blob: 6a2c34e6d962658d0a1532959a06b5ddebe65087 [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>
Linus Torvalds1da177e2005-04-16 15:20:36 -070035
Herbert Xu60476372007-04-09 11:59:39 -070036/* Don't change this without changing skb_csum_unnecessary! */
Linus Torvalds1da177e2005-04-16 15:20:36 -070037#define CHECKSUM_NONE 0
Herbert Xu60476372007-04-09 11:59:39 -070038#define CHECKSUM_UNNECESSARY 1
39#define CHECKSUM_COMPLETE 2
40#define CHECKSUM_PARTIAL 3
Linus Torvalds1da177e2005-04-16 15:20:36 -070041
42#define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
43 ~(SMP_CACHE_BYTES - 1))
David S. Millerfc910a22007-03-25 20:27:59 -070044#define SKB_WITH_OVERHEAD(X) \
Herbert Xudeea84b2007-10-21 16:27:46 -070045 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
David S. Millerfc910a22007-03-25 20:27:59 -070046#define SKB_MAX_ORDER(X, ORDER) \
47 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
Linus Torvalds1da177e2005-04-16 15:20:36 -070048#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
49#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
50
Eric Dumazet87fb4b72011-10-13 07:28:54 +000051/* return minimum truesize of one skb containing X bytes of data */
52#define SKB_TRUESIZE(X) ((X) + \
53 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
54 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
55
Linus Torvalds1da177e2005-04-16 15:20:36 -070056/* A. Checksumming of received packets by device.
57 *
58 * NONE: device failed to checksum this packet.
59 * skb->csum is undefined.
60 *
61 * UNNECESSARY: device parsed packet and wouldbe verified checksum.
62 * skb->csum is undefined.
63 * It is bad option, but, unfortunately, many of vendors do this.
64 * Apparently with secret goal to sell you new device, when you
65 * will add new protocol to your host. F.e. IPv6. 8)
66 *
Patrick McHardy84fa7932006-08-29 16:44:56 -070067 * COMPLETE: the most generic way. Device supplied checksum of _all_
Linus Torvalds1da177e2005-04-16 15:20:36 -070068 * the packet as seen by netif_rx in skb->csum.
69 * NOTE: Even if device supports only some protocols, but
Patrick McHardy84fa7932006-08-29 16:44:56 -070070 * is able to produce some skb->csum, it MUST use COMPLETE,
Linus Torvalds1da177e2005-04-16 15:20:36 -070071 * not UNNECESSARY.
72 *
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070073 * PARTIAL: identical to the case for output below. This may occur
74 * on a packet received directly from another Linux OS, e.g.,
75 * a virtualised Linux kernel on the same host. The packet can
76 * be treated in the same way as UNNECESSARY except that on
77 * output (i.e., forwarding) the checksum must be filled in
78 * by the OS or the hardware.
79 *
Linus Torvalds1da177e2005-04-16 15:20:36 -070080 * B. Checksumming on output.
81 *
82 * NONE: skb is checksummed by protocol or csum is not required.
83 *
Patrick McHardy84fa7932006-08-29 16:44:56 -070084 * PARTIAL: device is required to csum packet as seen by hard_start_xmit
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070085 * from skb->csum_start to the end and to record the checksum
86 * at skb->csum_start + skb->csum_offset.
Linus Torvalds1da177e2005-04-16 15:20:36 -070087 *
88 * Device must show its capabilities in dev->features, set
89 * at device setup time.
90 * NETIF_F_HW_CSUM - it is clever device, it is able to checksum
91 * everything.
Linus Torvalds1da177e2005-04-16 15:20:36 -070092 * NETIF_F_IP_CSUM - device is dumb. It is able to csum only
93 * TCP/UDP over IPv4. Sigh. Vendors like this
94 * way by an unknown reason. Though, see comment above
95 * about CHECKSUM_UNNECESSARY. 8)
Herbert Xuc6c6e3e2007-07-10 22:41:55 -070096 * NETIF_F_IPV6_CSUM about as dumb as the last one but does IPv6 instead.
Linus Torvalds1da177e2005-04-16 15:20:36 -070097 *
Yi Zou3af79302012-03-19 11:12:41 +000098 * UNNECESSARY: device will do per protocol specific csum. Protocol drivers
99 * that do not want net to perform the checksum calculation should use
100 * this flag in their outgoing skbs.
101 * NETIF_F_FCOE_CRC this indicates the device can do FCoE FC CRC
102 * offload. Correspondingly, the FCoE protocol driver
103 * stack should use CHECKSUM_UNNECESSARY.
104 *
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105 * Any questions? No questions, good. --ANK
106 */
107
Linus Torvalds1da177e2005-04-16 15:20:36 -0700108struct net_device;
David Howells716ea3a2007-04-02 20:19:53 -0700109struct scatterlist;
Jens Axboe9c55e012007-11-06 23:30:13 -0800110struct pipe_inode_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700111
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700112#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700113struct nf_conntrack {
114 atomic_t use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700115};
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700116#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700117
118#ifdef CONFIG_BRIDGE_NETFILTER
119struct nf_bridge_info {
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000120 atomic_t use;
121 unsigned int mask;
122 struct net_device *physindev;
123 struct net_device *physoutdev;
124 unsigned long data[32 / sizeof(unsigned long)];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700125};
126#endif
127
Linus Torvalds1da177e2005-04-16 15:20:36 -0700128struct sk_buff_head {
129 /* These two members must be first. */
130 struct sk_buff *next;
131 struct sk_buff *prev;
132
133 __u32 qlen;
134 spinlock_t lock;
135};
136
137struct sk_buff;
138
Ian Campbell9d4dde52011-12-22 23:39:14 +0000139/* To allow 64K frame to be packed as single skb without frag_list we
140 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
141 * buffers which do not start on a page boundary.
142 *
143 * Since GRO uses frags we allocate at least 16 regardless of page
144 * size.
Anton Blancharda715dea2011-03-27 14:57:26 +0000145 */
Ian Campbell9d4dde52011-12-22 23:39:14 +0000146#if (65536/PAGE_SIZE + 1) < 16
David S. Millereec00952011-03-29 23:34:08 -0700147#define MAX_SKB_FRAGS 16UL
Anton Blancharda715dea2011-03-27 14:57:26 +0000148#else
Ian Campbell9d4dde52011-12-22 23:39:14 +0000149#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
Anton Blancharda715dea2011-03-27 14:57:26 +0000150#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700151
152typedef struct skb_frag_struct skb_frag_t;
153
154struct skb_frag_struct {
Ian Campbella8605c62011-10-19 23:01:49 +0000155 struct {
156 struct page *p;
157 } page;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000158#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
David S. Millera309bb02007-07-30 18:47:03 -0700159 __u32 page_offset;
160 __u32 size;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000161#else
162 __u16 page_offset;
163 __u16 size;
164#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700165};
166
Eric Dumazet9e903e02011-10-18 21:00:24 +0000167static inline unsigned int skb_frag_size(const skb_frag_t *frag)
168{
169 return frag->size;
170}
171
172static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
173{
174 frag->size = size;
175}
176
177static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
178{
179 frag->size += delta;
180}
181
182static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
183{
184 frag->size -= delta;
185}
186
Patrick Ohlyac45f602009-02-12 05:03:37 +0000187#define HAVE_HW_TIME_STAMP
188
189/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800190 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000191 * @hwtstamp: hardware time stamp transformed into duration
192 * since arbitrary point in time
193 * @syststamp: hwtstamp transformed to system time base
194 *
195 * Software time stamps generated by ktime_get_real() are stored in
196 * skb->tstamp. The relation between the different kinds of time
197 * stamps is as follows:
198 *
199 * syststamp and tstamp can be compared against each other in
200 * arbitrary combinations. The accuracy of a
201 * syststamp/tstamp/"syststamp from other device" comparison is
202 * limited by the accuracy of the transformation into system time
203 * base. This depends on the device driver and its underlying
204 * hardware.
205 *
206 * hwtstamps can only be compared against other hwtstamps from
207 * the same device.
208 *
209 * This structure is attached to packets as part of the
210 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
211 */
212struct skb_shared_hwtstamps {
213 ktime_t hwtstamp;
214 ktime_t syststamp;
215};
216
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000217/* Definitions for tx_flags in struct skb_shared_info */
218enum {
219 /* generate hardware time stamp */
220 SKBTX_HW_TSTAMP = 1 << 0,
221
222 /* generate software time stamp */
223 SKBTX_SW_TSTAMP = 1 << 1,
224
225 /* device driver is going to provide hardware time stamp */
226 SKBTX_IN_PROGRESS = 1 << 2,
227
Shirley Maa6686f22011-07-06 12:22:12 +0000228 /* device driver supports TX zero-copy buffers */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000229 SKBTX_DEV_ZEROCOPY = 1 << 3,
Johannes Berg6e3e9392011-11-09 10:15:42 +0100230
231 /* generate wifi status information (where possible) */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000232 SKBTX_WIFI_STATUS = 1 << 4,
Shirley Maa6686f22011-07-06 12:22:12 +0000233};
234
235/*
236 * The callback notifies userspace to release buffers when skb DMA is done in
237 * lower device, the skb last reference should be 0 when calling this.
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000238 * The ctx field is used to track device context.
239 * The desc field is used to track userspace buffer index.
Shirley Maa6686f22011-07-06 12:22:12 +0000240 */
241struct ubuf_info {
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000242 void (*callback)(struct ubuf_info *);
243 void *ctx;
Shirley Maa6686f22011-07-06 12:22:12 +0000244 unsigned long desc;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000245};
246
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247/* This data is invariant across clones and lives at
248 * the end of the header data, ie. at skb->end.
249 */
250struct skb_shared_info {
Ian Campbell9f42f122012-01-05 07:13:39 +0000251 unsigned char nr_frags;
252 __u8 tx_flags;
Herbert Xu79671682006-06-22 02:40:14 -0700253 unsigned short gso_size;
254 /* Warning: this field is not always filled in (UFO)! */
255 unsigned short gso_segs;
256 unsigned short gso_type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000258 struct skb_shared_hwtstamps hwtstamps;
Ian Campbell9f42f122012-01-05 07:13:39 +0000259 __be32 ip6_frag_id;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700260
261 /*
262 * Warning : all fields before dataref are cleared in __alloc_skb()
263 */
264 atomic_t dataref;
265
Johann Baudy69e3c752009-05-18 22:11:22 -0700266 /* Intermediate layers must ensure that destructor_arg
267 * remains valid until skb destructor */
268 void * destructor_arg;
Shirley Maa6686f22011-07-06 12:22:12 +0000269
Eric Dumazetfed66382010-07-22 19:09:08 +0000270 /* must be last field, see pskb_expand_head() */
271 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700272};
273
274/* We divide dataref into two halves. The higher 16 bits hold references
275 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700276 * the entire skb->data. A clone of a headerless skb holds the length of
277 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278 *
279 * All users must obey the rule that the skb->data reference count must be
280 * greater than or equal to the payload reference count.
281 *
282 * Holding a reference to the payload part means that the user does not
283 * care about modifications to the header part of skb->data.
284 */
285#define SKB_DATAREF_SHIFT 16
286#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
287
David S. Millerd179cd12005-08-17 14:57:30 -0700288
289enum {
290 SKB_FCLONE_UNAVAILABLE,
291 SKB_FCLONE_ORIG,
292 SKB_FCLONE_CLONE,
293};
294
Herbert Xu79671682006-06-22 02:40:14 -0700295enum {
296 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700297 SKB_GSO_UDP = 1 << 1,
Herbert Xu576a30e2006-06-27 13:22:38 -0700298
299 /* This indicates the skb is from an untrusted source. */
300 SKB_GSO_DODGY = 1 << 2,
Michael Chanb0da85372006-06-29 12:30:00 -0700301
302 /* This indicates the tcp segment has CWR set. */
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700303 SKB_GSO_TCP_ECN = 1 << 3,
304
305 SKB_GSO_TCPV6 = 1 << 4,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800306
307 SKB_GSO_FCOE = 1 << 5,
Herbert Xu79671682006-06-22 02:40:14 -0700308};
309
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700310#if BITS_PER_LONG > 32
311#define NET_SKBUFF_DATA_USES_OFFSET 1
312#endif
313
314#ifdef NET_SKBUFF_DATA_USES_OFFSET
315typedef unsigned int sk_buff_data_t;
316#else
317typedef unsigned char *sk_buff_data_t;
318#endif
319
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100320#if defined(CONFIG_NF_DEFRAG_IPV4) || defined(CONFIG_NF_DEFRAG_IPV4_MODULE) || \
321 defined(CONFIG_NF_DEFRAG_IPV6) || defined(CONFIG_NF_DEFRAG_IPV6_MODULE)
322#define NET_SKBUFF_NF_DEFRAG_NEEDED 1
323#endif
324
Linus Torvalds1da177e2005-04-16 15:20:36 -0700325/**
326 * struct sk_buff - socket buffer
327 * @next: Next buffer in list
328 * @prev: Previous buffer in list
Herbert Xu325ed822005-10-03 13:57:23 -0700329 * @tstamp: Time we arrived
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700330 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700331 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700332 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000333 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700334 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700335 * @len: Length of actual data
336 * @data_len: Data length
337 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700338 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700339 * @csum: Checksum (must include start/offset pair)
340 * @csum_start: Offset from skb->head where checksumming should start
341 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700342 * @priority: Packet queueing priority
Martin Waitz67be2dd2005-05-01 08:59:26 -0700343 * @local_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700344 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700345 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700346 * @nohdr: Payload reference only, must not modify header
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700347 * @nfctinfo: Relationship of this skb to the connection
Linus Torvalds1da177e2005-04-16 15:20:36 -0700348 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700349 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700350 * @ipvs_property: skbuff is owned by ipvs
Randy Dunlap31729362008-02-18 20:52:13 -0800351 * @peeked: this packet has been seen already, so stats have been
352 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700353 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700354 * @protocol: Packet protocol from driver
355 * @destructor: Destruct function
356 * @nfct: Associated connection, if any
Randy Dunlap461ddf32005-11-20 21:25:15 -0800357 * @nfct_reasm: netfilter conntrack re-assembly pointer
Linus Torvalds1da177e2005-04-16 15:20:36 -0700358 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800359 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700360 * @tc_index: Traffic control index
361 * @tc_verd: traffic control verdict
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700362 * @rxhash: the packet hash computed on receive
363 * @queue_mapping: Queue mapping for multiqueue devices
Randy Dunlap553a5672008-04-20 10:51:01 -0700364 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700365 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Changli Gao4ca24622011-08-19 07:26:44 -0700366 * @l4_rxhash: indicate rxhash is a canonical 4-tuple hash over transport
367 * ports.
Johannes Berg6e3e9392011-11-09 10:15:42 +0100368 * @wifi_acked_valid: wifi_acked was set
369 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000370 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Randy Dunlapf4b8ea72006-06-22 16:00:11 -0700371 * @dma_cookie: a cookie to one of several possible DMA operations
372 * done by skb DMA functions
James Morris984bc162006-06-09 00:29:17 -0700373 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700374 * @mark: Generic packet mark
375 * @dropcount: total number of sk_receive_queue overflows
Patrick McHardy6aa895b02008-07-14 22:49:06 -0700376 * @vlan_tci: vlan tag control information
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700377 * @transport_header: Transport layer header
378 * @network_header: Network layer header
379 * @mac_header: Link layer header
380 * @tail: Tail pointer
381 * @end: End pointer
382 * @head: Head of buffer
383 * @data: Data head pointer
384 * @truesize: Buffer size
385 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700386 */
387
388struct sk_buff {
389 /* These two members must be first. */
390 struct sk_buff *next;
391 struct sk_buff *prev;
392
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -0700393 ktime_t tstamp;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000394
395 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700396 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700397
Linus Torvalds1da177e2005-04-16 15:20:36 -0700398 /*
399 * This is the control buffer. It is free to use for every
400 * layer. Please put your private variables there. If you
401 * want to keep them across layers you have to do a skb_clone()
402 * first. This is owned by whoever has the skb queued ATM.
403 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000404 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700405
Eric Dumazet7fee2262010-05-11 23:19:48 +0000406 unsigned long _skb_refdst;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000407#ifdef CONFIG_XFRM
408 struct sec_path *sp;
409#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700410 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700411 data_len;
412 __u16 mac_len,
413 hdr_len;
Al Viroff1dcad2006-11-20 18:07:29 -0800414 union {
415 __wsum csum;
Herbert Xu663ead32007-04-09 11:59:07 -0700416 struct {
417 __u16 csum_start;
418 __u16 csum_offset;
419 };
Al Viroff1dcad2006-11-20 18:07:29 -0800420 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 __u32 priority;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200422 kmemcheck_bitfield_begin(flags1);
Thomas Graf1cbb3382005-07-05 14:13:41 -0700423 __u8 local_df:1,
424 cloned:1,
425 ip_summed:2,
Harald Welte6869c4d2005-08-09 19:24:19 -0700426 nohdr:1,
427 nfctinfo:3;
David S. Millerd179cd12005-08-17 14:57:30 -0700428 __u8 pkt_type:3,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800429 fclone:2,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700430 ipvs_property:1,
Herbert Xua59322b2007-12-05 01:53:40 -0800431 peeked:1,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700432 nf_trace:1;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200433 kmemcheck_bitfield_end(flags1);
Eric Dumazet4ab408d2010-03-01 03:09:26 +0000434 __be16 protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435
436 void (*destructor)(struct sk_buff *skb);
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800437#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700438 struct nf_conntrack *nfct;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100439#endif
440#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800441 struct sk_buff *nfct_reasm;
442#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443#ifdef CONFIG_BRIDGE_NETFILTER
444 struct nf_bridge_info *nf_bridge;
445#endif
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700446
Eric Dumazet8964be42009-11-20 15:35:04 -0800447 int skb_iif;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000448
449 __u32 rxhash;
450
451 __u16 vlan_tci;
452
Linus Torvalds1da177e2005-04-16 15:20:36 -0700453#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700454 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700455#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700456 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700458#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200459
Eric Dumazet0a148422011-04-20 09:27:32 +0000460 __u16 queue_mapping;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200461 kmemcheck_bitfield_begin(flags2);
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400462#ifdef CONFIG_IPV6_NDISC_NODETYPE
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000463 __u8 ndisc_nodetype:2;
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400464#endif
Mel Gormanc93bdd02012-07-31 16:44:19 -0700465 __u8 pfmemalloc:1;
Tom Herbert3853b582010-11-21 13:17:29 +0000466 __u8 ooo_okay:1;
Tom Herbertbdeab992011-08-14 19:45:55 +0000467 __u8 l4_rxhash:1;
Johannes Berg6e3e9392011-11-09 10:15:42 +0100468 __u8 wifi_acked_valid:1;
469 __u8 wifi_acked:1;
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000470 __u8 no_fcs:1;
Eric Dumazetd3836f22012-04-27 00:33:38 +0000471 __u8 head_frag:1;
472 /* 8/10 bit hole (depending on ndisc_nodetype presence) */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200473 kmemcheck_bitfield_end(flags2);
474
Chris Leech97fc2f02006-05-23 17:55:33 -0700475#ifdef CONFIG_NET_DMA
476 dma_cookie_t dma_cookie;
477#endif
James Morris984bc162006-06-09 00:29:17 -0700478#ifdef CONFIG_NETWORK_SECMARK
479 __u32 secmark;
480#endif
Neil Horman3b885782009-10-12 13:26:31 -0700481 union {
482 __u32 mark;
483 __u32 dropcount;
Eric Dumazeta21d4572012-04-10 20:30:48 +0000484 __u32 avail_size;
Neil Horman3b885782009-10-12 13:26:31 -0700485 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700486
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700487 sk_buff_data_t transport_header;
488 sk_buff_data_t network_header;
489 sk_buff_data_t mac_header;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700491 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700492 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700493 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700494 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700495 unsigned int truesize;
496 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497};
498
499#ifdef __KERNEL__
500/*
501 * Handling routines are only of interest to the kernel
502 */
503#include <linux/slab.h>
504
Linus Torvalds1da177e2005-04-16 15:20:36 -0700505
Mel Gormanc93bdd02012-07-31 16:44:19 -0700506#define SKB_ALLOC_FCLONE 0x01
507#define SKB_ALLOC_RX 0x02
508
509/* Returns true if the skb was allocated from PFMEMALLOC reserves */
510static inline bool skb_pfmemalloc(const struct sk_buff *skb)
511{
512 return unlikely(skb->pfmemalloc);
513}
514
Eric Dumazet7fee2262010-05-11 23:19:48 +0000515/*
516 * skb might have a dst pointer attached, refcounted or not.
517 * _skb_refdst low order bit is set if refcount was _not_ taken
518 */
519#define SKB_DST_NOREF 1UL
520#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
521
522/**
523 * skb_dst - returns skb dst_entry
524 * @skb: buffer
525 *
526 * Returns skb dst_entry, regardless of reference taken or not.
527 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000528static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
529{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000530 /* If refdst was not refcounted, check we still are in a
531 * rcu_read_lock section
532 */
533 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
534 !rcu_read_lock_held() &&
535 !rcu_read_lock_bh_held());
536 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000537}
538
Eric Dumazet7fee2262010-05-11 23:19:48 +0000539/**
540 * skb_dst_set - sets skb dst
541 * @skb: buffer
542 * @dst: dst entry
543 *
544 * Sets skb dst, assuming a reference was taken on dst and should
545 * be released by skb_dst_drop()
546 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000547static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
548{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000549 skb->_skb_refdst = (unsigned long)dst;
550}
551
Eric Dumazet27b75c92010-10-15 05:44:11 +0000552extern void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst);
Eric Dumazet7fee2262010-05-11 23:19:48 +0000553
554/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300555 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000556 * @skb: buffer
557 */
558static inline bool skb_dst_is_noref(const struct sk_buff *skb)
559{
560 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000561}
562
Eric Dumazet511c3f92009-06-02 05:14:27 +0000563static inline struct rtable *skb_rtable(const struct sk_buff *skb)
564{
Eric Dumazetadf30902009-06-02 05:19:30 +0000565 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000566}
567
Jörn Engel231d06a2006-03-20 21:28:35 -0800568extern void kfree_skb(struct sk_buff *skb);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000569extern void consume_skb(struct sk_buff *skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700570extern void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000571extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000572
573extern void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
574extern bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
575 bool *fragstolen, int *delta_truesize);
576
David S. Millerd179cd12005-08-17 14:57:30 -0700577extern struct sk_buff *__alloc_skb(unsigned int size,
Mel Gormanc93bdd02012-07-31 16:44:19 -0700578 gfp_t priority, int flags, int node);
Eric Dumazetd3836f22012-04-27 00:33:38 +0000579extern struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700580static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100581 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700582{
Eric Dumazet564824b2010-10-11 19:05:25 +0000583 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700584}
585
586static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100587 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700588{
Mel Gormanc93bdd02012-07-31 16:44:19 -0700589 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700590}
591
Herbert Xue0053ec2007-10-14 00:37:52 -0700592extern struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
Michael S. Tsirkin48c83012011-08-31 08:03:29 +0000593extern int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700594extern struct sk_buff *skb_clone(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100595 gfp_t priority);
Victor Fusco86a76ca2005-07-08 14:57:47 -0700596extern struct sk_buff *skb_copy(const struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100597 gfp_t priority);
Eric Dumazet117632e2011-12-03 21:39:53 +0000598extern struct sk_buff *__pskb_copy(struct sk_buff *skb,
599 int headroom, gfp_t gfp_mask);
600
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601extern int pskb_expand_head(struct sk_buff *skb,
Victor Fusco86a76ca2005-07-08 14:57:47 -0700602 int nhead, int ntail,
Al Virodd0fc662005-10-07 07:46:04 +0100603 gfp_t gfp_mask);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604extern struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
605 unsigned int headroom);
606extern struct sk_buff *skb_copy_expand(const struct sk_buff *skb,
607 int newheadroom, int newtailroom,
Al Virodd0fc662005-10-07 07:46:04 +0100608 gfp_t priority);
David Howells716ea3a2007-04-02 20:19:53 -0700609extern int skb_to_sgvec(struct sk_buff *skb,
610 struct scatterlist *sg, int offset,
611 int len);
612extern int skb_cow_data(struct sk_buff *skb, int tailbits,
613 struct sk_buff **trailer);
Herbert Xu5b057c62006-06-23 02:06:41 -0700614extern int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000615#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700617extern int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
618 int getfrag(void *from, char *to, int offset,
619 int len,int odd, struct sk_buff *skb),
620 void *from, int length);
621
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800622struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700623 __u32 lower_offset;
624 __u32 upper_offset;
625 __u32 frag_idx;
626 __u32 stepped_offset;
627 struct sk_buff *root_skb;
628 struct sk_buff *cur_skb;
629 __u8 *frag_data;
630};
631
632extern void skb_prepare_seq_read(struct sk_buff *skb,
633 unsigned int from, unsigned int to,
634 struct skb_seq_state *st);
635extern unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
636 struct skb_seq_state *st);
637extern void skb_abort_seq_read(struct skb_seq_state *st);
638
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700639extern unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
640 unsigned int to, struct ts_config *config,
641 struct ts_state *state);
642
Tom Herbertbdeab992011-08-14 19:45:55 +0000643extern void __skb_get_rxhash(struct sk_buff *skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000644static inline __u32 skb_get_rxhash(struct sk_buff *skb)
645{
646 if (!skb->rxhash)
Tom Herbertbdeab992011-08-14 19:45:55 +0000647 __skb_get_rxhash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000648
649 return skb->rxhash;
650}
651
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700652#ifdef NET_SKBUFF_DATA_USES_OFFSET
653static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
654{
655 return skb->head + skb->end;
656}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000657
658static inline unsigned int skb_end_offset(const struct sk_buff *skb)
659{
660 return skb->end;
661}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700662#else
663static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
664{
665 return skb->end;
666}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000667
668static inline unsigned int skb_end_offset(const struct sk_buff *skb)
669{
670 return skb->end - skb->head;
671}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700672#endif
673
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700675#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700676
Patrick Ohlyac45f602009-02-12 05:03:37 +0000677static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
678{
679 return &skb_shinfo(skb)->hwtstamps;
680}
681
Linus Torvalds1da177e2005-04-16 15:20:36 -0700682/**
683 * skb_queue_empty - check if a queue is empty
684 * @list: queue head
685 *
686 * Returns true if the queue is empty, false otherwise.
687 */
688static inline int skb_queue_empty(const struct sk_buff_head *list)
689{
690 return list->next == (struct sk_buff *)list;
691}
692
693/**
David S. Millerfc7ebb22008-09-23 00:34:07 -0700694 * skb_queue_is_last - check if skb is the last entry in the queue
695 * @list: queue head
696 * @skb: buffer
697 *
698 * Returns true if @skb is the last buffer on the list.
699 */
700static inline bool skb_queue_is_last(const struct sk_buff_head *list,
701 const struct sk_buff *skb)
702{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000703 return skb->next == (struct sk_buff *)list;
David S. Millerfc7ebb22008-09-23 00:34:07 -0700704}
705
706/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800707 * skb_queue_is_first - check if skb is the first entry in the queue
708 * @list: queue head
709 * @skb: buffer
710 *
711 * Returns true if @skb is the first buffer on the list.
712 */
713static inline bool skb_queue_is_first(const struct sk_buff_head *list,
714 const struct sk_buff *skb)
715{
Eric Dumazeta02cec22010-09-22 20:43:57 +0000716 return skb->prev == (struct sk_buff *)list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800717}
718
719/**
David S. Miller249c8b42008-09-23 00:44:42 -0700720 * skb_queue_next - return the next packet in the queue
721 * @list: queue head
722 * @skb: current buffer
723 *
724 * Return the next packet in @list after @skb. It is only valid to
725 * call this if skb_queue_is_last() evaluates to false.
726 */
727static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
728 const struct sk_buff *skb)
729{
730 /* This BUG_ON may seem severe, but if we just return then we
731 * are going to dereference garbage.
732 */
733 BUG_ON(skb_queue_is_last(list, skb));
734 return skb->next;
735}
736
737/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800738 * skb_queue_prev - return the prev packet in the queue
739 * @list: queue head
740 * @skb: current buffer
741 *
742 * Return the prev packet in @list before @skb. It is only valid to
743 * call this if skb_queue_is_first() evaluates to false.
744 */
745static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
746 const struct sk_buff *skb)
747{
748 /* This BUG_ON may seem severe, but if we just return then we
749 * are going to dereference garbage.
750 */
751 BUG_ON(skb_queue_is_first(list, skb));
752 return skb->prev;
753}
754
755/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756 * skb_get - reference buffer
757 * @skb: buffer to reference
758 *
759 * Makes another reference to a socket buffer and returns a pointer
760 * to the buffer.
761 */
762static inline struct sk_buff *skb_get(struct sk_buff *skb)
763{
764 atomic_inc(&skb->users);
765 return skb;
766}
767
768/*
769 * If users == 1, we are the only owner and are can avoid redundant
770 * atomic change.
771 */
772
773/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774 * skb_cloned - is the buffer a clone
775 * @skb: buffer to check
776 *
777 * Returns true if the buffer was generated with skb_clone() and is
778 * one of multiple shared copies of the buffer. Cloned buffers are
779 * shared data so must not be written to under normal circumstances.
780 */
781static inline int skb_cloned(const struct sk_buff *skb)
782{
783 return skb->cloned &&
784 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
785}
786
787/**
788 * skb_header_cloned - is the header a clone
789 * @skb: buffer to check
790 *
791 * Returns true if modifying the header part of the buffer requires
792 * the data to be copied.
793 */
794static inline int skb_header_cloned(const struct sk_buff *skb)
795{
796 int dataref;
797
798 if (!skb->cloned)
799 return 0;
800
801 dataref = atomic_read(&skb_shinfo(skb)->dataref);
802 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
803 return dataref != 1;
804}
805
806/**
807 * skb_header_release - release reference to header
808 * @skb: buffer to operate on
809 *
810 * Drop a reference to the header part of the buffer. This is done
811 * by acquiring a payload reference. You must not read from the header
812 * part of skb->data after this.
813 */
814static inline void skb_header_release(struct sk_buff *skb)
815{
816 BUG_ON(skb->nohdr);
817 skb->nohdr = 1;
818 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
819}
820
821/**
822 * skb_shared - is the buffer shared
823 * @skb: buffer to check
824 *
825 * Returns true if more than one person has a reference to this
826 * buffer.
827 */
828static inline int skb_shared(const struct sk_buff *skb)
829{
830 return atomic_read(&skb->users) != 1;
831}
832
833/**
834 * skb_share_check - check if buffer is shared and if so clone it
835 * @skb: buffer to check
836 * @pri: priority for memory allocation
837 *
838 * If the buffer is shared the buffer is cloned and the old copy
839 * drops a reference. A new clone with a single reference is returned.
840 * If the buffer is not shared the original buffer is returned. When
841 * being called from interrupt status or with spinlocks held pri must
842 * be GFP_ATOMIC.
843 *
844 * NULL is returned on a memory allocation failure.
845 */
Eric Dumazet47061bc2012-08-03 20:54:15 +0000846static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700847{
848 might_sleep_if(pri & __GFP_WAIT);
849 if (skb_shared(skb)) {
850 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +0000851
852 if (likely(nskb))
853 consume_skb(skb);
854 else
855 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700856 skb = nskb;
857 }
858 return skb;
859}
860
861/*
862 * Copy shared buffers into a new sk_buff. We effectively do COW on
863 * packets to handle cases where we have a local reader and forward
864 * and a couple of other messy ones. The normal one is tcpdumping
865 * a packet thats being forwarded.
866 */
867
868/**
869 * skb_unshare - make a copy of a shared buffer
870 * @skb: buffer to check
871 * @pri: priority for memory allocation
872 *
873 * If the socket buffer is a clone then this function creates a new
874 * copy of the data, drops a reference count on the old copy and returns
875 * the new copy with the reference count at 1. If the buffer is not a clone
876 * the original buffer is returned. When called with a spinlock held or
877 * from interrupt state @pri must be %GFP_ATOMIC
878 *
879 * %NULL is returned on a memory allocation failure.
880 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -0700881static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +0100882 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883{
884 might_sleep_if(pri & __GFP_WAIT);
885 if (skb_cloned(skb)) {
886 struct sk_buff *nskb = skb_copy(skb, pri);
887 kfree_skb(skb); /* Free our shared copy */
888 skb = nskb;
889 }
890 return skb;
891}
892
893/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -0800894 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895 * @list_: list to peek at
896 *
897 * Peek an &sk_buff. Unlike most other operations you _MUST_
898 * be careful with this one. A peek leaves the buffer on the
899 * list and someone else may run off with it. You must hold
900 * the appropriate locks or have a private queue to do this.
901 *
902 * Returns %NULL for an empty list or a pointer to the head element.
903 * The reference count is not incremented and the reference is therefore
904 * volatile. Use with caution.
905 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -0400906static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907{
Eric Dumazet18d07002012-04-30 16:31:46 +0000908 struct sk_buff *skb = list_->next;
909
910 if (skb == (struct sk_buff *)list_)
911 skb = NULL;
912 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700913}
914
915/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +0000916 * skb_peek_next - peek skb following the given one from a queue
917 * @skb: skb to start from
918 * @list_: list to peek at
919 *
920 * Returns %NULL when the end of the list is met or a pointer to the
921 * next element. The reference count is not incremented and the
922 * reference is therefore volatile. Use with caution.
923 */
924static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
925 const struct sk_buff_head *list_)
926{
927 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +0000928
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +0000929 if (next == (struct sk_buff *)list_)
930 next = NULL;
931 return next;
932}
933
934/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -0800935 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -0700936 * @list_: list to peek at
937 *
938 * Peek an &sk_buff. Unlike most other operations you _MUST_
939 * be careful with this one. A peek leaves the buffer on the
940 * list and someone else may run off with it. You must hold
941 * the appropriate locks or have a private queue to do this.
942 *
943 * Returns %NULL for an empty list or a pointer to the tail element.
944 * The reference count is not incremented and the reference is therefore
945 * volatile. Use with caution.
946 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -0400947static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700948{
Eric Dumazet18d07002012-04-30 16:31:46 +0000949 struct sk_buff *skb = list_->prev;
950
951 if (skb == (struct sk_buff *)list_)
952 skb = NULL;
953 return skb;
954
Linus Torvalds1da177e2005-04-16 15:20:36 -0700955}
956
957/**
958 * skb_queue_len - get queue length
959 * @list_: list to measure
960 *
961 * Return the length of an &sk_buff queue.
962 */
963static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
964{
965 return list_->qlen;
966}
967
David S. Miller67fed452008-09-21 22:36:24 -0700968/**
969 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
970 * @list: queue to initialize
971 *
972 * This initializes only the list and queue length aspects of
973 * an sk_buff_head object. This allows to initialize the list
974 * aspects of an sk_buff_head without reinitializing things like
975 * the spinlock. It can also be used for on-stack sk_buff_head
976 * objects where the spinlock is known to not be used.
977 */
978static inline void __skb_queue_head_init(struct sk_buff_head *list)
979{
980 list->prev = list->next = (struct sk_buff *)list;
981 list->qlen = 0;
982}
983
Arjan van de Ven76f10ad2006-08-02 14:06:55 -0700984/*
985 * This function creates a split out lock class for each invocation;
986 * this is needed for now since a whole lot of users of the skb-queue
987 * infrastructure in drivers have different locking usage (in hardirq)
988 * than the networking core (in softirq only). In the long run either the
989 * network layer or drivers should need annotation to consolidate the
990 * main types of usage into 3 classes.
991 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700992static inline void skb_queue_head_init(struct sk_buff_head *list)
993{
994 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -0700995 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700996}
997
Pavel Emelianovc2ecba72007-04-17 12:45:31 -0700998static inline void skb_queue_head_init_class(struct sk_buff_head *list,
999 struct lock_class_key *class)
1000{
1001 skb_queue_head_init(list);
1002 lockdep_set_class(&list->lock, class);
1003}
1004
Linus Torvalds1da177e2005-04-16 15:20:36 -07001005/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001006 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007 *
1008 * The "__skb_xxxx()" functions are the non-atomic ones that
1009 * can only be called with interrupts disabled.
1010 */
Gerrit Renkerbf299272008-04-14 00:04:51 -07001011extern void skb_insert(struct sk_buff *old, struct sk_buff *newsk, struct sk_buff_head *list);
1012static inline void __skb_insert(struct sk_buff *newsk,
1013 struct sk_buff *prev, struct sk_buff *next,
1014 struct sk_buff_head *list)
1015{
1016 newsk->next = next;
1017 newsk->prev = prev;
1018 next->prev = prev->next = newsk;
1019 list->qlen++;
1020}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001021
David S. Miller67fed452008-09-21 22:36:24 -07001022static inline void __skb_queue_splice(const struct sk_buff_head *list,
1023 struct sk_buff *prev,
1024 struct sk_buff *next)
1025{
1026 struct sk_buff *first = list->next;
1027 struct sk_buff *last = list->prev;
1028
1029 first->prev = prev;
1030 prev->next = first;
1031
1032 last->next = next;
1033 next->prev = last;
1034}
1035
1036/**
1037 * skb_queue_splice - join two skb lists, this is designed for stacks
1038 * @list: the new list to add
1039 * @head: the place to add it in the first list
1040 */
1041static inline void skb_queue_splice(const struct sk_buff_head *list,
1042 struct sk_buff_head *head)
1043{
1044 if (!skb_queue_empty(list)) {
1045 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001046 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001047 }
1048}
1049
1050/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001051 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001052 * @list: the new list to add
1053 * @head: the place to add it in the first list
1054 *
1055 * The list at @list is reinitialised
1056 */
1057static inline void skb_queue_splice_init(struct sk_buff_head *list,
1058 struct sk_buff_head *head)
1059{
1060 if (!skb_queue_empty(list)) {
1061 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001062 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001063 __skb_queue_head_init(list);
1064 }
1065}
1066
1067/**
1068 * skb_queue_splice_tail - join two skb lists, each list being a queue
1069 * @list: the new list to add
1070 * @head: the place to add it in the first list
1071 */
1072static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1073 struct sk_buff_head *head)
1074{
1075 if (!skb_queue_empty(list)) {
1076 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001077 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001078 }
1079}
1080
1081/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001082 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001083 * @list: the new list to add
1084 * @head: the place to add it in the first list
1085 *
1086 * Each of the lists is a queue.
1087 * The list at @list is reinitialised
1088 */
1089static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1090 struct sk_buff_head *head)
1091{
1092 if (!skb_queue_empty(list)) {
1093 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001094 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001095 __skb_queue_head_init(list);
1096 }
1097}
1098
Linus Torvalds1da177e2005-04-16 15:20:36 -07001099/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001100 * __skb_queue_after - queue a buffer at the list head
1101 * @list: list to use
1102 * @prev: place after this buffer
1103 * @newsk: buffer to queue
1104 *
1105 * Queue a buffer int the middle of a list. This function takes no locks
1106 * and you must therefore hold required locks before calling it.
1107 *
1108 * A buffer cannot be placed on two lists at the same time.
1109 */
1110static inline void __skb_queue_after(struct sk_buff_head *list,
1111 struct sk_buff *prev,
1112 struct sk_buff *newsk)
1113{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001114 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001115}
1116
Gerrit Renker7de6c032008-04-14 00:05:09 -07001117extern void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1118 struct sk_buff_head *list);
1119
Gerrit Renkerf5572852008-04-14 00:05:28 -07001120static inline void __skb_queue_before(struct sk_buff_head *list,
1121 struct sk_buff *next,
1122 struct sk_buff *newsk)
1123{
1124 __skb_insert(newsk, next->prev, next, list);
1125}
1126
Stephen Hemminger300ce172005-10-30 13:47:34 -08001127/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001128 * __skb_queue_head - queue a buffer at the list head
1129 * @list: list to use
1130 * @newsk: buffer to queue
1131 *
1132 * Queue a buffer at the start of a list. This function takes no locks
1133 * and you must therefore hold required locks before calling it.
1134 *
1135 * A buffer cannot be placed on two lists at the same time.
1136 */
1137extern void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
1138static inline void __skb_queue_head(struct sk_buff_head *list,
1139 struct sk_buff *newsk)
1140{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001141 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001142}
1143
1144/**
1145 * __skb_queue_tail - queue a buffer at the list tail
1146 * @list: list to use
1147 * @newsk: buffer to queue
1148 *
1149 * Queue a buffer at the end of a list. This function takes no locks
1150 * and you must therefore hold required locks before calling it.
1151 *
1152 * A buffer cannot be placed on two lists at the same time.
1153 */
1154extern void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
1155static inline void __skb_queue_tail(struct sk_buff_head *list,
1156 struct sk_buff *newsk)
1157{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001158 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159}
1160
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162 * remove sk_buff from list. _Must_ be called atomically, and with
1163 * the list known..
1164 */
David S. Miller8728b832005-08-09 19:25:21 -07001165extern void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001166static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1167{
1168 struct sk_buff *next, *prev;
1169
1170 list->qlen--;
1171 next = skb->next;
1172 prev = skb->prev;
1173 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001174 next->prev = prev;
1175 prev->next = next;
1176}
1177
Gerrit Renkerf525c062008-04-14 00:04:12 -07001178/**
1179 * __skb_dequeue - remove from the head of the queue
1180 * @list: list to dequeue from
1181 *
1182 * Remove the head of the list. This function does not take any locks
1183 * so must be used with appropriate locks held only. The head item is
1184 * returned or %NULL if the list is empty.
1185 */
1186extern struct sk_buff *skb_dequeue(struct sk_buff_head *list);
1187static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1188{
1189 struct sk_buff *skb = skb_peek(list);
1190 if (skb)
1191 __skb_unlink(skb, list);
1192 return skb;
1193}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001194
1195/**
1196 * __skb_dequeue_tail - remove from the tail of the queue
1197 * @list: list to dequeue from
1198 *
1199 * Remove the tail of the list. This function does not take any locks
1200 * so must be used with appropriate locks held only. The tail item is
1201 * returned or %NULL if the list is empty.
1202 */
1203extern struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
1204static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1205{
1206 struct sk_buff *skb = skb_peek_tail(list);
1207 if (skb)
1208 __skb_unlink(skb, list);
1209 return skb;
1210}
1211
1212
David S. Millerbdcc0922012-03-07 20:53:36 -05001213static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001214{
1215 return skb->data_len;
1216}
1217
1218static inline unsigned int skb_headlen(const struct sk_buff *skb)
1219{
1220 return skb->len - skb->data_len;
1221}
1222
1223static inline int skb_pagelen(const struct sk_buff *skb)
1224{
1225 int i, len = 0;
1226
1227 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001228 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001229 return len + skb_headlen(skb);
1230}
1231
Ian Campbell131ea662011-08-19 06:25:00 +00001232/**
1233 * __skb_fill_page_desc - initialise a paged fragment in an skb
1234 * @skb: buffer containing fragment to be initialised
1235 * @i: paged fragment index to initialise
1236 * @page: the page to use for this fragment
1237 * @off: the offset to the data with @page
1238 * @size: the length of the data
1239 *
1240 * Initialises the @i'th fragment of @skb to point to &size bytes at
1241 * offset @off within @page.
1242 *
1243 * Does not take any additional reference on the fragment.
1244 */
1245static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1246 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001247{
1248 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1249
Mel Gormanc48a11c2012-07-31 16:44:23 -07001250 /*
1251 * Propagate page->pfmemalloc to the skb if we can. The problem is
1252 * that not all callers have unique ownership of the page. If
1253 * pfmemalloc is set, we check the mapping as a mapping implies
1254 * page->index is set (index and pfmemalloc share space).
1255 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1256 * do not lose pfmemalloc information as the pages would not be
1257 * allocated using __GFP_MEMALLOC.
1258 */
1259 if (page->pfmemalloc && !page->mapping)
1260 skb->pfmemalloc = true;
Ian Campbella8605c62011-10-19 23:01:49 +00001261 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001262 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001263 skb_frag_size_set(frag, size);
Ian Campbell131ea662011-08-19 06:25:00 +00001264}
1265
1266/**
1267 * skb_fill_page_desc - initialise a paged fragment in an skb
1268 * @skb: buffer containing fragment to be initialised
1269 * @i: paged fragment index to initialise
1270 * @page: the page to use for this fragment
1271 * @off: the offset to the data with @page
1272 * @size: the length of the data
1273 *
1274 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
1275 * @skb to point to &size bytes at offset @off within @page. In
1276 * addition updates @skb such that @i is the last fragment.
1277 *
1278 * Does not take any additional reference on the fragment.
1279 */
1280static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1281 struct page *page, int off, int size)
1282{
1283 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001284 skb_shinfo(skb)->nr_frags = i + 1;
1285}
1286
Peter Zijlstra654bed12008-10-07 14:22:33 -07001287extern void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page,
Eric Dumazet50269e12012-03-23 23:59:33 +00001288 int off, int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001289
Linus Torvalds1da177e2005-04-16 15:20:36 -07001290#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001291#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1293
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001294#ifdef NET_SKBUFF_DATA_USES_OFFSET
1295static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1296{
1297 return skb->head + skb->tail;
1298}
1299
1300static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1301{
1302 skb->tail = skb->data - skb->head;
1303}
1304
1305static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1306{
1307 skb_reset_tail_pointer(skb);
1308 skb->tail += offset;
1309}
1310#else /* NET_SKBUFF_DATA_USES_OFFSET */
1311static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1312{
1313 return skb->tail;
1314}
1315
1316static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1317{
1318 skb->tail = skb->data;
1319}
1320
1321static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1322{
1323 skb->tail = skb->data + offset;
1324}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001325
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001326#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1327
Linus Torvalds1da177e2005-04-16 15:20:36 -07001328/*
1329 * Add data to an sk_buff
1330 */
Ilpo Järvinen0dde3e12008-03-27 17:43:41 -07001331extern unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1333{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001334 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001335 SKB_LINEAR_ASSERT(skb);
1336 skb->tail += len;
1337 skb->len += len;
1338 return tmp;
1339}
1340
Ilpo Järvinenc2aa2702008-03-27 17:52:40 -07001341extern unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1343{
1344 skb->data -= len;
1345 skb->len += len;
1346 return skb->data;
1347}
1348
Ilpo Järvinen6be8ac22008-03-27 17:47:24 -07001349extern unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001350static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1351{
1352 skb->len -= len;
1353 BUG_ON(skb->len < skb->data_len);
1354 return skb->data += len;
1355}
1356
David S. Miller47d29642010-05-02 02:21:44 -07001357static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1358{
1359 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1360}
1361
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362extern unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
1363
1364static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1365{
1366 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001367 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001368 return NULL;
1369 skb->len -= len;
1370 return skb->data += len;
1371}
1372
1373static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1374{
1375 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1376}
1377
1378static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1379{
1380 if (likely(len <= skb_headlen(skb)))
1381 return 1;
1382 if (unlikely(len > skb->len))
1383 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001384 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385}
1386
1387/**
1388 * skb_headroom - bytes at buffer head
1389 * @skb: buffer to check
1390 *
1391 * Return the number of bytes of free space at the head of an &sk_buff.
1392 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001393static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394{
1395 return skb->data - skb->head;
1396}
1397
1398/**
1399 * skb_tailroom - bytes at buffer end
1400 * @skb: buffer to check
1401 *
1402 * Return the number of bytes of free space at the tail of an sk_buff
1403 */
1404static inline int skb_tailroom(const struct sk_buff *skb)
1405{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001406 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407}
1408
1409/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00001410 * skb_availroom - bytes at buffer end
1411 * @skb: buffer to check
1412 *
1413 * Return the number of bytes of free space at the tail of an sk_buff
1414 * allocated by sk_stream_alloc()
1415 */
1416static inline int skb_availroom(const struct sk_buff *skb)
1417{
1418 return skb_is_nonlinear(skb) ? 0 : skb->avail_size - skb->len;
1419}
1420
1421/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001422 * skb_reserve - adjust headroom
1423 * @skb: buffer to alter
1424 * @len: bytes to move
1425 *
1426 * Increase the headroom of an empty &sk_buff by reducing the tail
1427 * room. This is only allowed for an empty buffer.
1428 */
David S. Miller8243126c2006-01-17 02:54:21 -08001429static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001430{
1431 skb->data += len;
1432 skb->tail += len;
1433}
1434
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00001435static inline void skb_reset_mac_len(struct sk_buff *skb)
1436{
1437 skb->mac_len = skb->network_header - skb->mac_header;
1438}
1439
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001440#ifdef NET_SKBUFF_DATA_USES_OFFSET
1441static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1442{
1443 return skb->head + skb->transport_header;
1444}
1445
1446static inline void skb_reset_transport_header(struct sk_buff *skb)
1447{
1448 skb->transport_header = skb->data - skb->head;
1449}
1450
1451static inline void skb_set_transport_header(struct sk_buff *skb,
1452 const int offset)
1453{
1454 skb_reset_transport_header(skb);
1455 skb->transport_header += offset;
1456}
1457
1458static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1459{
1460 return skb->head + skb->network_header;
1461}
1462
1463static inline void skb_reset_network_header(struct sk_buff *skb)
1464{
1465 skb->network_header = skb->data - skb->head;
1466}
1467
1468static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1469{
1470 skb_reset_network_header(skb);
1471 skb->network_header += offset;
1472}
1473
1474static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1475{
1476 return skb->head + skb->mac_header;
1477}
1478
1479static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1480{
1481 return skb->mac_header != ~0U;
1482}
1483
1484static inline void skb_reset_mac_header(struct sk_buff *skb)
1485{
1486 skb->mac_header = skb->data - skb->head;
1487}
1488
1489static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1490{
1491 skb_reset_mac_header(skb);
1492 skb->mac_header += offset;
1493}
1494
1495#else /* NET_SKBUFF_DATA_USES_OFFSET */
1496
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001497static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1498{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001499 return skb->transport_header;
Arnaldo Carvalho de Melo9c702202007-04-25 18:04:18 -07001500}
1501
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001502static inline void skb_reset_transport_header(struct sk_buff *skb)
1503{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001504 skb->transport_header = skb->data;
Arnaldo Carvalho de Melobadff6d2007-03-13 13:06:52 -03001505}
1506
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001507static inline void skb_set_transport_header(struct sk_buff *skb,
1508 const int offset)
1509{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001510 skb->transport_header = skb->data + offset;
Arnaldo Carvalho de Melo967b05f2007-03-13 13:51:52 -03001511}
1512
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001513static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1514{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001515 return skb->network_header;
Arnaldo Carvalho de Melod56f90a2007-04-10 20:50:43 -07001516}
1517
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001518static inline void skb_reset_network_header(struct sk_buff *skb)
1519{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001520 skb->network_header = skb->data;
Arnaldo Carvalho de Meloc1d2bbe2007-04-10 20:45:18 -07001521}
1522
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001523static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1524{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001525 skb->network_header = skb->data + offset;
Arnaldo Carvalho de Meloc14d2452007-03-11 22:39:41 -03001526}
1527
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001528static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1529{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001530 return skb->mac_header;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001531}
1532
1533static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1534{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001535 return skb->mac_header != NULL;
Arnaldo Carvalho de Melo98e399f2007-03-19 15:33:04 -07001536}
1537
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001538static inline void skb_reset_mac_header(struct sk_buff *skb)
1539{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001540 skb->mac_header = skb->data;
Arnaldo Carvalho de Melo459a98e2007-03-19 15:30:44 -07001541}
1542
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001543static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1544{
Arnaldo Carvalho de Melob0e380b2007-04-10 21:21:55 -07001545 skb->mac_header = skb->data + offset;
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001546}
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001547#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1548
Eric Dumazet03606892012-02-23 10:55:02 +00001549static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1550{
1551 if (skb_mac_header_was_set(skb)) {
1552 const unsigned char *old_mac = skb_mac_header(skb);
1553
1554 skb_set_mac_header(skb, -skb->mac_len);
1555 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1556 }
1557}
1558
Michał Mirosław04fb4512010-12-14 15:24:08 +00001559static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1560{
1561 return skb->csum_start - skb_headroom(skb);
1562}
1563
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001564static inline int skb_transport_offset(const struct sk_buff *skb)
1565{
1566 return skb_transport_header(skb) - skb->data;
1567}
1568
1569static inline u32 skb_network_header_len(const struct sk_buff *skb)
1570{
1571 return skb->transport_header - skb->network_header;
1572}
1573
1574static inline int skb_network_offset(const struct sk_buff *skb)
1575{
1576 return skb_network_header(skb) - skb->data;
1577}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001578
Changli Gaof9599ce2010-08-04 04:43:44 +00001579static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1580{
1581 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1582}
1583
Linus Torvalds1da177e2005-04-16 15:20:36 -07001584/*
1585 * CPUs often take a performance hit when accessing unaligned memory
1586 * locations. The actual performance hit varies, it can be small if the
1587 * hardware handles it or large if we have to take an exception and fix it
1588 * in software.
1589 *
1590 * Since an ethernet header is 14 bytes network drivers often end up with
1591 * the IP header at an unaligned offset. The IP header can be aligned by
1592 * shifting the start of the packet by 2 bytes. Drivers should do this
1593 * with:
1594 *
Tobias Klauser8660c122009-07-13 22:48:16 +00001595 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001596 *
1597 * The downside to this alignment of the IP header is that the DMA is now
1598 * unaligned. On some architectures the cost of an unaligned DMA is high
1599 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00001600 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001601 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1602 * to be overridden.
1603 */
1604#ifndef NET_IP_ALIGN
1605#define NET_IP_ALIGN 2
1606#endif
1607
Anton Blanchard025be812006-03-31 02:27:06 -08001608/*
1609 * The networking layer reserves some headroom in skb data (via
1610 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1611 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08001612 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08001613 *
1614 * Unfortunately this headroom changes the DMA alignment of the resulting
1615 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1616 * on some architectures. An architecture can override this value,
1617 * perhaps setting it to a cacheline in size (since that will maintain
1618 * cacheline alignment of the DMA). It must be a power of 2.
1619 *
David S. Millerd6301d32009-02-08 19:24:13 -08001620 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08001621 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07001622 *
1623 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1624 * to reduce average number of cache lines per packet.
1625 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07001626 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08001627 */
1628#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07001629#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08001630#endif
1631
Herbert Xu3cc0e872006-06-09 16:13:38 -07001632extern int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001633
1634static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1635{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00001636 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07001637 WARN_ON(1);
1638 return;
1639 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001640 skb->len = len;
1641 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001642}
1643
Ilpo Järvinen419ae742008-03-27 17:54:01 -07001644extern void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001645
1646static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1647{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001648 if (skb->data_len)
1649 return ___pskb_trim(skb, len);
1650 __skb_trim(skb, len);
1651 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001652}
1653
1654static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1655{
1656 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1657}
1658
1659/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07001660 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1661 * @skb: buffer to alter
1662 * @len: new length
1663 *
1664 * This is identical to pskb_trim except that the caller knows that
1665 * the skb is not cloned so we should never get an error due to out-
1666 * of-memory.
1667 */
1668static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1669{
1670 int err = pskb_trim(skb, len);
1671 BUG_ON(err);
1672}
1673
1674/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001675 * skb_orphan - orphan a buffer
1676 * @skb: buffer to orphan
1677 *
1678 * If a buffer currently has an owner then we call the owner's
1679 * destructor function and make the @skb unowned. The buffer continues
1680 * to exist but is no longer charged to its former owner.
1681 */
1682static inline void skb_orphan(struct sk_buff *skb)
1683{
1684 if (skb->destructor)
1685 skb->destructor(skb);
1686 skb->destructor = NULL;
1687 skb->sk = NULL;
1688}
1689
1690/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00001691 * skb_orphan_frags - orphan the frags contained in a buffer
1692 * @skb: buffer to orphan frags from
1693 * @gfp_mask: allocation mask for replacement pages
1694 *
1695 * For each frag in the SKB which needs a destructor (i.e. has an
1696 * owner) create a copy of that frag and release the original
1697 * page by calling the destructor.
1698 */
1699static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1700{
1701 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1702 return 0;
1703 return skb_copy_ubufs(skb, gfp_mask);
1704}
1705
1706/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001707 * __skb_queue_purge - empty a list
1708 * @list: list to empty
1709 *
1710 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1711 * the list and one reference dropped. This function does not take the
1712 * list lock and the caller must hold the relevant locks to use it.
1713 */
1714extern void skb_queue_purge(struct sk_buff_head *list);
1715static inline void __skb_queue_purge(struct sk_buff_head *list)
1716{
1717 struct sk_buff *skb;
1718 while ((skb = __skb_dequeue(list)) != NULL)
1719 kfree_skb(skb);
1720}
1721
Eric Dumazet6f532612012-05-18 05:12:12 +00001722extern void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001723
Christoph Hellwig8af27452006-07-31 22:35:23 -07001724extern struct sk_buff *__netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00001725 unsigned int length,
1726 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07001727
1728/**
1729 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1730 * @dev: network device to receive on
1731 * @length: length to allocate
1732 *
1733 * Allocate a new &sk_buff and assign it a usage count of one. The
1734 * buffer has unspecified headroom built in. Users should allocate
1735 * the headroom they think they need without accounting for the
1736 * built in space. The built in space is used for optimisations.
1737 *
1738 * %NULL is returned if there is no free memory. Although this function
1739 * allocates memory it can be called from an interrupt.
1740 */
1741static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00001742 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07001743{
1744 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
1745}
1746
Eric Dumazet6f532612012-05-18 05:12:12 +00001747/* legacy helper around __netdev_alloc_skb() */
1748static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
1749 gfp_t gfp_mask)
1750{
1751 return __netdev_alloc_skb(NULL, length, gfp_mask);
1752}
1753
1754/* legacy helper around netdev_alloc_skb() */
1755static inline struct sk_buff *dev_alloc_skb(unsigned int length)
1756{
1757 return netdev_alloc_skb(NULL, length);
1758}
1759
1760
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001761static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
1762 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00001763{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001764 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00001765
1766 if (NET_IP_ALIGN && skb)
1767 skb_reserve(skb, NET_IP_ALIGN);
1768 return skb;
1769}
1770
Eric Dumazet4915a0d2011-07-11 20:08:34 -07001771static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
1772 unsigned int length)
1773{
1774 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
1775}
1776
Mel Gorman06140022012-07-31 16:44:24 -07001777/*
1778 * __skb_alloc_page - allocate pages for ps-rx on a skb and preserve pfmemalloc data
1779 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1780 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1781 * @order: size of the allocation
1782 *
1783 * Allocate a new page.
1784 *
1785 * %NULL is returned if there is no free memory.
1786*/
1787static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
1788 struct sk_buff *skb,
1789 unsigned int order)
1790{
1791 struct page *page;
1792
1793 gfp_mask |= __GFP_COLD;
1794
1795 if (!(gfp_mask & __GFP_NOMEMALLOC))
1796 gfp_mask |= __GFP_MEMALLOC;
1797
1798 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
1799 if (skb && page && page->pfmemalloc)
1800 skb->pfmemalloc = true;
1801
1802 return page;
1803}
1804
1805/**
1806 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
1807 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
1808 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
1809 *
1810 * Allocate a new page.
1811 *
1812 * %NULL is returned if there is no free memory.
1813 */
1814static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
1815 struct sk_buff *skb)
1816{
1817 return __skb_alloc_pages(gfp_mask, skb, 0);
1818}
1819
1820/**
1821 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
1822 * @page: The page that was allocated from skb_alloc_page
1823 * @skb: The skb that may need pfmemalloc set
1824 */
1825static inline void skb_propagate_pfmemalloc(struct page *page,
1826 struct sk_buff *skb)
1827{
1828 if (page && page->pfmemalloc)
1829 skb->pfmemalloc = true;
1830}
1831
Eric Dumazet564824b2010-10-11 19:05:25 +00001832/**
Ian Campbell131ea662011-08-19 06:25:00 +00001833 * skb_frag_page - retrieve the page refered to by a paged fragment
1834 * @frag: the paged fragment
1835 *
1836 * Returns the &struct page associated with @frag.
1837 */
1838static inline struct page *skb_frag_page(const skb_frag_t *frag)
1839{
Ian Campbella8605c62011-10-19 23:01:49 +00001840 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00001841}
1842
1843/**
1844 * __skb_frag_ref - take an addition reference on a paged fragment.
1845 * @frag: the paged fragment
1846 *
1847 * Takes an additional reference on the paged fragment @frag.
1848 */
1849static inline void __skb_frag_ref(skb_frag_t *frag)
1850{
1851 get_page(skb_frag_page(frag));
1852}
1853
1854/**
1855 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
1856 * @skb: the buffer
1857 * @f: the fragment offset.
1858 *
1859 * Takes an additional reference on the @f'th paged fragment of @skb.
1860 */
1861static inline void skb_frag_ref(struct sk_buff *skb, int f)
1862{
1863 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
1864}
1865
1866/**
1867 * __skb_frag_unref - release a reference on a paged fragment.
1868 * @frag: the paged fragment
1869 *
1870 * Releases a reference on the paged fragment @frag.
1871 */
1872static inline void __skb_frag_unref(skb_frag_t *frag)
1873{
1874 put_page(skb_frag_page(frag));
1875}
1876
1877/**
1878 * skb_frag_unref - release a reference on a paged fragment of an skb.
1879 * @skb: the buffer
1880 * @f: the fragment offset
1881 *
1882 * Releases a reference on the @f'th paged fragment of @skb.
1883 */
1884static inline void skb_frag_unref(struct sk_buff *skb, int f)
1885{
1886 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
1887}
1888
1889/**
1890 * skb_frag_address - gets the address of the data contained in a paged fragment
1891 * @frag: the paged fragment buffer
1892 *
1893 * Returns the address of the data within @frag. The page must already
1894 * be mapped.
1895 */
1896static inline void *skb_frag_address(const skb_frag_t *frag)
1897{
1898 return page_address(skb_frag_page(frag)) + frag->page_offset;
1899}
1900
1901/**
1902 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
1903 * @frag: the paged fragment buffer
1904 *
1905 * Returns the address of the data within @frag. Checks that the page
1906 * is mapped and returns %NULL otherwise.
1907 */
1908static inline void *skb_frag_address_safe(const skb_frag_t *frag)
1909{
1910 void *ptr = page_address(skb_frag_page(frag));
1911 if (unlikely(!ptr))
1912 return NULL;
1913
1914 return ptr + frag->page_offset;
1915}
1916
1917/**
1918 * __skb_frag_set_page - sets the page contained in a paged fragment
1919 * @frag: the paged fragment
1920 * @page: the page to set
1921 *
1922 * Sets the fragment @frag to contain @page.
1923 */
1924static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
1925{
Ian Campbella8605c62011-10-19 23:01:49 +00001926 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00001927}
1928
1929/**
1930 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
1931 * @skb: the buffer
1932 * @f: the fragment offset
1933 * @page: the page to set
1934 *
1935 * Sets the @f'th fragment of @skb to contain @page.
1936 */
1937static inline void skb_frag_set_page(struct sk_buff *skb, int f,
1938 struct page *page)
1939{
1940 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
1941}
1942
1943/**
1944 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00001945 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00001946 * @frag: the paged fragment to map
1947 * @offset: the offset within the fragment (starting at the
1948 * fragment's own offset)
1949 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00001950 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00001951 *
1952 * Maps the page associated with @frag to @device.
1953 */
1954static inline dma_addr_t skb_frag_dma_map(struct device *dev,
1955 const skb_frag_t *frag,
1956 size_t offset, size_t size,
1957 enum dma_data_direction dir)
1958{
1959 return dma_map_page(dev, skb_frag_page(frag),
1960 frag->page_offset + offset, size, dir);
1961}
1962
Eric Dumazet117632e2011-12-03 21:39:53 +00001963static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
1964 gfp_t gfp_mask)
1965{
1966 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
1967}
1968
Ian Campbell131ea662011-08-19 06:25:00 +00001969/**
Patrick McHardy334a8132007-06-25 04:35:20 -07001970 * skb_clone_writable - is the header of a clone writable
1971 * @skb: buffer to check
1972 * @len: length up to which to write
1973 *
1974 * Returns true if modifying the header part of the cloned buffer
1975 * does not requires the data to be copied.
1976 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001977static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07001978{
1979 return !skb_header_cloned(skb) &&
1980 skb_headroom(skb) + len <= skb->hdr_len;
1981}
1982
Herbert Xud9cc2042007-09-16 16:21:16 -07001983static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
1984 int cloned)
1985{
1986 int delta = 0;
1987
Herbert Xud9cc2042007-09-16 16:21:16 -07001988 if (headroom > skb_headroom(skb))
1989 delta = headroom - skb_headroom(skb);
1990
1991 if (delta || cloned)
1992 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
1993 GFP_ATOMIC);
1994 return 0;
1995}
1996
Patrick McHardy334a8132007-06-25 04:35:20 -07001997/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001998 * skb_cow - copy header of skb when it is required
1999 * @skb: buffer to cow
2000 * @headroom: needed headroom
2001 *
2002 * If the skb passed lacks sufficient headroom or its data part
2003 * is shared, data is reallocated. If reallocation fails, an error
2004 * is returned and original skb is not changed.
2005 *
2006 * The result is skb with writable area skb->head...skb->tail
2007 * and at least @headroom of space at head.
2008 */
2009static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2010{
Herbert Xud9cc2042007-09-16 16:21:16 -07002011 return __skb_cow(skb, headroom, skb_cloned(skb));
2012}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002013
Herbert Xud9cc2042007-09-16 16:21:16 -07002014/**
2015 * skb_cow_head - skb_cow but only making the head writable
2016 * @skb: buffer to cow
2017 * @headroom: needed headroom
2018 *
2019 * This function is identical to skb_cow except that we replace the
2020 * skb_cloned check by skb_header_cloned. It should be used when
2021 * you only need to push on some header and do not need to modify
2022 * the data.
2023 */
2024static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2025{
2026 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002027}
2028
2029/**
2030 * skb_padto - pad an skbuff up to a minimal size
2031 * @skb: buffer to pad
2032 * @len: minimal length
2033 *
2034 * Pads up a buffer to ensure the trailing bytes exist and are
2035 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002036 * is untouched. Otherwise it is extended. Returns zero on
2037 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002038 */
2039
Herbert Xu5b057c62006-06-23 02:06:41 -07002040static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002041{
2042 unsigned int size = skb->len;
2043 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002044 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002045 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002046}
2047
2048static inline int skb_add_data(struct sk_buff *skb,
2049 char __user *from, int copy)
2050{
2051 const int off = skb->len;
2052
2053 if (skb->ip_summed == CHECKSUM_NONE) {
2054 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08002055 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
Linus Torvalds1da177e2005-04-16 15:20:36 -07002056 copy, 0, &err);
2057 if (!err) {
2058 skb->csum = csum_block_add(skb->csum, csum, off);
2059 return 0;
2060 }
2061 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2062 return 0;
2063
2064 __skb_trim(skb, off);
2065 return -EFAULT;
2066}
2067
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002068static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2069 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002070{
2071 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002072 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002073
Ian Campbellea2ab692011-08-22 23:44:58 +00002074 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002075 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002076 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002077 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002078}
2079
Herbert Xu364c6ba2006-06-09 16:10:40 -07002080static inline int __skb_linearize(struct sk_buff *skb)
2081{
2082 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2083}
2084
Linus Torvalds1da177e2005-04-16 15:20:36 -07002085/**
2086 * skb_linearize - convert paged skb to linear one
2087 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002088 *
2089 * If there is no free memory -ENOMEM is returned, otherwise zero
2090 * is returned and the old skb data released.
2091 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002092static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002093{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002094 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2095}
2096
2097/**
2098 * skb_linearize_cow - make sure skb is linear and writable
2099 * @skb: buffer to process
2100 *
2101 * If there is no free memory -ENOMEM is returned, otherwise zero
2102 * is returned and the old skb data released.
2103 */
2104static inline int skb_linearize_cow(struct sk_buff *skb)
2105{
2106 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2107 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002108}
2109
2110/**
2111 * skb_postpull_rcsum - update checksum for received skb after pull
2112 * @skb: buffer to update
2113 * @start: start of data before pull
2114 * @len: length of data pulled
2115 *
2116 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002117 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2118 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002119 */
2120
2121static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002122 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002123{
Patrick McHardy84fa7932006-08-29 16:44:56 -07002124 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002125 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2126}
2127
Herbert Xucbb042f2006-03-20 22:43:56 -08002128unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2129
Linus Torvalds1da177e2005-04-16 15:20:36 -07002130/**
2131 * pskb_trim_rcsum - trim received skb and update checksum
2132 * @skb: buffer to trim
2133 * @len: new length
2134 *
2135 * This is exactly the same as pskb_trim except that it ensures the
2136 * checksum of received packets are still valid after the operation.
2137 */
2138
2139static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2140{
Stephen Hemminger0e4e4222005-09-08 12:32:03 -07002141 if (likely(len >= skb->len))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002142 return 0;
Patrick McHardy84fa7932006-08-29 16:44:56 -07002143 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002144 skb->ip_summed = CHECKSUM_NONE;
2145 return __pskb_trim(skb, len);
2146}
2147
Linus Torvalds1da177e2005-04-16 15:20:36 -07002148#define skb_queue_walk(queue, skb) \
2149 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002150 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002151 skb = skb->next)
2152
James Chapman46f89142007-04-30 00:07:31 -07002153#define skb_queue_walk_safe(queue, skb, tmp) \
2154 for (skb = (queue)->next, tmp = skb->next; \
2155 skb != (struct sk_buff *)(queue); \
2156 skb = tmp, tmp = skb->next)
2157
David S. Miller1164f522008-09-23 00:49:44 -07002158#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002159 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002160 skb = skb->next)
2161
2162#define skb_queue_walk_from_safe(queue, skb, tmp) \
2163 for (tmp = skb->next; \
2164 skb != (struct sk_buff *)(queue); \
2165 skb = tmp, tmp = skb->next)
2166
Stephen Hemminger300ce172005-10-30 13:47:34 -08002167#define skb_queue_reverse_walk(queue, skb) \
2168 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002169 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08002170 skb = skb->prev)
2171
David S. Miller686a2952011-01-20 22:47:32 -08002172#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2173 for (skb = (queue)->prev, tmp = skb->prev; \
2174 skb != (struct sk_buff *)(queue); \
2175 skb = tmp, tmp = skb->prev)
2176
2177#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2178 for (tmp = skb->prev; \
2179 skb != (struct sk_buff *)(queue); \
2180 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002181
David S. Miller21dc3302010-08-23 00:13:46 -07002182static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07002183{
2184 return skb_shinfo(skb)->frag_list != NULL;
2185}
2186
2187static inline void skb_frag_list_init(struct sk_buff *skb)
2188{
2189 skb_shinfo(skb)->frag_list = NULL;
2190}
2191
2192static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2193{
2194 frag->next = skb_shinfo(skb)->frag_list;
2195 skb_shinfo(skb)->frag_list = frag;
2196}
2197
2198#define skb_walk_frags(skb, iter) \
2199 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2200
Herbert Xua59322b2007-12-05 01:53:40 -08002201extern struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
Pavel Emelyanov3f518bf2012-02-21 07:30:58 +00002202 int *peeked, int *off, int *err);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002203extern struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags,
2204 int noblock, int *err);
2205extern unsigned int datagram_poll(struct file *file, struct socket *sock,
2206 struct poll_table_struct *wait);
2207extern int skb_copy_datagram_iovec(const struct sk_buff *from,
2208 int offset, struct iovec *to,
2209 int size);
Herbert Xufb286bb2005-11-10 13:01:24 -08002210extern int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002211 int hlen,
2212 struct iovec *iov);
Rusty Russelldb543c12008-08-15 15:13:53 -07002213extern int skb_copy_datagram_from_iovec(struct sk_buff *skb,
2214 int offset,
Michael S. Tsirkin6f26c9a2009-04-20 01:26:11 +00002215 const struct iovec *from,
2216 int from_offset,
Rusty Russelldb543c12008-08-15 15:13:53 -07002217 int len);
Michael S. Tsirkin0a1ec072009-04-20 01:25:46 +00002218extern int skb_copy_datagram_const_iovec(const struct sk_buff *from,
2219 int offset,
2220 const struct iovec *to,
2221 int to_offset,
2222 int size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002223extern void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
Eric Dumazet9d410c72009-10-30 05:03:53 +00002224extern void skb_free_datagram_locked(struct sock *sk,
2225 struct sk_buff *skb);
Herbert Xu27ab2562007-12-05 01:51:58 -08002226extern int skb_kill_datagram(struct sock *sk, struct sk_buff *skb,
Herbert Xu3305b802005-12-13 23:16:37 -08002227 unsigned int flags);
Al Viro2bbbc862006-11-14 21:37:14 -08002228extern __wsum skb_checksum(const struct sk_buff *skb, int offset,
2229 int len, __wsum csum);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002230extern int skb_copy_bits(const struct sk_buff *skb, int offset,
2231 void *to, int len);
Stephen Hemminger0c6fcc82007-04-20 16:40:01 -07002232extern int skb_store_bits(struct sk_buff *skb, int offset,
2233 const void *from, int len);
Al Viro81d77662006-11-14 21:37:33 -08002234extern __wsum skb_copy_and_csum_bits(const struct sk_buff *skb,
Linus Torvalds1da177e2005-04-16 15:20:36 -07002235 int offset, u8 *to, int len,
Al Viro81d77662006-11-14 21:37:33 -08002236 __wsum csum);
Jens Axboe9c55e012007-11-06 23:30:13 -08002237extern int skb_splice_bits(struct sk_buff *skb,
2238 unsigned int offset,
2239 struct pipe_inode_info *pipe,
2240 unsigned int len,
2241 unsigned int flags);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002242extern void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
2243extern void skb_split(struct sk_buff *skb,
2244 struct sk_buff *skb1, const u32 len);
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08002245extern int skb_shift(struct sk_buff *tgt, struct sk_buff *skb,
2246 int shiftlen);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002247
Michał Mirosławc8f44af2011-11-15 15:29:55 +00002248extern struct sk_buff *skb_segment(struct sk_buff *skb,
2249 netdev_features_t features);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002250
Linus Torvalds1da177e2005-04-16 15:20:36 -07002251static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2252 int len, void *buffer)
2253{
2254 int hlen = skb_headlen(skb);
2255
Patrick McHardy55820ee2005-07-05 14:08:10 -07002256 if (hlen - offset >= len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002257 return skb->data + offset;
2258
2259 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2260 return NULL;
2261
2262 return buffer;
2263}
2264
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03002265static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2266 void *to,
2267 const unsigned int len)
2268{
2269 memcpy(to, skb->data, len);
2270}
2271
2272static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2273 const int offset, void *to,
2274 const unsigned int len)
2275{
2276 memcpy(to, skb->data + offset, len);
2277}
2278
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03002279static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2280 const void *from,
2281 const unsigned int len)
2282{
2283 memcpy(skb->data, from, len);
2284}
2285
2286static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2287 const int offset,
2288 const void *from,
2289 const unsigned int len)
2290{
2291 memcpy(skb->data + offset, from, len);
2292}
2293
Linus Torvalds1da177e2005-04-16 15:20:36 -07002294extern void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002295
Patrick Ohlyac45f602009-02-12 05:03:37 +00002296static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2297{
2298 return skb->tstamp;
2299}
2300
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002301/**
2302 * skb_get_timestamp - get timestamp from a skb
2303 * @skb: skb to get stamp from
2304 * @stamp: pointer to struct timeval to store stamp in
2305 *
2306 * Timestamps are stored in the skb as offsets to a base timestamp.
2307 * This function converts the offset back to a struct timeval and stores
2308 * it in stamp.
2309 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00002310static inline void skb_get_timestamp(const struct sk_buff *skb,
2311 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002312{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002313 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002314}
2315
Patrick Ohlyac45f602009-02-12 05:03:37 +00002316static inline void skb_get_timestampns(const struct sk_buff *skb,
2317 struct timespec *stamp)
2318{
2319 *stamp = ktime_to_timespec(skb->tstamp);
2320}
2321
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002322static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002323{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002324 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002325}
2326
Stephen Hemminger164891a2007-04-23 22:26:16 -07002327static inline ktime_t net_timedelta(ktime_t t)
2328{
2329 return ktime_sub(ktime_get_real(), t);
2330}
2331
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07002332static inline ktime_t net_invalid_timestamp(void)
2333{
2334 return ktime_set(0, 0);
2335}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002336
Richard Cochranc1f19b52010-07-17 08:49:36 +00002337extern void skb_timestamping_init(void);
2338
2339#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2340
2341extern void skb_clone_tx_timestamp(struct sk_buff *skb);
2342extern bool skb_defer_rx_timestamp(struct sk_buff *skb);
2343
2344#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2345
2346static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2347{
2348}
2349
2350static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2351{
2352 return false;
2353}
2354
2355#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2356
2357/**
2358 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2359 *
Richard Cochranda92b192011-10-21 00:49:15 +00002360 * PHY drivers may accept clones of transmitted packets for
2361 * timestamping via their phy_driver.txtstamp method. These drivers
2362 * must call this function to return the skb back to the stack, with
2363 * or without a timestamp.
2364 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00002365 * @skb: clone of the the original outgoing packet
Richard Cochranda92b192011-10-21 00:49:15 +00002366 * @hwtstamps: hardware time stamps, may be NULL if not available
Richard Cochranc1f19b52010-07-17 08:49:36 +00002367 *
2368 */
2369void skb_complete_tx_timestamp(struct sk_buff *skb,
2370 struct skb_shared_hwtstamps *hwtstamps);
2371
Patrick Ohlyac45f602009-02-12 05:03:37 +00002372/**
2373 * skb_tstamp_tx - queue clone of skb with send time stamps
2374 * @orig_skb: the original outgoing packet
2375 * @hwtstamps: hardware time stamps, may be NULL if not available
2376 *
2377 * If the skb has a socket associated, then this function clones the
2378 * skb (thus sharing the actual data and optional structures), stores
2379 * the optional hardware time stamping information (if non NULL) or
2380 * generates a software time stamp (otherwise), then queues the clone
2381 * to the error queue of the socket. Errors are silently ignored.
2382 */
2383extern void skb_tstamp_tx(struct sk_buff *orig_skb,
2384 struct skb_shared_hwtstamps *hwtstamps);
2385
Richard Cochran4507a712010-07-17 08:48:28 +00002386static inline void sw_tx_timestamp(struct sk_buff *skb)
2387{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002388 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2389 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00002390 skb_tstamp_tx(skb, NULL);
2391}
2392
2393/**
2394 * skb_tx_timestamp() - Driver hook for transmit timestamping
2395 *
2396 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00002397 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00002398 *
2399 * @skb: A socket buffer.
2400 */
2401static inline void skb_tx_timestamp(struct sk_buff *skb)
2402{
Richard Cochranc1f19b52010-07-17 08:49:36 +00002403 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002404 sw_tx_timestamp(skb);
2405}
2406
Johannes Berg6e3e9392011-11-09 10:15:42 +01002407/**
2408 * skb_complete_wifi_ack - deliver skb with wifi status
2409 *
2410 * @skb: the original outgoing packet
2411 * @acked: ack status
2412 *
2413 */
2414void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2415
Herbert Xu759e5d02007-03-25 20:10:56 -07002416extern __sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
Al Virob51655b2006-11-14 21:40:42 -08002417extern __sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002418
Herbert Xu60476372007-04-09 11:59:39 -07002419static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2420{
2421 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2422}
2423
Herbert Xufb286bb2005-11-10 13:01:24 -08002424/**
2425 * skb_checksum_complete - Calculate checksum of an entire packet
2426 * @skb: packet to process
2427 *
2428 * This function calculates the checksum over the entire packet plus
2429 * the value of skb->csum. The latter can be used to supply the
2430 * checksum of a pseudo header as used by TCP/UDP. It returns the
2431 * checksum.
2432 *
2433 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2434 * this function can be used to verify that checksum on received
2435 * packets. In that case the function should return zero if the
2436 * checksum is correct. In particular, this function will return zero
2437 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2438 * hardware has already verified the correctness of the checksum.
2439 */
Al Viro4381ca32007-07-15 21:00:11 +01002440static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002441{
Herbert Xu60476372007-04-09 11:59:39 -07002442 return skb_csum_unnecessary(skb) ?
2443 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002444}
2445
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002446#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002447extern void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002448static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2449{
2450 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002451 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002452}
2453static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2454{
2455 if (nfct)
2456 atomic_inc(&nfct->use);
2457}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002458#endif
2459#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -08002460static inline void nf_conntrack_get_reasm(struct sk_buff *skb)
2461{
2462 if (skb)
2463 atomic_inc(&skb->users);
2464}
2465static inline void nf_conntrack_put_reasm(struct sk_buff *skb)
2466{
2467 if (skb)
2468 kfree_skb(skb);
2469}
2470#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002471#ifdef CONFIG_BRIDGE_NETFILTER
2472static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2473{
2474 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2475 kfree(nf_bridge);
2476}
2477static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2478{
2479 if (nf_bridge)
2480 atomic_inc(&nf_bridge->use);
2481}
2482#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08002483static inline void nf_reset(struct sk_buff *skb)
2484{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002485#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08002486 nf_conntrack_put(skb->nfct);
2487 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002488#endif
2489#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Patrick McHardya193a4a2006-03-20 19:23:05 -08002490 nf_conntrack_put_reasm(skb->nfct_reasm);
2491 skb->nfct_reasm = NULL;
2492#endif
2493#ifdef CONFIG_BRIDGE_NETFILTER
2494 nf_bridge_put(skb->nf_bridge);
2495 skb->nf_bridge = NULL;
2496#endif
2497}
2498
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002499/* Note: This doesn't put any conntrack and bridge info in dst. */
2500static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2501{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002502#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002503 dst->nfct = src->nfct;
2504 nf_conntrack_get(src->nfct);
2505 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002506#endif
2507#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002508 dst->nfct_reasm = src->nfct_reasm;
2509 nf_conntrack_get_reasm(src->nfct_reasm);
2510#endif
2511#ifdef CONFIG_BRIDGE_NETFILTER
2512 dst->nf_bridge = src->nf_bridge;
2513 nf_bridge_get(src->nf_bridge);
2514#endif
2515}
2516
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002517static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2518{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002519#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002520 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002521#endif
2522#ifdef NET_SKBUFF_NF_DEFRAG_NEEDED
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002523 nf_conntrack_put_reasm(dst->nfct_reasm);
2524#endif
2525#ifdef CONFIG_BRIDGE_NETFILTER
2526 nf_bridge_put(dst->nf_bridge);
2527#endif
2528 __nf_copy(dst, src);
2529}
2530
James Morris984bc162006-06-09 00:29:17 -07002531#ifdef CONFIG_NETWORK_SECMARK
2532static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2533{
2534 to->secmark = from->secmark;
2535}
2536
2537static inline void skb_init_secmark(struct sk_buff *skb)
2538{
2539 skb->secmark = 0;
2540}
2541#else
2542static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2543{ }
2544
2545static inline void skb_init_secmark(struct sk_buff *skb)
2546{ }
2547#endif
2548
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002549static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2550{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002551 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002552}
2553
Stephen Hemminger92477442009-03-21 13:39:26 -07002554static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002555{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002556 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002557}
2558
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002559static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2560{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002561 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002562}
2563
David S. Millerd5a9e242009-01-27 16:22:11 -08002564static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2565{
2566 skb->queue_mapping = rx_queue + 1;
2567}
2568
Stephen Hemminger92477442009-03-21 13:39:26 -07002569static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002570{
2571 return skb->queue_mapping - 1;
2572}
2573
Stephen Hemminger92477442009-03-21 13:39:26 -07002574static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002575{
Eric Dumazeta02cec22010-09-22 20:43:57 +00002576 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08002577}
2578
Vladislav Zolotarova3d22a62010-12-13 06:27:10 +00002579extern u16 __skb_tx_hash(const struct net_device *dev,
2580 const struct sk_buff *skb,
2581 unsigned int num_tx_queues);
Stephen Hemminger92477442009-03-21 13:39:26 -07002582
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002583#ifdef CONFIG_XFRM
2584static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2585{
2586 return skb->sp;
2587}
2588#else
2589static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2590{
2591 return NULL;
2592}
2593#endif
2594
David S. Millerbdcc0922012-03-07 20:53:36 -05002595static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07002596{
2597 return skb_shinfo(skb)->gso_size;
2598}
2599
David S. Millerbdcc0922012-03-07 20:53:36 -05002600static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02002601{
2602 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2603}
2604
Ben Hutchings4497b072008-06-19 16:22:28 -07002605extern void __skb_warn_lro_forwarding(const struct sk_buff *skb);
2606
2607static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2608{
2609 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2610 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002611 const struct skb_shared_info *shinfo = skb_shinfo(skb);
2612
Alexander Duyckb78462e2010-06-02 12:24:37 +00002613 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2614 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07002615 __skb_warn_lro_forwarding(skb);
2616 return true;
2617 }
2618 return false;
2619}
2620
Herbert Xu35fc92a2007-03-26 23:22:20 -07002621static inline void skb_forward_csum(struct sk_buff *skb)
2622{
2623 /* Unfortunately we don't support this one. Any brave souls? */
2624 if (skb->ip_summed == CHECKSUM_COMPLETE)
2625 skb->ip_summed = CHECKSUM_NONE;
2626}
2627
Eric Dumazetbc8acf22010-09-02 13:07:41 -07002628/**
2629 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2630 * @skb: skb to check
2631 *
2632 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2633 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2634 * use this helper, to document places where we make this assertion.
2635 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002636static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07002637{
2638#ifdef DEBUG
2639 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2640#endif
2641}
2642
Rusty Russellf35d9d82008-02-04 23:49:54 -05002643bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00002644
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00002645/**
2646 * skb_head_is_locked - Determine if the skb->head is locked down
2647 * @skb: skb to check
2648 *
2649 * The head on skbs build around a head frag can be removed if they are
2650 * not cloned. This function returns true if the skb head is locked down
2651 * due to either being allocated via kmalloc, or by being a clone with
2652 * multiple references to the head.
2653 */
2654static inline bool skb_head_is_locked(const struct sk_buff *skb)
2655{
2656 return !skb->head_frag || skb_cloned(skb);
2657}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002658#endif /* __KERNEL__ */
2659#endif /* _LINUX_SKBUFF_H */