blob: d8d397acb52c2469bb6978b76738cf4f1c9a5ac2 [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>
Eric Dumazet363ec39232014-02-26 14:02:11 -080035#include <linux/sched.h>
Jason Wang5203cd22013-03-26 23:11:21 +000036#include <net/flow_keys.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070037
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010038/* A. Checksumming of received packets by device.
39 *
40 * CHECKSUM_NONE:
41 *
42 * Device failed to checksum this packet e.g. due to lack of capabilities.
43 * The packet contains full (though not verified) checksum in packet but
44 * not in skb->csum. Thus, skb->csum is undefined in this case.
45 *
46 * CHECKSUM_UNNECESSARY:
47 *
48 * The hardware you're dealing with doesn't calculate the full checksum
49 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
50 * for specific protocols e.g. TCP/UDP/SCTP, then, for such packets it will
51 * set CHECKSUM_UNNECESSARY if their checksums are okay. skb->csum is still
52 * undefined in this case though. It is a bad option, but, unfortunately,
53 * nowadays most vendors do this. Apparently with the secret goal to sell
54 * you new devices, when you will add new protocol to your host, f.e. IPv6 8)
55 *
56 * CHECKSUM_COMPLETE:
57 *
58 * This is the most generic way. The device supplied checksum of the _whole_
59 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
60 * hardware doesn't need to parse L3/L4 headers to implement this.
61 *
62 * Note: Even if device supports only some protocols, but is able to produce
63 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
64 *
65 * CHECKSUM_PARTIAL:
66 *
67 * This is identical to the case for output below. This may occur on a packet
68 * received directly from another Linux OS, e.g., a virtualized Linux kernel
69 * on the same host. The packet can be treated in the same way as
70 * CHECKSUM_UNNECESSARY, except that on output (i.e., forwarding) the
71 * checksum must be filled in by the OS or the hardware.
72 *
73 * B. Checksumming on output.
74 *
75 * CHECKSUM_NONE:
76 *
77 * The skb was already checksummed by the protocol, or a checksum is not
78 * required.
79 *
80 * CHECKSUM_PARTIAL:
81 *
82 * The device is required to checksum the packet as seen by hard_start_xmit()
83 * from skb->csum_start up to the end, and to record/write the checksum at
84 * offset skb->csum_start + skb->csum_offset.
85 *
86 * The device must show its capabilities in dev->features, set up at device
87 * setup time, e.g. netdev_features.h:
88 *
89 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
90 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
91 * IPv4. Sigh. Vendors like this way for an unknown reason.
92 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
93 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
94 * NETIF_F_... - Well, you get the picture.
95 *
96 * CHECKSUM_UNNECESSARY:
97 *
98 * Normally, the device will do per protocol specific checksumming. Protocol
99 * implementations that do not want the NIC to perform the checksum
100 * calculation should use this flag in their outgoing skbs.
101 *
102 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
103 * offload. Correspondingly, the FCoE protocol driver
104 * stack should use CHECKSUM_UNNECESSARY.
105 *
106 * Any questions? No questions, good. --ANK
107 */
108
Herbert Xu60476372007-04-09 11:59:39 -0700109/* Don't change this without changing skb_csum_unnecessary! */
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100110#define CHECKSUM_NONE 0
111#define CHECKSUM_UNNECESSARY 1
112#define CHECKSUM_COMPLETE 2
113#define CHECKSUM_PARTIAL 3
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114
115#define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
116 ~(SMP_CACHE_BYTES - 1))
David S. Millerfc910a22007-03-25 20:27:59 -0700117#define SKB_WITH_OVERHEAD(X) \
Herbert Xudeea84b2007-10-21 16:27:46 -0700118 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
David S. Millerfc910a22007-03-25 20:27:59 -0700119#define SKB_MAX_ORDER(X, ORDER) \
120 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700121#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
122#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
123
Eric Dumazet87fb4b72011-10-13 07:28:54 +0000124/* return minimum truesize of one skb containing X bytes of data */
125#define SKB_TRUESIZE(X) ((X) + \
126 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
127 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
128
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129struct net_device;
David Howells716ea3a2007-04-02 20:19:53 -0700130struct scatterlist;
Jens Axboe9c55e012007-11-06 23:30:13 -0800131struct pipe_inode_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700132
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700133#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134struct nf_conntrack {
135 atomic_t use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136};
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700137#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700138
139#ifdef CONFIG_BRIDGE_NETFILTER
140struct nf_bridge_info {
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000141 atomic_t use;
142 unsigned int mask;
143 struct net_device *physindev;
144 struct net_device *physoutdev;
145 unsigned long data[32 / sizeof(unsigned long)];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146};
147#endif
148
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149struct sk_buff_head {
150 /* These two members must be first. */
151 struct sk_buff *next;
152 struct sk_buff *prev;
153
154 __u32 qlen;
155 spinlock_t lock;
156};
157
158struct sk_buff;
159
Ian Campbell9d4dde52011-12-22 23:39:14 +0000160/* To allow 64K frame to be packed as single skb without frag_list we
161 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
162 * buffers which do not start on a page boundary.
163 *
164 * Since GRO uses frags we allocate at least 16 regardless of page
165 * size.
Anton Blancharda715dea2011-03-27 14:57:26 +0000166 */
Ian Campbell9d4dde52011-12-22 23:39:14 +0000167#if (65536/PAGE_SIZE + 1) < 16
David S. Millereec00952011-03-29 23:34:08 -0700168#define MAX_SKB_FRAGS 16UL
Anton Blancharda715dea2011-03-27 14:57:26 +0000169#else
Ian Campbell9d4dde52011-12-22 23:39:14 +0000170#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
Anton Blancharda715dea2011-03-27 14:57:26 +0000171#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172
173typedef struct skb_frag_struct skb_frag_t;
174
175struct skb_frag_struct {
Ian Campbella8605c62011-10-19 23:01:49 +0000176 struct {
177 struct page *p;
178 } page;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000179#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
David S. Millera309bb02007-07-30 18:47:03 -0700180 __u32 page_offset;
181 __u32 size;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000182#else
183 __u16 page_offset;
184 __u16 size;
185#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186};
187
Eric Dumazet9e903e02011-10-18 21:00:24 +0000188static inline unsigned int skb_frag_size(const skb_frag_t *frag)
189{
190 return frag->size;
191}
192
193static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
194{
195 frag->size = size;
196}
197
198static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
199{
200 frag->size += delta;
201}
202
203static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
204{
205 frag->size -= delta;
206}
207
Patrick Ohlyac45f602009-02-12 05:03:37 +0000208#define HAVE_HW_TIME_STAMP
209
210/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800211 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000212 * @hwtstamp: hardware time stamp transformed into duration
213 * since arbitrary point in time
214 * @syststamp: hwtstamp transformed to system time base
215 *
216 * Software time stamps generated by ktime_get_real() are stored in
217 * skb->tstamp. The relation between the different kinds of time
218 * stamps is as follows:
219 *
220 * syststamp and tstamp can be compared against each other in
221 * arbitrary combinations. The accuracy of a
222 * syststamp/tstamp/"syststamp from other device" comparison is
223 * limited by the accuracy of the transformation into system time
224 * base. This depends on the device driver and its underlying
225 * hardware.
226 *
227 * hwtstamps can only be compared against other hwtstamps from
228 * the same device.
229 *
230 * This structure is attached to packets as part of the
231 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
232 */
233struct skb_shared_hwtstamps {
234 ktime_t hwtstamp;
235 ktime_t syststamp;
236};
237
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000238/* Definitions for tx_flags in struct skb_shared_info */
239enum {
240 /* generate hardware time stamp */
241 SKBTX_HW_TSTAMP = 1 << 0,
242
243 /* generate software time stamp */
244 SKBTX_SW_TSTAMP = 1 << 1,
245
246 /* device driver is going to provide hardware time stamp */
247 SKBTX_IN_PROGRESS = 1 << 2,
248
Shirley Maa6686f22011-07-06 12:22:12 +0000249 /* device driver supports TX zero-copy buffers */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000250 SKBTX_DEV_ZEROCOPY = 1 << 3,
Johannes Berg6e3e9392011-11-09 10:15:42 +0100251
252 /* generate wifi status information (where possible) */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000253 SKBTX_WIFI_STATUS = 1 << 4,
Pravin B Shelarc9af6db2013-02-11 09:27:41 +0000254
255 /* This indicates at least one fragment might be overwritten
256 * (as in vmsplice(), sendfile() ...)
257 * If we need to compute a TX checksum, we'll need to copy
258 * all frags to avoid possible bad checksum
259 */
260 SKBTX_SHARED_FRAG = 1 << 5,
Shirley Maa6686f22011-07-06 12:22:12 +0000261};
262
263/*
264 * The callback notifies userspace to release buffers when skb DMA is done in
265 * lower device, the skb last reference should be 0 when calling this.
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000266 * The zerocopy_success argument is true if zero copy transmit occurred,
267 * false on data copy or out of memory error caused by data copy attempt.
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000268 * The ctx field is used to track device context.
269 * The desc field is used to track userspace buffer index.
Shirley Maa6686f22011-07-06 12:22:12 +0000270 */
271struct ubuf_info {
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000272 void (*callback)(struct ubuf_info *, bool zerocopy_success);
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000273 void *ctx;
Shirley Maa6686f22011-07-06 12:22:12 +0000274 unsigned long desc;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000275};
276
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277/* This data is invariant across clones and lives at
278 * the end of the header data, ie. at skb->end.
279 */
280struct skb_shared_info {
Ian Campbell9f42f122012-01-05 07:13:39 +0000281 unsigned char nr_frags;
282 __u8 tx_flags;
Herbert Xu79671682006-06-22 02:40:14 -0700283 unsigned short gso_size;
284 /* Warning: this field is not always filled in (UFO)! */
285 unsigned short gso_segs;
286 unsigned short gso_type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000288 struct skb_shared_hwtstamps hwtstamps;
Ian Campbell9f42f122012-01-05 07:13:39 +0000289 __be32 ip6_frag_id;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700290
291 /*
292 * Warning : all fields before dataref are cleared in __alloc_skb()
293 */
294 atomic_t dataref;
295
Johann Baudy69e3c752009-05-18 22:11:22 -0700296 /* Intermediate layers must ensure that destructor_arg
297 * remains valid until skb destructor */
298 void * destructor_arg;
Shirley Maa6686f22011-07-06 12:22:12 +0000299
Eric Dumazetfed66382010-07-22 19:09:08 +0000300 /* must be last field, see pskb_expand_head() */
301 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302};
303
304/* We divide dataref into two halves. The higher 16 bits hold references
305 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700306 * the entire skb->data. A clone of a headerless skb holds the length of
307 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308 *
309 * All users must obey the rule that the skb->data reference count must be
310 * greater than or equal to the payload reference count.
311 *
312 * Holding a reference to the payload part means that the user does not
313 * care about modifications to the header part of skb->data.
314 */
315#define SKB_DATAREF_SHIFT 16
316#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
317
David S. Millerd179cd12005-08-17 14:57:30 -0700318
319enum {
320 SKB_FCLONE_UNAVAILABLE,
321 SKB_FCLONE_ORIG,
322 SKB_FCLONE_CLONE,
323};
324
Herbert Xu79671682006-06-22 02:40:14 -0700325enum {
326 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700327 SKB_GSO_UDP = 1 << 1,
Herbert Xu576a30e2006-06-27 13:22:38 -0700328
329 /* This indicates the skb is from an untrusted source. */
330 SKB_GSO_DODGY = 1 << 2,
Michael Chanb0da85372006-06-29 12:30:00 -0700331
332 /* This indicates the tcp segment has CWR set. */
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700333 SKB_GSO_TCP_ECN = 1 << 3,
334
335 SKB_GSO_TCPV6 = 1 << 4,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800336
337 SKB_GSO_FCOE = 1 << 5,
Pravin B Shelar68c33162013-02-14 14:02:41 +0000338
339 SKB_GSO_GRE = 1 << 6,
Pravin B Shelar73136262013-03-07 13:21:51 +0000340
Eric Dumazetcb32f512013-10-19 11:42:57 -0700341 SKB_GSO_IPIP = 1 << 7,
Simon Horman0d89d202013-05-23 21:02:52 +0000342
Eric Dumazet61c1db72013-10-20 20:47:30 -0700343 SKB_GSO_SIT = 1 << 8,
Eric Dumazetcb32f512013-10-19 11:42:57 -0700344
Eric Dumazet61c1db72013-10-20 20:47:30 -0700345 SKB_GSO_UDP_TUNNEL = 1 << 9,
346
347 SKB_GSO_MPLS = 1 << 10,
Herbert Xu79671682006-06-22 02:40:14 -0700348};
349
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700350#if BITS_PER_LONG > 32
351#define NET_SKBUFF_DATA_USES_OFFSET 1
352#endif
353
354#ifdef NET_SKBUFF_DATA_USES_OFFSET
355typedef unsigned int sk_buff_data_t;
356#else
357typedef unsigned char *sk_buff_data_t;
358#endif
359
Eric Dumazet363ec39232014-02-26 14:02:11 -0800360/**
361 * struct skb_mstamp - multi resolution time stamps
362 * @stamp_us: timestamp in us resolution
363 * @stamp_jiffies: timestamp in jiffies
364 */
365struct skb_mstamp {
366 union {
367 u64 v64;
368 struct {
369 u32 stamp_us;
370 u32 stamp_jiffies;
371 };
372 };
373};
374
375/**
376 * skb_mstamp_get - get current timestamp
377 * @cl: place to store timestamps
378 */
379static inline void skb_mstamp_get(struct skb_mstamp *cl)
380{
381 u64 val = local_clock();
382
383 do_div(val, NSEC_PER_USEC);
384 cl->stamp_us = (u32)val;
385 cl->stamp_jiffies = (u32)jiffies;
386}
387
388/**
389 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
390 * @t1: pointer to newest sample
391 * @t0: pointer to oldest sample
392 */
393static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
394 const struct skb_mstamp *t0)
395{
396 s32 delta_us = t1->stamp_us - t0->stamp_us;
397 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
398
399 /* If delta_us is negative, this might be because interval is too big,
400 * or local_clock() drift is too big : fallback using jiffies.
401 */
402 if (delta_us <= 0 ||
403 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
404
405 delta_us = jiffies_to_usecs(delta_jiffies);
406
407 return delta_us;
408}
409
410
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411/**
412 * struct sk_buff - socket buffer
413 * @next: Next buffer in list
414 * @prev: Previous buffer in list
Eric Dumazet363ec39232014-02-26 14:02:11 -0800415 * @tstamp: Time we arrived/left
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700416 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700418 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000419 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700420 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 * @len: Length of actual data
422 * @data_len: Data length
423 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700424 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700425 * @csum: Checksum (must include start/offset pair)
426 * @csum_start: Offset from skb->head where checksumming should start
427 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700428 * @priority: Packet queueing priority
WANG Cong60ff7462014-05-04 16:39:18 -0700429 * @ignore_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700431 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432 * @nohdr: Payload reference only, must not modify header
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700433 * @nfctinfo: Relationship of this skb to the connection
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700435 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700436 * @ipvs_property: skbuff is owned by ipvs
Randy Dunlap31729362008-02-18 20:52:13 -0800437 * @peeked: this packet has been seen already, so stats have been
438 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700439 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700440 * @protocol: Packet protocol from driver
441 * @destructor: Destruct function
442 * @nfct: Associated connection, if any
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800444 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445 * @tc_index: Traffic control index
446 * @tc_verd: traffic control verdict
Tom Herbert61b905d2014-03-24 15:34:47 -0700447 * @hash: the packet hash
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700448 * @queue_mapping: Queue mapping for multiqueue devices
Randy Dunlap553a5672008-04-20 10:51:01 -0700449 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700450 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Tom Herbert61b905d2014-03-24 15:34:47 -0700451 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
Changli Gao4ca24622011-08-19 07:26:44 -0700452 * ports.
Johannes Berg6e3e9392011-11-09 10:15:42 +0100453 * @wifi_acked_valid: wifi_acked was set
454 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000455 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Randy Dunlapf4b8ea72006-06-22 16:00:11 -0700456 * @dma_cookie: a cookie to one of several possible DMA operations
457 * done by skb DMA functions
Eliezer Tamir06021292013-06-10 11:39:50 +0300458 * @napi_id: id of the NAPI struct this skb came from
James Morris984bc162006-06-09 00:29:17 -0700459 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700460 * @mark: Generic packet mark
461 * @dropcount: total number of sk_receive_queue overflows
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000462 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b02008-07-14 22:49:06 -0700463 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000464 * @inner_protocol: Protocol (encapsulation)
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000465 * @inner_transport_header: Inner transport layer header (encapsulation)
466 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000467 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700468 * @transport_header: Transport layer header
469 * @network_header: Network layer header
470 * @mac_header: Link layer header
471 * @tail: Tail pointer
472 * @end: End pointer
473 * @head: Head of buffer
474 * @data: Data head pointer
475 * @truesize: Buffer size
476 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477 */
478
479struct sk_buff {
480 /* These two members must be first. */
481 struct sk_buff *next;
482 struct sk_buff *prev;
483
Eric Dumazet363ec39232014-02-26 14:02:11 -0800484 union {
485 ktime_t tstamp;
486 struct skb_mstamp skb_mstamp;
487 };
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000488
489 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492 /*
493 * This is the control buffer. It is free to use for every
494 * layer. Please put your private variables there. If you
495 * want to keep them across layers you have to do a skb_clone()
496 * first. This is owned by whoever has the skb queued ATM.
497 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000498 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499
Eric Dumazet7fee2262010-05-11 23:19:48 +0000500 unsigned long _skb_refdst;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000501#ifdef CONFIG_XFRM
502 struct sec_path *sp;
503#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700505 data_len;
506 __u16 mac_len,
507 hdr_len;
Al Viroff1dcad2006-11-20 18:07:29 -0800508 union {
509 __wsum csum;
Herbert Xu663ead32007-04-09 11:59:07 -0700510 struct {
511 __u16 csum_start;
512 __u16 csum_offset;
513 };
Al Viroff1dcad2006-11-20 18:07:29 -0800514 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515 __u32 priority;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200516 kmemcheck_bitfield_begin(flags1);
WANG Cong60ff7462014-05-04 16:39:18 -0700517 __u8 ignore_df:1,
Thomas Graf1cbb3382005-07-05 14:13:41 -0700518 cloned:1,
519 ip_summed:2,
Harald Welte6869c4d2005-08-09 19:24:19 -0700520 nohdr:1,
521 nfctinfo:3;
David S. Millerd179cd12005-08-17 14:57:30 -0700522 __u8 pkt_type:3,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800523 fclone:2,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700524 ipvs_property:1,
Herbert Xua59322b2007-12-05 01:53:40 -0800525 peeked:1,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700526 nf_trace:1;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200527 kmemcheck_bitfield_end(flags1);
Eric Dumazet4ab408d2010-03-01 03:09:26 +0000528 __be16 protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529
530 void (*destructor)(struct sk_buff *skb);
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800531#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700532 struct nf_conntrack *nfct;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100533#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534#ifdef CONFIG_BRIDGE_NETFILTER
535 struct nf_bridge_info *nf_bridge;
536#endif
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700537
Eric Dumazet8964be42009-11-20 15:35:04 -0800538 int skb_iif;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000539
Tom Herbert61b905d2014-03-24 15:34:47 -0700540 __u32 hash;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000541
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000542 __be16 vlan_proto;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000543 __u16 vlan_tci;
544
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700546 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700548 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200551
Eric Dumazet0a148422011-04-20 09:27:32 +0000552 __u16 queue_mapping;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200553 kmemcheck_bitfield_begin(flags2);
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400554#ifdef CONFIG_IPV6_NDISC_NODETYPE
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000555 __u8 ndisc_nodetype:2;
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400556#endif
Mel Gormanc93bdd02012-07-31 16:44:19 -0700557 __u8 pfmemalloc:1;
Tom Herbert3853b582010-11-21 13:17:29 +0000558 __u8 ooo_okay:1;
Tom Herbert61b905d2014-03-24 15:34:47 -0700559 __u8 l4_hash:1;
Johannes Berg6e3e9392011-11-09 10:15:42 +0100560 __u8 wifi_acked_valid:1;
561 __u8 wifi_acked:1;
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000562 __u8 no_fcs:1;
Eric Dumazetd3836f22012-04-27 00:33:38 +0000563 __u8 head_frag:1;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000564 /* Encapsulation protocol and NIC drivers should use
565 * this flag to indicate to each other if the skb contains
566 * encapsulated packet or not and maybe use the inner packet
567 * headers if needed
568 */
569 __u8 encapsulation:1;
Tom Herbert7e2b10c2014-06-04 17:20:02 -0700570 __u8 encap_hdr_csum:1;
571 /* 5/7 bit hole (depending on ndisc_nodetype presence) */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200572 kmemcheck_bitfield_end(flags2);
573
Cong Wange0d10952013-08-01 11:10:25 +0800574#if defined CONFIG_NET_DMA || defined CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir06021292013-06-10 11:39:50 +0300575 union {
576 unsigned int napi_id;
577 dma_cookie_t dma_cookie;
578 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700579#endif
James Morris984bc162006-06-09 00:29:17 -0700580#ifdef CONFIG_NETWORK_SECMARK
581 __u32 secmark;
582#endif
Neil Horman3b885782009-10-12 13:26:31 -0700583 union {
584 __u32 mark;
585 __u32 dropcount;
Eric Dumazet16fad692013-03-14 05:40:32 +0000586 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700587 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700588
Simon Horman0d89d202013-05-23 21:02:52 +0000589 __be16 inner_protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000590 __u16 inner_transport_header;
591 __u16 inner_network_header;
592 __u16 inner_mac_header;
593 __u16 transport_header;
594 __u16 network_header;
595 __u16 mac_header;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700596 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700597 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700598 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700599 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700600 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700601 unsigned int truesize;
602 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603};
604
605#ifdef __KERNEL__
606/*
607 * Handling routines are only of interest to the kernel
608 */
609#include <linux/slab.h>
610
Linus Torvalds1da177e2005-04-16 15:20:36 -0700611
Mel Gormanc93bdd02012-07-31 16:44:19 -0700612#define SKB_ALLOC_FCLONE 0x01
613#define SKB_ALLOC_RX 0x02
614
615/* Returns true if the skb was allocated from PFMEMALLOC reserves */
616static inline bool skb_pfmemalloc(const struct sk_buff *skb)
617{
618 return unlikely(skb->pfmemalloc);
619}
620
Eric Dumazet7fee2262010-05-11 23:19:48 +0000621/*
622 * skb might have a dst pointer attached, refcounted or not.
623 * _skb_refdst low order bit is set if refcount was _not_ taken
624 */
625#define SKB_DST_NOREF 1UL
626#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
627
628/**
629 * skb_dst - returns skb dst_entry
630 * @skb: buffer
631 *
632 * Returns skb dst_entry, regardless of reference taken or not.
633 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000634static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
635{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000636 /* If refdst was not refcounted, check we still are in a
637 * rcu_read_lock section
638 */
639 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
640 !rcu_read_lock_held() &&
641 !rcu_read_lock_bh_held());
642 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000643}
644
Eric Dumazet7fee2262010-05-11 23:19:48 +0000645/**
646 * skb_dst_set - sets skb dst
647 * @skb: buffer
648 * @dst: dst entry
649 *
650 * Sets skb dst, assuming a reference was taken on dst and should
651 * be released by skb_dst_drop()
652 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000653static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
654{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000655 skb->_skb_refdst = (unsigned long)dst;
656}
657
Joe Perches7965bd42013-09-26 14:48:15 -0700658void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
659 bool force);
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200660
661/**
662 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
663 * @skb: buffer
664 * @dst: dst entry
665 *
666 * Sets skb dst, assuming a reference was not taken on dst.
667 * If dst entry is cached, we do not take reference and dst_release
668 * will be avoided by refdst_drop. If dst entry is not cached, we take
669 * reference, so that last dst_release can destroy the dst immediately.
670 */
671static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
672{
673 __skb_dst_set_noref(skb, dst, false);
674}
675
676/**
677 * skb_dst_set_noref_force - sets skb dst, without taking reference
678 * @skb: buffer
679 * @dst: dst entry
680 *
681 * Sets skb dst, assuming a reference was not taken on dst.
682 * No reference is taken and no dst_release will be called. While for
683 * cached dsts deferred reclaim is a basic feature, for entries that are
684 * not cached it is caller's job to guarantee that last dst_release for
685 * provided dst happens when nobody uses it, eg. after a RCU grace period.
686 */
687static inline void skb_dst_set_noref_force(struct sk_buff *skb,
688 struct dst_entry *dst)
689{
690 __skb_dst_set_noref(skb, dst, true);
691}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000692
693/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300694 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000695 * @skb: buffer
696 */
697static inline bool skb_dst_is_noref(const struct sk_buff *skb)
698{
699 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000700}
701
Eric Dumazet511c3f92009-06-02 05:14:27 +0000702static inline struct rtable *skb_rtable(const struct sk_buff *skb)
703{
Eric Dumazetadf30902009-06-02 05:19:30 +0000704 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000705}
706
Joe Perches7965bd42013-09-26 14:48:15 -0700707void kfree_skb(struct sk_buff *skb);
708void kfree_skb_list(struct sk_buff *segs);
709void skb_tx_error(struct sk_buff *skb);
710void consume_skb(struct sk_buff *skb);
711void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000712extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000713
Joe Perches7965bd42013-09-26 14:48:15 -0700714void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
715bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
716 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000717
Joe Perches7965bd42013-09-26 14:48:15 -0700718struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
719 int node);
720struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700721static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100722 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700723{
Eric Dumazet564824b2010-10-11 19:05:25 +0000724 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700725}
726
727static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100728 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700729{
Mel Gormanc93bdd02012-07-31 16:44:19 -0700730 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700731}
732
Joe Perches7965bd42013-09-26 14:48:15 -0700733struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
Patrick McHardy0ebd0ac2013-04-17 06:46:58 +0000734static inline struct sk_buff *alloc_skb_head(gfp_t priority)
735{
736 return __alloc_skb_head(priority, -1);
737}
738
Joe Perches7965bd42013-09-26 14:48:15 -0700739struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
740int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
741struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
742struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
743struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom, gfp_t gfp_mask);
Eric Dumazet117632e2011-12-03 21:39:53 +0000744
Joe Perches7965bd42013-09-26 14:48:15 -0700745int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
746struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
747 unsigned int headroom);
748struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
749 int newtailroom, gfp_t priority);
Fan Du25a91d82014-01-18 09:54:23 +0800750int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
751 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -0700752int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
753 int len);
754int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
755int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000756#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700757
Joe Perches7965bd42013-09-26 14:48:15 -0700758int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
759 int getfrag(void *from, char *to, int offset,
760 int len, int odd, struct sk_buff *skb),
761 void *from, int length);
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700762
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800763struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700764 __u32 lower_offset;
765 __u32 upper_offset;
766 __u32 frag_idx;
767 __u32 stepped_offset;
768 struct sk_buff *root_skb;
769 struct sk_buff *cur_skb;
770 __u8 *frag_data;
771};
772
Joe Perches7965bd42013-09-26 14:48:15 -0700773void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
774 unsigned int to, struct skb_seq_state *st);
775unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
776 struct skb_seq_state *st);
777void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -0700778
Joe Perches7965bd42013-09-26 14:48:15 -0700779unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
780 unsigned int to, struct ts_config *config,
781 struct ts_state *state);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700782
Tom Herbert09323cc2013-12-15 22:16:19 -0800783/*
784 * Packet hash types specify the type of hash in skb_set_hash.
785 *
786 * Hash types refer to the protocol layer addresses which are used to
787 * construct a packet's hash. The hashes are used to differentiate or identify
788 * flows of the protocol layer for the hash type. Hash types are either
789 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
790 *
791 * Properties of hashes:
792 *
793 * 1) Two packets in different flows have different hash values
794 * 2) Two packets in the same flow should have the same hash value
795 *
796 * A hash at a higher layer is considered to be more specific. A driver should
797 * set the most specific hash possible.
798 *
799 * A driver cannot indicate a more specific hash than the layer at which a hash
800 * was computed. For instance an L3 hash cannot be set as an L4 hash.
801 *
802 * A driver may indicate a hash level which is less specific than the
803 * actual layer the hash was computed on. For instance, a hash computed
804 * at L4 may be considered an L3 hash. This should only be done if the
805 * driver can't unambiguously determine that the HW computed the hash at
806 * the higher layer. Note that the "should" in the second property above
807 * permits this.
808 */
809enum pkt_hash_types {
810 PKT_HASH_TYPE_NONE, /* Undefined type */
811 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
812 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
813 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
814};
815
816static inline void
817skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
818{
Tom Herbert61b905d2014-03-24 15:34:47 -0700819 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
820 skb->hash = hash;
Tom Herbert09323cc2013-12-15 22:16:19 -0800821}
822
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800823void __skb_get_hash(struct sk_buff *skb);
824static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000825{
Tom Herbert61b905d2014-03-24 15:34:47 -0700826 if (!skb->l4_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800827 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000828
Tom Herbert61b905d2014-03-24 15:34:47 -0700829 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000830}
831
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800832static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
833{
Tom Herbert61b905d2014-03-24 15:34:47 -0700834 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800835}
836
Tom Herbert7539fad2013-12-15 22:12:18 -0800837static inline void skb_clear_hash(struct sk_buff *skb)
838{
Tom Herbert61b905d2014-03-24 15:34:47 -0700839 skb->hash = 0;
840 skb->l4_hash = 0;
Tom Herbert7539fad2013-12-15 22:12:18 -0800841}
842
843static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
844{
Tom Herbert61b905d2014-03-24 15:34:47 -0700845 if (!skb->l4_hash)
Tom Herbert7539fad2013-12-15 22:12:18 -0800846 skb_clear_hash(skb);
847}
848
Tom Herbert3df7a742013-12-15 22:16:29 -0800849static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
850{
Tom Herbert61b905d2014-03-24 15:34:47 -0700851 to->hash = from->hash;
852 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -0800853};
854
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700855#ifdef NET_SKBUFF_DATA_USES_OFFSET
856static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
857{
858 return skb->head + skb->end;
859}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000860
861static inline unsigned int skb_end_offset(const struct sk_buff *skb)
862{
863 return skb->end;
864}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700865#else
866static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
867{
868 return skb->end;
869}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000870
871static inline unsigned int skb_end_offset(const struct sk_buff *skb)
872{
873 return skb->end - skb->head;
874}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700875#endif
876
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700878#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700879
Patrick Ohlyac45f602009-02-12 05:03:37 +0000880static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
881{
882 return &skb_shinfo(skb)->hwtstamps;
883}
884
Linus Torvalds1da177e2005-04-16 15:20:36 -0700885/**
886 * skb_queue_empty - check if a queue is empty
887 * @list: queue head
888 *
889 * Returns true if the queue is empty, false otherwise.
890 */
891static inline int skb_queue_empty(const struct sk_buff_head *list)
892{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100893 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700894}
895
896/**
David S. Millerfc7ebb22008-09-23 00:34:07 -0700897 * skb_queue_is_last - check if skb is the last entry in the queue
898 * @list: queue head
899 * @skb: buffer
900 *
901 * Returns true if @skb is the last buffer on the list.
902 */
903static inline bool skb_queue_is_last(const struct sk_buff_head *list,
904 const struct sk_buff *skb)
905{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100906 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -0700907}
908
909/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800910 * skb_queue_is_first - check if skb is the first entry in the queue
911 * @list: queue head
912 * @skb: buffer
913 *
914 * Returns true if @skb is the first buffer on the list.
915 */
916static inline bool skb_queue_is_first(const struct sk_buff_head *list,
917 const struct sk_buff *skb)
918{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100919 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800920}
921
922/**
David S. Miller249c8b42008-09-23 00:44:42 -0700923 * skb_queue_next - return the next packet in the queue
924 * @list: queue head
925 * @skb: current buffer
926 *
927 * Return the next packet in @list after @skb. It is only valid to
928 * call this if skb_queue_is_last() evaluates to false.
929 */
930static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
931 const struct sk_buff *skb)
932{
933 /* This BUG_ON may seem severe, but if we just return then we
934 * are going to dereference garbage.
935 */
936 BUG_ON(skb_queue_is_last(list, skb));
937 return skb->next;
938}
939
940/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800941 * skb_queue_prev - return the prev packet in the queue
942 * @list: queue head
943 * @skb: current buffer
944 *
945 * Return the prev packet in @list before @skb. It is only valid to
946 * call this if skb_queue_is_first() evaluates to false.
947 */
948static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
949 const struct sk_buff *skb)
950{
951 /* This BUG_ON may seem severe, but if we just return then we
952 * are going to dereference garbage.
953 */
954 BUG_ON(skb_queue_is_first(list, skb));
955 return skb->prev;
956}
957
958/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700959 * skb_get - reference buffer
960 * @skb: buffer to reference
961 *
962 * Makes another reference to a socket buffer and returns a pointer
963 * to the buffer.
964 */
965static inline struct sk_buff *skb_get(struct sk_buff *skb)
966{
967 atomic_inc(&skb->users);
968 return skb;
969}
970
971/*
972 * If users == 1, we are the only owner and are can avoid redundant
973 * atomic change.
974 */
975
976/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700977 * skb_cloned - is the buffer a clone
978 * @skb: buffer to check
979 *
980 * Returns true if the buffer was generated with skb_clone() and is
981 * one of multiple shared copies of the buffer. Cloned buffers are
982 * shared data so must not be written to under normal circumstances.
983 */
984static inline int skb_cloned(const struct sk_buff *skb)
985{
986 return skb->cloned &&
987 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
988}
989
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +0000990static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
991{
992 might_sleep_if(pri & __GFP_WAIT);
993
994 if (skb_cloned(skb))
995 return pskb_expand_head(skb, 0, 0, pri);
996
997 return 0;
998}
999
Linus Torvalds1da177e2005-04-16 15:20:36 -07001000/**
1001 * skb_header_cloned - is the header a clone
1002 * @skb: buffer to check
1003 *
1004 * Returns true if modifying the header part of the buffer requires
1005 * the data to be copied.
1006 */
1007static inline int skb_header_cloned(const struct sk_buff *skb)
1008{
1009 int dataref;
1010
1011 if (!skb->cloned)
1012 return 0;
1013
1014 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1015 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1016 return dataref != 1;
1017}
1018
1019/**
1020 * skb_header_release - release reference to header
1021 * @skb: buffer to operate on
1022 *
1023 * Drop a reference to the header part of the buffer. This is done
1024 * by acquiring a payload reference. You must not read from the header
1025 * part of skb->data after this.
1026 */
1027static inline void skb_header_release(struct sk_buff *skb)
1028{
1029 BUG_ON(skb->nohdr);
1030 skb->nohdr = 1;
1031 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1032}
1033
1034/**
1035 * skb_shared - is the buffer shared
1036 * @skb: buffer to check
1037 *
1038 * Returns true if more than one person has a reference to this
1039 * buffer.
1040 */
1041static inline int skb_shared(const struct sk_buff *skb)
1042{
1043 return atomic_read(&skb->users) != 1;
1044}
1045
1046/**
1047 * skb_share_check - check if buffer is shared and if so clone it
1048 * @skb: buffer to check
1049 * @pri: priority for memory allocation
1050 *
1051 * If the buffer is shared the buffer is cloned and the old copy
1052 * drops a reference. A new clone with a single reference is returned.
1053 * If the buffer is not shared the original buffer is returned. When
1054 * being called from interrupt status or with spinlocks held pri must
1055 * be GFP_ATOMIC.
1056 *
1057 * NULL is returned on a memory allocation failure.
1058 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001059static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001060{
1061 might_sleep_if(pri & __GFP_WAIT);
1062 if (skb_shared(skb)) {
1063 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001064
1065 if (likely(nskb))
1066 consume_skb(skb);
1067 else
1068 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001069 skb = nskb;
1070 }
1071 return skb;
1072}
1073
1074/*
1075 * Copy shared buffers into a new sk_buff. We effectively do COW on
1076 * packets to handle cases where we have a local reader and forward
1077 * and a couple of other messy ones. The normal one is tcpdumping
1078 * a packet thats being forwarded.
1079 */
1080
1081/**
1082 * skb_unshare - make a copy of a shared buffer
1083 * @skb: buffer to check
1084 * @pri: priority for memory allocation
1085 *
1086 * If the socket buffer is a clone then this function creates a new
1087 * copy of the data, drops a reference count on the old copy and returns
1088 * the new copy with the reference count at 1. If the buffer is not a clone
1089 * the original buffer is returned. When called with a spinlock held or
1090 * from interrupt state @pri must be %GFP_ATOMIC
1091 *
1092 * %NULL is returned on a memory allocation failure.
1093 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001094static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001095 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001096{
1097 might_sleep_if(pri & __GFP_WAIT);
1098 if (skb_cloned(skb)) {
1099 struct sk_buff *nskb = skb_copy(skb, pri);
1100 kfree_skb(skb); /* Free our shared copy */
1101 skb = nskb;
1102 }
1103 return skb;
1104}
1105
1106/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001107 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108 * @list_: list to peek at
1109 *
1110 * Peek an &sk_buff. Unlike most other operations you _MUST_
1111 * be careful with this one. A peek leaves the buffer on the
1112 * list and someone else may run off with it. You must hold
1113 * the appropriate locks or have a private queue to do this.
1114 *
1115 * Returns %NULL for an empty list or a pointer to the head element.
1116 * The reference count is not incremented and the reference is therefore
1117 * volatile. Use with caution.
1118 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001119static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001120{
Eric Dumazet18d07002012-04-30 16:31:46 +00001121 struct sk_buff *skb = list_->next;
1122
1123 if (skb == (struct sk_buff *)list_)
1124 skb = NULL;
1125 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001126}
1127
1128/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001129 * skb_peek_next - peek skb following the given one from a queue
1130 * @skb: skb to start from
1131 * @list_: list to peek at
1132 *
1133 * Returns %NULL when the end of the list is met or a pointer to the
1134 * next element. The reference count is not incremented and the
1135 * reference is therefore volatile. Use with caution.
1136 */
1137static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1138 const struct sk_buff_head *list_)
1139{
1140 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001141
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001142 if (next == (struct sk_buff *)list_)
1143 next = NULL;
1144 return next;
1145}
1146
1147/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001148 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001149 * @list_: list to peek at
1150 *
1151 * Peek an &sk_buff. Unlike most other operations you _MUST_
1152 * be careful with this one. A peek leaves the buffer on the
1153 * list and someone else may run off with it. You must hold
1154 * the appropriate locks or have a private queue to do this.
1155 *
1156 * Returns %NULL for an empty list or a pointer to the tail element.
1157 * The reference count is not incremented and the reference is therefore
1158 * volatile. Use with caution.
1159 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001160static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161{
Eric Dumazet18d07002012-04-30 16:31:46 +00001162 struct sk_buff *skb = list_->prev;
1163
1164 if (skb == (struct sk_buff *)list_)
1165 skb = NULL;
1166 return skb;
1167
Linus Torvalds1da177e2005-04-16 15:20:36 -07001168}
1169
1170/**
1171 * skb_queue_len - get queue length
1172 * @list_: list to measure
1173 *
1174 * Return the length of an &sk_buff queue.
1175 */
1176static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1177{
1178 return list_->qlen;
1179}
1180
David S. Miller67fed452008-09-21 22:36:24 -07001181/**
1182 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1183 * @list: queue to initialize
1184 *
1185 * This initializes only the list and queue length aspects of
1186 * an sk_buff_head object. This allows to initialize the list
1187 * aspects of an sk_buff_head without reinitializing things like
1188 * the spinlock. It can also be used for on-stack sk_buff_head
1189 * objects where the spinlock is known to not be used.
1190 */
1191static inline void __skb_queue_head_init(struct sk_buff_head *list)
1192{
1193 list->prev = list->next = (struct sk_buff *)list;
1194 list->qlen = 0;
1195}
1196
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001197/*
1198 * This function creates a split out lock class for each invocation;
1199 * this is needed for now since a whole lot of users of the skb-queue
1200 * infrastructure in drivers have different locking usage (in hardirq)
1201 * than the networking core (in softirq only). In the long run either the
1202 * network layer or drivers should need annotation to consolidate the
1203 * main types of usage into 3 classes.
1204 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001205static inline void skb_queue_head_init(struct sk_buff_head *list)
1206{
1207 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001208 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209}
1210
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001211static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1212 struct lock_class_key *class)
1213{
1214 skb_queue_head_init(list);
1215 lockdep_set_class(&list->lock, class);
1216}
1217
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001219 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001220 *
1221 * The "__skb_xxxx()" functions are the non-atomic ones that
1222 * can only be called with interrupts disabled.
1223 */
Joe Perches7965bd42013-09-26 14:48:15 -07001224void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1225 struct sk_buff_head *list);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001226static inline void __skb_insert(struct sk_buff *newsk,
1227 struct sk_buff *prev, struct sk_buff *next,
1228 struct sk_buff_head *list)
1229{
1230 newsk->next = next;
1231 newsk->prev = prev;
1232 next->prev = prev->next = newsk;
1233 list->qlen++;
1234}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001235
David S. Miller67fed452008-09-21 22:36:24 -07001236static inline void __skb_queue_splice(const struct sk_buff_head *list,
1237 struct sk_buff *prev,
1238 struct sk_buff *next)
1239{
1240 struct sk_buff *first = list->next;
1241 struct sk_buff *last = list->prev;
1242
1243 first->prev = prev;
1244 prev->next = first;
1245
1246 last->next = next;
1247 next->prev = last;
1248}
1249
1250/**
1251 * skb_queue_splice - join two skb lists, this is designed for stacks
1252 * @list: the new list to add
1253 * @head: the place to add it in the first list
1254 */
1255static inline void skb_queue_splice(const struct sk_buff_head *list,
1256 struct sk_buff_head *head)
1257{
1258 if (!skb_queue_empty(list)) {
1259 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001260 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001261 }
1262}
1263
1264/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001265 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001266 * @list: the new list to add
1267 * @head: the place to add it in the first list
1268 *
1269 * The list at @list is reinitialised
1270 */
1271static inline void skb_queue_splice_init(struct sk_buff_head *list,
1272 struct sk_buff_head *head)
1273{
1274 if (!skb_queue_empty(list)) {
1275 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001276 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001277 __skb_queue_head_init(list);
1278 }
1279}
1280
1281/**
1282 * skb_queue_splice_tail - join two skb lists, each list being a queue
1283 * @list: the new list to add
1284 * @head: the place to add it in the first list
1285 */
1286static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1287 struct sk_buff_head *head)
1288{
1289 if (!skb_queue_empty(list)) {
1290 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001291 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001292 }
1293}
1294
1295/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001296 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001297 * @list: the new list to add
1298 * @head: the place to add it in the first list
1299 *
1300 * Each of the lists is a queue.
1301 * The list at @list is reinitialised
1302 */
1303static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1304 struct sk_buff_head *head)
1305{
1306 if (!skb_queue_empty(list)) {
1307 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001308 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001309 __skb_queue_head_init(list);
1310 }
1311}
1312
Linus Torvalds1da177e2005-04-16 15:20:36 -07001313/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001314 * __skb_queue_after - queue a buffer at the list head
1315 * @list: list to use
1316 * @prev: place after this buffer
1317 * @newsk: buffer to queue
1318 *
1319 * Queue a buffer int the middle of a list. This function takes no locks
1320 * and you must therefore hold required locks before calling it.
1321 *
1322 * A buffer cannot be placed on two lists at the same time.
1323 */
1324static inline void __skb_queue_after(struct sk_buff_head *list,
1325 struct sk_buff *prev,
1326 struct sk_buff *newsk)
1327{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001328 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001329}
1330
Joe Perches7965bd42013-09-26 14:48:15 -07001331void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1332 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001333
Gerrit Renkerf5572852008-04-14 00:05:28 -07001334static inline void __skb_queue_before(struct sk_buff_head *list,
1335 struct sk_buff *next,
1336 struct sk_buff *newsk)
1337{
1338 __skb_insert(newsk, next->prev, next, list);
1339}
1340
Stephen Hemminger300ce172005-10-30 13:47:34 -08001341/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001342 * __skb_queue_head - queue a buffer at the list head
1343 * @list: list to use
1344 * @newsk: buffer to queue
1345 *
1346 * Queue a buffer at the start of a list. This function takes no locks
1347 * and you must therefore hold required locks before calling it.
1348 *
1349 * A buffer cannot be placed on two lists at the same time.
1350 */
Joe Perches7965bd42013-09-26 14:48:15 -07001351void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352static inline void __skb_queue_head(struct sk_buff_head *list,
1353 struct sk_buff *newsk)
1354{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001355 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356}
1357
1358/**
1359 * __skb_queue_tail - queue a buffer at the list tail
1360 * @list: list to use
1361 * @newsk: buffer to queue
1362 *
1363 * Queue a buffer at the end of a list. This function takes no locks
1364 * and you must therefore hold required locks before calling it.
1365 *
1366 * A buffer cannot be placed on two lists at the same time.
1367 */
Joe Perches7965bd42013-09-26 14:48:15 -07001368void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001369static inline void __skb_queue_tail(struct sk_buff_head *list,
1370 struct sk_buff *newsk)
1371{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001372 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001373}
1374
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376 * remove sk_buff from list. _Must_ be called atomically, and with
1377 * the list known..
1378 */
Joe Perches7965bd42013-09-26 14:48:15 -07001379void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001380static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1381{
1382 struct sk_buff *next, *prev;
1383
1384 list->qlen--;
1385 next = skb->next;
1386 prev = skb->prev;
1387 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001388 next->prev = prev;
1389 prev->next = next;
1390}
1391
Gerrit Renkerf525c062008-04-14 00:04:12 -07001392/**
1393 * __skb_dequeue - remove from the head of the queue
1394 * @list: list to dequeue from
1395 *
1396 * Remove the head of the list. This function does not take any locks
1397 * so must be used with appropriate locks held only. The head item is
1398 * returned or %NULL if the list is empty.
1399 */
Joe Perches7965bd42013-09-26 14:48:15 -07001400struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Gerrit Renkerf525c062008-04-14 00:04:12 -07001401static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1402{
1403 struct sk_buff *skb = skb_peek(list);
1404 if (skb)
1405 __skb_unlink(skb, list);
1406 return skb;
1407}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001408
1409/**
1410 * __skb_dequeue_tail - remove from the tail of the queue
1411 * @list: list to dequeue from
1412 *
1413 * Remove the tail of the list. This function does not take any locks
1414 * so must be used with appropriate locks held only. The tail item is
1415 * returned or %NULL if the list is empty.
1416 */
Joe Perches7965bd42013-09-26 14:48:15 -07001417struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1419{
1420 struct sk_buff *skb = skb_peek_tail(list);
1421 if (skb)
1422 __skb_unlink(skb, list);
1423 return skb;
1424}
1425
1426
David S. Millerbdcc0922012-03-07 20:53:36 -05001427static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428{
1429 return skb->data_len;
1430}
1431
1432static inline unsigned int skb_headlen(const struct sk_buff *skb)
1433{
1434 return skb->len - skb->data_len;
1435}
1436
1437static inline int skb_pagelen(const struct sk_buff *skb)
1438{
1439 int i, len = 0;
1440
1441 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001442 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001443 return len + skb_headlen(skb);
1444}
1445
Ian Campbell131ea662011-08-19 06:25:00 +00001446/**
1447 * __skb_fill_page_desc - initialise a paged fragment in an skb
1448 * @skb: buffer containing fragment to be initialised
1449 * @i: paged fragment index to initialise
1450 * @page: the page to use for this fragment
1451 * @off: the offset to the data with @page
1452 * @size: the length of the data
1453 *
1454 * Initialises the @i'th fragment of @skb to point to &size bytes at
1455 * offset @off within @page.
1456 *
1457 * Does not take any additional reference on the fragment.
1458 */
1459static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1460 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001461{
1462 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1463
Mel Gormanc48a11c2012-07-31 16:44:23 -07001464 /*
1465 * Propagate page->pfmemalloc to the skb if we can. The problem is
1466 * that not all callers have unique ownership of the page. If
1467 * pfmemalloc is set, we check the mapping as a mapping implies
1468 * page->index is set (index and pfmemalloc share space).
1469 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1470 * do not lose pfmemalloc information as the pages would not be
1471 * allocated using __GFP_MEMALLOC.
1472 */
Ian Campbella8605c62011-10-19 23:01:49 +00001473 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001475 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001476
1477 page = compound_head(page);
1478 if (page->pfmemalloc && !page->mapping)
1479 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00001480}
1481
1482/**
1483 * skb_fill_page_desc - initialise a paged fragment in an skb
1484 * @skb: buffer containing fragment to be initialised
1485 * @i: paged fragment index to initialise
1486 * @page: the page to use for this fragment
1487 * @off: the offset to the data with @page
1488 * @size: the length of the data
1489 *
1490 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01001491 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00001492 * addition updates @skb such that @i is the last fragment.
1493 *
1494 * Does not take any additional reference on the fragment.
1495 */
1496static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1497 struct page *page, int off, int size)
1498{
1499 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001500 skb_shinfo(skb)->nr_frags = i + 1;
1501}
1502
Joe Perches7965bd42013-09-26 14:48:15 -07001503void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1504 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001505
Jason Wangf8e617e2013-11-01 14:07:47 +08001506void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1507 unsigned int truesize);
1508
Linus Torvalds1da177e2005-04-16 15:20:36 -07001509#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001510#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001511#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1512
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001513#ifdef NET_SKBUFF_DATA_USES_OFFSET
1514static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1515{
1516 return skb->head + skb->tail;
1517}
1518
1519static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1520{
1521 skb->tail = skb->data - skb->head;
1522}
1523
1524static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1525{
1526 skb_reset_tail_pointer(skb);
1527 skb->tail += offset;
1528}
Simon Horman7cc46192013-05-28 20:34:29 +00001529
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001530#else /* NET_SKBUFF_DATA_USES_OFFSET */
1531static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1532{
1533 return skb->tail;
1534}
1535
1536static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1537{
1538 skb->tail = skb->data;
1539}
1540
1541static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1542{
1543 skb->tail = skb->data + offset;
1544}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001545
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001546#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1547
Linus Torvalds1da177e2005-04-16 15:20:36 -07001548/*
1549 * Add data to an sk_buff
1550 */
Mathias Krause0c7ddf32013-11-07 14:18:24 +01001551unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001552unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001553static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1554{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001555 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001556 SKB_LINEAR_ASSERT(skb);
1557 skb->tail += len;
1558 skb->len += len;
1559 return tmp;
1560}
1561
Joe Perches7965bd42013-09-26 14:48:15 -07001562unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001563static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1564{
1565 skb->data -= len;
1566 skb->len += len;
1567 return skb->data;
1568}
1569
Joe Perches7965bd42013-09-26 14:48:15 -07001570unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001571static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1572{
1573 skb->len -= len;
1574 BUG_ON(skb->len < skb->data_len);
1575 return skb->data += len;
1576}
1577
David S. Miller47d29642010-05-02 02:21:44 -07001578static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1579{
1580 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1581}
1582
Joe Perches7965bd42013-09-26 14:48:15 -07001583unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001584
1585static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1586{
1587 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001588 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001589 return NULL;
1590 skb->len -= len;
1591 return skb->data += len;
1592}
1593
1594static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1595{
1596 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1597}
1598
1599static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1600{
1601 if (likely(len <= skb_headlen(skb)))
1602 return 1;
1603 if (unlikely(len > skb->len))
1604 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001605 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001606}
1607
1608/**
1609 * skb_headroom - bytes at buffer head
1610 * @skb: buffer to check
1611 *
1612 * Return the number of bytes of free space at the head of an &sk_buff.
1613 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001614static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001615{
1616 return skb->data - skb->head;
1617}
1618
1619/**
1620 * skb_tailroom - bytes at buffer end
1621 * @skb: buffer to check
1622 *
1623 * Return the number of bytes of free space at the tail of an sk_buff
1624 */
1625static inline int skb_tailroom(const struct sk_buff *skb)
1626{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001627 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001628}
1629
1630/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00001631 * skb_availroom - bytes at buffer end
1632 * @skb: buffer to check
1633 *
1634 * Return the number of bytes of free space at the tail of an sk_buff
1635 * allocated by sk_stream_alloc()
1636 */
1637static inline int skb_availroom(const struct sk_buff *skb)
1638{
Eric Dumazet16fad692013-03-14 05:40:32 +00001639 if (skb_is_nonlinear(skb))
1640 return 0;
1641
1642 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00001643}
1644
1645/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001646 * skb_reserve - adjust headroom
1647 * @skb: buffer to alter
1648 * @len: bytes to move
1649 *
1650 * Increase the headroom of an empty &sk_buff by reducing the tail
1651 * room. This is only allowed for an empty buffer.
1652 */
David S. Miller8243126c2006-01-17 02:54:21 -08001653static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001654{
1655 skb->data += len;
1656 skb->tail += len;
1657}
1658
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001659static inline void skb_reset_inner_headers(struct sk_buff *skb)
1660{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001661 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001662 skb->inner_network_header = skb->network_header;
1663 skb->inner_transport_header = skb->transport_header;
1664}
1665
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00001666static inline void skb_reset_mac_len(struct sk_buff *skb)
1667{
1668 skb->mac_len = skb->network_header - skb->mac_header;
1669}
1670
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001671static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1672 *skb)
1673{
1674 return skb->head + skb->inner_transport_header;
1675}
1676
1677static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1678{
1679 skb->inner_transport_header = skb->data - skb->head;
1680}
1681
1682static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1683 const int offset)
1684{
1685 skb_reset_inner_transport_header(skb);
1686 skb->inner_transport_header += offset;
1687}
1688
1689static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1690{
1691 return skb->head + skb->inner_network_header;
1692}
1693
1694static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1695{
1696 skb->inner_network_header = skb->data - skb->head;
1697}
1698
1699static inline void skb_set_inner_network_header(struct sk_buff *skb,
1700 const int offset)
1701{
1702 skb_reset_inner_network_header(skb);
1703 skb->inner_network_header += offset;
1704}
1705
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001706static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1707{
1708 return skb->head + skb->inner_mac_header;
1709}
1710
1711static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1712{
1713 skb->inner_mac_header = skb->data - skb->head;
1714}
1715
1716static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1717 const int offset)
1718{
1719 skb_reset_inner_mac_header(skb);
1720 skb->inner_mac_header += offset;
1721}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001722static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1723{
Cong Wang35d04612013-05-29 15:16:05 +08001724 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001725}
1726
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001727static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1728{
1729 return skb->head + skb->transport_header;
1730}
1731
1732static inline void skb_reset_transport_header(struct sk_buff *skb)
1733{
1734 skb->transport_header = skb->data - skb->head;
1735}
1736
1737static inline void skb_set_transport_header(struct sk_buff *skb,
1738 const int offset)
1739{
1740 skb_reset_transport_header(skb);
1741 skb->transport_header += offset;
1742}
1743
1744static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1745{
1746 return skb->head + skb->network_header;
1747}
1748
1749static inline void skb_reset_network_header(struct sk_buff *skb)
1750{
1751 skb->network_header = skb->data - skb->head;
1752}
1753
1754static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1755{
1756 skb_reset_network_header(skb);
1757 skb->network_header += offset;
1758}
1759
1760static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1761{
1762 return skb->head + skb->mac_header;
1763}
1764
1765static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1766{
Cong Wang35d04612013-05-29 15:16:05 +08001767 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001768}
1769
1770static inline void skb_reset_mac_header(struct sk_buff *skb)
1771{
1772 skb->mac_header = skb->data - skb->head;
1773}
1774
1775static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1776{
1777 skb_reset_mac_header(skb);
1778 skb->mac_header += offset;
1779}
1780
Timo Teräs0e3da5b2013-12-16 11:02:09 +02001781static inline void skb_pop_mac_header(struct sk_buff *skb)
1782{
1783 skb->mac_header = skb->network_header;
1784}
1785
Ying Xuefbbdb8f2013-03-27 16:46:06 +00001786static inline void skb_probe_transport_header(struct sk_buff *skb,
1787 const int offset_hint)
1788{
1789 struct flow_keys keys;
1790
1791 if (skb_transport_header_was_set(skb))
1792 return;
1793 else if (skb_flow_dissect(skb, &keys))
1794 skb_set_transport_header(skb, keys.thoff);
1795 else
1796 skb_set_transport_header(skb, offset_hint);
1797}
1798
Eric Dumazet03606892012-02-23 10:55:02 +00001799static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1800{
1801 if (skb_mac_header_was_set(skb)) {
1802 const unsigned char *old_mac = skb_mac_header(skb);
1803
1804 skb_set_mac_header(skb, -skb->mac_len);
1805 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1806 }
1807}
1808
Michał Mirosław04fb4512010-12-14 15:24:08 +00001809static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1810{
1811 return skb->csum_start - skb_headroom(skb);
1812}
1813
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001814static inline int skb_transport_offset(const struct sk_buff *skb)
1815{
1816 return skb_transport_header(skb) - skb->data;
1817}
1818
1819static inline u32 skb_network_header_len(const struct sk_buff *skb)
1820{
1821 return skb->transport_header - skb->network_header;
1822}
1823
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001824static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1825{
1826 return skb->inner_transport_header - skb->inner_network_header;
1827}
1828
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001829static inline int skb_network_offset(const struct sk_buff *skb)
1830{
1831 return skb_network_header(skb) - skb->data;
1832}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001833
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001834static inline int skb_inner_network_offset(const struct sk_buff *skb)
1835{
1836 return skb_inner_network_header(skb) - skb->data;
1837}
1838
Changli Gaof9599ce2010-08-04 04:43:44 +00001839static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1840{
1841 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1842}
1843
Linus Torvalds1da177e2005-04-16 15:20:36 -07001844/*
1845 * CPUs often take a performance hit when accessing unaligned memory
1846 * locations. The actual performance hit varies, it can be small if the
1847 * hardware handles it or large if we have to take an exception and fix it
1848 * in software.
1849 *
1850 * Since an ethernet header is 14 bytes network drivers often end up with
1851 * the IP header at an unaligned offset. The IP header can be aligned by
1852 * shifting the start of the packet by 2 bytes. Drivers should do this
1853 * with:
1854 *
Tobias Klauser8660c122009-07-13 22:48:16 +00001855 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001856 *
1857 * The downside to this alignment of the IP header is that the DMA is now
1858 * unaligned. On some architectures the cost of an unaligned DMA is high
1859 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00001860 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001861 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1862 * to be overridden.
1863 */
1864#ifndef NET_IP_ALIGN
1865#define NET_IP_ALIGN 2
1866#endif
1867
Anton Blanchard025be812006-03-31 02:27:06 -08001868/*
1869 * The networking layer reserves some headroom in skb data (via
1870 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1871 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08001872 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08001873 *
1874 * Unfortunately this headroom changes the DMA alignment of the resulting
1875 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1876 * on some architectures. An architecture can override this value,
1877 * perhaps setting it to a cacheline in size (since that will maintain
1878 * cacheline alignment of the DMA). It must be a power of 2.
1879 *
David S. Millerd6301d32009-02-08 19:24:13 -08001880 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08001881 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07001882 *
1883 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1884 * to reduce average number of cache lines per packet.
1885 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07001886 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08001887 */
1888#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07001889#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08001890#endif
1891
Joe Perches7965bd42013-09-26 14:48:15 -07001892int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001893
1894static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1895{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00001896 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07001897 WARN_ON(1);
1898 return;
1899 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001900 skb->len = len;
1901 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001902}
1903
Joe Perches7965bd42013-09-26 14:48:15 -07001904void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001905
1906static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1907{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001908 if (skb->data_len)
1909 return ___pskb_trim(skb, len);
1910 __skb_trim(skb, len);
1911 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001912}
1913
1914static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1915{
1916 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1917}
1918
1919/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07001920 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1921 * @skb: buffer to alter
1922 * @len: new length
1923 *
1924 * This is identical to pskb_trim except that the caller knows that
1925 * the skb is not cloned so we should never get an error due to out-
1926 * of-memory.
1927 */
1928static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1929{
1930 int err = pskb_trim(skb, len);
1931 BUG_ON(err);
1932}
1933
1934/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001935 * skb_orphan - orphan a buffer
1936 * @skb: buffer to orphan
1937 *
1938 * If a buffer currently has an owner then we call the owner's
1939 * destructor function and make the @skb unowned. The buffer continues
1940 * to exist but is no longer charged to its former owner.
1941 */
1942static inline void skb_orphan(struct sk_buff *skb)
1943{
Eric Dumazetc34a7612013-07-30 16:11:15 -07001944 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001945 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07001946 skb->destructor = NULL;
1947 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07001948 } else {
1949 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07001950 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001951}
1952
1953/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00001954 * skb_orphan_frags - orphan the frags contained in a buffer
1955 * @skb: buffer to orphan frags from
1956 * @gfp_mask: allocation mask for replacement pages
1957 *
1958 * For each frag in the SKB which needs a destructor (i.e. has an
1959 * owner) create a copy of that frag and release the original
1960 * page by calling the destructor.
1961 */
1962static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1963{
1964 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1965 return 0;
1966 return skb_copy_ubufs(skb, gfp_mask);
1967}
1968
1969/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001970 * __skb_queue_purge - empty a list
1971 * @list: list to empty
1972 *
1973 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1974 * the list and one reference dropped. This function does not take the
1975 * list lock and the caller must hold the relevant locks to use it.
1976 */
Joe Perches7965bd42013-09-26 14:48:15 -07001977void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001978static inline void __skb_queue_purge(struct sk_buff_head *list)
1979{
1980 struct sk_buff *skb;
1981 while ((skb = __skb_dequeue(list)) != NULL)
1982 kfree_skb(skb);
1983}
1984
Alexander Duycke5e67302013-02-08 10:17:15 +00001985#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
1986#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
1987#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
1988
Joe Perches7965bd42013-09-26 14:48:15 -07001989void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001990
Joe Perches7965bd42013-09-26 14:48:15 -07001991struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
1992 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07001993
1994/**
1995 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1996 * @dev: network device to receive on
1997 * @length: length to allocate
1998 *
1999 * Allocate a new &sk_buff and assign it a usage count of one. The
2000 * buffer has unspecified headroom built in. Users should allocate
2001 * the headroom they think they need without accounting for the
2002 * built in space. The built in space is used for optimisations.
2003 *
2004 * %NULL is returned if there is no free memory. Although this function
2005 * allocates memory it can be called from an interrupt.
2006 */
2007static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002008 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002009{
2010 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2011}
2012
Eric Dumazet6f532612012-05-18 05:12:12 +00002013/* legacy helper around __netdev_alloc_skb() */
2014static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2015 gfp_t gfp_mask)
2016{
2017 return __netdev_alloc_skb(NULL, length, gfp_mask);
2018}
2019
2020/* legacy helper around netdev_alloc_skb() */
2021static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2022{
2023 return netdev_alloc_skb(NULL, length);
2024}
2025
2026
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002027static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2028 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002029{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002030 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002031
2032 if (NET_IP_ALIGN && skb)
2033 skb_reserve(skb, NET_IP_ALIGN);
2034 return skb;
2035}
2036
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002037static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2038 unsigned int length)
2039{
2040 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2041}
2042
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002043/**
2044 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
Mel Gorman06140022012-07-31 16:44:24 -07002045 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2046 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2047 * @order: size of the allocation
2048 *
2049 * Allocate a new page.
2050 *
2051 * %NULL is returned if there is no free memory.
2052*/
2053static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
2054 struct sk_buff *skb,
2055 unsigned int order)
2056{
2057 struct page *page;
2058
2059 gfp_mask |= __GFP_COLD;
2060
2061 if (!(gfp_mask & __GFP_NOMEMALLOC))
2062 gfp_mask |= __GFP_MEMALLOC;
2063
2064 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2065 if (skb && page && page->pfmemalloc)
2066 skb->pfmemalloc = true;
2067
2068 return page;
2069}
2070
2071/**
2072 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2073 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2074 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2075 *
2076 * Allocate a new page.
2077 *
2078 * %NULL is returned if there is no free memory.
2079 */
2080static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
2081 struct sk_buff *skb)
2082{
2083 return __skb_alloc_pages(gfp_mask, skb, 0);
2084}
2085
2086/**
2087 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2088 * @page: The page that was allocated from skb_alloc_page
2089 * @skb: The skb that may need pfmemalloc set
2090 */
2091static inline void skb_propagate_pfmemalloc(struct page *page,
2092 struct sk_buff *skb)
2093{
2094 if (page && page->pfmemalloc)
2095 skb->pfmemalloc = true;
2096}
2097
Eric Dumazet564824b2010-10-11 19:05:25 +00002098/**
Masanari Iidae2278672014-02-18 22:54:36 +09002099 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002100 * @frag: the paged fragment
2101 *
2102 * Returns the &struct page associated with @frag.
2103 */
2104static inline struct page *skb_frag_page(const skb_frag_t *frag)
2105{
Ian Campbella8605c62011-10-19 23:01:49 +00002106 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00002107}
2108
2109/**
2110 * __skb_frag_ref - take an addition reference on a paged fragment.
2111 * @frag: the paged fragment
2112 *
2113 * Takes an additional reference on the paged fragment @frag.
2114 */
2115static inline void __skb_frag_ref(skb_frag_t *frag)
2116{
2117 get_page(skb_frag_page(frag));
2118}
2119
2120/**
2121 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2122 * @skb: the buffer
2123 * @f: the fragment offset.
2124 *
2125 * Takes an additional reference on the @f'th paged fragment of @skb.
2126 */
2127static inline void skb_frag_ref(struct sk_buff *skb, int f)
2128{
2129 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2130}
2131
2132/**
2133 * __skb_frag_unref - release a reference on a paged fragment.
2134 * @frag: the paged fragment
2135 *
2136 * Releases a reference on the paged fragment @frag.
2137 */
2138static inline void __skb_frag_unref(skb_frag_t *frag)
2139{
2140 put_page(skb_frag_page(frag));
2141}
2142
2143/**
2144 * skb_frag_unref - release a reference on a paged fragment of an skb.
2145 * @skb: the buffer
2146 * @f: the fragment offset
2147 *
2148 * Releases a reference on the @f'th paged fragment of @skb.
2149 */
2150static inline void skb_frag_unref(struct sk_buff *skb, int f)
2151{
2152 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2153}
2154
2155/**
2156 * skb_frag_address - gets the address of the data contained in a paged fragment
2157 * @frag: the paged fragment buffer
2158 *
2159 * Returns the address of the data within @frag. The page must already
2160 * be mapped.
2161 */
2162static inline void *skb_frag_address(const skb_frag_t *frag)
2163{
2164 return page_address(skb_frag_page(frag)) + frag->page_offset;
2165}
2166
2167/**
2168 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2169 * @frag: the paged fragment buffer
2170 *
2171 * Returns the address of the data within @frag. Checks that the page
2172 * is mapped and returns %NULL otherwise.
2173 */
2174static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2175{
2176 void *ptr = page_address(skb_frag_page(frag));
2177 if (unlikely(!ptr))
2178 return NULL;
2179
2180 return ptr + frag->page_offset;
2181}
2182
2183/**
2184 * __skb_frag_set_page - sets the page contained in a paged fragment
2185 * @frag: the paged fragment
2186 * @page: the page to set
2187 *
2188 * Sets the fragment @frag to contain @page.
2189 */
2190static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2191{
Ian Campbella8605c62011-10-19 23:01:49 +00002192 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002193}
2194
2195/**
2196 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2197 * @skb: the buffer
2198 * @f: the fragment offset
2199 * @page: the page to set
2200 *
2201 * Sets the @f'th fragment of @skb to contain @page.
2202 */
2203static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2204 struct page *page)
2205{
2206 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2207}
2208
Eric Dumazet400dfd32013-10-17 16:27:07 -07002209bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2210
Ian Campbell131ea662011-08-19 06:25:00 +00002211/**
2212 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002213 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002214 * @frag: the paged fragment to map
2215 * @offset: the offset within the fragment (starting at the
2216 * fragment's own offset)
2217 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002218 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00002219 *
2220 * Maps the page associated with @frag to @device.
2221 */
2222static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2223 const skb_frag_t *frag,
2224 size_t offset, size_t size,
2225 enum dma_data_direction dir)
2226{
2227 return dma_map_page(dev, skb_frag_page(frag),
2228 frag->page_offset + offset, size, dir);
2229}
2230
Eric Dumazet117632e2011-12-03 21:39:53 +00002231static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2232 gfp_t gfp_mask)
2233{
2234 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2235}
2236
Ian Campbell131ea662011-08-19 06:25:00 +00002237/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002238 * skb_clone_writable - is the header of a clone writable
2239 * @skb: buffer to check
2240 * @len: length up to which to write
2241 *
2242 * Returns true if modifying the header part of the cloned buffer
2243 * does not requires the data to be copied.
2244 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002245static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002246{
2247 return !skb_header_cloned(skb) &&
2248 skb_headroom(skb) + len <= skb->hdr_len;
2249}
2250
Herbert Xud9cc2042007-09-16 16:21:16 -07002251static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2252 int cloned)
2253{
2254 int delta = 0;
2255
Herbert Xud9cc2042007-09-16 16:21:16 -07002256 if (headroom > skb_headroom(skb))
2257 delta = headroom - skb_headroom(skb);
2258
2259 if (delta || cloned)
2260 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2261 GFP_ATOMIC);
2262 return 0;
2263}
2264
Patrick McHardy334a8132007-06-25 04:35:20 -07002265/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002266 * skb_cow - copy header of skb when it is required
2267 * @skb: buffer to cow
2268 * @headroom: needed headroom
2269 *
2270 * If the skb passed lacks sufficient headroom or its data part
2271 * is shared, data is reallocated. If reallocation fails, an error
2272 * is returned and original skb is not changed.
2273 *
2274 * The result is skb with writable area skb->head...skb->tail
2275 * and at least @headroom of space at head.
2276 */
2277static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2278{
Herbert Xud9cc2042007-09-16 16:21:16 -07002279 return __skb_cow(skb, headroom, skb_cloned(skb));
2280}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002281
Herbert Xud9cc2042007-09-16 16:21:16 -07002282/**
2283 * skb_cow_head - skb_cow but only making the head writable
2284 * @skb: buffer to cow
2285 * @headroom: needed headroom
2286 *
2287 * This function is identical to skb_cow except that we replace the
2288 * skb_cloned check by skb_header_cloned. It should be used when
2289 * you only need to push on some header and do not need to modify
2290 * the data.
2291 */
2292static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2293{
2294 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002295}
2296
2297/**
2298 * skb_padto - pad an skbuff up to a minimal size
2299 * @skb: buffer to pad
2300 * @len: minimal length
2301 *
2302 * Pads up a buffer to ensure the trailing bytes exist and are
2303 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002304 * is untouched. Otherwise it is extended. Returns zero on
2305 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002306 */
2307
Herbert Xu5b057c62006-06-23 02:06:41 -07002308static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002309{
2310 unsigned int size = skb->len;
2311 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002312 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002313 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002314}
2315
2316static inline int skb_add_data(struct sk_buff *skb,
2317 char __user *from, int copy)
2318{
2319 const int off = skb->len;
2320
2321 if (skb->ip_summed == CHECKSUM_NONE) {
2322 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08002323 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
Linus Torvalds1da177e2005-04-16 15:20:36 -07002324 copy, 0, &err);
2325 if (!err) {
2326 skb->csum = csum_block_add(skb->csum, csum, off);
2327 return 0;
2328 }
2329 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2330 return 0;
2331
2332 __skb_trim(skb, off);
2333 return -EFAULT;
2334}
2335
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002336static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2337 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002338{
2339 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002340 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002341
Ian Campbellea2ab692011-08-22 23:44:58 +00002342 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002343 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002344 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002345 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002346}
2347
Herbert Xu364c6ba2006-06-09 16:10:40 -07002348static inline int __skb_linearize(struct sk_buff *skb)
2349{
2350 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2351}
2352
Linus Torvalds1da177e2005-04-16 15:20:36 -07002353/**
2354 * skb_linearize - convert paged skb to linear one
2355 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002356 *
2357 * If there is no free memory -ENOMEM is returned, otherwise zero
2358 * is returned and the old skb data released.
2359 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002360static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002361{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002362 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2363}
2364
2365/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00002366 * skb_has_shared_frag - can any frag be overwritten
2367 * @skb: buffer to test
2368 *
2369 * Return true if the skb has at least one frag that might be modified
2370 * by an external entity (as in vmsplice()/sendfile())
2371 */
2372static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2373{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00002374 return skb_is_nonlinear(skb) &&
2375 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00002376}
2377
2378/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07002379 * skb_linearize_cow - make sure skb is linear and writable
2380 * @skb: buffer to process
2381 *
2382 * If there is no free memory -ENOMEM is returned, otherwise zero
2383 * is returned and the old skb data released.
2384 */
2385static inline int skb_linearize_cow(struct sk_buff *skb)
2386{
2387 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2388 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002389}
2390
2391/**
2392 * skb_postpull_rcsum - update checksum for received skb after pull
2393 * @skb: buffer to update
2394 * @start: start of data before pull
2395 * @len: length of data pulled
2396 *
2397 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002398 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2399 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002400 */
2401
2402static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002403 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002404{
Patrick McHardy84fa7932006-08-29 16:44:56 -07002405 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002406 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2407}
2408
Herbert Xucbb042f2006-03-20 22:43:56 -08002409unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2410
David S. Miller7ce5a272013-12-02 17:26:05 -05002411/**
2412 * pskb_trim_rcsum - trim received skb and update checksum
2413 * @skb: buffer to trim
2414 * @len: new length
2415 *
2416 * This is exactly the same as pskb_trim except that it ensures the
2417 * checksum of received packets are still valid after the operation.
2418 */
2419
2420static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2421{
2422 if (likely(len >= skb->len))
2423 return 0;
2424 if (skb->ip_summed == CHECKSUM_COMPLETE)
2425 skb->ip_summed = CHECKSUM_NONE;
2426 return __pskb_trim(skb, len);
2427}
2428
Linus Torvalds1da177e2005-04-16 15:20:36 -07002429#define skb_queue_walk(queue, skb) \
2430 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002431 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002432 skb = skb->next)
2433
James Chapman46f89142007-04-30 00:07:31 -07002434#define skb_queue_walk_safe(queue, skb, tmp) \
2435 for (skb = (queue)->next, tmp = skb->next; \
2436 skb != (struct sk_buff *)(queue); \
2437 skb = tmp, tmp = skb->next)
2438
David S. Miller1164f522008-09-23 00:49:44 -07002439#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002440 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002441 skb = skb->next)
2442
2443#define skb_queue_walk_from_safe(queue, skb, tmp) \
2444 for (tmp = skb->next; \
2445 skb != (struct sk_buff *)(queue); \
2446 skb = tmp, tmp = skb->next)
2447
Stephen Hemminger300ce172005-10-30 13:47:34 -08002448#define skb_queue_reverse_walk(queue, skb) \
2449 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002450 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08002451 skb = skb->prev)
2452
David S. Miller686a2952011-01-20 22:47:32 -08002453#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2454 for (skb = (queue)->prev, tmp = skb->prev; \
2455 skb != (struct sk_buff *)(queue); \
2456 skb = tmp, tmp = skb->prev)
2457
2458#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2459 for (tmp = skb->prev; \
2460 skb != (struct sk_buff *)(queue); \
2461 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002462
David S. Miller21dc3302010-08-23 00:13:46 -07002463static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07002464{
2465 return skb_shinfo(skb)->frag_list != NULL;
2466}
2467
2468static inline void skb_frag_list_init(struct sk_buff *skb)
2469{
2470 skb_shinfo(skb)->frag_list = NULL;
2471}
2472
2473static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2474{
2475 frag->next = skb_shinfo(skb)->frag_list;
2476 skb_shinfo(skb)->frag_list = frag;
2477}
2478
2479#define skb_walk_frags(skb, iter) \
2480 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2481
Joe Perches7965bd42013-09-26 14:48:15 -07002482struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2483 int *peeked, int *off, int *err);
2484struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2485 int *err);
2486unsigned int datagram_poll(struct file *file, struct socket *sock,
2487 struct poll_table_struct *wait);
2488int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2489 struct iovec *to, int size);
2490int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2491 struct iovec *iov);
2492int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2493 const struct iovec *from, int from_offset,
2494 int len);
2495int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2496 int offset, size_t count);
2497int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2498 const struct iovec *to, int to_offset,
2499 int size);
2500void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2501void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2502int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07002503int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2504int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2505__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2506 int len, __wsum csum);
2507int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2508 struct pipe_inode_info *pipe, unsigned int len,
2509 unsigned int flags);
2510void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01002511unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00002512int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2513 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07002514void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2515int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2516void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Florian Westphalde960aa2014-01-26 10:58:16 +01002517unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -07002518struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002519
Daniel Borkmann2817a332013-10-30 11:50:51 +01002520struct skb_checksum_ops {
2521 __wsum (*update)(const void *mem, int len, __wsum wsum);
2522 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2523};
2524
2525__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2526 __wsum csum, const struct skb_checksum_ops *ops);
2527__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2528 __wsum csum);
2529
Linus Torvalds1da177e2005-04-16 15:20:36 -07002530static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2531 int len, void *buffer)
2532{
2533 int hlen = skb_headlen(skb);
2534
Patrick McHardy55820ee2005-07-05 14:08:10 -07002535 if (hlen - offset >= len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002536 return skb->data + offset;
2537
2538 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2539 return NULL;
2540
2541 return buffer;
2542}
2543
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01002544/**
2545 * skb_needs_linearize - check if we need to linearize a given skb
2546 * depending on the given device features.
2547 * @skb: socket buffer to check
2548 * @features: net device features
2549 *
2550 * Returns true if either:
2551 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2552 * 2. skb is fragmented and the device does not support SG.
2553 */
2554static inline bool skb_needs_linearize(struct sk_buff *skb,
2555 netdev_features_t features)
2556{
2557 return skb_is_nonlinear(skb) &&
2558 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2559 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2560}
2561
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03002562static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2563 void *to,
2564 const unsigned int len)
2565{
2566 memcpy(to, skb->data, len);
2567}
2568
2569static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2570 const int offset, void *to,
2571 const unsigned int len)
2572{
2573 memcpy(to, skb->data + offset, len);
2574}
2575
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03002576static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2577 const void *from,
2578 const unsigned int len)
2579{
2580 memcpy(skb->data, from, len);
2581}
2582
2583static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2584 const int offset,
2585 const void *from,
2586 const unsigned int len)
2587{
2588 memcpy(skb->data + offset, from, len);
2589}
2590
Joe Perches7965bd42013-09-26 14:48:15 -07002591void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002592
Patrick Ohlyac45f602009-02-12 05:03:37 +00002593static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2594{
2595 return skb->tstamp;
2596}
2597
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002598/**
2599 * skb_get_timestamp - get timestamp from a skb
2600 * @skb: skb to get stamp from
2601 * @stamp: pointer to struct timeval to store stamp in
2602 *
2603 * Timestamps are stored in the skb as offsets to a base timestamp.
2604 * This function converts the offset back to a struct timeval and stores
2605 * it in stamp.
2606 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00002607static inline void skb_get_timestamp(const struct sk_buff *skb,
2608 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002609{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002610 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002611}
2612
Patrick Ohlyac45f602009-02-12 05:03:37 +00002613static inline void skb_get_timestampns(const struct sk_buff *skb,
2614 struct timespec *stamp)
2615{
2616 *stamp = ktime_to_timespec(skb->tstamp);
2617}
2618
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002619static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002620{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002621 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002622}
2623
Stephen Hemminger164891a2007-04-23 22:26:16 -07002624static inline ktime_t net_timedelta(ktime_t t)
2625{
2626 return ktime_sub(ktime_get_real(), t);
2627}
2628
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07002629static inline ktime_t net_invalid_timestamp(void)
2630{
2631 return ktime_set(0, 0);
2632}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002633
Richard Cochranc1f19b52010-07-17 08:49:36 +00002634#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2635
Joe Perches7965bd42013-09-26 14:48:15 -07002636void skb_clone_tx_timestamp(struct sk_buff *skb);
2637bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00002638
2639#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2640
2641static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2642{
2643}
2644
2645static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2646{
2647 return false;
2648}
2649
2650#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2651
2652/**
2653 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2654 *
Richard Cochranda92b192011-10-21 00:49:15 +00002655 * PHY drivers may accept clones of transmitted packets for
2656 * timestamping via their phy_driver.txtstamp method. These drivers
2657 * must call this function to return the skb back to the stack, with
2658 * or without a timestamp.
2659 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00002660 * @skb: clone of the the original outgoing packet
Richard Cochranda92b192011-10-21 00:49:15 +00002661 * @hwtstamps: hardware time stamps, may be NULL if not available
Richard Cochranc1f19b52010-07-17 08:49:36 +00002662 *
2663 */
2664void skb_complete_tx_timestamp(struct sk_buff *skb,
2665 struct skb_shared_hwtstamps *hwtstamps);
2666
Patrick Ohlyac45f602009-02-12 05:03:37 +00002667/**
2668 * skb_tstamp_tx - queue clone of skb with send time stamps
2669 * @orig_skb: the original outgoing packet
2670 * @hwtstamps: hardware time stamps, may be NULL if not available
2671 *
2672 * If the skb has a socket associated, then this function clones the
2673 * skb (thus sharing the actual data and optional structures), stores
2674 * the optional hardware time stamping information (if non NULL) or
2675 * generates a software time stamp (otherwise), then queues the clone
2676 * to the error queue of the socket. Errors are silently ignored.
2677 */
Joe Perches7965bd42013-09-26 14:48:15 -07002678void skb_tstamp_tx(struct sk_buff *orig_skb,
2679 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00002680
Richard Cochran4507a712010-07-17 08:48:28 +00002681static inline void sw_tx_timestamp(struct sk_buff *skb)
2682{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002683 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2684 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00002685 skb_tstamp_tx(skb, NULL);
2686}
2687
2688/**
2689 * skb_tx_timestamp() - Driver hook for transmit timestamping
2690 *
2691 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00002692 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00002693 *
David S. Miller73409f32013-12-27 13:04:33 -05002694 * Specifically, one should make absolutely sure that this function is
2695 * called before TX completion of this packet can trigger. Otherwise
2696 * the packet could potentially already be freed.
2697 *
Richard Cochran4507a712010-07-17 08:48:28 +00002698 * @skb: A socket buffer.
2699 */
2700static inline void skb_tx_timestamp(struct sk_buff *skb)
2701{
Richard Cochranc1f19b52010-07-17 08:49:36 +00002702 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002703 sw_tx_timestamp(skb);
2704}
2705
Johannes Berg6e3e9392011-11-09 10:15:42 +01002706/**
2707 * skb_complete_wifi_ack - deliver skb with wifi status
2708 *
2709 * @skb: the original outgoing packet
2710 * @acked: ack status
2711 *
2712 */
2713void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2714
Joe Perches7965bd42013-09-26 14:48:15 -07002715__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2716__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002717
Herbert Xu60476372007-04-09 11:59:39 -07002718static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2719{
2720 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2721}
2722
Herbert Xufb286bb2005-11-10 13:01:24 -08002723/**
2724 * skb_checksum_complete - Calculate checksum of an entire packet
2725 * @skb: packet to process
2726 *
2727 * This function calculates the checksum over the entire packet plus
2728 * the value of skb->csum. The latter can be used to supply the
2729 * checksum of a pseudo header as used by TCP/UDP. It returns the
2730 * checksum.
2731 *
2732 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2733 * this function can be used to verify that checksum on received
2734 * packets. In that case the function should return zero if the
2735 * checksum is correct. In particular, this function will return zero
2736 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2737 * hardware has already verified the correctness of the checksum.
2738 */
Al Viro4381ca32007-07-15 21:00:11 +01002739static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002740{
Herbert Xu60476372007-04-09 11:59:39 -07002741 return skb_csum_unnecessary(skb) ?
2742 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002743}
2744
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002745/* Check if we need to perform checksum complete validation.
2746 *
2747 * Returns true if checksum complete is needed, false otherwise
2748 * (either checksum is unnecessary or zero checksum is allowed).
2749 */
2750static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
2751 bool zero_okay,
2752 __sum16 check)
2753{
2754 if (skb_csum_unnecessary(skb)) {
2755 return false;
2756 } else if (zero_okay && !check) {
2757 skb->ip_summed = CHECKSUM_UNNECESSARY;
2758 return false;
2759 }
2760
2761 return true;
2762}
2763
2764/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
2765 * in checksum_init.
2766 */
2767#define CHECKSUM_BREAK 76
2768
2769/* Validate (init) checksum based on checksum complete.
2770 *
2771 * Return values:
2772 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
2773 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
2774 * checksum is stored in skb->csum for use in __skb_checksum_complete
2775 * non-zero: value of invalid checksum
2776 *
2777 */
2778static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
2779 bool complete,
2780 __wsum psum)
2781{
2782 if (skb->ip_summed == CHECKSUM_COMPLETE) {
2783 if (!csum_fold(csum_add(psum, skb->csum))) {
2784 skb->ip_summed = CHECKSUM_UNNECESSARY;
2785 return 0;
2786 }
2787 }
2788
2789 skb->csum = psum;
2790
2791 if (complete || skb->len <= CHECKSUM_BREAK)
2792 return __skb_checksum_complete(skb);
2793
2794 return 0;
2795}
2796
2797static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
2798{
2799 return 0;
2800}
2801
2802/* Perform checksum validate (init). Note that this is a macro since we only
2803 * want to calculate the pseudo header which is an input function if necessary.
2804 * First we try to validate without any computation (checksum unnecessary) and
2805 * then calculate based on checksum complete calling the function to compute
2806 * pseudo header.
2807 *
2808 * Return values:
2809 * 0: checksum is validated or try to in skb_checksum_complete
2810 * non-zero: value of invalid checksum
2811 */
2812#define __skb_checksum_validate(skb, proto, complete, \
2813 zero_okay, check, compute_pseudo) \
2814({ \
2815 __sum16 __ret = 0; \
2816 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
2817 __ret = __skb_checksum_validate_complete(skb, \
2818 complete, compute_pseudo(skb, proto)); \
2819 __ret; \
2820})
2821
2822#define skb_checksum_init(skb, proto, compute_pseudo) \
2823 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
2824
2825#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
2826 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
2827
2828#define skb_checksum_validate(skb, proto, compute_pseudo) \
2829 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
2830
2831#define skb_checksum_validate_zero_check(skb, proto, check, \
2832 compute_pseudo) \
2833 __skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
2834
2835#define skb_checksum_simple_validate(skb) \
2836 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
2837
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002838#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07002839void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002840static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2841{
2842 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002843 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002844}
2845static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2846{
2847 if (nfct)
2848 atomic_inc(&nfct->use);
2849}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002850#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002851#ifdef CONFIG_BRIDGE_NETFILTER
2852static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2853{
2854 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2855 kfree(nf_bridge);
2856}
2857static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2858{
2859 if (nf_bridge)
2860 atomic_inc(&nf_bridge->use);
2861}
2862#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08002863static inline void nf_reset(struct sk_buff *skb)
2864{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002865#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08002866 nf_conntrack_put(skb->nfct);
2867 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002868#endif
Patrick McHardya193a4a2006-03-20 19:23:05 -08002869#ifdef CONFIG_BRIDGE_NETFILTER
2870 nf_bridge_put(skb->nf_bridge);
2871 skb->nf_bridge = NULL;
2872#endif
2873}
2874
Patrick McHardy124dff02013-04-05 20:42:05 +02002875static inline void nf_reset_trace(struct sk_buff *skb)
2876{
Florian Westphal478b3602014-02-15 23:48:45 +01002877#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00002878 skb->nf_trace = 0;
2879#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002880}
2881
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002882/* Note: This doesn't put any conntrack and bridge info in dst. */
2883static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2884{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002885#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002886 dst->nfct = src->nfct;
2887 nf_conntrack_get(src->nfct);
2888 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002889#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002890#ifdef CONFIG_BRIDGE_NETFILTER
2891 dst->nf_bridge = src->nf_bridge;
2892 nf_bridge_get(src->nf_bridge);
2893#endif
Florian Westphal478b3602014-02-15 23:48:45 +01002894#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
2895 dst->nf_trace = src->nf_trace;
2896#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002897}
2898
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002899static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2900{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002901#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002902 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002903#endif
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002904#ifdef CONFIG_BRIDGE_NETFILTER
2905 nf_bridge_put(dst->nf_bridge);
2906#endif
2907 __nf_copy(dst, src);
2908}
2909
James Morris984bc162006-06-09 00:29:17 -07002910#ifdef CONFIG_NETWORK_SECMARK
2911static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2912{
2913 to->secmark = from->secmark;
2914}
2915
2916static inline void skb_init_secmark(struct sk_buff *skb)
2917{
2918 skb->secmark = 0;
2919}
2920#else
2921static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2922{ }
2923
2924static inline void skb_init_secmark(struct sk_buff *skb)
2925{ }
2926#endif
2927
Eric W. Biederman574f7192014-04-01 12:20:24 -07002928static inline bool skb_irq_freeable(const struct sk_buff *skb)
2929{
2930 return !skb->destructor &&
2931#if IS_ENABLED(CONFIG_XFRM)
2932 !skb->sp &&
2933#endif
2934#if IS_ENABLED(CONFIG_NF_CONNTRACK)
2935 !skb->nfct &&
2936#endif
2937 !skb->_skb_refdst &&
2938 !skb_has_frag_list(skb);
2939}
2940
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002941static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2942{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002943 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002944}
2945
Stephen Hemminger92477442009-03-21 13:39:26 -07002946static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002947{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002948 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002949}
2950
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002951static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2952{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002953 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002954}
2955
David S. Millerd5a9e242009-01-27 16:22:11 -08002956static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2957{
2958 skb->queue_mapping = rx_queue + 1;
2959}
2960
Stephen Hemminger92477442009-03-21 13:39:26 -07002961static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002962{
2963 return skb->queue_mapping - 1;
2964}
2965
Stephen Hemminger92477442009-03-21 13:39:26 -07002966static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002967{
Eric Dumazeta02cec22010-09-22 20:43:57 +00002968 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08002969}
2970
Joe Perches7965bd42013-09-26 14:48:15 -07002971u16 __skb_tx_hash(const struct net_device *dev, const struct sk_buff *skb,
2972 unsigned int num_tx_queues);
Stephen Hemminger92477442009-03-21 13:39:26 -07002973
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04002974static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2975{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002976#ifdef CONFIG_XFRM
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002977 return skb->sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002978#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002979 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002980#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04002981}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002982
Pravin B Shelar68c33162013-02-14 14:02:41 +00002983/* Keeps track of mac header offset relative to skb->head.
2984 * It is useful for TSO of Tunneling protocol. e.g. GRE.
2985 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07002986 * tunnel skb it points to outer mac header.
2987 * Keeps track of level of encapsulation of network headers.
2988 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00002989struct skb_gso_cb {
Eric Dumazet3347c962013-10-19 11:42:56 -07002990 int mac_offset;
2991 int encap_level;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07002992 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00002993};
2994#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
2995
2996static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
2997{
2998 return (skb_mac_header(inner_skb) - inner_skb->head) -
2999 SKB_GSO_CB(inner_skb)->mac_offset;
3000}
3001
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00003002static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3003{
3004 int new_headroom, headroom;
3005 int ret;
3006
3007 headroom = skb_headroom(skb);
3008 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3009 if (ret)
3010 return ret;
3011
3012 new_headroom = skb_headroom(skb);
3013 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3014 return 0;
3015}
3016
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003017/* Compute the checksum for a gso segment. First compute the checksum value
3018 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3019 * then add in skb->csum (checksum from csum_start to end of packet).
3020 * skb->csum and csum_start are then updated to reflect the checksum of the
3021 * resultant packet starting from the transport header-- the resultant checksum
3022 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3023 * header.
3024 */
3025static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3026{
3027 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3028 skb_transport_offset(skb);
3029 __u16 csum;
3030
3031 csum = csum_fold(csum_partial(skb_transport_header(skb),
3032 plen, skb->csum));
3033 skb->csum = res;
3034 SKB_GSO_CB(skb)->csum_start -= plen;
3035
3036 return csum;
3037}
3038
David S. Millerbdcc0922012-03-07 20:53:36 -05003039static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07003040{
3041 return skb_shinfo(skb)->gso_size;
3042}
3043
Eric Dumazet36a8f392013-09-29 01:21:32 -07003044/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05003045static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02003046{
3047 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3048}
3049
Joe Perches7965bd42013-09-26 14:48:15 -07003050void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07003051
3052static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3053{
3054 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3055 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003056 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3057
Alexander Duyckb78462e2010-06-02 12:24:37 +00003058 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3059 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07003060 __skb_warn_lro_forwarding(skb);
3061 return true;
3062 }
3063 return false;
3064}
3065
Herbert Xu35fc92a2007-03-26 23:22:20 -07003066static inline void skb_forward_csum(struct sk_buff *skb)
3067{
3068 /* Unfortunately we don't support this one. Any brave souls? */
3069 if (skb->ip_summed == CHECKSUM_COMPLETE)
3070 skb->ip_summed = CHECKSUM_NONE;
3071}
3072
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003073/**
3074 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3075 * @skb: skb to check
3076 *
3077 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3078 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3079 * use this helper, to document places where we make this assertion.
3080 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003081static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003082{
3083#ifdef DEBUG
3084 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3085#endif
3086}
3087
Rusty Russellf35d9d82008-02-04 23:49:54 -05003088bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00003089
Paul Durranted1f50c2014-01-09 10:02:46 +00003090int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3091
Daniel Borkmannf77668d2013-03-19 06:39:30 +00003092u32 __skb_get_poff(const struct sk_buff *skb);
3093
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00003094/**
3095 * skb_head_is_locked - Determine if the skb->head is locked down
3096 * @skb: skb to check
3097 *
3098 * The head on skbs build around a head frag can be removed if they are
3099 * not cloned. This function returns true if the skb head is locked down
3100 * due to either being allocated via kmalloc, or by being a clone with
3101 * multiple references to the head.
3102 */
3103static inline bool skb_head_is_locked(const struct sk_buff *skb)
3104{
3105 return !skb->head_frag || skb_cloned(skb);
3106}
Florian Westphalfe6cc552014-02-13 23:09:12 +01003107
3108/**
3109 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3110 *
3111 * @skb: GSO skb
3112 *
3113 * skb_gso_network_seglen is used to determine the real size of the
3114 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3115 *
3116 * The MAC/L2 header is not accounted for.
3117 */
3118static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3119{
3120 unsigned int hdr_len = skb_transport_header(skb) -
3121 skb_network_header(skb);
3122 return hdr_len + skb_gso_transport_seglen(skb);
3123}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003124#endif /* __KERNEL__ */
3125#endif /* _LINUX_SKBUFF_H */