns %NULL otherwise.
2110 */
2111static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2112{
2113 void *ptr = page_address(skb_frag_page(frag));
2114 if (unlikely(!ptr))
2115 return NULL;
2116
2117 return ptr + frag->page_offset;
2118}
2119
2120/**
2121 * __skb_frag_set_page - sets the page contained in a paged fragment
2122 * @frag: the paged fragment
2123 * @page: the page to set
2124 *
2125 * Sets the fragment @frag to contain @page.
2126 */
2127static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2128{
Ian Campbella8605c62011-10-19 23:01:49 +00002129 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002130}
2131
2132/**
2133 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2134 * @skb: the buffer
2135 * @f: the fragment offset
2136 * @page: the page to set
2137 *
2138 * Sets the @f'th fragment of @skb to contain @page.
2139 */
2140static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2141 struct page *page)
2142{
2143 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2144}
2145
Eric Dumazet400dfd32013-10-17 16:27:07 -07002146bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2147
Ian Campbell131ea662011-08-19 06:25:00 +00002148/**
2149 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002150 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002151 * @frag: the paged fragment to map
2152 * @offset: the offset within the fragment (starting at the
2153 * fragment's own offset)
2154 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002155 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00002156 *
2157 * Maps the page associated with @frag to @device.
2158 */
2159static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2160 const skb_frag_t *frag,
2161 size_t offset, size_t size,
2162 enum dma_data_direction dir)
2163{
2164 return dma_map_page(dev, skb_frag_page(frag),
2165 frag->page_offset + offset, size, dir);
2166}
2167
Eric Dumazet117632e2011-12-03 21:39:53 +00002168static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2169 gfp_t gfp_mask)
2170{
2171 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2172}
2173
Ian Campbell131ea662011-08-19 06:25:00 +00002174/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002175 * skb_clone_writable - is the header of a clone writable
2176 * @skb: buffer to check
2177 * @len: length up to which to write
2178 *
2179 * Returns true if modifying the header part of the cloned buffer
2180 * does not requires the data to be copied.
2181 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002182static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002183{
2184 return !skb_header_cloned(skb) &&
2185 skb_headroom(skb) + len <= skb->hdr_len;
2186}
2187
Herbert Xud9cc2042007-09-16 16:21:16 -07002188static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2189 int cloned)
2190{
2191 int delta = 0;
2192
Herbert Xud9cc2042007-09-16 16:21:16 -07002193 if (headroom > skb_headroom(skb))
2194 delta = headroom - skb_headroom(skb);
2195
2196 if (delta || cloned)
2197 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2198 GFP_ATOMIC);
2199 return 0;
2200}
2201
Patrick McHardy334a8132007-06-25 04:35:20 -07002202/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002203 * skb_cow - copy header of skb when it is required
2204 * @skb: buffer to cow
2205 * @headroom: needed headroom
2206 *
2207 * If the skb passed lacks sufficient headroom or its data part
2208 * is shared, data is reallocated. If reallocation fails, an error
2209 * is returned and original skb is not changed.
2210 *
2211 * The result is skb with writable area skb->head...skb->tail
2212 * and at least @headroom of space at head.
2213 */
2214static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2215{
Herbert Xud9cc2042007-09-16 16:21:16 -07002216 return __skb_cow(skb, headroom, skb_cloned(skb));
2217}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002218
Herbert Xud9cc2042007-09-16 16:21:16 -07002219/**
2220 * skb_cow_head - skb_cow but only making the head writable
2221 * @skb: buffer to cow
2222 * @headroom: needed headroom
2223 *
2224 * This function is identical to skb_cow except that we replace the
2225 * skb_cloned check by skb_header_cloned. It should be used when
2226 * you only need to push on some header and do not need to modify
2227 * the data.
2228 */
2229static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2230{
2231 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002232}
2233
2234/**
2235 * skb_padto - pad an skbuff up to a minimal size
2236 * @skb: buffer to pad
2237 * @len: minimal length
2238 *
2239 * Pads up a buffer to ensure the trailing bytes exist and are
2240 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002241 * is untouched. Otherwise it is extended. Returns zero on
2242 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002243 */
2244
Herbert Xu5b057c62006-06-23 02:06:41 -07002245static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002246{
2247 unsigned int size = skb->len;
2248 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002249 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002250 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002251}
2252
2253static inline int skb_add_data(struct sk_buff *skb,
2254 char __user *from, int copy)
2255{
2256 const int off = skb->len;
2257
2258 if (skb->ip_summed == CHECKSUM_NONE) {
2259 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08002260 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
Linus Torvalds1da177e2005-04-16 15:20:36 -07002261 copy, 0, &err);
2262 if (!err) {
2263 skb->csum = csum_block_add(skb->csum, csum, off);
2264 return 0;
2265 }
2266 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2267 return 0;
2268
2269 __skb_trim(skb, off);
2270 return -EFAULT;
2271}
2272
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002273static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2274 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002275{
2276 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002277 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002278
Ian Campbellea2ab692011-08-22 23:44:58 +00002279 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002280 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002281 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002282 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002283}
2284
Herbert Xu364c6ba2006-06-09 16:10:40 -07002285static inline int __skb_linearize(struct sk_buff *skb)
2286{
2287 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2288}
2289
Linus Torvalds1da177e2005-04-16 15:20:36 -07002290/**
2291 * skb_linearize - convert paged skb to linear one
2292 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002293 *
2294 * If there is no free memory -ENOMEM is returned, otherwise zero
2295 * is returned and the old skb data released.
2296 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002297static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002298{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002299 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2300}
2301
2302/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00002303 * skb_has_shared_frag - can any frag be overwritten
2304 * @skb: buffer to test
2305 *
2306 * Return true if the skb has at least one frag that might be modified
2307 * by an external entity (as in vmsplice()/sendfile())
2308 */
2309static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2310{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00002311 return skb_is_nonlinear(skb) &&
2312 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00002313}
2314
2315/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07002316 * skb_linearize_cow - make sure skb is linear and writable
2317 * @skb: buffer to process
2318 *
2319 * If there is no free memory -ENOMEM is returned, otherwise zero
2320 * is returned and the old skb data released.
2321 */
2322static inline int skb_linearize_cow(struct sk_buff *skb)
2323{
2324 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2325 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002326}
2327
2328/**
2329 * skb_postpull_rcsum - update checksum for received skb after pull
2330 * @skb: buffer to update
2331 * @start: start of data before pull
2332 * @len: length of data pulled
2333 *
2334 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002335 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2336 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002337 */
2338
2339static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002340 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002341{
Patrick McHardy84fa7932006-08-29 16:44:56 -07002342 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002343 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2344}
2345
Herbert Xucbb042f2006-03-20 22:43:56 -08002346unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2347
David S. Miller7ce5a272013-12-02 17:26:05 -05002348/**
2349 * pskb_trim_rcsum - trim received skb and update checksum
2350 * @skb: buffer to trim
2351 * @len: new length
2352 *
2353 * This is exactly the same as pskb_trim except that it ensures the
2354 * checksum of received packets are still valid after the operation.
2355 */
2356
2357static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2358{
2359 if (likely(len >= skb->len))
2360 return 0;
2361 if (skb->ip_summed == CHECKSUM_COMPLETE)
2362 skb->ip_summed = CHECKSUM_NONE;
2363 return __pskb_trim(skb, len);
2364}
2365
Linus Torvalds1da177e2005-04-16 15:20:36 -07002366#define skb_queue_walk(queue, skb) \
2367 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002368 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002369 skb = skb->next)
2370
James Chapman46f89142007-04-30 00:07:31 -07002371#define skb_queue_walk_safe(queue, skb, tmp) \
2372 for (skb = (queue)->next, tmp = skb->next; \
2373 skb != (struct sk_buff *)(queue); \
2374 skb = tmp, tmp = skb->next)
2375
David S. Miller1164f522008-09-23 00:49:44 -07002376#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002377 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002378 skb = skb->next)
2379
2380#define skb_queue_walk_from_safe(queue, skb, tmp) \
2381 for (tmp = skb->next; \
2382 skb != (struct sk_buff *)(queue); \
2383 skb = tmp, tmp = skb->next)
2384
Stephen Hemminger300ce172005-10-30 13:47:34 -08002385#define skb_queue_reverse_walk(queue, skb) \
2386 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002387 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08002388 skb = skb->prev)
2389
David S. Miller686a2952011-01-20 22:47:32 -08002390#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2391 for (skb = (queue)->prev, tmp = skb->prev; \
2392 skb != (struct sk_buff *)(queue); \
2393 skb = tmp, tmp = skb->prev)
2394
2395#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2396 for (tmp = skb->prev; \
2397 skb != (struct sk_buff *)(queue); \
2398 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002399
David S. Miller21dc3302010-08-23 00:13:46 -07002400static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07002401{
2402 return skb_shinfo(skb)->frag_list != NULL;
2403}
2404
2405static inline void skb_frag_list_init(struct sk_buff *skb)
2406{
2407 skb_shinfo(skb)->frag_list = NULL;
2408}
2409
2410static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2411{
2412 frag->next = skb_shinfo(skb)->frag_list;
2413 skb_shinfo(skb)->frag_list = frag;
2414}
2415
2416#define skb_walk_frags(skb, iter) \
2417 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2418
Joe Perches7965bd42013-09-26 14:48:15 -07002419struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2420 int *peeked, int *off, int *err);
2421struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2422 int *err);
2423unsigned int datagram_poll(struct file *file, struct socket *sock,
2424 struct poll_table_struct *wait);
2425int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2426 struct iovec *to, int size);
2427int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2428 struct iovec *iov);
2429int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2430 const struct iovec *from, int from_offset,
2431 int len);
2432int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2433 int offset, size_t count);
2434int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2435 const struct iovec *to, int to_offset,
2436 int size);
2437void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2438void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2439int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07002440int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2441int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2442__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2443 int len, __wsum csum);
2444int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2445 struct pipe_inode_info *pipe, unsigned int len,
2446 unsigned int flags);
2447void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01002448unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
2449void skb_zerocopy(struct sk_buff *to, const struct sk_buff *from,
2450 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07002451void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2452int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2453void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Joe Perches7965bd42013-09-26 14:48:15 -07002454struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002455
Daniel Borkmann2817a332013-10-30 11:50:51 +01002456struct skb_checksum_ops {
2457 __wsum (*update)(const void *mem, int len, __wsum wsum);
2458 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2459};
2460
2461__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2462 __wsum csum, const struct skb_checksum_ops *ops);
2463__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2464 __wsum csum);
2465
Linus Torvalds1da177e2005-04-16 15:20:36 -07002466static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2467 int len, void *buffer)
2468{
2469 int hlen = skb_headlen(skb);
2470
Patrick McHardy55820ee2005-07-05 14:08:10 -07002471 if (hlen - offset >= len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002472 return skb->data + offset;
2473
2474 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2475 return NULL;
2476
2477 return buffer;
2478}
2479
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01002480/**
2481 * skb_needs_linearize - check if we need to linearize a given skb
2482 * depending on the given device features.
2483 * @skb: socket buffer to check
2484 * @features: net device features
2485 *
2486 * Returns true if either:
2487 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2488 * 2. skb is fragmented and the device does not support SG.
2489 */
2490static inline bool skb_needs_linearize(struct sk_buff *skb,
2491 netdev_features_t features)
2492{
2493 return skb_is_nonlinear(skb) &&
2494 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2495 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2496}
2497
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03002498static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2499 void *to,
2500 const unsigned int len)
2501{
2502 memcpy(to, skb->data, len);
2503}
2504
2505static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2506 const int offset, void *to,
2507 const unsigned int len)
2508{
2509 memcpy(to, skb->data + offset, len);
2510}
2511
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03002512static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2513 const void *from,
2514 const unsigned int len)
2515{
2516 memcpy(skb->data, from, len);
2517}
2518
2519static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2520 const int offset,
2521 const void *from,
2522 const unsigned int len)
2523{
2524 memcpy(skb->data + offset, from, len);
2525}
2526
Joe Perches7965bd42013-09-26 14:48:15 -07002527void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002528
Patrick Ohlyac45f602009-02-12 05:03:37 +00002529static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2530{
2531 return skb->tstamp;
2532}
2533
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002534/**
2535 * skb_get_timestamp - get timestamp from a skb
2536 * @skb: skb to get stamp from
2537 * @stamp: pointer to struct timeval to store stamp in
2538 *
2539 * Timestamps are stored in the skb as offsets to a base timestamp.
2540 * This function converts the offset back to a struct timeval and stores
2541 * it in stamp.
2542 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00002543static inline void skb_get_timestamp(const struct sk_buff *skb,
2544 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002545{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002546 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002547}
2548
Patrick Ohlyac45f602009-02-12 05:03:37 +00002549static inline void skb_get_timestampns(const struct sk_buff *skb,
2550 struct timespec *stamp)
2551{
2552 *stamp = ktime_to_timespec(skb->tstamp);
2553}
2554
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002555static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002556{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002557 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002558}
2559
Stephen Hemminger164891a2007-04-23 22:26:16 -07002560static inline ktime_t net_timedelta(ktime_t t)
2561{
2562 return ktime_sub(ktime_get_real(), t);
2563}
2564
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07002565static inline ktime_t net_invalid_timestamp(void)
2566{
2567 return ktime_set(0, 0);
2568}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002569
Joe Perches7965bd42013-09-26 14:48:15 -07002570void skb_timestamping_init(void);
Richard Cochranc1f19b52010-07-17 08:49:36 +00002571
2572#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2573
Joe Perches7965bd42013-09-26 14:48:15 -07002574void skb_clone_tx_timestamp(struct sk_buff *skb);
2575bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00002576
2577#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2578
2579static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2580{
2581}
2582
2583static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2584{
2585 return false;
2586}
2587
2588#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2589
2590/**
2591 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2592 *
Richard Cochranda92b192011-10-21 00:49:15 +00002593 * PHY drivers may accept clones of transmitted packets for
2594 * timestamping via their phy_driver.txtstamp method. These drivers
2595 * must call this function to return the skb back to the stack, with
2596 * or without a timestamp.
2597 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00002598 * @skb: clone of the the original outgoing packet
Richard Cochranda92b192011-10-21 00:49:15 +00002599 * @hwtstamps: hardware time stamps, may be NULL if not available
Richard Cochranc1f19b52010-07-17 08:49:36 +00002600 *
2601 */
2602void skb_complete_tx_timestamp(struct sk_buff *skb,
2603 struct skb_shared_hwtstamps *hwtstamps);
2604
Patrick Ohlyac45f602009-02-12 05:03:37 +00002605/**
2606 * skb_tstamp_tx - queue clone of skb with send time stamps
2607 * @orig_skb: the original outgoing packet
2608 * @hwtstamps: hardware time stamps, may be NULL if not available
2609 *
2610 * If the skb has a socket associated, then this function clones the
2611 * skb (thus sharing the actual data and optional structures), stores
2612 * the optional hardware time stamping information (if non NULL) or
2613 * generates a software time stamp (otherwise), then queues the clone
2614 * to the error queue of the socket. Errors are silently ignored.
2615 */
Joe Perches7965bd42013-09-26 14:48:15 -07002616void skb_tstamp_tx(struct sk_buff *orig_skb,
2617 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00002618
Richard Cochran4507a712010-07-17 08:48:28 +00002619static inline void sw_tx_timestamp(struct sk_buff *skb)
2620{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002621 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2622 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00002623 skb_tstamp_tx(skb, NULL);
2624}
2625
2626/**
2627 * skb_tx_timestamp() - Driver hook for transmit timestamping
2628 *
2629 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00002630 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00002631 *
David S. Miller73409f32013-12-27 13:04:33 -05002632 * Specifically, one should make absolutely sure that this function is
2633 * called before TX completion of this packet can trigger. Otherwise
2634 * the packet could potentially already be freed.
2635 *
Richard Cochran4507a712010-07-17 08:48:28 +00002636 * @skb: A socket buffer.
2637 */
2638static inline void skb_tx_timestamp(struct sk_buff *skb)
2639{
Richard Cochranc1f19b52010-07-17 08:49:36 +00002640 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002641 sw_tx_timestamp(skb);
2642}
2643
Johannes Berg6e3e9392011-11-09 10:15:42 +01002644/**
2645 * skb_complete_wifi_ack - deliver skb with wifi status
2646 *
2647 * @skb: the original outgoing packet
2648 * @acked: ack status
2649 *
2650 */
2651void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2652
Joe Perches7965bd42013-09-26 14:48:15 -07002653__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2654__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002655
Herbert Xu60476372007-04-09 11:59:39 -07002656static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2657{
2658 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2659}
2660
Herbert Xufb286bb2005-11-10 13:01:24 -08002661/**
2662 * skb_checksum_complete - Calculate checksum of an entire packet
2663 * @skb: packet to process
2664 *
2665 * This function calculates the checksum over the entire packet plus
2666 * the value of skb->csum. The latter can be used to supply the
2667 * checksum of a pseudo header as used by TCP/UDP. It returns the
2668 * checksum.
2669 *
2670 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2671 * this function can be used to verify that checksum on received
2672 * packets. In that case the function should return zero if the
2673 * checksum is correct. In particular, this function will return zero
2674 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2675 * hardware has already verified the correctness of the checksum.
2676 */
Al Viro4381ca32007-07-15 21:00:11 +01002677static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002678{
Herbert Xu60476372007-04-09 11:59:39 -07002679 return skb_csum_unnecessary(skb) ?
2680 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002681}
2682
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002683#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07002684void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002685static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2686{
2687 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002688 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002689}
2690static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2691{
2692 if (nfct)
2693 atomic_inc(&nfct->use);
2694}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002695#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002696#ifdef CONFIG_BRIDGE_NETFILTER
2697static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2698{
2699 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2700 kfree(nf_bridge);
2701}
2702static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2703{
2704 if (nf_bridge)
2705 atomic_inc(&nf_bridge->use);
2706}
2707#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08002708static inline void nf_reset(struct sk_buff *skb)
2709{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002710#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08002711 nf_conntrack_put(skb->nfct);
2712 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002713#endif
Patrick McHardya193a4a2006-03-20 19:23:05 -08002714#ifdef CONFIG_BRIDGE_NETFILTER
2715 nf_bridge_put(skb->nf_bridge);
2716 skb->nf_bridge = NULL;
2717#endif
2718}
2719
Patrick McHardy124dff02013-04-05 20:42:05 +02002720static inline void nf_reset_trace(struct sk_buff *skb)
2721{
Gao feng130549fe2013-03-21 19:48:41 +00002722#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE)
2723 skb->nf_trace = 0;
2724#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002725}
2726
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002727/* Note: This doesn't put any conntrack and bridge info in dst. */
2728static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2729{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002730#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002731 dst->nfct = src->nfct;
2732 nf_conntrack_get(src->nfct);
2733 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002734#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002735#ifdef CONFIG_BRIDGE_NETFILTER
2736 dst->nf_bridge = src->nf_bridge;
2737 nf_bridge_get(src->nf_bridge);
2738#endif
2739}
2740
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002741static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2742{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002743#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002744 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002745#endif
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002746#ifdef CONFIG_BRIDGE_NETFILTER
2747 nf_bridge_put(dst->nf_bridge);
2748#endif
2749 __nf_copy(dst, src);
2750}
2751
James Morris984bc162006-06-09 00:29:17 -07002752#ifdef CONFIG_NETWORK_SECMARK
2753static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2754{
2755 to->secmark = from->secmark;
2756}
2757
2758static inline void skb_init_secmark(struct sk_buff *skb)
2759{
2760 skb->secmark = 0;
2761}
2762#else
2763static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2764{ }
2765
2766static inline void skb_init_secmark(struct sk_buff *skb)
2767{ }
2768#endif
2769
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002770static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2771{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002772 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002773}
2774
Stephen Hemminger92477442009-03-21 13:39:26 -07002775static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002776{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002777 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002778}
2779
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002780static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2781{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002782 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002783}
2784
David S. Millerd5a9e242009-01-27 16:22:11 -08002785static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2786{
2787 skb->queue_mapping = rx_queue + 1;
2788}
2789
Stephen Hemminger92477442009-03-21 13:39:26 -07002790static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002791{
2792 return skb->queue_mapping - 1;
2793}
2794
Stephen Hemminger92477442009-03-21 13:39:26 -07002795static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002796{
Eric Dumazeta02cec22010-09-22 20:43:57 +00002797 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08002798}
2799
Joe Perches7965bd42013-09-26 14:48:15 -07002800u16 __skb_tx_hash(const struct net_device *dev, const struct sk_buff *skb,
2801 unsigned int num_tx_queues);
Stephen Hemminger92477442009-03-21 13:39:26 -07002802
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04002803static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2804{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002805#ifdef CONFIG_XFRM
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002806 return skb->sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002807#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002808 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002809#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04002810}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002811
Pravin B Shelar68c33162013-02-14 14:02:41 +00002812/* Keeps track of mac header offset relative to skb->head.
2813 * It is useful for TSO of Tunneling protocol. e.g. GRE.
2814 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07002815 * tunnel skb it points to outer mac header.
2816 * Keeps track of level of encapsulation of network headers.
2817 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00002818struct skb_gso_cb {
Eric Dumazet3347c962013-10-19 11:42:56 -07002819 int mac_offset;
2820 int encap_level;
Pravin B Shelar68c33162013-02-14 14:02:41 +00002821};
2822#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
2823
2824static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
2825{
2826 return (skb_mac_header(inner_skb) - inner_skb->head) -
2827 SKB_GSO_CB(inner_skb)->mac_offset;
2828}
2829
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00002830static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
2831{
2832 int new_headroom, headroom;
2833 int ret;
2834
2835 headroom = skb_headroom(skb);
2836 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
2837 if (ret)
2838 return ret;
2839
2840 new_headroom = skb_headroom(skb);
2841 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
2842 return 0;
2843}
2844
David S. Millerbdcc0922012-03-07 20:53:36 -05002845static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07002846{
2847 return skb_shinfo(skb)->gso_size;
2848}
2849
Eric Dumazet36a8f392013-09-29 01:21:32 -07002850/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05002851static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02002852{
2853 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2854}
2855
Joe Perches7965bd42013-09-26 14:48:15 -07002856void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07002857
2858static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2859{
2860 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2861 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002862 const struct skb_shared_info *shinfo = skb_shinfo(skb);
2863
Alexander Duyckb78462e2010-06-02 12:24:37 +00002864 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2865 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07002866 __skb_warn_lro_forwarding(skb);
2867 return true;
2868 }
2869 return false;
2870}
2871
Herbert Xu35fc92a2007-03-26 23:22:20 -07002872static inline void skb_forward_csum(struct sk_buff *skb)
2873{
2874 /* Unfortunately we don't support this one. Any brave souls? */
2875 if (skb->ip_summed == CHECKSUM_COMPLETE)
2876 skb->ip_summed = CHECKSUM_NONE;
2877}
2878
Eric Dumazetbc8acf22010-09-02 13:07:41 -07002879/**
2880 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2881 * @skb: skb to check
2882 *
2883 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2884 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2885 * use this helper, to document places where we make this assertion.
2886 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002887static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07002888{
2889#ifdef DEBUG
2890 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2891#endif
2892}
2893
Rusty Russellf35d9d82008-02-04 23:49:54 -05002894bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00002895
Daniel Borkmannf77668d2013-03-19 06:39:30 +00002896u32 __skb_get_poff(const struct sk_buff *skb);
2897
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00002898/**
2899 * skb_head_is_locked - Determine if the skb->head is locked down
2900 * @skb: skb to check
2901 *
2902 * The head on skbs build around a head frag can be removed if they are
2903 * not cloned. This function returns true if the skb head is locked down
2904 * due to either being allocated via kmalloc, or by being a clone with
2905 * multiple references to the head.
2906 */
2907static inline bool skb_head_is_locked(const struct sk_buff *skb)
2908{
2909 return !skb->head_frag || skb_cloned(skb);
2910}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002911#endif /* __KERNEL__ */
2912#endif /* _LINUX_SKBUFF_H */