blob: c705808bef9ce8b7c92e7436a3fc1b8dfa13ccc0 [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,
Tom Herbert0f4f4ff2014-06-04 17:20:16 -0700348
349 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
Tom Herbert4749c092014-06-04 17:20:23 -0700350
351 SKB_GSO_GRE_CSUM = 1 << 12,
Herbert Xu79671682006-06-22 02:40:14 -0700352};
353
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700354#if BITS_PER_LONG > 32
355#define NET_SKBUFF_DATA_USES_OFFSET 1
356#endif
357
358#ifdef NET_SKBUFF_DATA_USES_OFFSET
359typedef unsigned int sk_buff_data_t;
360#else
361typedef unsigned char *sk_buff_data_t;
362#endif
363
Eric Dumazet363ec39232014-02-26 14:02:11 -0800364/**
365 * struct skb_mstamp - multi resolution time stamps
366 * @stamp_us: timestamp in us resolution
367 * @stamp_jiffies: timestamp in jiffies
368 */
369struct skb_mstamp {
370 union {
371 u64 v64;
372 struct {
373 u32 stamp_us;
374 u32 stamp_jiffies;
375 };
376 };
377};
378
379/**
380 * skb_mstamp_get - get current timestamp
381 * @cl: place to store timestamps
382 */
383static inline void skb_mstamp_get(struct skb_mstamp *cl)
384{
385 u64 val = local_clock();
386
387 do_div(val, NSEC_PER_USEC);
388 cl->stamp_us = (u32)val;
389 cl->stamp_jiffies = (u32)jiffies;
390}
391
392/**
393 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
394 * @t1: pointer to newest sample
395 * @t0: pointer to oldest sample
396 */
397static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
398 const struct skb_mstamp *t0)
399{
400 s32 delta_us = t1->stamp_us - t0->stamp_us;
401 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
402
403 /* If delta_us is negative, this might be because interval is too big,
404 * or local_clock() drift is too big : fallback using jiffies.
405 */
406 if (delta_us <= 0 ||
407 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
408
409 delta_us = jiffies_to_usecs(delta_jiffies);
410
411 return delta_us;
412}
413
414
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415/**
416 * struct sk_buff - socket buffer
417 * @next: Next buffer in list
418 * @prev: Previous buffer in list
Eric Dumazet363ec39232014-02-26 14:02:11 -0800419 * @tstamp: Time we arrived/left
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700420 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700422 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000423 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700424 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425 * @len: Length of actual data
426 * @data_len: Data length
427 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700428 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700429 * @csum: Checksum (must include start/offset pair)
430 * @csum_start: Offset from skb->head where checksumming should start
431 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700432 * @priority: Packet queueing priority
WANG Cong60ff7462014-05-04 16:39:18 -0700433 * @ignore_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700435 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436 * @nohdr: Payload reference only, must not modify header
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700437 * @nfctinfo: Relationship of this skb to the connection
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700439 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700440 * @ipvs_property: skbuff is owned by ipvs
Randy Dunlap31729362008-02-18 20:52:13 -0800441 * @peeked: this packet has been seen already, so stats have been
442 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700443 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700444 * @protocol: Packet protocol from driver
445 * @destructor: Destruct function
446 * @nfct: Associated connection, if any
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800448 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449 * @tc_index: Traffic control index
450 * @tc_verd: traffic control verdict
Tom Herbert61b905d2014-03-24 15:34:47 -0700451 * @hash: the packet hash
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700452 * @queue_mapping: Queue mapping for multiqueue devices
Randy Dunlap553a5672008-04-20 10:51:01 -0700453 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700454 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Tom Herbert61b905d2014-03-24 15:34:47 -0700455 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
Changli Gao4ca24622011-08-19 07:26:44 -0700456 * ports.
Johannes Berg6e3e9392011-11-09 10:15:42 +0100457 * @wifi_acked_valid: wifi_acked was set
458 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000459 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Randy Dunlapf4b8ea72006-06-22 16:00:11 -0700460 * @dma_cookie: a cookie to one of several possible DMA operations
461 * done by skb DMA functions
Eliezer Tamir06021292013-06-10 11:39:50 +0300462 * @napi_id: id of the NAPI struct this skb came from
James Morris984bc162006-06-09 00:29:17 -0700463 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700464 * @mark: Generic packet mark
465 * @dropcount: total number of sk_receive_queue overflows
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000466 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b02008-07-14 22:49:06 -0700467 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000468 * @inner_protocol: Protocol (encapsulation)
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000469 * @inner_transport_header: Inner transport layer header (encapsulation)
470 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000471 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700472 * @transport_header: Transport layer header
473 * @network_header: Network layer header
474 * @mac_header: Link layer header
475 * @tail: Tail pointer
476 * @end: End pointer
477 * @head: Head of buffer
478 * @data: Data head pointer
479 * @truesize: Buffer size
480 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481 */
482
483struct sk_buff {
484 /* These two members must be first. */
485 struct sk_buff *next;
486 struct sk_buff *prev;
487
Eric Dumazet363ec39232014-02-26 14:02:11 -0800488 union {
489 ktime_t tstamp;
490 struct skb_mstamp skb_mstamp;
491 };
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000492
493 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496 /*
497 * This is the control buffer. It is free to use for every
498 * layer. Please put your private variables there. If you
499 * want to keep them across layers you have to do a skb_clone()
500 * first. This is owned by whoever has the skb queued ATM.
501 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000502 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503
Eric Dumazet7fee2262010-05-11 23:19:48 +0000504 unsigned long _skb_refdst;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000505#ifdef CONFIG_XFRM
506 struct sec_path *sp;
507#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700509 data_len;
510 __u16 mac_len,
511 hdr_len;
Al Viroff1dcad2006-11-20 18:07:29 -0800512 union {
513 __wsum csum;
Herbert Xu663ead32007-04-09 11:59:07 -0700514 struct {
515 __u16 csum_start;
516 __u16 csum_offset;
517 };
Al Viroff1dcad2006-11-20 18:07:29 -0800518 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 __u32 priority;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200520 kmemcheck_bitfield_begin(flags1);
WANG Cong60ff7462014-05-04 16:39:18 -0700521 __u8 ignore_df:1,
Thomas Graf1cbb3382005-07-05 14:13:41 -0700522 cloned:1,
523 ip_summed:2,
Harald Welte6869c4d2005-08-09 19:24:19 -0700524 nohdr:1,
525 nfctinfo:3;
David S. Millerd179cd12005-08-17 14:57:30 -0700526 __u8 pkt_type:3,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800527 fclone:2,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700528 ipvs_property:1,
Herbert Xua59322b2007-12-05 01:53:40 -0800529 peeked:1,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700530 nf_trace:1;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200531 kmemcheck_bitfield_end(flags1);
Eric Dumazet4ab408d2010-03-01 03:09:26 +0000532 __be16 protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533
534 void (*destructor)(struct sk_buff *skb);
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800535#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700536 struct nf_conntrack *nfct;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100537#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538#ifdef CONFIG_BRIDGE_NETFILTER
539 struct nf_bridge_info *nf_bridge;
540#endif
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700541
Eric Dumazet8964be42009-11-20 15:35:04 -0800542 int skb_iif;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000543
Tom Herbert61b905d2014-03-24 15:34:47 -0700544 __u32 hash;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000545
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000546 __be16 vlan_proto;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000547 __u16 vlan_tci;
548
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700550 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700552 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200555
Eric Dumazet0a148422011-04-20 09:27:32 +0000556 __u16 queue_mapping;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200557 kmemcheck_bitfield_begin(flags2);
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400558#ifdef CONFIG_IPV6_NDISC_NODETYPE
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000559 __u8 ndisc_nodetype:2;
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400560#endif
Mel Gormanc93bdd02012-07-31 16:44:19 -0700561 __u8 pfmemalloc:1;
Tom Herbert3853b582010-11-21 13:17:29 +0000562 __u8 ooo_okay:1;
Tom Herbert61b905d2014-03-24 15:34:47 -0700563 __u8 l4_hash:1;
Johannes Berg6e3e9392011-11-09 10:15:42 +0100564 __u8 wifi_acked_valid:1;
565 __u8 wifi_acked:1;
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000566 __u8 no_fcs:1;
Eric Dumazetd3836f22012-04-27 00:33:38 +0000567 __u8 head_frag:1;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000568 /* Encapsulation protocol and NIC drivers should use
569 * this flag to indicate to each other if the skb contains
570 * encapsulated packet or not and maybe use the inner packet
571 * headers if needed
572 */
573 __u8 encapsulation:1;
Tom Herbert7e2b10c2014-06-04 17:20:02 -0700574 __u8 encap_hdr_csum:1;
575 /* 5/7 bit hole (depending on ndisc_nodetype presence) */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200576 kmemcheck_bitfield_end(flags2);
577
Cong Wange0d10952013-08-01 11:10:25 +0800578#if defined CONFIG_NET_DMA || defined CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir06021292013-06-10 11:39:50 +0300579 union {
580 unsigned int napi_id;
581 dma_cookie_t dma_cookie;
582 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700583#endif
James Morris984bc162006-06-09 00:29:17 -0700584#ifdef CONFIG_NETWORK_SECMARK
585 __u32 secmark;
586#endif
Neil Horman3b885782009-10-12 13:26:31 -0700587 union {
588 __u32 mark;
589 __u32 dropcount;
Eric Dumazet16fad692013-03-14 05:40:32 +0000590 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700591 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700592
Simon Horman0d89d202013-05-23 21:02:52 +0000593 __be16 inner_protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000594 __u16 inner_transport_header;
595 __u16 inner_network_header;
596 __u16 inner_mac_header;
597 __u16 transport_header;
598 __u16 network_header;
599 __u16 mac_header;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700601 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700602 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700604 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700605 unsigned int truesize;
606 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607};
608
609#ifdef __KERNEL__
610/*
611 * Handling routines are only of interest to the kernel
612 */
613#include <linux/slab.h>
614
Linus Torvalds1da177e2005-04-16 15:20:36 -0700615
Mel Gormanc93bdd02012-07-31 16:44:19 -0700616#define SKB_ALLOC_FCLONE 0x01
617#define SKB_ALLOC_RX 0x02
618
619/* Returns true if the skb was allocated from PFMEMALLOC reserves */
620static inline bool skb_pfmemalloc(const struct sk_buff *skb)
621{
622 return unlikely(skb->pfmemalloc);
623}
624
Eric Dumazet7fee2262010-05-11 23:19:48 +0000625/*
626 * skb might have a dst pointer attached, refcounted or not.
627 * _skb_refdst low order bit is set if refcount was _not_ taken
628 */
629#define SKB_DST_NOREF 1UL
630#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
631
632/**
633 * skb_dst - returns skb dst_entry
634 * @skb: buffer
635 *
636 * Returns skb dst_entry, regardless of reference taken or not.
637 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000638static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
639{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000640 /* If refdst was not refcounted, check we still are in a
641 * rcu_read_lock section
642 */
643 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
644 !rcu_read_lock_held() &&
645 !rcu_read_lock_bh_held());
646 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000647}
648
Eric Dumazet7fee2262010-05-11 23:19:48 +0000649/**
650 * skb_dst_set - sets skb dst
651 * @skb: buffer
652 * @dst: dst entry
653 *
654 * Sets skb dst, assuming a reference was taken on dst and should
655 * be released by skb_dst_drop()
656 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000657static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
658{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000659 skb->_skb_refdst = (unsigned long)dst;
660}
661
Joe Perches7965bd42013-09-26 14:48:15 -0700662void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
663 bool force);
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200664
665/**
666 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
667 * @skb: buffer
668 * @dst: dst entry
669 *
670 * Sets skb dst, assuming a reference was not taken on dst.
671 * If dst entry is cached, we do not take reference and dst_release
672 * will be avoided by refdst_drop. If dst entry is not cached, we take
673 * reference, so that last dst_release can destroy the dst immediately.
674 */
675static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
676{
677 __skb_dst_set_noref(skb, dst, false);
678}
679
680/**
681 * skb_dst_set_noref_force - sets skb dst, without taking reference
682 * @skb: buffer
683 * @dst: dst entry
684 *
685 * Sets skb dst, assuming a reference was not taken on dst.
686 * No reference is taken and no dst_release will be called. While for
687 * cached dsts deferred reclaim is a basic feature, for entries that are
688 * not cached it is caller's job to guarantee that last dst_release for
689 * provided dst happens when nobody uses it, eg. after a RCU grace period.
690 */
691static inline void skb_dst_set_noref_force(struct sk_buff *skb,
692 struct dst_entry *dst)
693{
694 __skb_dst_set_noref(skb, dst, true);
695}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000696
697/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300698 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000699 * @skb: buffer
700 */
701static inline bool skb_dst_is_noref(const struct sk_buff *skb)
702{
703 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000704}
705
Eric Dumazet511c3f92009-06-02 05:14:27 +0000706static inline struct rtable *skb_rtable(const struct sk_buff *skb)
707{
Eric Dumazetadf30902009-06-02 05:19:30 +0000708 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000709}
710
Joe Perches7965bd42013-09-26 14:48:15 -0700711void kfree_skb(struct sk_buff *skb);
712void kfree_skb_list(struct sk_buff *segs);
713void skb_tx_error(struct sk_buff *skb);
714void consume_skb(struct sk_buff *skb);
715void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000716extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000717
Joe Perches7965bd42013-09-26 14:48:15 -0700718void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
719bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
720 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000721
Joe Perches7965bd42013-09-26 14:48:15 -0700722struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
723 int node);
724struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700725static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100726 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700727{
Eric Dumazet564824b2010-10-11 19:05:25 +0000728 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700729}
730
731static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100732 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700733{
Mel Gormanc93bdd02012-07-31 16:44:19 -0700734 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700735}
736
Joe Perches7965bd42013-09-26 14:48:15 -0700737struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
Patrick McHardy0ebd0ac2013-04-17 06:46:58 +0000738static inline struct sk_buff *alloc_skb_head(gfp_t priority)
739{
740 return __alloc_skb_head(priority, -1);
741}
742
Joe Perches7965bd42013-09-26 14:48:15 -0700743struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
744int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
745struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
746struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
747struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom, gfp_t gfp_mask);
Eric Dumazet117632e2011-12-03 21:39:53 +0000748
Joe Perches7965bd42013-09-26 14:48:15 -0700749int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
750struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
751 unsigned int headroom);
752struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
753 int newtailroom, gfp_t priority);
Fan Du25a91d82014-01-18 09:54:23 +0800754int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
755 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -0700756int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
757 int len);
758int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
759int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000760#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761
Joe Perches7965bd42013-09-26 14:48:15 -0700762int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
763 int getfrag(void *from, char *to, int offset,
764 int len, int odd, struct sk_buff *skb),
765 void *from, int length);
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700766
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800767struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700768 __u32 lower_offset;
769 __u32 upper_offset;
770 __u32 frag_idx;
771 __u32 stepped_offset;
772 struct sk_buff *root_skb;
773 struct sk_buff *cur_skb;
774 __u8 *frag_data;
775};
776
Joe Perches7965bd42013-09-26 14:48:15 -0700777void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
778 unsigned int to, struct skb_seq_state *st);
779unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
780 struct skb_seq_state *st);
781void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -0700782
Joe Perches7965bd42013-09-26 14:48:15 -0700783unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
784 unsigned int to, struct ts_config *config,
785 struct ts_state *state);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700786
Tom Herbert09323cc2013-12-15 22:16:19 -0800787/*
788 * Packet hash types specify the type of hash in skb_set_hash.
789 *
790 * Hash types refer to the protocol layer addresses which are used to
791 * construct a packet's hash. The hashes are used to differentiate or identify
792 * flows of the protocol layer for the hash type. Hash types are either
793 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
794 *
795 * Properties of hashes:
796 *
797 * 1) Two packets in different flows have different hash values
798 * 2) Two packets in the same flow should have the same hash value
799 *
800 * A hash at a higher layer is considered to be more specific. A driver should
801 * set the most specific hash possible.
802 *
803 * A driver cannot indicate a more specific hash than the layer at which a hash
804 * was computed. For instance an L3 hash cannot be set as an L4 hash.
805 *
806 * A driver may indicate a hash level which is less specific than the
807 * actual layer the hash was computed on. For instance, a hash computed
808 * at L4 may be considered an L3 hash. This should only be done if the
809 * driver can't unambiguously determine that the HW computed the hash at
810 * the higher layer. Note that the "should" in the second property above
811 * permits this.
812 */
813enum pkt_hash_types {
814 PKT_HASH_TYPE_NONE, /* Undefined type */
815 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
816 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
817 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
818};
819
820static inline void
821skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
822{
Tom Herbert61b905d2014-03-24 15:34:47 -0700823 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
824 skb->hash = hash;
Tom Herbert09323cc2013-12-15 22:16:19 -0800825}
826
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800827void __skb_get_hash(struct sk_buff *skb);
828static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000829{
Tom Herbert61b905d2014-03-24 15:34:47 -0700830 if (!skb->l4_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800831 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000832
Tom Herbert61b905d2014-03-24 15:34:47 -0700833 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000834}
835
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800836static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
837{
Tom Herbert61b905d2014-03-24 15:34:47 -0700838 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800839}
840
Tom Herbert7539fad2013-12-15 22:12:18 -0800841static inline void skb_clear_hash(struct sk_buff *skb)
842{
Tom Herbert61b905d2014-03-24 15:34:47 -0700843 skb->hash = 0;
844 skb->l4_hash = 0;
Tom Herbert7539fad2013-12-15 22:12:18 -0800845}
846
847static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
848{
Tom Herbert61b905d2014-03-24 15:34:47 -0700849 if (!skb->l4_hash)
Tom Herbert7539fad2013-12-15 22:12:18 -0800850 skb_clear_hash(skb);
851}
852
Tom Herbert3df7a742013-12-15 22:16:29 -0800853static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
854{
Tom Herbert61b905d2014-03-24 15:34:47 -0700855 to->hash = from->hash;
856 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -0800857};
858
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700859#ifdef NET_SKBUFF_DATA_USES_OFFSET
860static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
861{
862 return skb->head + skb->end;
863}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000864
865static inline unsigned int skb_end_offset(const struct sk_buff *skb)
866{
867 return skb->end;
868}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700869#else
870static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
871{
872 return skb->end;
873}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000874
875static inline unsigned int skb_end_offset(const struct sk_buff *skb)
876{
877 return skb->end - skb->head;
878}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700879#endif
880
Linus Torvalds1da177e2005-04-16 15:20:36 -0700881/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700882#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700883
Patrick Ohlyac45f602009-02-12 05:03:37 +0000884static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
885{
886 return &skb_shinfo(skb)->hwtstamps;
887}
888
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889/**
890 * skb_queue_empty - check if a queue is empty
891 * @list: queue head
892 *
893 * Returns true if the queue is empty, false otherwise.
894 */
895static inline int skb_queue_empty(const struct sk_buff_head *list)
896{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100897 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898}
899
900/**
David S. Millerfc7ebb22008-09-23 00:34:07 -0700901 * skb_queue_is_last - check if skb is the last entry in the queue
902 * @list: queue head
903 * @skb: buffer
904 *
905 * Returns true if @skb is the last buffer on the list.
906 */
907static inline bool skb_queue_is_last(const struct sk_buff_head *list,
908 const struct sk_buff *skb)
909{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100910 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -0700911}
912
913/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800914 * skb_queue_is_first - check if skb is the first entry in the queue
915 * @list: queue head
916 * @skb: buffer
917 *
918 * Returns true if @skb is the first buffer on the list.
919 */
920static inline bool skb_queue_is_first(const struct sk_buff_head *list,
921 const struct sk_buff *skb)
922{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100923 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800924}
925
926/**
David S. Miller249c8b42008-09-23 00:44:42 -0700927 * skb_queue_next - return the next packet in the queue
928 * @list: queue head
929 * @skb: current buffer
930 *
931 * Return the next packet in @list after @skb. It is only valid to
932 * call this if skb_queue_is_last() evaluates to false.
933 */
934static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
935 const struct sk_buff *skb)
936{
937 /* This BUG_ON may seem severe, but if we just return then we
938 * are going to dereference garbage.
939 */
940 BUG_ON(skb_queue_is_last(list, skb));
941 return skb->next;
942}
943
944/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800945 * skb_queue_prev - return the prev packet in the queue
946 * @list: queue head
947 * @skb: current buffer
948 *
949 * Return the prev packet in @list before @skb. It is only valid to
950 * call this if skb_queue_is_first() evaluates to false.
951 */
952static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
953 const struct sk_buff *skb)
954{
955 /* This BUG_ON may seem severe, but if we just return then we
956 * are going to dereference garbage.
957 */
958 BUG_ON(skb_queue_is_first(list, skb));
959 return skb->prev;
960}
961
962/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963 * skb_get - reference buffer
964 * @skb: buffer to reference
965 *
966 * Makes another reference to a socket buffer and returns a pointer
967 * to the buffer.
968 */
969static inline struct sk_buff *skb_get(struct sk_buff *skb)
970{
971 atomic_inc(&skb->users);
972 return skb;
973}
974
975/*
976 * If users == 1, we are the only owner and are can avoid redundant
977 * atomic change.
978 */
979
980/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700981 * skb_cloned - is the buffer a clone
982 * @skb: buffer to check
983 *
984 * Returns true if the buffer was generated with skb_clone() and is
985 * one of multiple shared copies of the buffer. Cloned buffers are
986 * shared data so must not be written to under normal circumstances.
987 */
988static inline int skb_cloned(const struct sk_buff *skb)
989{
990 return skb->cloned &&
991 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
992}
993
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +0000994static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
995{
996 might_sleep_if(pri & __GFP_WAIT);
997
998 if (skb_cloned(skb))
999 return pskb_expand_head(skb, 0, 0, pri);
1000
1001 return 0;
1002}
1003
Linus Torvalds1da177e2005-04-16 15:20:36 -07001004/**
1005 * skb_header_cloned - is the header a clone
1006 * @skb: buffer to check
1007 *
1008 * Returns true if modifying the header part of the buffer requires
1009 * the data to be copied.
1010 */
1011static inline int skb_header_cloned(const struct sk_buff *skb)
1012{
1013 int dataref;
1014
1015 if (!skb->cloned)
1016 return 0;
1017
1018 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1019 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1020 return dataref != 1;
1021}
1022
1023/**
1024 * skb_header_release - release reference to header
1025 * @skb: buffer to operate on
1026 *
1027 * Drop a reference to the header part of the buffer. This is done
1028 * by acquiring a payload reference. You must not read from the header
1029 * part of skb->data after this.
1030 */
1031static inline void skb_header_release(struct sk_buff *skb)
1032{
1033 BUG_ON(skb->nohdr);
1034 skb->nohdr = 1;
1035 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1036}
1037
1038/**
1039 * skb_shared - is the buffer shared
1040 * @skb: buffer to check
1041 *
1042 * Returns true if more than one person has a reference to this
1043 * buffer.
1044 */
1045static inline int skb_shared(const struct sk_buff *skb)
1046{
1047 return atomic_read(&skb->users) != 1;
1048}
1049
1050/**
1051 * skb_share_check - check if buffer is shared and if so clone it
1052 * @skb: buffer to check
1053 * @pri: priority for memory allocation
1054 *
1055 * If the buffer is shared the buffer is cloned and the old copy
1056 * drops a reference. A new clone with a single reference is returned.
1057 * If the buffer is not shared the original buffer is returned. When
1058 * being called from interrupt status or with spinlocks held pri must
1059 * be GFP_ATOMIC.
1060 *
1061 * NULL is returned on a memory allocation failure.
1062 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001063static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001064{
1065 might_sleep_if(pri & __GFP_WAIT);
1066 if (skb_shared(skb)) {
1067 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001068
1069 if (likely(nskb))
1070 consume_skb(skb);
1071 else
1072 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001073 skb = nskb;
1074 }
1075 return skb;
1076}
1077
1078/*
1079 * Copy shared buffers into a new sk_buff. We effectively do COW on
1080 * packets to handle cases where we have a local reader and forward
1081 * and a couple of other messy ones. The normal one is tcpdumping
1082 * a packet thats being forwarded.
1083 */
1084
1085/**
1086 * skb_unshare - make a copy of a shared buffer
1087 * @skb: buffer to check
1088 * @pri: priority for memory allocation
1089 *
1090 * If the socket buffer is a clone then this function creates a new
1091 * copy of the data, drops a reference count on the old copy and returns
1092 * the new copy with the reference count at 1. If the buffer is not a clone
1093 * the original buffer is returned. When called with a spinlock held or
1094 * from interrupt state @pri must be %GFP_ATOMIC
1095 *
1096 * %NULL is returned on a memory allocation failure.
1097 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001098static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001099 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001100{
1101 might_sleep_if(pri & __GFP_WAIT);
1102 if (skb_cloned(skb)) {
1103 struct sk_buff *nskb = skb_copy(skb, pri);
1104 kfree_skb(skb); /* Free our shared copy */
1105 skb = nskb;
1106 }
1107 return skb;
1108}
1109
1110/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001111 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001112 * @list_: list to peek at
1113 *
1114 * Peek an &sk_buff. Unlike most other operations you _MUST_
1115 * be careful with this one. A peek leaves the buffer on the
1116 * list and someone else may run off with it. You must hold
1117 * the appropriate locks or have a private queue to do this.
1118 *
1119 * Returns %NULL for an empty list or a pointer to the head element.
1120 * The reference count is not incremented and the reference is therefore
1121 * volatile. Use with caution.
1122 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001123static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001124{
Eric Dumazet18d07002012-04-30 16:31:46 +00001125 struct sk_buff *skb = list_->next;
1126
1127 if (skb == (struct sk_buff *)list_)
1128 skb = NULL;
1129 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130}
1131
1132/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001133 * skb_peek_next - peek skb following the given one from a queue
1134 * @skb: skb to start from
1135 * @list_: list to peek at
1136 *
1137 * Returns %NULL when the end of the list is met or a pointer to the
1138 * next element. The reference count is not incremented and the
1139 * reference is therefore volatile. Use with caution.
1140 */
1141static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1142 const struct sk_buff_head *list_)
1143{
1144 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001145
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001146 if (next == (struct sk_buff *)list_)
1147 next = NULL;
1148 return next;
1149}
1150
1151/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001152 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001153 * @list_: list to peek at
1154 *
1155 * Peek an &sk_buff. Unlike most other operations you _MUST_
1156 * be careful with this one. A peek leaves the buffer on the
1157 * list and someone else may run off with it. You must hold
1158 * the appropriate locks or have a private queue to do this.
1159 *
1160 * Returns %NULL for an empty list or a pointer to the tail element.
1161 * The reference count is not incremented and the reference is therefore
1162 * volatile. Use with caution.
1163 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001164static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001165{
Eric Dumazet18d07002012-04-30 16:31:46 +00001166 struct sk_buff *skb = list_->prev;
1167
1168 if (skb == (struct sk_buff *)list_)
1169 skb = NULL;
1170 return skb;
1171
Linus Torvalds1da177e2005-04-16 15:20:36 -07001172}
1173
1174/**
1175 * skb_queue_len - get queue length
1176 * @list_: list to measure
1177 *
1178 * Return the length of an &sk_buff queue.
1179 */
1180static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1181{
1182 return list_->qlen;
1183}
1184
David S. Miller67fed452008-09-21 22:36:24 -07001185/**
1186 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1187 * @list: queue to initialize
1188 *
1189 * This initializes only the list and queue length aspects of
1190 * an sk_buff_head object. This allows to initialize the list
1191 * aspects of an sk_buff_head without reinitializing things like
1192 * the spinlock. It can also be used for on-stack sk_buff_head
1193 * objects where the spinlock is known to not be used.
1194 */
1195static inline void __skb_queue_head_init(struct sk_buff_head *list)
1196{
1197 list->prev = list->next = (struct sk_buff *)list;
1198 list->qlen = 0;
1199}
1200
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001201/*
1202 * This function creates a split out lock class for each invocation;
1203 * this is needed for now since a whole lot of users of the skb-queue
1204 * infrastructure in drivers have different locking usage (in hardirq)
1205 * than the networking core (in softirq only). In the long run either the
1206 * network layer or drivers should need annotation to consolidate the
1207 * main types of usage into 3 classes.
1208 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209static inline void skb_queue_head_init(struct sk_buff_head *list)
1210{
1211 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001212 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001213}
1214
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001215static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1216 struct lock_class_key *class)
1217{
1218 skb_queue_head_init(list);
1219 lockdep_set_class(&list->lock, class);
1220}
1221
Linus Torvalds1da177e2005-04-16 15:20:36 -07001222/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001223 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001224 *
1225 * The "__skb_xxxx()" functions are the non-atomic ones that
1226 * can only be called with interrupts disabled.
1227 */
Joe Perches7965bd42013-09-26 14:48:15 -07001228void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1229 struct sk_buff_head *list);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001230static inline void __skb_insert(struct sk_buff *newsk,
1231 struct sk_buff *prev, struct sk_buff *next,
1232 struct sk_buff_head *list)
1233{
1234 newsk->next = next;
1235 newsk->prev = prev;
1236 next->prev = prev->next = newsk;
1237 list->qlen++;
1238}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001239
David S. Miller67fed452008-09-21 22:36:24 -07001240static inline void __skb_queue_splice(const struct sk_buff_head *list,
1241 struct sk_buff *prev,
1242 struct sk_buff *next)
1243{
1244 struct sk_buff *first = list->next;
1245 struct sk_buff *last = list->prev;
1246
1247 first->prev = prev;
1248 prev->next = first;
1249
1250 last->next = next;
1251 next->prev = last;
1252}
1253
1254/**
1255 * skb_queue_splice - join two skb lists, this is designed for stacks
1256 * @list: the new list to add
1257 * @head: the place to add it in the first list
1258 */
1259static inline void skb_queue_splice(const struct sk_buff_head *list,
1260 struct sk_buff_head *head)
1261{
1262 if (!skb_queue_empty(list)) {
1263 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001264 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001265 }
1266}
1267
1268/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001269 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001270 * @list: the new list to add
1271 * @head: the place to add it in the first list
1272 *
1273 * The list at @list is reinitialised
1274 */
1275static inline void skb_queue_splice_init(struct sk_buff_head *list,
1276 struct sk_buff_head *head)
1277{
1278 if (!skb_queue_empty(list)) {
1279 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001280 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001281 __skb_queue_head_init(list);
1282 }
1283}
1284
1285/**
1286 * skb_queue_splice_tail - join two skb lists, each list being a queue
1287 * @list: the new list to add
1288 * @head: the place to add it in the first list
1289 */
1290static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1291 struct sk_buff_head *head)
1292{
1293 if (!skb_queue_empty(list)) {
1294 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001295 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001296 }
1297}
1298
1299/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001300 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001301 * @list: the new list to add
1302 * @head: the place to add it in the first list
1303 *
1304 * Each of the lists is a queue.
1305 * The list at @list is reinitialised
1306 */
1307static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1308 struct sk_buff_head *head)
1309{
1310 if (!skb_queue_empty(list)) {
1311 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001312 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001313 __skb_queue_head_init(list);
1314 }
1315}
1316
Linus Torvalds1da177e2005-04-16 15:20:36 -07001317/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001318 * __skb_queue_after - queue a buffer at the list head
1319 * @list: list to use
1320 * @prev: place after this buffer
1321 * @newsk: buffer to queue
1322 *
1323 * Queue a buffer int the middle of a list. This function takes no locks
1324 * and you must therefore hold required locks before calling it.
1325 *
1326 * A buffer cannot be placed on two lists at the same time.
1327 */
1328static inline void __skb_queue_after(struct sk_buff_head *list,
1329 struct sk_buff *prev,
1330 struct sk_buff *newsk)
1331{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001332 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001333}
1334
Joe Perches7965bd42013-09-26 14:48:15 -07001335void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1336 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001337
Gerrit Renkerf5572852008-04-14 00:05:28 -07001338static inline void __skb_queue_before(struct sk_buff_head *list,
1339 struct sk_buff *next,
1340 struct sk_buff *newsk)
1341{
1342 __skb_insert(newsk, next->prev, next, list);
1343}
1344
Stephen Hemminger300ce172005-10-30 13:47:34 -08001345/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001346 * __skb_queue_head - queue a buffer at the list head
1347 * @list: list to use
1348 * @newsk: buffer to queue
1349 *
1350 * Queue a buffer at the start of a list. This function takes no locks
1351 * and you must therefore hold required locks before calling it.
1352 *
1353 * A buffer cannot be placed on two lists at the same time.
1354 */
Joe Perches7965bd42013-09-26 14:48:15 -07001355void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001356static inline void __skb_queue_head(struct sk_buff_head *list,
1357 struct sk_buff *newsk)
1358{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001359 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001360}
1361
1362/**
1363 * __skb_queue_tail - queue a buffer at the list tail
1364 * @list: list to use
1365 * @newsk: buffer to queue
1366 *
1367 * Queue a buffer at the end of a list. This function takes no locks
1368 * and you must therefore hold required locks before calling it.
1369 *
1370 * A buffer cannot be placed on two lists at the same time.
1371 */
Joe Perches7965bd42013-09-26 14:48:15 -07001372void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001373static inline void __skb_queue_tail(struct sk_buff_head *list,
1374 struct sk_buff *newsk)
1375{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001376 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001377}
1378
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001380 * remove sk_buff from list. _Must_ be called atomically, and with
1381 * the list known..
1382 */
Joe Perches7965bd42013-09-26 14:48:15 -07001383void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001384static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1385{
1386 struct sk_buff *next, *prev;
1387
1388 list->qlen--;
1389 next = skb->next;
1390 prev = skb->prev;
1391 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392 next->prev = prev;
1393 prev->next = next;
1394}
1395
Gerrit Renkerf525c062008-04-14 00:04:12 -07001396/**
1397 * __skb_dequeue - remove from the head of the queue
1398 * @list: list to dequeue from
1399 *
1400 * Remove the head of the list. This function does not take any locks
1401 * so must be used with appropriate locks held only. The head item is
1402 * returned or %NULL if the list is empty.
1403 */
Joe Perches7965bd42013-09-26 14:48:15 -07001404struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Gerrit Renkerf525c062008-04-14 00:04:12 -07001405static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1406{
1407 struct sk_buff *skb = skb_peek(list);
1408 if (skb)
1409 __skb_unlink(skb, list);
1410 return skb;
1411}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412
1413/**
1414 * __skb_dequeue_tail - remove from the tail of the queue
1415 * @list: list to dequeue from
1416 *
1417 * Remove the tail of the list. This function does not take any locks
1418 * so must be used with appropriate locks held only. The tail item is
1419 * returned or %NULL if the list is empty.
1420 */
Joe Perches7965bd42013-09-26 14:48:15 -07001421struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001422static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1423{
1424 struct sk_buff *skb = skb_peek_tail(list);
1425 if (skb)
1426 __skb_unlink(skb, list);
1427 return skb;
1428}
1429
1430
David S. Millerbdcc0922012-03-07 20:53:36 -05001431static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001432{
1433 return skb->data_len;
1434}
1435
1436static inline unsigned int skb_headlen(const struct sk_buff *skb)
1437{
1438 return skb->len - skb->data_len;
1439}
1440
1441static inline int skb_pagelen(const struct sk_buff *skb)
1442{
1443 int i, len = 0;
1444
1445 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001446 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001447 return len + skb_headlen(skb);
1448}
1449
Ian Campbell131ea662011-08-19 06:25:00 +00001450/**
1451 * __skb_fill_page_desc - initialise a paged fragment in an skb
1452 * @skb: buffer containing fragment to be initialised
1453 * @i: paged fragment index to initialise
1454 * @page: the page to use for this fragment
1455 * @off: the offset to the data with @page
1456 * @size: the length of the data
1457 *
1458 * Initialises the @i'th fragment of @skb to point to &size bytes at
1459 * offset @off within @page.
1460 *
1461 * Does not take any additional reference on the fragment.
1462 */
1463static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1464 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001465{
1466 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1467
Mel Gormanc48a11c2012-07-31 16:44:23 -07001468 /*
1469 * Propagate page->pfmemalloc to the skb if we can. The problem is
1470 * that not all callers have unique ownership of the page. If
1471 * pfmemalloc is set, we check the mapping as a mapping implies
1472 * page->index is set (index and pfmemalloc share space).
1473 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1474 * do not lose pfmemalloc information as the pages would not be
1475 * allocated using __GFP_MEMALLOC.
1476 */
Ian Campbella8605c62011-10-19 23:01:49 +00001477 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001479 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001480
1481 page = compound_head(page);
1482 if (page->pfmemalloc && !page->mapping)
1483 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00001484}
1485
1486/**
1487 * skb_fill_page_desc - initialise a paged fragment in an skb
1488 * @skb: buffer containing fragment to be initialised
1489 * @i: paged fragment index to initialise
1490 * @page: the page to use for this fragment
1491 * @off: the offset to the data with @page
1492 * @size: the length of the data
1493 *
1494 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01001495 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00001496 * addition updates @skb such that @i is the last fragment.
1497 *
1498 * Does not take any additional reference on the fragment.
1499 */
1500static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1501 struct page *page, int off, int size)
1502{
1503 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001504 skb_shinfo(skb)->nr_frags = i + 1;
1505}
1506
Joe Perches7965bd42013-09-26 14:48:15 -07001507void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1508 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001509
Jason Wangf8e617e2013-11-01 14:07:47 +08001510void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1511 unsigned int truesize);
1512
Linus Torvalds1da177e2005-04-16 15:20:36 -07001513#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001514#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001515#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1516
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001517#ifdef NET_SKBUFF_DATA_USES_OFFSET
1518static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1519{
1520 return skb->head + skb->tail;
1521}
1522
1523static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1524{
1525 skb->tail = skb->data - skb->head;
1526}
1527
1528static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1529{
1530 skb_reset_tail_pointer(skb);
1531 skb->tail += offset;
1532}
Simon Horman7cc46192013-05-28 20:34:29 +00001533
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001534#else /* NET_SKBUFF_DATA_USES_OFFSET */
1535static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1536{
1537 return skb->tail;
1538}
1539
1540static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1541{
1542 skb->tail = skb->data;
1543}
1544
1545static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1546{
1547 skb->tail = skb->data + offset;
1548}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001549
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001550#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1551
Linus Torvalds1da177e2005-04-16 15:20:36 -07001552/*
1553 * Add data to an sk_buff
1554 */
Mathias Krause0c7ddf32013-11-07 14:18:24 +01001555unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001556unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001557static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1558{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001559 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001560 SKB_LINEAR_ASSERT(skb);
1561 skb->tail += len;
1562 skb->len += len;
1563 return tmp;
1564}
1565
Joe Perches7965bd42013-09-26 14:48:15 -07001566unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1568{
1569 skb->data -= len;
1570 skb->len += len;
1571 return skb->data;
1572}
1573
Joe Perches7965bd42013-09-26 14:48:15 -07001574unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001575static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1576{
1577 skb->len -= len;
1578 BUG_ON(skb->len < skb->data_len);
1579 return skb->data += len;
1580}
1581
David S. Miller47d29642010-05-02 02:21:44 -07001582static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1583{
1584 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1585}
1586
Joe Perches7965bd42013-09-26 14:48:15 -07001587unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001588
1589static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1590{
1591 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001592 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001593 return NULL;
1594 skb->len -= len;
1595 return skb->data += len;
1596}
1597
1598static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1599{
1600 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1601}
1602
1603static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1604{
1605 if (likely(len <= skb_headlen(skb)))
1606 return 1;
1607 if (unlikely(len > skb->len))
1608 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001609 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001610}
1611
1612/**
1613 * skb_headroom - bytes at buffer head
1614 * @skb: buffer to check
1615 *
1616 * Return the number of bytes of free space at the head of an &sk_buff.
1617 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001618static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001619{
1620 return skb->data - skb->head;
1621}
1622
1623/**
1624 * skb_tailroom - bytes at buffer end
1625 * @skb: buffer to check
1626 *
1627 * Return the number of bytes of free space at the tail of an sk_buff
1628 */
1629static inline int skb_tailroom(const struct sk_buff *skb)
1630{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001631 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001632}
1633
1634/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00001635 * skb_availroom - bytes at buffer end
1636 * @skb: buffer to check
1637 *
1638 * Return the number of bytes of free space at the tail of an sk_buff
1639 * allocated by sk_stream_alloc()
1640 */
1641static inline int skb_availroom(const struct sk_buff *skb)
1642{
Eric Dumazet16fad692013-03-14 05:40:32 +00001643 if (skb_is_nonlinear(skb))
1644 return 0;
1645
1646 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00001647}
1648
1649/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001650 * skb_reserve - adjust headroom
1651 * @skb: buffer to alter
1652 * @len: bytes to move
1653 *
1654 * Increase the headroom of an empty &sk_buff by reducing the tail
1655 * room. This is only allowed for an empty buffer.
1656 */
David S. Miller8243126c2006-01-17 02:54:21 -08001657static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001658{
1659 skb->data += len;
1660 skb->tail += len;
1661}
1662
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001663static inline void skb_reset_inner_headers(struct sk_buff *skb)
1664{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001665 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001666 skb->inner_network_header = skb->network_header;
1667 skb->inner_transport_header = skb->transport_header;
1668}
1669
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00001670static inline void skb_reset_mac_len(struct sk_buff *skb)
1671{
1672 skb->mac_len = skb->network_header - skb->mac_header;
1673}
1674
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001675static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1676 *skb)
1677{
1678 return skb->head + skb->inner_transport_header;
1679}
1680
1681static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1682{
1683 skb->inner_transport_header = skb->data - skb->head;
1684}
1685
1686static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1687 const int offset)
1688{
1689 skb_reset_inner_transport_header(skb);
1690 skb->inner_transport_header += offset;
1691}
1692
1693static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1694{
1695 return skb->head + skb->inner_network_header;
1696}
1697
1698static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1699{
1700 skb->inner_network_header = skb->data - skb->head;
1701}
1702
1703static inline void skb_set_inner_network_header(struct sk_buff *skb,
1704 const int offset)
1705{
1706 skb_reset_inner_network_header(skb);
1707 skb->inner_network_header += offset;
1708}
1709
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001710static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1711{
1712 return skb->head + skb->inner_mac_header;
1713}
1714
1715static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1716{
1717 skb->inner_mac_header = skb->data - skb->head;
1718}
1719
1720static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1721 const int offset)
1722{
1723 skb_reset_inner_mac_header(skb);
1724 skb->inner_mac_header += offset;
1725}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001726static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1727{
Cong Wang35d04612013-05-29 15:16:05 +08001728 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001729}
1730
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001731static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1732{
1733 return skb->head + skb->transport_header;
1734}
1735
1736static inline void skb_reset_transport_header(struct sk_buff *skb)
1737{
1738 skb->transport_header = skb->data - skb->head;
1739}
1740
1741static inline void skb_set_transport_header(struct sk_buff *skb,
1742 const int offset)
1743{
1744 skb_reset_transport_header(skb);
1745 skb->transport_header += offset;
1746}
1747
1748static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1749{
1750 return skb->head + skb->network_header;
1751}
1752
1753static inline void skb_reset_network_header(struct sk_buff *skb)
1754{
1755 skb->network_header = skb->data - skb->head;
1756}
1757
1758static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1759{
1760 skb_reset_network_header(skb);
1761 skb->network_header += offset;
1762}
1763
1764static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1765{
1766 return skb->head + skb->mac_header;
1767}
1768
1769static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1770{
Cong Wang35d04612013-05-29 15:16:05 +08001771 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001772}
1773
1774static inline void skb_reset_mac_header(struct sk_buff *skb)
1775{
1776 skb->mac_header = skb->data - skb->head;
1777}
1778
1779static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1780{
1781 skb_reset_mac_header(skb);
1782 skb->mac_header += offset;
1783}
1784
Timo Teräs0e3da5b2013-12-16 11:02:09 +02001785static inline void skb_pop_mac_header(struct sk_buff *skb)
1786{
1787 skb->mac_header = skb->network_header;
1788}
1789
Ying Xuefbbdb8f2013-03-27 16:46:06 +00001790static inline void skb_probe_transport_header(struct sk_buff *skb,
1791 const int offset_hint)
1792{
1793 struct flow_keys keys;
1794
1795 if (skb_transport_header_was_set(skb))
1796 return;
1797 else if (skb_flow_dissect(skb, &keys))
1798 skb_set_transport_header(skb, keys.thoff);
1799 else
1800 skb_set_transport_header(skb, offset_hint);
1801}
1802
Eric Dumazet03606892012-02-23 10:55:02 +00001803static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1804{
1805 if (skb_mac_header_was_set(skb)) {
1806 const unsigned char *old_mac = skb_mac_header(skb);
1807
1808 skb_set_mac_header(skb, -skb->mac_len);
1809 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1810 }
1811}
1812
Michał Mirosław04fb4512010-12-14 15:24:08 +00001813static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1814{
1815 return skb->csum_start - skb_headroom(skb);
1816}
1817
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001818static inline int skb_transport_offset(const struct sk_buff *skb)
1819{
1820 return skb_transport_header(skb) - skb->data;
1821}
1822
1823static inline u32 skb_network_header_len(const struct sk_buff *skb)
1824{
1825 return skb->transport_header - skb->network_header;
1826}
1827
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001828static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1829{
1830 return skb->inner_transport_header - skb->inner_network_header;
1831}
1832
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001833static inline int skb_network_offset(const struct sk_buff *skb)
1834{
1835 return skb_network_header(skb) - skb->data;
1836}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001837
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001838static inline int skb_inner_network_offset(const struct sk_buff *skb)
1839{
1840 return skb_inner_network_header(skb) - skb->data;
1841}
1842
Changli Gaof9599ce2010-08-04 04:43:44 +00001843static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1844{
1845 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1846}
1847
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848/*
1849 * CPUs often take a performance hit when accessing unaligned memory
1850 * locations. The actual performance hit varies, it can be small if the
1851 * hardware handles it or large if we have to take an exception and fix it
1852 * in software.
1853 *
1854 * Since an ethernet header is 14 bytes network drivers often end up with
1855 * the IP header at an unaligned offset. The IP header can be aligned by
1856 * shifting the start of the packet by 2 bytes. Drivers should do this
1857 * with:
1858 *
Tobias Klauser8660c122009-07-13 22:48:16 +00001859 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001860 *
1861 * The downside to this alignment of the IP header is that the DMA is now
1862 * unaligned. On some architectures the cost of an unaligned DMA is high
1863 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00001864 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001865 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1866 * to be overridden.
1867 */
1868#ifndef NET_IP_ALIGN
1869#define NET_IP_ALIGN 2
1870#endif
1871
Anton Blanchard025be812006-03-31 02:27:06 -08001872/*
1873 * The networking layer reserves some headroom in skb data (via
1874 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1875 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08001876 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08001877 *
1878 * Unfortunately this headroom changes the DMA alignment of the resulting
1879 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1880 * on some architectures. An architecture can override this value,
1881 * perhaps setting it to a cacheline in size (since that will maintain
1882 * cacheline alignment of the DMA). It must be a power of 2.
1883 *
David S. Millerd6301d32009-02-08 19:24:13 -08001884 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08001885 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07001886 *
1887 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1888 * to reduce average number of cache lines per packet.
1889 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07001890 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08001891 */
1892#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07001893#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08001894#endif
1895
Joe Perches7965bd42013-09-26 14:48:15 -07001896int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001897
1898static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1899{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00001900 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07001901 WARN_ON(1);
1902 return;
1903 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001904 skb->len = len;
1905 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001906}
1907
Joe Perches7965bd42013-09-26 14:48:15 -07001908void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001909
1910static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1911{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001912 if (skb->data_len)
1913 return ___pskb_trim(skb, len);
1914 __skb_trim(skb, len);
1915 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001916}
1917
1918static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1919{
1920 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1921}
1922
1923/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07001924 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1925 * @skb: buffer to alter
1926 * @len: new length
1927 *
1928 * This is identical to pskb_trim except that the caller knows that
1929 * the skb is not cloned so we should never get an error due to out-
1930 * of-memory.
1931 */
1932static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1933{
1934 int err = pskb_trim(skb, len);
1935 BUG_ON(err);
1936}
1937
1938/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001939 * skb_orphan - orphan a buffer
1940 * @skb: buffer to orphan
1941 *
1942 * If a buffer currently has an owner then we call the owner's
1943 * destructor function and make the @skb unowned. The buffer continues
1944 * to exist but is no longer charged to its former owner.
1945 */
1946static inline void skb_orphan(struct sk_buff *skb)
1947{
Eric Dumazetc34a7612013-07-30 16:11:15 -07001948 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001949 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07001950 skb->destructor = NULL;
1951 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07001952 } else {
1953 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07001954 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001955}
1956
1957/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00001958 * skb_orphan_frags - orphan the frags contained in a buffer
1959 * @skb: buffer to orphan frags from
1960 * @gfp_mask: allocation mask for replacement pages
1961 *
1962 * For each frag in the SKB which needs a destructor (i.e. has an
1963 * owner) create a copy of that frag and release the original
1964 * page by calling the destructor.
1965 */
1966static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1967{
1968 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1969 return 0;
1970 return skb_copy_ubufs(skb, gfp_mask);
1971}
1972
1973/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001974 * __skb_queue_purge - empty a list
1975 * @list: list to empty
1976 *
1977 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1978 * the list and one reference dropped. This function does not take the
1979 * list lock and the caller must hold the relevant locks to use it.
1980 */
Joe Perches7965bd42013-09-26 14:48:15 -07001981void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001982static inline void __skb_queue_purge(struct sk_buff_head *list)
1983{
1984 struct sk_buff *skb;
1985 while ((skb = __skb_dequeue(list)) != NULL)
1986 kfree_skb(skb);
1987}
1988
Alexander Duycke5e67302013-02-08 10:17:15 +00001989#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
1990#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
1991#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
1992
Joe Perches7965bd42013-09-26 14:48:15 -07001993void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001994
Joe Perches7965bd42013-09-26 14:48:15 -07001995struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
1996 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07001997
1998/**
1999 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2000 * @dev: network device to receive on
2001 * @length: length to allocate
2002 *
2003 * Allocate a new &sk_buff and assign it a usage count of one. The
2004 * buffer has unspecified headroom built in. Users should allocate
2005 * the headroom they think they need without accounting for the
2006 * built in space. The built in space is used for optimisations.
2007 *
2008 * %NULL is returned if there is no free memory. Although this function
2009 * allocates memory it can be called from an interrupt.
2010 */
2011static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002012 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002013{
2014 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2015}
2016
Eric Dumazet6f532612012-05-18 05:12:12 +00002017/* legacy helper around __netdev_alloc_skb() */
2018static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2019 gfp_t gfp_mask)
2020{
2021 return __netdev_alloc_skb(NULL, length, gfp_mask);
2022}
2023
2024/* legacy helper around netdev_alloc_skb() */
2025static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2026{
2027 return netdev_alloc_skb(NULL, length);
2028}
2029
2030
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002031static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2032 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002033{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002034 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002035
2036 if (NET_IP_ALIGN && skb)
2037 skb_reserve(skb, NET_IP_ALIGN);
2038 return skb;
2039}
2040
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002041static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2042 unsigned int length)
2043{
2044 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2045}
2046
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002047/**
2048 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
Mel Gorman06140022012-07-31 16:44:24 -07002049 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2050 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2051 * @order: size of the allocation
2052 *
2053 * Allocate a new page.
2054 *
2055 * %NULL is returned if there is no free memory.
2056*/
2057static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
2058 struct sk_buff *skb,
2059 unsigned int order)
2060{
2061 struct page *page;
2062
2063 gfp_mask |= __GFP_COLD;
2064
2065 if (!(gfp_mask & __GFP_NOMEMALLOC))
2066 gfp_mask |= __GFP_MEMALLOC;
2067
2068 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2069 if (skb && page && page->pfmemalloc)
2070 skb->pfmemalloc = true;
2071
2072 return page;
2073}
2074
2075/**
2076 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2077 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2078 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2079 *
2080 * Allocate a new page.
2081 *
2082 * %NULL is returned if there is no free memory.
2083 */
2084static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
2085 struct sk_buff *skb)
2086{
2087 return __skb_alloc_pages(gfp_mask, skb, 0);
2088}
2089
2090/**
2091 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2092 * @page: The page that was allocated from skb_alloc_page
2093 * @skb: The skb that may need pfmemalloc set
2094 */
2095static inline void skb_propagate_pfmemalloc(struct page *page,
2096 struct sk_buff *skb)
2097{
2098 if (page && page->pfmemalloc)
2099 skb->pfmemalloc = true;
2100}
2101
Eric Dumazet564824b2010-10-11 19:05:25 +00002102/**
Masanari Iidae2278672014-02-18 22:54:36 +09002103 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002104 * @frag: the paged fragment
2105 *
2106 * Returns the &struct page associated with @frag.
2107 */
2108static inline struct page *skb_frag_page(const skb_frag_t *frag)
2109{
Ian Campbella8605c62011-10-19 23:01:49 +00002110 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00002111}
2112
2113/**
2114 * __skb_frag_ref - take an addition reference on a paged fragment.
2115 * @frag: the paged fragment
2116 *
2117 * Takes an additional reference on the paged fragment @frag.
2118 */
2119static inline void __skb_frag_ref(skb_frag_t *frag)
2120{
2121 get_page(skb_frag_page(frag));
2122}
2123
2124/**
2125 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2126 * @skb: the buffer
2127 * @f: the fragment offset.
2128 *
2129 * Takes an additional reference on the @f'th paged fragment of @skb.
2130 */
2131static inline void skb_frag_ref(struct sk_buff *skb, int f)
2132{
2133 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2134}
2135
2136/**
2137 * __skb_frag_unref - release a reference on a paged fragment.
2138 * @frag: the paged fragment
2139 *
2140 * Releases a reference on the paged fragment @frag.
2141 */
2142static inline void __skb_frag_unref(skb_frag_t *frag)
2143{
2144 put_page(skb_frag_page(frag));
2145}
2146
2147/**
2148 * skb_frag_unref - release a reference on a paged fragment of an skb.
2149 * @skb: the buffer
2150 * @f: the fragment offset
2151 *
2152 * Releases a reference on the @f'th paged fragment of @skb.
2153 */
2154static inline void skb_frag_unref(struct sk_buff *skb, int f)
2155{
2156 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2157}
2158
2159/**
2160 * skb_frag_address - gets the address of the data contained in a paged fragment
2161 * @frag: the paged fragment buffer
2162 *
2163 * Returns the address of the data within @frag. The page must already
2164 * be mapped.
2165 */
2166static inline void *skb_frag_address(const skb_frag_t *frag)
2167{
2168 return page_address(skb_frag_page(frag)) + frag->page_offset;
2169}
2170
2171/**
2172 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2173 * @frag: the paged fragment buffer
2174 *
2175 * Returns the address of the data within @frag. Checks that the page
2176 * is mapped and returns %NULL otherwise.
2177 */
2178static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2179{
2180 void *ptr = page_address(skb_frag_page(frag));
2181 if (unlikely(!ptr))
2182 return NULL;
2183
2184 return ptr + frag->page_offset;
2185}
2186
2187/**
2188 * __skb_frag_set_page - sets the page contained in a paged fragment
2189 * @frag: the paged fragment
2190 * @page: the page to set
2191 *
2192 * Sets the fragment @frag to contain @page.
2193 */
2194static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2195{
Ian Campbella8605c62011-10-19 23:01:49 +00002196 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002197}
2198
2199/**
2200 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2201 * @skb: the buffer
2202 * @f: the fragment offset
2203 * @page: the page to set
2204 *
2205 * Sets the @f'th fragment of @skb to contain @page.
2206 */
2207static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2208 struct page *page)
2209{
2210 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2211}
2212
Eric Dumazet400dfd32013-10-17 16:27:07 -07002213bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2214
Ian Campbell131ea662011-08-19 06:25:00 +00002215/**
2216 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002217 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002218 * @frag: the paged fragment to map
2219 * @offset: the offset within the fragment (starting at the
2220 * fragment's own offset)
2221 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002222 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00002223 *
2224 * Maps the page associated with @frag to @device.
2225 */
2226static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2227 const skb_frag_t *frag,
2228 size_t offset, size_t size,
2229 enum dma_data_direction dir)
2230{
2231 return dma_map_page(dev, skb_frag_page(frag),
2232 frag->page_offset + offset, size, dir);
2233}
2234
Eric Dumazet117632e2011-12-03 21:39:53 +00002235static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2236 gfp_t gfp_mask)
2237{
2238 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2239}
2240
Ian Campbell131ea662011-08-19 06:25:00 +00002241/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002242 * skb_clone_writable - is the header of a clone writable
2243 * @skb: buffer to check
2244 * @len: length up to which to write
2245 *
2246 * Returns true if modifying the header part of the cloned buffer
2247 * does not requires the data to be copied.
2248 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002249static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002250{
2251 return !skb_header_cloned(skb) &&
2252 skb_headroom(skb) + len <= skb->hdr_len;
2253}
2254
Herbert Xud9cc2042007-09-16 16:21:16 -07002255static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2256 int cloned)
2257{
2258 int delta = 0;
2259
Herbert Xud9cc2042007-09-16 16:21:16 -07002260 if (headroom > skb_headroom(skb))
2261 delta = headroom - skb_headroom(skb);
2262
2263 if (delta || cloned)
2264 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2265 GFP_ATOMIC);
2266 return 0;
2267}
2268
Patrick McHardy334a8132007-06-25 04:35:20 -07002269/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002270 * skb_cow - copy header of skb when it is required
2271 * @skb: buffer to cow
2272 * @headroom: needed headroom
2273 *
2274 * If the skb passed lacks sufficient headroom or its data part
2275 * is shared, data is reallocated. If reallocation fails, an error
2276 * is returned and original skb is not changed.
2277 *
2278 * The result is skb with writable area skb->head...skb->tail
2279 * and at least @headroom of space at head.
2280 */
2281static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2282{
Herbert Xud9cc2042007-09-16 16:21:16 -07002283 return __skb_cow(skb, headroom, skb_cloned(skb));
2284}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002285
Herbert Xud9cc2042007-09-16 16:21:16 -07002286/**
2287 * skb_cow_head - skb_cow but only making the head writable
2288 * @skb: buffer to cow
2289 * @headroom: needed headroom
2290 *
2291 * This function is identical to skb_cow except that we replace the
2292 * skb_cloned check by skb_header_cloned. It should be used when
2293 * you only need to push on some header and do not need to modify
2294 * the data.
2295 */
2296static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2297{
2298 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002299}
2300
2301/**
2302 * skb_padto - pad an skbuff up to a minimal size
2303 * @skb: buffer to pad
2304 * @len: minimal length
2305 *
2306 * Pads up a buffer to ensure the trailing bytes exist and are
2307 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002308 * is untouched. Otherwise it is extended. Returns zero on
2309 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002310 */
2311
Herbert Xu5b057c62006-06-23 02:06:41 -07002312static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002313{
2314 unsigned int size = skb->len;
2315 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002316 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002317 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002318}
2319
2320static inline int skb_add_data(struct sk_buff *skb,
2321 char __user *from, int copy)
2322{
2323 const int off = skb->len;
2324
2325 if (skb->ip_summed == CHECKSUM_NONE) {
2326 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08002327 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
Linus Torvalds1da177e2005-04-16 15:20:36 -07002328 copy, 0, &err);
2329 if (!err) {
2330 skb->csum = csum_block_add(skb->csum, csum, off);
2331 return 0;
2332 }
2333 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2334 return 0;
2335
2336 __skb_trim(skb, off);
2337 return -EFAULT;
2338}
2339
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002340static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2341 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002342{
2343 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002344 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002345
Ian Campbellea2ab692011-08-22 23:44:58 +00002346 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002347 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002348 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002349 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002350}
2351
Herbert Xu364c6ba2006-06-09 16:10:40 -07002352static inline int __skb_linearize(struct sk_buff *skb)
2353{
2354 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2355}
2356
Linus Torvalds1da177e2005-04-16 15:20:36 -07002357/**
2358 * skb_linearize - convert paged skb to linear one
2359 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002360 *
2361 * If there is no free memory -ENOMEM is returned, otherwise zero
2362 * is returned and the old skb data released.
2363 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002364static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002365{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002366 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2367}
2368
2369/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00002370 * skb_has_shared_frag - can any frag be overwritten
2371 * @skb: buffer to test
2372 *
2373 * Return true if the skb has at least one frag that might be modified
2374 * by an external entity (as in vmsplice()/sendfile())
2375 */
2376static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2377{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00002378 return skb_is_nonlinear(skb) &&
2379 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00002380}
2381
2382/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07002383 * skb_linearize_cow - make sure skb is linear and writable
2384 * @skb: buffer to process
2385 *
2386 * If there is no free memory -ENOMEM is returned, otherwise zero
2387 * is returned and the old skb data released.
2388 */
2389static inline int skb_linearize_cow(struct sk_buff *skb)
2390{
2391 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2392 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002393}
2394
2395/**
2396 * skb_postpull_rcsum - update checksum for received skb after pull
2397 * @skb: buffer to update
2398 * @start: start of data before pull
2399 * @len: length of data pulled
2400 *
2401 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002402 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2403 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002404 */
2405
2406static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002407 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002408{
Patrick McHardy84fa7932006-08-29 16:44:56 -07002409 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002410 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2411}
2412
Herbert Xucbb042f2006-03-20 22:43:56 -08002413unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2414
David S. Miller7ce5a272013-12-02 17:26:05 -05002415/**
2416 * pskb_trim_rcsum - trim received skb and update checksum
2417 * @skb: buffer to trim
2418 * @len: new length
2419 *
2420 * This is exactly the same as pskb_trim except that it ensures the
2421 * checksum of received packets are still valid after the operation.
2422 */
2423
2424static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2425{
2426 if (likely(len >= skb->len))
2427 return 0;
2428 if (skb->ip_summed == CHECKSUM_COMPLETE)
2429 skb->ip_summed = CHECKSUM_NONE;
2430 return __pskb_trim(skb, len);
2431}
2432
Linus Torvalds1da177e2005-04-16 15:20:36 -07002433#define skb_queue_walk(queue, skb) \
2434 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002435 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002436 skb = skb->next)
2437
James Chapman46f89142007-04-30 00:07:31 -07002438#define skb_queue_walk_safe(queue, skb, tmp) \
2439 for (skb = (queue)->next, tmp = skb->next; \
2440 skb != (struct sk_buff *)(queue); \
2441 skb = tmp, tmp = skb->next)
2442
David S. Miller1164f522008-09-23 00:49:44 -07002443#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002444 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002445 skb = skb->next)
2446
2447#define skb_queue_walk_from_safe(queue, skb, tmp) \
2448 for (tmp = skb->next; \
2449 skb != (struct sk_buff *)(queue); \
2450 skb = tmp, tmp = skb->next)
2451
Stephen Hemminger300ce172005-10-30 13:47:34 -08002452#define skb_queue_reverse_walk(queue, skb) \
2453 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002454 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08002455 skb = skb->prev)
2456
David S. Miller686a2952011-01-20 22:47:32 -08002457#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2458 for (skb = (queue)->prev, tmp = skb->prev; \
2459 skb != (struct sk_buff *)(queue); \
2460 skb = tmp, tmp = skb->prev)
2461
2462#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2463 for (tmp = skb->prev; \
2464 skb != (struct sk_buff *)(queue); \
2465 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002466
David S. Miller21dc3302010-08-23 00:13:46 -07002467static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07002468{
2469 return skb_shinfo(skb)->frag_list != NULL;
2470}
2471
2472static inline void skb_frag_list_init(struct sk_buff *skb)
2473{
2474 skb_shinfo(skb)->frag_list = NULL;
2475}
2476
2477static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2478{
2479 frag->next = skb_shinfo(skb)->frag_list;
2480 skb_shinfo(skb)->frag_list = frag;
2481}
2482
2483#define skb_walk_frags(skb, iter) \
2484 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2485
Joe Perches7965bd42013-09-26 14:48:15 -07002486struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2487 int *peeked, int *off, int *err);
2488struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2489 int *err);
2490unsigned int datagram_poll(struct file *file, struct socket *sock,
2491 struct poll_table_struct *wait);
2492int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2493 struct iovec *to, int size);
2494int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2495 struct iovec *iov);
2496int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2497 const struct iovec *from, int from_offset,
2498 int len);
2499int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2500 int offset, size_t count);
2501int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2502 const struct iovec *to, int to_offset,
2503 int size);
2504void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2505void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2506int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07002507int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2508int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2509__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2510 int len, __wsum csum);
2511int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2512 struct pipe_inode_info *pipe, unsigned int len,
2513 unsigned int flags);
2514void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01002515unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00002516int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2517 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07002518void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2519int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2520void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Florian Westphalde960aa2014-01-26 10:58:16 +01002521unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -07002522struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002523
Daniel Borkmann2817a332013-10-30 11:50:51 +01002524struct skb_checksum_ops {
2525 __wsum (*update)(const void *mem, int len, __wsum wsum);
2526 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2527};
2528
2529__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2530 __wsum csum, const struct skb_checksum_ops *ops);
2531__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2532 __wsum csum);
2533
Linus Torvalds1da177e2005-04-16 15:20:36 -07002534static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2535 int len, void *buffer)
2536{
2537 int hlen = skb_headlen(skb);
2538
Patrick McHardy55820ee2005-07-05 14:08:10 -07002539 if (hlen - offset >= len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002540 return skb->data + offset;
2541
2542 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2543 return NULL;
2544
2545 return buffer;
2546}
2547
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01002548/**
2549 * skb_needs_linearize - check if we need to linearize a given skb
2550 * depending on the given device features.
2551 * @skb: socket buffer to check
2552 * @features: net device features
2553 *
2554 * Returns true if either:
2555 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2556 * 2. skb is fragmented and the device does not support SG.
2557 */
2558static inline bool skb_needs_linearize(struct sk_buff *skb,
2559 netdev_features_t features)
2560{
2561 return skb_is_nonlinear(skb) &&
2562 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2563 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2564}
2565
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03002566static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2567 void *to,
2568 const unsigned int len)
2569{
2570 memcpy(to, skb->data, len);
2571}
2572
2573static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2574 const int offset, void *to,
2575 const unsigned int len)
2576{
2577 memcpy(to, skb->data + offset, len);
2578}
2579
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03002580static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2581 const void *from,
2582 const unsigned int len)
2583{
2584 memcpy(skb->data, from, len);
2585}
2586
2587static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2588 const int offset,
2589 const void *from,
2590 const unsigned int len)
2591{
2592 memcpy(skb->data + offset, from, len);
2593}
2594
Joe Perches7965bd42013-09-26 14:48:15 -07002595void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002596
Patrick Ohlyac45f602009-02-12 05:03:37 +00002597static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2598{
2599 return skb->tstamp;
2600}
2601
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002602/**
2603 * skb_get_timestamp - get timestamp from a skb
2604 * @skb: skb to get stamp from
2605 * @stamp: pointer to struct timeval to store stamp in
2606 *
2607 * Timestamps are stored in the skb as offsets to a base timestamp.
2608 * This function converts the offset back to a struct timeval and stores
2609 * it in stamp.
2610 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00002611static inline void skb_get_timestamp(const struct sk_buff *skb,
2612 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002613{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002614 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002615}
2616
Patrick Ohlyac45f602009-02-12 05:03:37 +00002617static inline void skb_get_timestampns(const struct sk_buff *skb,
2618 struct timespec *stamp)
2619{
2620 *stamp = ktime_to_timespec(skb->tstamp);
2621}
2622
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002623static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002624{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002625 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002626}
2627
Stephen Hemminger164891a2007-04-23 22:26:16 -07002628static inline ktime_t net_timedelta(ktime_t t)
2629{
2630 return ktime_sub(ktime_get_real(), t);
2631}
2632
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07002633static inline ktime_t net_invalid_timestamp(void)
2634{
2635 return ktime_set(0, 0);
2636}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002637
Richard Cochranc1f19b52010-07-17 08:49:36 +00002638#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2639
Joe Perches7965bd42013-09-26 14:48:15 -07002640void skb_clone_tx_timestamp(struct sk_buff *skb);
2641bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00002642
2643#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2644
2645static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2646{
2647}
2648
2649static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2650{
2651 return false;
2652}
2653
2654#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2655
2656/**
2657 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2658 *
Richard Cochranda92b192011-10-21 00:49:15 +00002659 * PHY drivers may accept clones of transmitted packets for
2660 * timestamping via their phy_driver.txtstamp method. These drivers
2661 * must call this function to return the skb back to the stack, with
2662 * or without a timestamp.
2663 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00002664 * @skb: clone of the the original outgoing packet
Richard Cochranda92b192011-10-21 00:49:15 +00002665 * @hwtstamps: hardware time stamps, may be NULL if not available
Richard Cochranc1f19b52010-07-17 08:49:36 +00002666 *
2667 */
2668void skb_complete_tx_timestamp(struct sk_buff *skb,
2669 struct skb_shared_hwtstamps *hwtstamps);
2670
Patrick Ohlyac45f602009-02-12 05:03:37 +00002671/**
2672 * skb_tstamp_tx - queue clone of skb with send time stamps
2673 * @orig_skb: the original outgoing packet
2674 * @hwtstamps: hardware time stamps, may be NULL if not available
2675 *
2676 * If the skb has a socket associated, then this function clones the
2677 * skb (thus sharing the actual data and optional structures), stores
2678 * the optional hardware time stamping information (if non NULL) or
2679 * generates a software time stamp (otherwise), then queues the clone
2680 * to the error queue of the socket. Errors are silently ignored.
2681 */
Joe Perches7965bd42013-09-26 14:48:15 -07002682void skb_tstamp_tx(struct sk_buff *orig_skb,
2683 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00002684
Richard Cochran4507a712010-07-17 08:48:28 +00002685static inline void sw_tx_timestamp(struct sk_buff *skb)
2686{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002687 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2688 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00002689 skb_tstamp_tx(skb, NULL);
2690}
2691
2692/**
2693 * skb_tx_timestamp() - Driver hook for transmit timestamping
2694 *
2695 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00002696 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00002697 *
David S. Miller73409f32013-12-27 13:04:33 -05002698 * Specifically, one should make absolutely sure that this function is
2699 * called before TX completion of this packet can trigger. Otherwise
2700 * the packet could potentially already be freed.
2701 *
Richard Cochran4507a712010-07-17 08:48:28 +00002702 * @skb: A socket buffer.
2703 */
2704static inline void skb_tx_timestamp(struct sk_buff *skb)
2705{
Richard Cochranc1f19b52010-07-17 08:49:36 +00002706 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002707 sw_tx_timestamp(skb);
2708}
2709
Johannes Berg6e3e9392011-11-09 10:15:42 +01002710/**
2711 * skb_complete_wifi_ack - deliver skb with wifi status
2712 *
2713 * @skb: the original outgoing packet
2714 * @acked: ack status
2715 *
2716 */
2717void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2718
Joe Perches7965bd42013-09-26 14:48:15 -07002719__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2720__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002721
Herbert Xu60476372007-04-09 11:59:39 -07002722static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2723{
2724 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2725}
2726
Herbert Xufb286bb2005-11-10 13:01:24 -08002727/**
2728 * skb_checksum_complete - Calculate checksum of an entire packet
2729 * @skb: packet to process
2730 *
2731 * This function calculates the checksum over the entire packet plus
2732 * the value of skb->csum. The latter can be used to supply the
2733 * checksum of a pseudo header as used by TCP/UDP. It returns the
2734 * checksum.
2735 *
2736 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2737 * this function can be used to verify that checksum on received
2738 * packets. In that case the function should return zero if the
2739 * checksum is correct. In particular, this function will return zero
2740 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2741 * hardware has already verified the correctness of the checksum.
2742 */
Al Viro4381ca32007-07-15 21:00:11 +01002743static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002744{
Herbert Xu60476372007-04-09 11:59:39 -07002745 return skb_csum_unnecessary(skb) ?
2746 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002747}
2748
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002749/* Check if we need to perform checksum complete validation.
2750 *
2751 * Returns true if checksum complete is needed, false otherwise
2752 * (either checksum is unnecessary or zero checksum is allowed).
2753 */
2754static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
2755 bool zero_okay,
2756 __sum16 check)
2757{
2758 if (skb_csum_unnecessary(skb)) {
2759 return false;
2760 } else if (zero_okay && !check) {
2761 skb->ip_summed = CHECKSUM_UNNECESSARY;
2762 return false;
2763 }
2764
2765 return true;
2766}
2767
2768/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
2769 * in checksum_init.
2770 */
2771#define CHECKSUM_BREAK 76
2772
2773/* Validate (init) checksum based on checksum complete.
2774 *
2775 * Return values:
2776 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
2777 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
2778 * checksum is stored in skb->csum for use in __skb_checksum_complete
2779 * non-zero: value of invalid checksum
2780 *
2781 */
2782static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
2783 bool complete,
2784 __wsum psum)
2785{
2786 if (skb->ip_summed == CHECKSUM_COMPLETE) {
2787 if (!csum_fold(csum_add(psum, skb->csum))) {
2788 skb->ip_summed = CHECKSUM_UNNECESSARY;
2789 return 0;
2790 }
2791 }
2792
2793 skb->csum = psum;
2794
2795 if (complete || skb->len <= CHECKSUM_BREAK)
2796 return __skb_checksum_complete(skb);
2797
2798 return 0;
2799}
2800
2801static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
2802{
2803 return 0;
2804}
2805
2806/* Perform checksum validate (init). Note that this is a macro since we only
2807 * want to calculate the pseudo header which is an input function if necessary.
2808 * First we try to validate without any computation (checksum unnecessary) and
2809 * then calculate based on checksum complete calling the function to compute
2810 * pseudo header.
2811 *
2812 * Return values:
2813 * 0: checksum is validated or try to in skb_checksum_complete
2814 * non-zero: value of invalid checksum
2815 */
2816#define __skb_checksum_validate(skb, proto, complete, \
2817 zero_okay, check, compute_pseudo) \
2818({ \
2819 __sum16 __ret = 0; \
2820 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
2821 __ret = __skb_checksum_validate_complete(skb, \
2822 complete, compute_pseudo(skb, proto)); \
2823 __ret; \
2824})
2825
2826#define skb_checksum_init(skb, proto, compute_pseudo) \
2827 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
2828
2829#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
2830 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
2831
2832#define skb_checksum_validate(skb, proto, compute_pseudo) \
2833 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
2834
2835#define skb_checksum_validate_zero_check(skb, proto, check, \
2836 compute_pseudo) \
2837 __skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
2838
2839#define skb_checksum_simple_validate(skb) \
2840 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
2841
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002842#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07002843void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002844static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2845{
2846 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002847 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002848}
2849static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2850{
2851 if (nfct)
2852 atomic_inc(&nfct->use);
2853}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002854#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002855#ifdef CONFIG_BRIDGE_NETFILTER
2856static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2857{
2858 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2859 kfree(nf_bridge);
2860}
2861static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2862{
2863 if (nf_bridge)
2864 atomic_inc(&nf_bridge->use);
2865}
2866#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08002867static inline void nf_reset(struct sk_buff *skb)
2868{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002869#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08002870 nf_conntrack_put(skb->nfct);
2871 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002872#endif
Patrick McHardya193a4a2006-03-20 19:23:05 -08002873#ifdef CONFIG_BRIDGE_NETFILTER
2874 nf_bridge_put(skb->nf_bridge);
2875 skb->nf_bridge = NULL;
2876#endif
2877}
2878
Patrick McHardy124dff02013-04-05 20:42:05 +02002879static inline void nf_reset_trace(struct sk_buff *skb)
2880{
Florian Westphal478b3602014-02-15 23:48:45 +01002881#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00002882 skb->nf_trace = 0;
2883#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002884}
2885
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002886/* Note: This doesn't put any conntrack and bridge info in dst. */
2887static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2888{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002889#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002890 dst->nfct = src->nfct;
2891 nf_conntrack_get(src->nfct);
2892 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002893#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002894#ifdef CONFIG_BRIDGE_NETFILTER
2895 dst->nf_bridge = src->nf_bridge;
2896 nf_bridge_get(src->nf_bridge);
2897#endif
Florian Westphal478b3602014-02-15 23:48:45 +01002898#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
2899 dst->nf_trace = src->nf_trace;
2900#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002901}
2902
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002903static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2904{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002905#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002906 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002907#endif
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002908#ifdef CONFIG_BRIDGE_NETFILTER
2909 nf_bridge_put(dst->nf_bridge);
2910#endif
2911 __nf_copy(dst, src);
2912}
2913
James Morris984bc162006-06-09 00:29:17 -07002914#ifdef CONFIG_NETWORK_SECMARK
2915static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2916{
2917 to->secmark = from->secmark;
2918}
2919
2920static inline void skb_init_secmark(struct sk_buff *skb)
2921{
2922 skb->secmark = 0;
2923}
2924#else
2925static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2926{ }
2927
2928static inline void skb_init_secmark(struct sk_buff *skb)
2929{ }
2930#endif
2931
Eric W. Biederman574f7192014-04-01 12:20:24 -07002932static inline bool skb_irq_freeable(const struct sk_buff *skb)
2933{
2934 return !skb->destructor &&
2935#if IS_ENABLED(CONFIG_XFRM)
2936 !skb->sp &&
2937#endif
2938#if IS_ENABLED(CONFIG_NF_CONNTRACK)
2939 !skb->nfct &&
2940#endif
2941 !skb->_skb_refdst &&
2942 !skb_has_frag_list(skb);
2943}
2944
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002945static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2946{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002947 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002948}
2949
Stephen Hemminger92477442009-03-21 13:39:26 -07002950static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002951{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002952 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002953}
2954
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002955static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2956{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002957 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002958}
2959
David S. Millerd5a9e242009-01-27 16:22:11 -08002960static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2961{
2962 skb->queue_mapping = rx_queue + 1;
2963}
2964
Stephen Hemminger92477442009-03-21 13:39:26 -07002965static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002966{
2967 return skb->queue_mapping - 1;
2968}
2969
Stephen Hemminger92477442009-03-21 13:39:26 -07002970static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002971{
Eric Dumazeta02cec22010-09-22 20:43:57 +00002972 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08002973}
2974
Joe Perches7965bd42013-09-26 14:48:15 -07002975u16 __skb_tx_hash(const struct net_device *dev, const struct sk_buff *skb,
2976 unsigned int num_tx_queues);
Stephen Hemminger92477442009-03-21 13:39:26 -07002977
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04002978static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2979{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002980#ifdef CONFIG_XFRM
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002981 return skb->sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002982#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002983 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002984#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04002985}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002986
Pravin B Shelar68c33162013-02-14 14:02:41 +00002987/* Keeps track of mac header offset relative to skb->head.
2988 * It is useful for TSO of Tunneling protocol. e.g. GRE.
2989 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07002990 * tunnel skb it points to outer mac header.
2991 * Keeps track of level of encapsulation of network headers.
2992 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00002993struct skb_gso_cb {
Eric Dumazet3347c962013-10-19 11:42:56 -07002994 int mac_offset;
2995 int encap_level;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07002996 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00002997};
2998#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
2999
3000static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3001{
3002 return (skb_mac_header(inner_skb) - inner_skb->head) -
3003 SKB_GSO_CB(inner_skb)->mac_offset;
3004}
3005
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00003006static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3007{
3008 int new_headroom, headroom;
3009 int ret;
3010
3011 headroom = skb_headroom(skb);
3012 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3013 if (ret)
3014 return ret;
3015
3016 new_headroom = skb_headroom(skb);
3017 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3018 return 0;
3019}
3020
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003021/* Compute the checksum for a gso segment. First compute the checksum value
3022 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3023 * then add in skb->csum (checksum from csum_start to end of packet).
3024 * skb->csum and csum_start are then updated to reflect the checksum of the
3025 * resultant packet starting from the transport header-- the resultant checksum
3026 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3027 * header.
3028 */
3029static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3030{
3031 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3032 skb_transport_offset(skb);
3033 __u16 csum;
3034
3035 csum = csum_fold(csum_partial(skb_transport_header(skb),
3036 plen, skb->csum));
3037 skb->csum = res;
3038 SKB_GSO_CB(skb)->csum_start -= plen;
3039
3040 return csum;
3041}
3042
David S. Millerbdcc0922012-03-07 20:53:36 -05003043static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07003044{
3045 return skb_shinfo(skb)->gso_size;
3046}
3047
Eric Dumazet36a8f392013-09-29 01:21:32 -07003048/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05003049static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02003050{
3051 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3052}
3053
Joe Perches7965bd42013-09-26 14:48:15 -07003054void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07003055
3056static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3057{
3058 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3059 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003060 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3061
Alexander Duyckb78462e2010-06-02 12:24:37 +00003062 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3063 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07003064 __skb_warn_lro_forwarding(skb);
3065 return true;
3066 }
3067 return false;
3068}
3069
Herbert Xu35fc92a2007-03-26 23:22:20 -07003070static inline void skb_forward_csum(struct sk_buff *skb)
3071{
3072 /* Unfortunately we don't support this one. Any brave souls? */
3073 if (skb->ip_summed == CHECKSUM_COMPLETE)
3074 skb->ip_summed = CHECKSUM_NONE;
3075}
3076
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003077/**
3078 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3079 * @skb: skb to check
3080 *
3081 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3082 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3083 * use this helper, to document places where we make this assertion.
3084 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003085static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003086{
3087#ifdef DEBUG
3088 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3089#endif
3090}
3091
Rusty Russellf35d9d82008-02-04 23:49:54 -05003092bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00003093
Paul Durranted1f50c2014-01-09 10:02:46 +00003094int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3095
Daniel Borkmannf77668d2013-03-19 06:39:30 +00003096u32 __skb_get_poff(const struct sk_buff *skb);
3097
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00003098/**
3099 * skb_head_is_locked - Determine if the skb->head is locked down
3100 * @skb: skb to check
3101 *
3102 * The head on skbs build around a head frag can be removed if they are
3103 * not cloned. This function returns true if the skb head is locked down
3104 * due to either being allocated via kmalloc, or by being a clone with
3105 * multiple references to the head.
3106 */
3107static inline bool skb_head_is_locked(const struct sk_buff *skb)
3108{
3109 return !skb->head_frag || skb_cloned(skb);
3110}
Florian Westphalfe6cc552014-02-13 23:09:12 +01003111
3112/**
3113 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3114 *
3115 * @skb: GSO skb
3116 *
3117 * skb_gso_network_seglen is used to determine the real size of the
3118 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3119 *
3120 * The MAC/L2 header is not accounted for.
3121 */
3122static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3123{
3124 unsigned int hdr_len = skb_transport_header(skb) -
3125 skb_network_header(skb);
3126 return hdr_len + skb_gso_transport_seglen(skb);
3127}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003128#endif /* __KERNEL__ */
3129#endif /* _LINUX_SKBUFF_H */