blob: 1f50bfe2243dcd993c0ecc06d17cc5ed371c67c4 [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 McHardy6aa895b2008-07-14 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);
Octavian Purdilabad93e92014-06-12 01:36:26 +0300747struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
748 gfp_t gfp_mask, bool fclone);
749static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
750 gfp_t gfp_mask)
751{
752 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
753}
Eric Dumazet117632e2011-12-03 21:39:53 +0000754
Joe Perches7965bd42013-09-26 14:48:15 -0700755int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
756struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
757 unsigned int headroom);
758struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
759 int newtailroom, gfp_t priority);
Fan Du25a91d82014-01-18 09:54:23 +0800760int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
761 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -0700762int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
763 int len);
764int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
765int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000766#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700767
Joe Perches7965bd42013-09-26 14:48:15 -0700768int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
769 int getfrag(void *from, char *to, int offset,
770 int len, int odd, struct sk_buff *skb),
771 void *from, int length);
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700772
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800773struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700774 __u32 lower_offset;
775 __u32 upper_offset;
776 __u32 frag_idx;
777 __u32 stepped_offset;
778 struct sk_buff *root_skb;
779 struct sk_buff *cur_skb;
780 __u8 *frag_data;
781};
782
Joe Perches7965bd42013-09-26 14:48:15 -0700783void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
784 unsigned int to, struct skb_seq_state *st);
785unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
786 struct skb_seq_state *st);
787void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -0700788
Joe Perches7965bd42013-09-26 14:48:15 -0700789unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
790 unsigned int to, struct ts_config *config,
791 struct ts_state *state);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700792
Tom Herbert09323cc2013-12-15 22:16:19 -0800793/*
794 * Packet hash types specify the type of hash in skb_set_hash.
795 *
796 * Hash types refer to the protocol layer addresses which are used to
797 * construct a packet's hash. The hashes are used to differentiate or identify
798 * flows of the protocol layer for the hash type. Hash types are either
799 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
800 *
801 * Properties of hashes:
802 *
803 * 1) Two packets in different flows have different hash values
804 * 2) Two packets in the same flow should have the same hash value
805 *
806 * A hash at a higher layer is considered to be more specific. A driver should
807 * set the most specific hash possible.
808 *
809 * A driver cannot indicate a more specific hash than the layer at which a hash
810 * was computed. For instance an L3 hash cannot be set as an L4 hash.
811 *
812 * A driver may indicate a hash level which is less specific than the
813 * actual layer the hash was computed on. For instance, a hash computed
814 * at L4 may be considered an L3 hash. This should only be done if the
815 * driver can't unambiguously determine that the HW computed the hash at
816 * the higher layer. Note that the "should" in the second property above
817 * permits this.
818 */
819enum pkt_hash_types {
820 PKT_HASH_TYPE_NONE, /* Undefined type */
821 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
822 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
823 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
824};
825
826static inline void
827skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
828{
Tom Herbert61b905d2014-03-24 15:34:47 -0700829 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
830 skb->hash = hash;
Tom Herbert09323cc2013-12-15 22:16:19 -0800831}
832
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800833void __skb_get_hash(struct sk_buff *skb);
834static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000835{
Tom Herbert61b905d2014-03-24 15:34:47 -0700836 if (!skb->l4_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800837 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000838
Tom Herbert61b905d2014-03-24 15:34:47 -0700839 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000840}
841
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800842static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
843{
Tom Herbert61b905d2014-03-24 15:34:47 -0700844 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800845}
846
Tom Herbert7539fad2013-12-15 22:12:18 -0800847static inline void skb_clear_hash(struct sk_buff *skb)
848{
Tom Herbert61b905d2014-03-24 15:34:47 -0700849 skb->hash = 0;
850 skb->l4_hash = 0;
Tom Herbert7539fad2013-12-15 22:12:18 -0800851}
852
853static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
854{
Tom Herbert61b905d2014-03-24 15:34:47 -0700855 if (!skb->l4_hash)
Tom Herbert7539fad2013-12-15 22:12:18 -0800856 skb_clear_hash(skb);
857}
858
Tom Herbert3df7a742013-12-15 22:16:29 -0800859static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
860{
Tom Herbert61b905d2014-03-24 15:34:47 -0700861 to->hash = from->hash;
862 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -0800863};
864
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700865#ifdef NET_SKBUFF_DATA_USES_OFFSET
866static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
867{
868 return skb->head + skb->end;
869}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000870
871static inline unsigned int skb_end_offset(const struct sk_buff *skb)
872{
873 return skb->end;
874}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700875#else
876static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
877{
878 return skb->end;
879}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000880
881static inline unsigned int skb_end_offset(const struct sk_buff *skb)
882{
883 return skb->end - skb->head;
884}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700885#endif
886
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700888#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889
Patrick Ohlyac45f602009-02-12 05:03:37 +0000890static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
891{
892 return &skb_shinfo(skb)->hwtstamps;
893}
894
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895/**
896 * skb_queue_empty - check if a queue is empty
897 * @list: queue head
898 *
899 * Returns true if the queue is empty, false otherwise.
900 */
901static inline int skb_queue_empty(const struct sk_buff_head *list)
902{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100903 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700904}
905
906/**
David S. Millerfc7ebb22008-09-23 00:34:07 -0700907 * skb_queue_is_last - check if skb is the last entry in the queue
908 * @list: queue head
909 * @skb: buffer
910 *
911 * Returns true if @skb is the last buffer on the list.
912 */
913static inline bool skb_queue_is_last(const struct sk_buff_head *list,
914 const struct sk_buff *skb)
915{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100916 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -0700917}
918
919/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800920 * skb_queue_is_first - check if skb is the first entry in the queue
921 * @list: queue head
922 * @skb: buffer
923 *
924 * Returns true if @skb is the first buffer on the list.
925 */
926static inline bool skb_queue_is_first(const struct sk_buff_head *list,
927 const struct sk_buff *skb)
928{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100929 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800930}
931
932/**
David S. Miller249c8b42008-09-23 00:44:42 -0700933 * skb_queue_next - return the next packet in the queue
934 * @list: queue head
935 * @skb: current buffer
936 *
937 * Return the next packet in @list after @skb. It is only valid to
938 * call this if skb_queue_is_last() evaluates to false.
939 */
940static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
941 const struct sk_buff *skb)
942{
943 /* This BUG_ON may seem severe, but if we just return then we
944 * are going to dereference garbage.
945 */
946 BUG_ON(skb_queue_is_last(list, skb));
947 return skb->next;
948}
949
950/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800951 * skb_queue_prev - return the prev packet in the queue
952 * @list: queue head
953 * @skb: current buffer
954 *
955 * Return the prev packet in @list before @skb. It is only valid to
956 * call this if skb_queue_is_first() evaluates to false.
957 */
958static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
959 const struct sk_buff *skb)
960{
961 /* This BUG_ON may seem severe, but if we just return then we
962 * are going to dereference garbage.
963 */
964 BUG_ON(skb_queue_is_first(list, skb));
965 return skb->prev;
966}
967
968/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700969 * skb_get - reference buffer
970 * @skb: buffer to reference
971 *
972 * Makes another reference to a socket buffer and returns a pointer
973 * to the buffer.
974 */
975static inline struct sk_buff *skb_get(struct sk_buff *skb)
976{
977 atomic_inc(&skb->users);
978 return skb;
979}
980
981/*
982 * If users == 1, we are the only owner and are can avoid redundant
983 * atomic change.
984 */
985
986/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700987 * skb_cloned - is the buffer a clone
988 * @skb: buffer to check
989 *
990 * Returns true if the buffer was generated with skb_clone() and is
991 * one of multiple shared copies of the buffer. Cloned buffers are
992 * shared data so must not be written to under normal circumstances.
993 */
994static inline int skb_cloned(const struct sk_buff *skb)
995{
996 return skb->cloned &&
997 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
998}
999
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001000static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1001{
1002 might_sleep_if(pri & __GFP_WAIT);
1003
1004 if (skb_cloned(skb))
1005 return pskb_expand_head(skb, 0, 0, pri);
1006
1007 return 0;
1008}
1009
Linus Torvalds1da177e2005-04-16 15:20:36 -07001010/**
1011 * skb_header_cloned - is the header a clone
1012 * @skb: buffer to check
1013 *
1014 * Returns true if modifying the header part of the buffer requires
1015 * the data to be copied.
1016 */
1017static inline int skb_header_cloned(const struct sk_buff *skb)
1018{
1019 int dataref;
1020
1021 if (!skb->cloned)
1022 return 0;
1023
1024 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1025 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1026 return dataref != 1;
1027}
1028
1029/**
1030 * skb_header_release - release reference to header
1031 * @skb: buffer to operate on
1032 *
1033 * Drop a reference to the header part of the buffer. This is done
1034 * by acquiring a payload reference. You must not read from the header
1035 * part of skb->data after this.
1036 */
1037static inline void skb_header_release(struct sk_buff *skb)
1038{
1039 BUG_ON(skb->nohdr);
1040 skb->nohdr = 1;
1041 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1042}
1043
1044/**
1045 * skb_shared - is the buffer shared
1046 * @skb: buffer to check
1047 *
1048 * Returns true if more than one person has a reference to this
1049 * buffer.
1050 */
1051static inline int skb_shared(const struct sk_buff *skb)
1052{
1053 return atomic_read(&skb->users) != 1;
1054}
1055
1056/**
1057 * skb_share_check - check if buffer is shared and if so clone it
1058 * @skb: buffer to check
1059 * @pri: priority for memory allocation
1060 *
1061 * If the buffer is shared the buffer is cloned and the old copy
1062 * drops a reference. A new clone with a single reference is returned.
1063 * If the buffer is not shared the original buffer is returned. When
1064 * being called from interrupt status or with spinlocks held pri must
1065 * be GFP_ATOMIC.
1066 *
1067 * NULL is returned on a memory allocation failure.
1068 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001069static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001070{
1071 might_sleep_if(pri & __GFP_WAIT);
1072 if (skb_shared(skb)) {
1073 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001074
1075 if (likely(nskb))
1076 consume_skb(skb);
1077 else
1078 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001079 skb = nskb;
1080 }
1081 return skb;
1082}
1083
1084/*
1085 * Copy shared buffers into a new sk_buff. We effectively do COW on
1086 * packets to handle cases where we have a local reader and forward
1087 * and a couple of other messy ones. The normal one is tcpdumping
1088 * a packet thats being forwarded.
1089 */
1090
1091/**
1092 * skb_unshare - make a copy of a shared buffer
1093 * @skb: buffer to check
1094 * @pri: priority for memory allocation
1095 *
1096 * If the socket buffer is a clone then this function creates a new
1097 * copy of the data, drops a reference count on the old copy and returns
1098 * the new copy with the reference count at 1. If the buffer is not a clone
1099 * the original buffer is returned. When called with a spinlock held or
1100 * from interrupt state @pri must be %GFP_ATOMIC
1101 *
1102 * %NULL is returned on a memory allocation failure.
1103 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001104static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001105 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001106{
1107 might_sleep_if(pri & __GFP_WAIT);
1108 if (skb_cloned(skb)) {
1109 struct sk_buff *nskb = skb_copy(skb, pri);
1110 kfree_skb(skb); /* Free our shared copy */
1111 skb = nskb;
1112 }
1113 return skb;
1114}
1115
1116/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001117 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001118 * @list_: list to peek at
1119 *
1120 * Peek an &sk_buff. Unlike most other operations you _MUST_
1121 * be careful with this one. A peek leaves the buffer on the
1122 * list and someone else may run off with it. You must hold
1123 * the appropriate locks or have a private queue to do this.
1124 *
1125 * Returns %NULL for an empty list or a pointer to the head element.
1126 * The reference count is not incremented and the reference is therefore
1127 * volatile. Use with caution.
1128 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001129static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001130{
Eric Dumazet18d07002012-04-30 16:31:46 +00001131 struct sk_buff *skb = list_->next;
1132
1133 if (skb == (struct sk_buff *)list_)
1134 skb = NULL;
1135 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001136}
1137
1138/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001139 * skb_peek_next - peek skb following the given one from a queue
1140 * @skb: skb to start from
1141 * @list_: list to peek at
1142 *
1143 * Returns %NULL when the end of the list is met or a pointer to the
1144 * next element. The reference count is not incremented and the
1145 * reference is therefore volatile. Use with caution.
1146 */
1147static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1148 const struct sk_buff_head *list_)
1149{
1150 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001151
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001152 if (next == (struct sk_buff *)list_)
1153 next = NULL;
1154 return next;
1155}
1156
1157/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001158 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001159 * @list_: list to peek at
1160 *
1161 * Peek an &sk_buff. Unlike most other operations you _MUST_
1162 * be careful with this one. A peek leaves the buffer on the
1163 * list and someone else may run off with it. You must hold
1164 * the appropriate locks or have a private queue to do this.
1165 *
1166 * Returns %NULL for an empty list or a pointer to the tail element.
1167 * The reference count is not incremented and the reference is therefore
1168 * volatile. Use with caution.
1169 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001170static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001171{
Eric Dumazet18d07002012-04-30 16:31:46 +00001172 struct sk_buff *skb = list_->prev;
1173
1174 if (skb == (struct sk_buff *)list_)
1175 skb = NULL;
1176 return skb;
1177
Linus Torvalds1da177e2005-04-16 15:20:36 -07001178}
1179
1180/**
1181 * skb_queue_len - get queue length
1182 * @list_: list to measure
1183 *
1184 * Return the length of an &sk_buff queue.
1185 */
1186static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1187{
1188 return list_->qlen;
1189}
1190
David S. Miller67fed452008-09-21 22:36:24 -07001191/**
1192 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1193 * @list: queue to initialize
1194 *
1195 * This initializes only the list and queue length aspects of
1196 * an sk_buff_head object. This allows to initialize the list
1197 * aspects of an sk_buff_head without reinitializing things like
1198 * the spinlock. It can also be used for on-stack sk_buff_head
1199 * objects where the spinlock is known to not be used.
1200 */
1201static inline void __skb_queue_head_init(struct sk_buff_head *list)
1202{
1203 list->prev = list->next = (struct sk_buff *)list;
1204 list->qlen = 0;
1205}
1206
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001207/*
1208 * This function creates a split out lock class for each invocation;
1209 * this is needed for now since a whole lot of users of the skb-queue
1210 * infrastructure in drivers have different locking usage (in hardirq)
1211 * than the networking core (in softirq only). In the long run either the
1212 * network layer or drivers should need annotation to consolidate the
1213 * main types of usage into 3 classes.
1214 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001215static inline void skb_queue_head_init(struct sk_buff_head *list)
1216{
1217 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001218 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001219}
1220
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001221static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1222 struct lock_class_key *class)
1223{
1224 skb_queue_head_init(list);
1225 lockdep_set_class(&list->lock, class);
1226}
1227
Linus Torvalds1da177e2005-04-16 15:20:36 -07001228/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001229 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230 *
1231 * The "__skb_xxxx()" functions are the non-atomic ones that
1232 * can only be called with interrupts disabled.
1233 */
Joe Perches7965bd42013-09-26 14:48:15 -07001234void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1235 struct sk_buff_head *list);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001236static inline void __skb_insert(struct sk_buff *newsk,
1237 struct sk_buff *prev, struct sk_buff *next,
1238 struct sk_buff_head *list)
1239{
1240 newsk->next = next;
1241 newsk->prev = prev;
1242 next->prev = prev->next = newsk;
1243 list->qlen++;
1244}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001245
David S. Miller67fed452008-09-21 22:36:24 -07001246static inline void __skb_queue_splice(const struct sk_buff_head *list,
1247 struct sk_buff *prev,
1248 struct sk_buff *next)
1249{
1250 struct sk_buff *first = list->next;
1251 struct sk_buff *last = list->prev;
1252
1253 first->prev = prev;
1254 prev->next = first;
1255
1256 last->next = next;
1257 next->prev = last;
1258}
1259
1260/**
1261 * skb_queue_splice - join two skb lists, this is designed for stacks
1262 * @list: the new list to add
1263 * @head: the place to add it in the first list
1264 */
1265static inline void skb_queue_splice(const struct sk_buff_head *list,
1266 struct sk_buff_head *head)
1267{
1268 if (!skb_queue_empty(list)) {
1269 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001270 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001271 }
1272}
1273
1274/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001275 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001276 * @list: the new list to add
1277 * @head: the place to add it in the first list
1278 *
1279 * The list at @list is reinitialised
1280 */
1281static inline void skb_queue_splice_init(struct sk_buff_head *list,
1282 struct sk_buff_head *head)
1283{
1284 if (!skb_queue_empty(list)) {
1285 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001286 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001287 __skb_queue_head_init(list);
1288 }
1289}
1290
1291/**
1292 * skb_queue_splice_tail - join two skb lists, each list being a queue
1293 * @list: the new list to add
1294 * @head: the place to add it in the first list
1295 */
1296static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1297 struct sk_buff_head *head)
1298{
1299 if (!skb_queue_empty(list)) {
1300 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001301 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001302 }
1303}
1304
1305/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001306 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001307 * @list: the new list to add
1308 * @head: the place to add it in the first list
1309 *
1310 * Each of the lists is a queue.
1311 * The list at @list is reinitialised
1312 */
1313static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1314 struct sk_buff_head *head)
1315{
1316 if (!skb_queue_empty(list)) {
1317 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001318 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001319 __skb_queue_head_init(list);
1320 }
1321}
1322
Linus Torvalds1da177e2005-04-16 15:20:36 -07001323/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001324 * __skb_queue_after - queue a buffer at the list head
1325 * @list: list to use
1326 * @prev: place after this buffer
1327 * @newsk: buffer to queue
1328 *
1329 * Queue a buffer int the middle of a list. This function takes no locks
1330 * and you must therefore hold required locks before calling it.
1331 *
1332 * A buffer cannot be placed on two lists at the same time.
1333 */
1334static inline void __skb_queue_after(struct sk_buff_head *list,
1335 struct sk_buff *prev,
1336 struct sk_buff *newsk)
1337{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001338 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001339}
1340
Joe Perches7965bd42013-09-26 14:48:15 -07001341void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1342 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001343
Gerrit Renkerf5572852008-04-14 00:05:28 -07001344static inline void __skb_queue_before(struct sk_buff_head *list,
1345 struct sk_buff *next,
1346 struct sk_buff *newsk)
1347{
1348 __skb_insert(newsk, next->prev, next, list);
1349}
1350
Stephen Hemminger300ce172005-10-30 13:47:34 -08001351/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352 * __skb_queue_head - queue a buffer at the list head
1353 * @list: list to use
1354 * @newsk: buffer to queue
1355 *
1356 * Queue a buffer at the start of a list. This function takes no locks
1357 * and you must therefore hold required locks before calling it.
1358 *
1359 * A buffer cannot be placed on two lists at the same time.
1360 */
Joe Perches7965bd42013-09-26 14:48:15 -07001361void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001362static inline void __skb_queue_head(struct sk_buff_head *list,
1363 struct sk_buff *newsk)
1364{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001365 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001366}
1367
1368/**
1369 * __skb_queue_tail - queue a buffer at the list tail
1370 * @list: list to use
1371 * @newsk: buffer to queue
1372 *
1373 * Queue a buffer at the end of a list. This function takes no locks
1374 * and you must therefore hold required locks before calling it.
1375 *
1376 * A buffer cannot be placed on two lists at the same time.
1377 */
Joe Perches7965bd42013-09-26 14:48:15 -07001378void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379static inline void __skb_queue_tail(struct sk_buff_head *list,
1380 struct sk_buff *newsk)
1381{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001382 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001383}
1384
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001386 * remove sk_buff from list. _Must_ be called atomically, and with
1387 * the list known..
1388 */
Joe Perches7965bd42013-09-26 14:48:15 -07001389void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001390static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1391{
1392 struct sk_buff *next, *prev;
1393
1394 list->qlen--;
1395 next = skb->next;
1396 prev = skb->prev;
1397 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398 next->prev = prev;
1399 prev->next = next;
1400}
1401
Gerrit Renkerf525c062008-04-14 00:04:12 -07001402/**
1403 * __skb_dequeue - remove from the head of the queue
1404 * @list: list to dequeue from
1405 *
1406 * Remove the head of the list. This function does not take any locks
1407 * so must be used with appropriate locks held only. The head item is
1408 * returned or %NULL if the list is empty.
1409 */
Joe Perches7965bd42013-09-26 14:48:15 -07001410struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Gerrit Renkerf525c062008-04-14 00:04:12 -07001411static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1412{
1413 struct sk_buff *skb = skb_peek(list);
1414 if (skb)
1415 __skb_unlink(skb, list);
1416 return skb;
1417}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001418
1419/**
1420 * __skb_dequeue_tail - remove from the tail of the queue
1421 * @list: list to dequeue from
1422 *
1423 * Remove the tail of the list. This function does not take any locks
1424 * so must be used with appropriate locks held only. The tail item is
1425 * returned or %NULL if the list is empty.
1426 */
Joe Perches7965bd42013-09-26 14:48:15 -07001427struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1429{
1430 struct sk_buff *skb = skb_peek_tail(list);
1431 if (skb)
1432 __skb_unlink(skb, list);
1433 return skb;
1434}
1435
1436
David S. Millerbdcc0922012-03-07 20:53:36 -05001437static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001438{
1439 return skb->data_len;
1440}
1441
1442static inline unsigned int skb_headlen(const struct sk_buff *skb)
1443{
1444 return skb->len - skb->data_len;
1445}
1446
1447static inline int skb_pagelen(const struct sk_buff *skb)
1448{
1449 int i, len = 0;
1450
1451 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001452 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001453 return len + skb_headlen(skb);
1454}
1455
Ian Campbell131ea662011-08-19 06:25:00 +00001456/**
1457 * __skb_fill_page_desc - initialise a paged fragment in an skb
1458 * @skb: buffer containing fragment to be initialised
1459 * @i: paged fragment index to initialise
1460 * @page: the page to use for this fragment
1461 * @off: the offset to the data with @page
1462 * @size: the length of the data
1463 *
1464 * Initialises the @i'th fragment of @skb to point to &size bytes at
1465 * offset @off within @page.
1466 *
1467 * Does not take any additional reference on the fragment.
1468 */
1469static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1470 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001471{
1472 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1473
Mel Gormanc48a11c2012-07-31 16:44:23 -07001474 /*
1475 * Propagate page->pfmemalloc to the skb if we can. The problem is
1476 * that not all callers have unique ownership of the page. If
1477 * pfmemalloc is set, we check the mapping as a mapping implies
1478 * page->index is set (index and pfmemalloc share space).
1479 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1480 * do not lose pfmemalloc information as the pages would not be
1481 * allocated using __GFP_MEMALLOC.
1482 */
Ian Campbella8605c62011-10-19 23:01:49 +00001483 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001484 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001485 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001486
1487 page = compound_head(page);
1488 if (page->pfmemalloc && !page->mapping)
1489 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00001490}
1491
1492/**
1493 * skb_fill_page_desc - initialise a paged fragment in an skb
1494 * @skb: buffer containing fragment to be initialised
1495 * @i: paged fragment index to initialise
1496 * @page: the page to use for this fragment
1497 * @off: the offset to the data with @page
1498 * @size: the length of the data
1499 *
1500 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01001501 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00001502 * addition updates @skb such that @i is the last fragment.
1503 *
1504 * Does not take any additional reference on the fragment.
1505 */
1506static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1507 struct page *page, int off, int size)
1508{
1509 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001510 skb_shinfo(skb)->nr_frags = i + 1;
1511}
1512
Joe Perches7965bd42013-09-26 14:48:15 -07001513void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1514 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001515
Jason Wangf8e617e2013-11-01 14:07:47 +08001516void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1517 unsigned int truesize);
1518
Linus Torvalds1da177e2005-04-16 15:20:36 -07001519#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001520#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001521#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1522
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001523#ifdef NET_SKBUFF_DATA_USES_OFFSET
1524static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1525{
1526 return skb->head + skb->tail;
1527}
1528
1529static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1530{
1531 skb->tail = skb->data - skb->head;
1532}
1533
1534static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1535{
1536 skb_reset_tail_pointer(skb);
1537 skb->tail += offset;
1538}
Simon Horman7cc46192013-05-28 20:34:29 +00001539
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001540#else /* NET_SKBUFF_DATA_USES_OFFSET */
1541static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1542{
1543 return skb->tail;
1544}
1545
1546static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1547{
1548 skb->tail = skb->data;
1549}
1550
1551static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1552{
1553 skb->tail = skb->data + offset;
1554}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001555
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001556#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1557
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558/*
1559 * Add data to an sk_buff
1560 */
Mathias Krause0c7ddf32013-11-07 14:18:24 +01001561unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001562unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001563static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1564{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001565 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001566 SKB_LINEAR_ASSERT(skb);
1567 skb->tail += len;
1568 skb->len += len;
1569 return tmp;
1570}
1571
Joe Perches7965bd42013-09-26 14:48:15 -07001572unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001573static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1574{
1575 skb->data -= len;
1576 skb->len += len;
1577 return skb->data;
1578}
1579
Joe Perches7965bd42013-09-26 14:48:15 -07001580unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001581static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1582{
1583 skb->len -= len;
1584 BUG_ON(skb->len < skb->data_len);
1585 return skb->data += len;
1586}
1587
David S. Miller47d29642010-05-02 02:21:44 -07001588static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1589{
1590 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1591}
1592
Joe Perches7965bd42013-09-26 14:48:15 -07001593unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001594
1595static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1596{
1597 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001598 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001599 return NULL;
1600 skb->len -= len;
1601 return skb->data += len;
1602}
1603
1604static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1605{
1606 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1607}
1608
1609static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1610{
1611 if (likely(len <= skb_headlen(skb)))
1612 return 1;
1613 if (unlikely(len > skb->len))
1614 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001615 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001616}
1617
1618/**
1619 * skb_headroom - bytes at buffer head
1620 * @skb: buffer to check
1621 *
1622 * Return the number of bytes of free space at the head of an &sk_buff.
1623 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001624static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001625{
1626 return skb->data - skb->head;
1627}
1628
1629/**
1630 * skb_tailroom - bytes at buffer end
1631 * @skb: buffer to check
1632 *
1633 * Return the number of bytes of free space at the tail of an sk_buff
1634 */
1635static inline int skb_tailroom(const struct sk_buff *skb)
1636{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001637 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001638}
1639
1640/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00001641 * skb_availroom - bytes at buffer end
1642 * @skb: buffer to check
1643 *
1644 * Return the number of bytes of free space at the tail of an sk_buff
1645 * allocated by sk_stream_alloc()
1646 */
1647static inline int skb_availroom(const struct sk_buff *skb)
1648{
Eric Dumazet16fad692013-03-14 05:40:32 +00001649 if (skb_is_nonlinear(skb))
1650 return 0;
1651
1652 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00001653}
1654
1655/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001656 * skb_reserve - adjust headroom
1657 * @skb: buffer to alter
1658 * @len: bytes to move
1659 *
1660 * Increase the headroom of an empty &sk_buff by reducing the tail
1661 * room. This is only allowed for an empty buffer.
1662 */
David S. Miller8243126c2006-01-17 02:54:21 -08001663static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001664{
1665 skb->data += len;
1666 skb->tail += len;
1667}
1668
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001669static inline void skb_reset_inner_headers(struct sk_buff *skb)
1670{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001671 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001672 skb->inner_network_header = skb->network_header;
1673 skb->inner_transport_header = skb->transport_header;
1674}
1675
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00001676static inline void skb_reset_mac_len(struct sk_buff *skb)
1677{
1678 skb->mac_len = skb->network_header - skb->mac_header;
1679}
1680
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001681static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1682 *skb)
1683{
1684 return skb->head + skb->inner_transport_header;
1685}
1686
1687static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1688{
1689 skb->inner_transport_header = skb->data - skb->head;
1690}
1691
1692static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1693 const int offset)
1694{
1695 skb_reset_inner_transport_header(skb);
1696 skb->inner_transport_header += offset;
1697}
1698
1699static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1700{
1701 return skb->head + skb->inner_network_header;
1702}
1703
1704static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1705{
1706 skb->inner_network_header = skb->data - skb->head;
1707}
1708
1709static inline void skb_set_inner_network_header(struct sk_buff *skb,
1710 const int offset)
1711{
1712 skb_reset_inner_network_header(skb);
1713 skb->inner_network_header += offset;
1714}
1715
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001716static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1717{
1718 return skb->head + skb->inner_mac_header;
1719}
1720
1721static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1722{
1723 skb->inner_mac_header = skb->data - skb->head;
1724}
1725
1726static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1727 const int offset)
1728{
1729 skb_reset_inner_mac_header(skb);
1730 skb->inner_mac_header += offset;
1731}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001732static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1733{
Cong Wang35d04612013-05-29 15:16:05 +08001734 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001735}
1736
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001737static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1738{
1739 return skb->head + skb->transport_header;
1740}
1741
1742static inline void skb_reset_transport_header(struct sk_buff *skb)
1743{
1744 skb->transport_header = skb->data - skb->head;
1745}
1746
1747static inline void skb_set_transport_header(struct sk_buff *skb,
1748 const int offset)
1749{
1750 skb_reset_transport_header(skb);
1751 skb->transport_header += offset;
1752}
1753
1754static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1755{
1756 return skb->head + skb->network_header;
1757}
1758
1759static inline void skb_reset_network_header(struct sk_buff *skb)
1760{
1761 skb->network_header = skb->data - skb->head;
1762}
1763
1764static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1765{
1766 skb_reset_network_header(skb);
1767 skb->network_header += offset;
1768}
1769
1770static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1771{
1772 return skb->head + skb->mac_header;
1773}
1774
1775static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1776{
Cong Wang35d04612013-05-29 15:16:05 +08001777 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001778}
1779
1780static inline void skb_reset_mac_header(struct sk_buff *skb)
1781{
1782 skb->mac_header = skb->data - skb->head;
1783}
1784
1785static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1786{
1787 skb_reset_mac_header(skb);
1788 skb->mac_header += offset;
1789}
1790
Timo Teräs0e3da5b2013-12-16 11:02:09 +02001791static inline void skb_pop_mac_header(struct sk_buff *skb)
1792{
1793 skb->mac_header = skb->network_header;
1794}
1795
Ying Xuefbbdb8f2013-03-27 16:46:06 +00001796static inline void skb_probe_transport_header(struct sk_buff *skb,
1797 const int offset_hint)
1798{
1799 struct flow_keys keys;
1800
1801 if (skb_transport_header_was_set(skb))
1802 return;
1803 else if (skb_flow_dissect(skb, &keys))
1804 skb_set_transport_header(skb, keys.thoff);
1805 else
1806 skb_set_transport_header(skb, offset_hint);
1807}
1808
Eric Dumazet03606892012-02-23 10:55:02 +00001809static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1810{
1811 if (skb_mac_header_was_set(skb)) {
1812 const unsigned char *old_mac = skb_mac_header(skb);
1813
1814 skb_set_mac_header(skb, -skb->mac_len);
1815 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1816 }
1817}
1818
Michał Mirosław04fb4512010-12-14 15:24:08 +00001819static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1820{
1821 return skb->csum_start - skb_headroom(skb);
1822}
1823
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001824static inline int skb_transport_offset(const struct sk_buff *skb)
1825{
1826 return skb_transport_header(skb) - skb->data;
1827}
1828
1829static inline u32 skb_network_header_len(const struct sk_buff *skb)
1830{
1831 return skb->transport_header - skb->network_header;
1832}
1833
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001834static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1835{
1836 return skb->inner_transport_header - skb->inner_network_header;
1837}
1838
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001839static inline int skb_network_offset(const struct sk_buff *skb)
1840{
1841 return skb_network_header(skb) - skb->data;
1842}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001843
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001844static inline int skb_inner_network_offset(const struct sk_buff *skb)
1845{
1846 return skb_inner_network_header(skb) - skb->data;
1847}
1848
Changli Gaof9599ce2010-08-04 04:43:44 +00001849static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1850{
1851 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1852}
1853
Linus Torvalds1da177e2005-04-16 15:20:36 -07001854/*
1855 * CPUs often take a performance hit when accessing unaligned memory
1856 * locations. The actual performance hit varies, it can be small if the
1857 * hardware handles it or large if we have to take an exception and fix it
1858 * in software.
1859 *
1860 * Since an ethernet header is 14 bytes network drivers often end up with
1861 * the IP header at an unaligned offset. The IP header can be aligned by
1862 * shifting the start of the packet by 2 bytes. Drivers should do this
1863 * with:
1864 *
Tobias Klauser8660c122009-07-13 22:48:16 +00001865 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001866 *
1867 * The downside to this alignment of the IP header is that the DMA is now
1868 * unaligned. On some architectures the cost of an unaligned DMA is high
1869 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00001870 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001871 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1872 * to be overridden.
1873 */
1874#ifndef NET_IP_ALIGN
1875#define NET_IP_ALIGN 2
1876#endif
1877
Anton Blanchard025be812006-03-31 02:27:06 -08001878/*
1879 * The networking layer reserves some headroom in skb data (via
1880 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1881 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08001882 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08001883 *
1884 * Unfortunately this headroom changes the DMA alignment of the resulting
1885 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1886 * on some architectures. An architecture can override this value,
1887 * perhaps setting it to a cacheline in size (since that will maintain
1888 * cacheline alignment of the DMA). It must be a power of 2.
1889 *
David S. Millerd6301d32009-02-08 19:24:13 -08001890 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08001891 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07001892 *
1893 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1894 * to reduce average number of cache lines per packet.
1895 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07001896 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08001897 */
1898#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07001899#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08001900#endif
1901
Joe Perches7965bd42013-09-26 14:48:15 -07001902int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001903
1904static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1905{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00001906 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07001907 WARN_ON(1);
1908 return;
1909 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001910 skb->len = len;
1911 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001912}
1913
Joe Perches7965bd42013-09-26 14:48:15 -07001914void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001915
1916static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1917{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001918 if (skb->data_len)
1919 return ___pskb_trim(skb, len);
1920 __skb_trim(skb, len);
1921 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001922}
1923
1924static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1925{
1926 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1927}
1928
1929/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07001930 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1931 * @skb: buffer to alter
1932 * @len: new length
1933 *
1934 * This is identical to pskb_trim except that the caller knows that
1935 * the skb is not cloned so we should never get an error due to out-
1936 * of-memory.
1937 */
1938static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1939{
1940 int err = pskb_trim(skb, len);
1941 BUG_ON(err);
1942}
1943
1944/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001945 * skb_orphan - orphan a buffer
1946 * @skb: buffer to orphan
1947 *
1948 * If a buffer currently has an owner then we call the owner's
1949 * destructor function and make the @skb unowned. The buffer continues
1950 * to exist but is no longer charged to its former owner.
1951 */
1952static inline void skb_orphan(struct sk_buff *skb)
1953{
Eric Dumazetc34a7612013-07-30 16:11:15 -07001954 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001955 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07001956 skb->destructor = NULL;
1957 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07001958 } else {
1959 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07001960 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001961}
1962
1963/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00001964 * skb_orphan_frags - orphan the frags contained in a buffer
1965 * @skb: buffer to orphan frags from
1966 * @gfp_mask: allocation mask for replacement pages
1967 *
1968 * For each frag in the SKB which needs a destructor (i.e. has an
1969 * owner) create a copy of that frag and release the original
1970 * page by calling the destructor.
1971 */
1972static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1973{
1974 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1975 return 0;
1976 return skb_copy_ubufs(skb, gfp_mask);
1977}
1978
1979/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001980 * __skb_queue_purge - empty a list
1981 * @list: list to empty
1982 *
1983 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1984 * the list and one reference dropped. This function does not take the
1985 * list lock and the caller must hold the relevant locks to use it.
1986 */
Joe Perches7965bd42013-09-26 14:48:15 -07001987void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001988static inline void __skb_queue_purge(struct sk_buff_head *list)
1989{
1990 struct sk_buff *skb;
1991 while ((skb = __skb_dequeue(list)) != NULL)
1992 kfree_skb(skb);
1993}
1994
Alexander Duycke5e67302013-02-08 10:17:15 +00001995#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
1996#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
1997#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
1998
Joe Perches7965bd42013-09-26 14:48:15 -07001999void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002000
Joe Perches7965bd42013-09-26 14:48:15 -07002001struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2002 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07002003
2004/**
2005 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2006 * @dev: network device to receive on
2007 * @length: length to allocate
2008 *
2009 * Allocate a new &sk_buff and assign it a usage count of one. The
2010 * buffer has unspecified headroom built in. Users should allocate
2011 * the headroom they think they need without accounting for the
2012 * built in space. The built in space is used for optimisations.
2013 *
2014 * %NULL is returned if there is no free memory. Although this function
2015 * allocates memory it can be called from an interrupt.
2016 */
2017static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002018 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002019{
2020 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2021}
2022
Eric Dumazet6f532612012-05-18 05:12:12 +00002023/* legacy helper around __netdev_alloc_skb() */
2024static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2025 gfp_t gfp_mask)
2026{
2027 return __netdev_alloc_skb(NULL, length, gfp_mask);
2028}
2029
2030/* legacy helper around netdev_alloc_skb() */
2031static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2032{
2033 return netdev_alloc_skb(NULL, length);
2034}
2035
2036
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002037static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2038 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002039{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002040 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002041
2042 if (NET_IP_ALIGN && skb)
2043 skb_reserve(skb, NET_IP_ALIGN);
2044 return skb;
2045}
2046
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002047static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2048 unsigned int length)
2049{
2050 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2051}
2052
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002053/**
2054 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
Mel Gorman06140022012-07-31 16:44:24 -07002055 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2056 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2057 * @order: size of the allocation
2058 *
2059 * Allocate a new page.
2060 *
2061 * %NULL is returned if there is no free memory.
2062*/
2063static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
2064 struct sk_buff *skb,
2065 unsigned int order)
2066{
2067 struct page *page;
2068
2069 gfp_mask |= __GFP_COLD;
2070
2071 if (!(gfp_mask & __GFP_NOMEMALLOC))
2072 gfp_mask |= __GFP_MEMALLOC;
2073
2074 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2075 if (skb && page && page->pfmemalloc)
2076 skb->pfmemalloc = true;
2077
2078 return page;
2079}
2080
2081/**
2082 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2083 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2084 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2085 *
2086 * Allocate a new page.
2087 *
2088 * %NULL is returned if there is no free memory.
2089 */
2090static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
2091 struct sk_buff *skb)
2092{
2093 return __skb_alloc_pages(gfp_mask, skb, 0);
2094}
2095
2096/**
2097 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2098 * @page: The page that was allocated from skb_alloc_page
2099 * @skb: The skb that may need pfmemalloc set
2100 */
2101static inline void skb_propagate_pfmemalloc(struct page *page,
2102 struct sk_buff *skb)
2103{
2104 if (page && page->pfmemalloc)
2105 skb->pfmemalloc = true;
2106}
2107
Eric Dumazet564824b2010-10-11 19:05:25 +00002108/**
Masanari Iidae2278672014-02-18 22:54:36 +09002109 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002110 * @frag: the paged fragment
2111 *
2112 * Returns the &struct page associated with @frag.
2113 */
2114static inline struct page *skb_frag_page(const skb_frag_t *frag)
2115{
Ian Campbella8605c62011-10-19 23:01:49 +00002116 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00002117}
2118
2119/**
2120 * __skb_frag_ref - take an addition reference on a paged fragment.
2121 * @frag: the paged fragment
2122 *
2123 * Takes an additional reference on the paged fragment @frag.
2124 */
2125static inline void __skb_frag_ref(skb_frag_t *frag)
2126{
2127 get_page(skb_frag_page(frag));
2128}
2129
2130/**
2131 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2132 * @skb: the buffer
2133 * @f: the fragment offset.
2134 *
2135 * Takes an additional reference on the @f'th paged fragment of @skb.
2136 */
2137static inline void skb_frag_ref(struct sk_buff *skb, int f)
2138{
2139 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2140}
2141
2142/**
2143 * __skb_frag_unref - release a reference on a paged fragment.
2144 * @frag: the paged fragment
2145 *
2146 * Releases a reference on the paged fragment @frag.
2147 */
2148static inline void __skb_frag_unref(skb_frag_t *frag)
2149{
2150 put_page(skb_frag_page(frag));
2151}
2152
2153/**
2154 * skb_frag_unref - release a reference on a paged fragment of an skb.
2155 * @skb: the buffer
2156 * @f: the fragment offset
2157 *
2158 * Releases a reference on the @f'th paged fragment of @skb.
2159 */
2160static inline void skb_frag_unref(struct sk_buff *skb, int f)
2161{
2162 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2163}
2164
2165/**
2166 * skb_frag_address - gets the address of the data contained in a paged fragment
2167 * @frag: the paged fragment buffer
2168 *
2169 * Returns the address of the data within @frag. The page must already
2170 * be mapped.
2171 */
2172static inline void *skb_frag_address(const skb_frag_t *frag)
2173{
2174 return page_address(skb_frag_page(frag)) + frag->page_offset;
2175}
2176
2177/**
2178 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2179 * @frag: the paged fragment buffer
2180 *
2181 * Returns the address of the data within @frag. Checks that the page
2182 * is mapped and returns %NULL otherwise.
2183 */
2184static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2185{
2186 void *ptr = page_address(skb_frag_page(frag));
2187 if (unlikely(!ptr))
2188 return NULL;
2189
2190 return ptr + frag->page_offset;
2191}
2192
2193/**
2194 * __skb_frag_set_page - sets the page contained in a paged fragment
2195 * @frag: the paged fragment
2196 * @page: the page to set
2197 *
2198 * Sets the fragment @frag to contain @page.
2199 */
2200static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2201{
Ian Campbella8605c62011-10-19 23:01:49 +00002202 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002203}
2204
2205/**
2206 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2207 * @skb: the buffer
2208 * @f: the fragment offset
2209 * @page: the page to set
2210 *
2211 * Sets the @f'th fragment of @skb to contain @page.
2212 */
2213static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2214 struct page *page)
2215{
2216 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2217}
2218
Eric Dumazet400dfd32013-10-17 16:27:07 -07002219bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2220
Ian Campbell131ea662011-08-19 06:25:00 +00002221/**
2222 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002223 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002224 * @frag: the paged fragment to map
2225 * @offset: the offset within the fragment (starting at the
2226 * fragment's own offset)
2227 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002228 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00002229 *
2230 * Maps the page associated with @frag to @device.
2231 */
2232static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2233 const skb_frag_t *frag,
2234 size_t offset, size_t size,
2235 enum dma_data_direction dir)
2236{
2237 return dma_map_page(dev, skb_frag_page(frag),
2238 frag->page_offset + offset, size, dir);
2239}
2240
Eric Dumazet117632e2011-12-03 21:39:53 +00002241static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2242 gfp_t gfp_mask)
2243{
2244 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2245}
2246
Octavian Purdilabad93e92014-06-12 01:36:26 +03002247
2248static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2249 gfp_t gfp_mask)
2250{
2251 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2252}
2253
2254
Ian Campbell131ea662011-08-19 06:25:00 +00002255/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002256 * skb_clone_writable - is the header of a clone writable
2257 * @skb: buffer to check
2258 * @len: length up to which to write
2259 *
2260 * Returns true if modifying the header part of the cloned buffer
2261 * does not requires the data to be copied.
2262 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002263static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002264{
2265 return !skb_header_cloned(skb) &&
2266 skb_headroom(skb) + len <= skb->hdr_len;
2267}
2268
Herbert Xud9cc2042007-09-16 16:21:16 -07002269static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2270 int cloned)
2271{
2272 int delta = 0;
2273
Herbert Xud9cc2042007-09-16 16:21:16 -07002274 if (headroom > skb_headroom(skb))
2275 delta = headroom - skb_headroom(skb);
2276
2277 if (delta || cloned)
2278 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2279 GFP_ATOMIC);
2280 return 0;
2281}
2282
Patrick McHardy334a8132007-06-25 04:35:20 -07002283/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002284 * skb_cow - copy header of skb when it is required
2285 * @skb: buffer to cow
2286 * @headroom: needed headroom
2287 *
2288 * If the skb passed lacks sufficient headroom or its data part
2289 * is shared, data is reallocated. If reallocation fails, an error
2290 * is returned and original skb is not changed.
2291 *
2292 * The result is skb with writable area skb->head...skb->tail
2293 * and at least @headroom of space at head.
2294 */
2295static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2296{
Herbert Xud9cc2042007-09-16 16:21:16 -07002297 return __skb_cow(skb, headroom, skb_cloned(skb));
2298}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002299
Herbert Xud9cc2042007-09-16 16:21:16 -07002300/**
2301 * skb_cow_head - skb_cow but only making the head writable
2302 * @skb: buffer to cow
2303 * @headroom: needed headroom
2304 *
2305 * This function is identical to skb_cow except that we replace the
2306 * skb_cloned check by skb_header_cloned. It should be used when
2307 * you only need to push on some header and do not need to modify
2308 * the data.
2309 */
2310static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2311{
2312 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002313}
2314
2315/**
2316 * skb_padto - pad an skbuff up to a minimal size
2317 * @skb: buffer to pad
2318 * @len: minimal length
2319 *
2320 * Pads up a buffer to ensure the trailing bytes exist and are
2321 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002322 * is untouched. Otherwise it is extended. Returns zero on
2323 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002324 */
2325
Herbert Xu5b057c62006-06-23 02:06:41 -07002326static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002327{
2328 unsigned int size = skb->len;
2329 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002330 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002331 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002332}
2333
2334static inline int skb_add_data(struct sk_buff *skb,
2335 char __user *from, int copy)
2336{
2337 const int off = skb->len;
2338
2339 if (skb->ip_summed == CHECKSUM_NONE) {
2340 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08002341 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
Linus Torvalds1da177e2005-04-16 15:20:36 -07002342 copy, 0, &err);
2343 if (!err) {
2344 skb->csum = csum_block_add(skb->csum, csum, off);
2345 return 0;
2346 }
2347 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2348 return 0;
2349
2350 __skb_trim(skb, off);
2351 return -EFAULT;
2352}
2353
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002354static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2355 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002356{
2357 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002358 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002359
Ian Campbellea2ab692011-08-22 23:44:58 +00002360 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002361 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002362 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002363 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002364}
2365
Herbert Xu364c6ba2006-06-09 16:10:40 -07002366static inline int __skb_linearize(struct sk_buff *skb)
2367{
2368 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2369}
2370
Linus Torvalds1da177e2005-04-16 15:20:36 -07002371/**
2372 * skb_linearize - convert paged skb to linear one
2373 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002374 *
2375 * If there is no free memory -ENOMEM is returned, otherwise zero
2376 * is returned and the old skb data released.
2377 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002378static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002379{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002380 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2381}
2382
2383/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00002384 * skb_has_shared_frag - can any frag be overwritten
2385 * @skb: buffer to test
2386 *
2387 * Return true if the skb has at least one frag that might be modified
2388 * by an external entity (as in vmsplice()/sendfile())
2389 */
2390static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2391{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00002392 return skb_is_nonlinear(skb) &&
2393 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00002394}
2395
2396/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07002397 * skb_linearize_cow - make sure skb is linear and writable
2398 * @skb: buffer to process
2399 *
2400 * If there is no free memory -ENOMEM is returned, otherwise zero
2401 * is returned and the old skb data released.
2402 */
2403static inline int skb_linearize_cow(struct sk_buff *skb)
2404{
2405 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2406 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002407}
2408
2409/**
2410 * skb_postpull_rcsum - update checksum for received skb after pull
2411 * @skb: buffer to update
2412 * @start: start of data before pull
2413 * @len: length of data pulled
2414 *
2415 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002416 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2417 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002418 */
2419
2420static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002421 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002422{
Patrick McHardy84fa7932006-08-29 16:44:56 -07002423 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002424 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2425}
2426
Herbert Xucbb042f2006-03-20 22:43:56 -08002427unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2428
David S. Miller7ce5a272013-12-02 17:26:05 -05002429/**
2430 * pskb_trim_rcsum - trim received skb and update checksum
2431 * @skb: buffer to trim
2432 * @len: new length
2433 *
2434 * This is exactly the same as pskb_trim except that it ensures the
2435 * checksum of received packets are still valid after the operation.
2436 */
2437
2438static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2439{
2440 if (likely(len >= skb->len))
2441 return 0;
2442 if (skb->ip_summed == CHECKSUM_COMPLETE)
2443 skb->ip_summed = CHECKSUM_NONE;
2444 return __pskb_trim(skb, len);
2445}
2446
Linus Torvalds1da177e2005-04-16 15:20:36 -07002447#define skb_queue_walk(queue, skb) \
2448 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002449 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002450 skb = skb->next)
2451
James Chapman46f89142007-04-30 00:07:31 -07002452#define skb_queue_walk_safe(queue, skb, tmp) \
2453 for (skb = (queue)->next, tmp = skb->next; \
2454 skb != (struct sk_buff *)(queue); \
2455 skb = tmp, tmp = skb->next)
2456
David S. Miller1164f522008-09-23 00:49:44 -07002457#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002458 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002459 skb = skb->next)
2460
2461#define skb_queue_walk_from_safe(queue, skb, tmp) \
2462 for (tmp = skb->next; \
2463 skb != (struct sk_buff *)(queue); \
2464 skb = tmp, tmp = skb->next)
2465
Stephen Hemminger300ce172005-10-30 13:47:34 -08002466#define skb_queue_reverse_walk(queue, skb) \
2467 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002468 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08002469 skb = skb->prev)
2470
David S. Miller686a2952011-01-20 22:47:32 -08002471#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2472 for (skb = (queue)->prev, tmp = skb->prev; \
2473 skb != (struct sk_buff *)(queue); \
2474 skb = tmp, tmp = skb->prev)
2475
2476#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2477 for (tmp = skb->prev; \
2478 skb != (struct sk_buff *)(queue); \
2479 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002480
David S. Miller21dc3302010-08-23 00:13:46 -07002481static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07002482{
2483 return skb_shinfo(skb)->frag_list != NULL;
2484}
2485
2486static inline void skb_frag_list_init(struct sk_buff *skb)
2487{
2488 skb_shinfo(skb)->frag_list = NULL;
2489}
2490
2491static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2492{
2493 frag->next = skb_shinfo(skb)->frag_list;
2494 skb_shinfo(skb)->frag_list = frag;
2495}
2496
2497#define skb_walk_frags(skb, iter) \
2498 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2499
Joe Perches7965bd42013-09-26 14:48:15 -07002500struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2501 int *peeked, int *off, int *err);
2502struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2503 int *err);
2504unsigned int datagram_poll(struct file *file, struct socket *sock,
2505 struct poll_table_struct *wait);
2506int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2507 struct iovec *to, int size);
2508int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2509 struct iovec *iov);
2510int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2511 const struct iovec *from, int from_offset,
2512 int len);
2513int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2514 int offset, size_t count);
2515int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2516 const struct iovec *to, int to_offset,
2517 int size);
2518void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2519void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2520int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07002521int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2522int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2523__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2524 int len, __wsum csum);
2525int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2526 struct pipe_inode_info *pipe, unsigned int len,
2527 unsigned int flags);
2528void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01002529unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00002530int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2531 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07002532void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2533int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2534void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Florian Westphalde960aa2014-01-26 10:58:16 +01002535unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -07002536struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002537
Daniel Borkmann2817a332013-10-30 11:50:51 +01002538struct skb_checksum_ops {
2539 __wsum (*update)(const void *mem, int len, __wsum wsum);
2540 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2541};
2542
2543__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2544 __wsum csum, const struct skb_checksum_ops *ops);
2545__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2546 __wsum csum);
2547
Linus Torvalds1da177e2005-04-16 15:20:36 -07002548static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2549 int len, void *buffer)
2550{
2551 int hlen = skb_headlen(skb);
2552
Patrick McHardy55820ee2005-07-05 14:08:10 -07002553 if (hlen - offset >= len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002554 return skb->data + offset;
2555
2556 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2557 return NULL;
2558
2559 return buffer;
2560}
2561
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01002562/**
2563 * skb_needs_linearize - check if we need to linearize a given skb
2564 * depending on the given device features.
2565 * @skb: socket buffer to check
2566 * @features: net device features
2567 *
2568 * Returns true if either:
2569 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2570 * 2. skb is fragmented and the device does not support SG.
2571 */
2572static inline bool skb_needs_linearize(struct sk_buff *skb,
2573 netdev_features_t features)
2574{
2575 return skb_is_nonlinear(skb) &&
2576 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2577 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2578}
2579
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03002580static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2581 void *to,
2582 const unsigned int len)
2583{
2584 memcpy(to, skb->data, len);
2585}
2586
2587static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2588 const int offset, void *to,
2589 const unsigned int len)
2590{
2591 memcpy(to, skb->data + offset, len);
2592}
2593
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03002594static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2595 const void *from,
2596 const unsigned int len)
2597{
2598 memcpy(skb->data, from, len);
2599}
2600
2601static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2602 const int offset,
2603 const void *from,
2604 const unsigned int len)
2605{
2606 memcpy(skb->data + offset, from, len);
2607}
2608
Joe Perches7965bd42013-09-26 14:48:15 -07002609void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002610
Patrick Ohlyac45f602009-02-12 05:03:37 +00002611static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2612{
2613 return skb->tstamp;
2614}
2615
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002616/**
2617 * skb_get_timestamp - get timestamp from a skb
2618 * @skb: skb to get stamp from
2619 * @stamp: pointer to struct timeval to store stamp in
2620 *
2621 * Timestamps are stored in the skb as offsets to a base timestamp.
2622 * This function converts the offset back to a struct timeval and stores
2623 * it in stamp.
2624 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00002625static inline void skb_get_timestamp(const struct sk_buff *skb,
2626 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002627{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002628 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002629}
2630
Patrick Ohlyac45f602009-02-12 05:03:37 +00002631static inline void skb_get_timestampns(const struct sk_buff *skb,
2632 struct timespec *stamp)
2633{
2634 *stamp = ktime_to_timespec(skb->tstamp);
2635}
2636
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002637static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002638{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002639 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002640}
2641
Stephen Hemminger164891a2007-04-23 22:26:16 -07002642static inline ktime_t net_timedelta(ktime_t t)
2643{
2644 return ktime_sub(ktime_get_real(), t);
2645}
2646
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07002647static inline ktime_t net_invalid_timestamp(void)
2648{
2649 return ktime_set(0, 0);
2650}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002651
Richard Cochranc1f19b52010-07-17 08:49:36 +00002652#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2653
Joe Perches7965bd42013-09-26 14:48:15 -07002654void skb_clone_tx_timestamp(struct sk_buff *skb);
2655bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00002656
2657#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2658
2659static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2660{
2661}
2662
2663static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2664{
2665 return false;
2666}
2667
2668#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2669
2670/**
2671 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2672 *
Richard Cochranda92b192011-10-21 00:49:15 +00002673 * PHY drivers may accept clones of transmitted packets for
2674 * timestamping via their phy_driver.txtstamp method. These drivers
2675 * must call this function to return the skb back to the stack, with
2676 * or without a timestamp.
2677 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00002678 * @skb: clone of the the original outgoing packet
Richard Cochranda92b192011-10-21 00:49:15 +00002679 * @hwtstamps: hardware time stamps, may be NULL if not available
Richard Cochranc1f19b52010-07-17 08:49:36 +00002680 *
2681 */
2682void skb_complete_tx_timestamp(struct sk_buff *skb,
2683 struct skb_shared_hwtstamps *hwtstamps);
2684
Patrick Ohlyac45f602009-02-12 05:03:37 +00002685/**
2686 * skb_tstamp_tx - queue clone of skb with send time stamps
2687 * @orig_skb: the original outgoing packet
2688 * @hwtstamps: hardware time stamps, may be NULL if not available
2689 *
2690 * If the skb has a socket associated, then this function clones the
2691 * skb (thus sharing the actual data and optional structures), stores
2692 * the optional hardware time stamping information (if non NULL) or
2693 * generates a software time stamp (otherwise), then queues the clone
2694 * to the error queue of the socket. Errors are silently ignored.
2695 */
Joe Perches7965bd42013-09-26 14:48:15 -07002696void skb_tstamp_tx(struct sk_buff *orig_skb,
2697 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00002698
Richard Cochran4507a712010-07-17 08:48:28 +00002699static inline void sw_tx_timestamp(struct sk_buff *skb)
2700{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002701 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2702 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00002703 skb_tstamp_tx(skb, NULL);
2704}
2705
2706/**
2707 * skb_tx_timestamp() - Driver hook for transmit timestamping
2708 *
2709 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00002710 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00002711 *
David S. Miller73409f32013-12-27 13:04:33 -05002712 * Specifically, one should make absolutely sure that this function is
2713 * called before TX completion of this packet can trigger. Otherwise
2714 * the packet could potentially already be freed.
2715 *
Richard Cochran4507a712010-07-17 08:48:28 +00002716 * @skb: A socket buffer.
2717 */
2718static inline void skb_tx_timestamp(struct sk_buff *skb)
2719{
Richard Cochranc1f19b52010-07-17 08:49:36 +00002720 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002721 sw_tx_timestamp(skb);
2722}
2723
Johannes Berg6e3e9392011-11-09 10:15:42 +01002724/**
2725 * skb_complete_wifi_ack - deliver skb with wifi status
2726 *
2727 * @skb: the original outgoing packet
2728 * @acked: ack status
2729 *
2730 */
2731void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2732
Joe Perches7965bd42013-09-26 14:48:15 -07002733__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2734__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002735
Herbert Xu60476372007-04-09 11:59:39 -07002736static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2737{
2738 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2739}
2740
Herbert Xufb286bb2005-11-10 13:01:24 -08002741/**
2742 * skb_checksum_complete - Calculate checksum of an entire packet
2743 * @skb: packet to process
2744 *
2745 * This function calculates the checksum over the entire packet plus
2746 * the value of skb->csum. The latter can be used to supply the
2747 * checksum of a pseudo header as used by TCP/UDP. It returns the
2748 * checksum.
2749 *
2750 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2751 * this function can be used to verify that checksum on received
2752 * packets. In that case the function should return zero if the
2753 * checksum is correct. In particular, this function will return zero
2754 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2755 * hardware has already verified the correctness of the checksum.
2756 */
Al Viro4381ca32007-07-15 21:00:11 +01002757static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002758{
Herbert Xu60476372007-04-09 11:59:39 -07002759 return skb_csum_unnecessary(skb) ?
2760 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002761}
2762
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002763/* Check if we need to perform checksum complete validation.
2764 *
2765 * Returns true if checksum complete is needed, false otherwise
2766 * (either checksum is unnecessary or zero checksum is allowed).
2767 */
2768static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
2769 bool zero_okay,
2770 __sum16 check)
2771{
2772 if (skb_csum_unnecessary(skb)) {
2773 return false;
2774 } else if (zero_okay && !check) {
2775 skb->ip_summed = CHECKSUM_UNNECESSARY;
2776 return false;
2777 }
2778
2779 return true;
2780}
2781
2782/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
2783 * in checksum_init.
2784 */
2785#define CHECKSUM_BREAK 76
2786
2787/* Validate (init) checksum based on checksum complete.
2788 *
2789 * Return values:
2790 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
2791 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
2792 * checksum is stored in skb->csum for use in __skb_checksum_complete
2793 * non-zero: value of invalid checksum
2794 *
2795 */
2796static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
2797 bool complete,
2798 __wsum psum)
2799{
2800 if (skb->ip_summed == CHECKSUM_COMPLETE) {
2801 if (!csum_fold(csum_add(psum, skb->csum))) {
2802 skb->ip_summed = CHECKSUM_UNNECESSARY;
2803 return 0;
2804 }
2805 }
2806
2807 skb->csum = psum;
2808
2809 if (complete || skb->len <= CHECKSUM_BREAK)
2810 return __skb_checksum_complete(skb);
2811
2812 return 0;
2813}
2814
2815static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
2816{
2817 return 0;
2818}
2819
2820/* Perform checksum validate (init). Note that this is a macro since we only
2821 * want to calculate the pseudo header which is an input function if necessary.
2822 * First we try to validate without any computation (checksum unnecessary) and
2823 * then calculate based on checksum complete calling the function to compute
2824 * pseudo header.
2825 *
2826 * Return values:
2827 * 0: checksum is validated or try to in skb_checksum_complete
2828 * non-zero: value of invalid checksum
2829 */
2830#define __skb_checksum_validate(skb, proto, complete, \
2831 zero_okay, check, compute_pseudo) \
2832({ \
2833 __sum16 __ret = 0; \
2834 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
2835 __ret = __skb_checksum_validate_complete(skb, \
2836 complete, compute_pseudo(skb, proto)); \
2837 __ret; \
2838})
2839
2840#define skb_checksum_init(skb, proto, compute_pseudo) \
2841 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
2842
2843#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
2844 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
2845
2846#define skb_checksum_validate(skb, proto, compute_pseudo) \
2847 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
2848
2849#define skb_checksum_validate_zero_check(skb, proto, check, \
2850 compute_pseudo) \
2851 __skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
2852
2853#define skb_checksum_simple_validate(skb) \
2854 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
2855
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002856#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07002857void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002858static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2859{
2860 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002861 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002862}
2863static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2864{
2865 if (nfct)
2866 atomic_inc(&nfct->use);
2867}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002868#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002869#ifdef CONFIG_BRIDGE_NETFILTER
2870static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2871{
2872 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2873 kfree(nf_bridge);
2874}
2875static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2876{
2877 if (nf_bridge)
2878 atomic_inc(&nf_bridge->use);
2879}
2880#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08002881static inline void nf_reset(struct sk_buff *skb)
2882{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002883#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08002884 nf_conntrack_put(skb->nfct);
2885 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002886#endif
Patrick McHardya193a4a2006-03-20 19:23:05 -08002887#ifdef CONFIG_BRIDGE_NETFILTER
2888 nf_bridge_put(skb->nf_bridge);
2889 skb->nf_bridge = NULL;
2890#endif
2891}
2892
Patrick McHardy124dff02013-04-05 20:42:05 +02002893static inline void nf_reset_trace(struct sk_buff *skb)
2894{
Florian Westphal478b3602014-02-15 23:48:45 +01002895#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00002896 skb->nf_trace = 0;
2897#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002898}
2899
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002900/* Note: This doesn't put any conntrack and bridge info in dst. */
2901static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2902{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002903#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002904 dst->nfct = src->nfct;
2905 nf_conntrack_get(src->nfct);
2906 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002907#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002908#ifdef CONFIG_BRIDGE_NETFILTER
2909 dst->nf_bridge = src->nf_bridge;
2910 nf_bridge_get(src->nf_bridge);
2911#endif
Florian Westphal478b3602014-02-15 23:48:45 +01002912#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
2913 dst->nf_trace = src->nf_trace;
2914#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002915}
2916
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002917static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2918{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002919#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002920 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002921#endif
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002922#ifdef CONFIG_BRIDGE_NETFILTER
2923 nf_bridge_put(dst->nf_bridge);
2924#endif
2925 __nf_copy(dst, src);
2926}
2927
James Morris984bc162006-06-09 00:29:17 -07002928#ifdef CONFIG_NETWORK_SECMARK
2929static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2930{
2931 to->secmark = from->secmark;
2932}
2933
2934static inline void skb_init_secmark(struct sk_buff *skb)
2935{
2936 skb->secmark = 0;
2937}
2938#else
2939static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2940{ }
2941
2942static inline void skb_init_secmark(struct sk_buff *skb)
2943{ }
2944#endif
2945
Eric W. Biederman574f7192014-04-01 12:20:24 -07002946static inline bool skb_irq_freeable(const struct sk_buff *skb)
2947{
2948 return !skb->destructor &&
2949#if IS_ENABLED(CONFIG_XFRM)
2950 !skb->sp &&
2951#endif
2952#if IS_ENABLED(CONFIG_NF_CONNTRACK)
2953 !skb->nfct &&
2954#endif
2955 !skb->_skb_refdst &&
2956 !skb_has_frag_list(skb);
2957}
2958
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002959static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2960{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002961 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002962}
2963
Stephen Hemminger92477442009-03-21 13:39:26 -07002964static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002965{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002966 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002967}
2968
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002969static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2970{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002971 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002972}
2973
David S. Millerd5a9e242009-01-27 16:22:11 -08002974static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2975{
2976 skb->queue_mapping = rx_queue + 1;
2977}
2978
Stephen Hemminger92477442009-03-21 13:39:26 -07002979static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002980{
2981 return skb->queue_mapping - 1;
2982}
2983
Stephen Hemminger92477442009-03-21 13:39:26 -07002984static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002985{
Eric Dumazeta02cec22010-09-22 20:43:57 +00002986 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08002987}
2988
Joe Perches7965bd42013-09-26 14:48:15 -07002989u16 __skb_tx_hash(const struct net_device *dev, const struct sk_buff *skb,
2990 unsigned int num_tx_queues);
Stephen Hemminger92477442009-03-21 13:39:26 -07002991
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04002992static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2993{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002994#ifdef CONFIG_XFRM
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002995 return skb->sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002996#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002997 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002998#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04002999}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003000
Pravin B Shelar68c33162013-02-14 14:02:41 +00003001/* Keeps track of mac header offset relative to skb->head.
3002 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3003 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07003004 * tunnel skb it points to outer mac header.
3005 * Keeps track of level of encapsulation of network headers.
3006 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00003007struct skb_gso_cb {
Eric Dumazet3347c962013-10-19 11:42:56 -07003008 int mac_offset;
3009 int encap_level;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003010 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00003011};
3012#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3013
3014static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3015{
3016 return (skb_mac_header(inner_skb) - inner_skb->head) -
3017 SKB_GSO_CB(inner_skb)->mac_offset;
3018}
3019
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00003020static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3021{
3022 int new_headroom, headroom;
3023 int ret;
3024
3025 headroom = skb_headroom(skb);
3026 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3027 if (ret)
3028 return ret;
3029
3030 new_headroom = skb_headroom(skb);
3031 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3032 return 0;
3033}
3034
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003035/* Compute the checksum for a gso segment. First compute the checksum value
3036 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3037 * then add in skb->csum (checksum from csum_start to end of packet).
3038 * skb->csum and csum_start are then updated to reflect the checksum of the
3039 * resultant packet starting from the transport header-- the resultant checksum
3040 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3041 * header.
3042 */
3043static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3044{
3045 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3046 skb_transport_offset(skb);
3047 __u16 csum;
3048
3049 csum = csum_fold(csum_partial(skb_transport_header(skb),
3050 plen, skb->csum));
3051 skb->csum = res;
3052 SKB_GSO_CB(skb)->csum_start -= plen;
3053
3054 return csum;
3055}
3056
David S. Millerbdcc0922012-03-07 20:53:36 -05003057static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07003058{
3059 return skb_shinfo(skb)->gso_size;
3060}
3061
Eric Dumazet36a8f392013-09-29 01:21:32 -07003062/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05003063static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02003064{
3065 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3066}
3067
Joe Perches7965bd42013-09-26 14:48:15 -07003068void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07003069
3070static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3071{
3072 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3073 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003074 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3075
Alexander Duyckb78462e2010-06-02 12:24:37 +00003076 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3077 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07003078 __skb_warn_lro_forwarding(skb);
3079 return true;
3080 }
3081 return false;
3082}
3083
Herbert Xu35fc92a2007-03-26 23:22:20 -07003084static inline void skb_forward_csum(struct sk_buff *skb)
3085{
3086 /* Unfortunately we don't support this one. Any brave souls? */
3087 if (skb->ip_summed == CHECKSUM_COMPLETE)
3088 skb->ip_summed = CHECKSUM_NONE;
3089}
3090
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003091/**
3092 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3093 * @skb: skb to check
3094 *
3095 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3096 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3097 * use this helper, to document places where we make this assertion.
3098 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003099static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003100{
3101#ifdef DEBUG
3102 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3103#endif
3104}
3105
Rusty Russellf35d9d82008-02-04 23:49:54 -05003106bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00003107
Paul Durranted1f50c2014-01-09 10:02:46 +00003108int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3109
Daniel Borkmannf77668d2013-03-19 06:39:30 +00003110u32 __skb_get_poff(const struct sk_buff *skb);
3111
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00003112/**
3113 * skb_head_is_locked - Determine if the skb->head is locked down
3114 * @skb: skb to check
3115 *
3116 * The head on skbs build around a head frag can be removed if they are
3117 * not cloned. This function returns true if the skb head is locked down
3118 * due to either being allocated via kmalloc, or by being a clone with
3119 * multiple references to the head.
3120 */
3121static inline bool skb_head_is_locked(const struct sk_buff *skb)
3122{
3123 return !skb->head_frag || skb_cloned(skb);
3124}
Florian Westphalfe6cc552014-02-13 23:09:12 +01003125
3126/**
3127 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3128 *
3129 * @skb: GSO skb
3130 *
3131 * skb_gso_network_seglen is used to determine the real size of the
3132 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3133 *
3134 * The MAC/L2 header is not accounted for.
3135 */
3136static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3137{
3138 unsigned int hdr_len = skb_transport_header(skb) -
3139 skb_network_header(skb);
3140 return hdr_len + skb_gso_transport_seglen(skb);
3141}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003142#endif /* __KERNEL__ */
3143#endif /* _LINUX_SKBUFF_H */