blob: 5b5cd3189c985f89a055e35d8d1c67c63879d8fc [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * Definitions for the 'struct sk_buff' memory handlers.
3 *
4 * Authors:
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14#ifndef _LINUX_SKBUFF_H
15#define _LINUX_SKBUFF_H
16
Linus Torvalds1da177e2005-04-16 15:20:36 -070017#include <linux/kernel.h>
Vegard Nossumfe55f6d2008-08-30 12:16:35 +020018#include <linux/kmemcheck.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070019#include <linux/compiler.h>
20#include <linux/time.h>
Paul Gortmaker187f1882011-11-23 20:12:59 -050021#include <linux/bug.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070022#include <linux/cache.h>
23
Arun Sharma600634972011-07-26 16:09:06 -070024#include <linux/atomic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <asm/types.h>
26#include <linux/spinlock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <linux/net.h>
Thomas Graf3fc7e8a2005-06-23 21:00:17 -070028#include <linux/textsearch.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070029#include <net/checksum.h>
Al Viroa80958f2006-12-04 20:41:19 +000030#include <linux/rcupdate.h>
Chris Leech97fc2f02006-05-23 17:55:33 -070031#include <linux/dmaengine.h>
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -070032#include <linux/hrtimer.h>
Ian Campbell131ea662011-08-19 06:25:00 +000033#include <linux/dma-mapping.h>
Michał Mirosławc8f44af2011-11-15 15:29:55 +000034#include <linux/netdev_features.h>
Eric Dumazet363ec39232014-02-26 14:02:11 -080035#include <linux/sched.h>
Jason Wang5203cd22013-03-26 23:11:21 +000036#include <net/flow_keys.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070037
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010038/* A. Checksumming of received packets by device.
39 *
40 * CHECKSUM_NONE:
41 *
42 * Device failed to checksum this packet e.g. due to lack of capabilities.
43 * The packet contains full (though not verified) checksum in packet but
44 * not in skb->csum. Thus, skb->csum is undefined in this case.
45 *
46 * CHECKSUM_UNNECESSARY:
47 *
48 * The hardware you're dealing with doesn't calculate the full checksum
49 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
50 * for specific protocols e.g. TCP/UDP/SCTP, then, for such packets it will
51 * set CHECKSUM_UNNECESSARY if their checksums are okay. skb->csum is still
52 * undefined in this case though. It is a bad option, but, unfortunately,
53 * nowadays most vendors do this. Apparently with the secret goal to sell
54 * you new devices, when you will add new protocol to your host, f.e. IPv6 8)
55 *
56 * CHECKSUM_COMPLETE:
57 *
58 * This is the most generic way. The device supplied checksum of the _whole_
59 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
60 * hardware doesn't need to parse L3/L4 headers to implement this.
61 *
62 * Note: Even if device supports only some protocols, but is able to produce
63 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
64 *
65 * CHECKSUM_PARTIAL:
66 *
67 * This is identical to the case for output below. This may occur on a packet
68 * received directly from another Linux OS, e.g., a virtualized Linux kernel
69 * on the same host. The packet can be treated in the same way as
70 * CHECKSUM_UNNECESSARY, except that on output (i.e., forwarding) the
71 * checksum must be filled in by the OS or the hardware.
72 *
73 * B. Checksumming on output.
74 *
75 * CHECKSUM_NONE:
76 *
77 * The skb was already checksummed by the protocol, or a checksum is not
78 * required.
79 *
80 * CHECKSUM_PARTIAL:
81 *
82 * The device is required to checksum the packet as seen by hard_start_xmit()
83 * from skb->csum_start up to the end, and to record/write the checksum at
84 * offset skb->csum_start + skb->csum_offset.
85 *
86 * The device must show its capabilities in dev->features, set up at device
87 * setup time, e.g. netdev_features.h:
88 *
89 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
90 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
91 * IPv4. Sigh. Vendors like this way for an unknown reason.
92 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
93 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
94 * NETIF_F_... - Well, you get the picture.
95 *
96 * CHECKSUM_UNNECESSARY:
97 *
98 * Normally, the device will do per protocol specific checksumming. Protocol
99 * implementations that do not want the NIC to perform the checksum
100 * calculation should use this flag in their outgoing skbs.
101 *
102 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
103 * offload. Correspondingly, the FCoE protocol driver
104 * stack should use CHECKSUM_UNNECESSARY.
105 *
106 * Any questions? No questions, good. --ANK
107 */
108
Herbert Xu60476372007-04-09 11:59:39 -0700109/* Don't change this without changing skb_csum_unnecessary! */
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100110#define CHECKSUM_NONE 0
111#define CHECKSUM_UNNECESSARY 1
112#define CHECKSUM_COMPLETE 2
113#define CHECKSUM_PARTIAL 3
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114
115#define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
116 ~(SMP_CACHE_BYTES - 1))
David S. Millerfc910a22007-03-25 20:27:59 -0700117#define SKB_WITH_OVERHEAD(X) \
Herbert Xudeea84b2007-10-21 16:27:46 -0700118 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
David S. Millerfc910a22007-03-25 20:27:59 -0700119#define SKB_MAX_ORDER(X, ORDER) \
120 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700121#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
122#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
123
Eric Dumazet87fb4b72011-10-13 07:28:54 +0000124/* return minimum truesize of one skb containing X bytes of data */
125#define SKB_TRUESIZE(X) ((X) + \
126 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
127 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
128
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129struct net_device;
David Howells716ea3a2007-04-02 20:19:53 -0700130struct scatterlist;
Jens Axboe9c55e012007-11-06 23:30:13 -0800131struct pipe_inode_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700132
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700133#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134struct nf_conntrack {
135 atomic_t use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136};
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700137#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700138
139#ifdef CONFIG_BRIDGE_NETFILTER
140struct nf_bridge_info {
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000141 atomic_t use;
142 unsigned int mask;
143 struct net_device *physindev;
144 struct net_device *physoutdev;
145 unsigned long data[32 / sizeof(unsigned long)];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146};
147#endif
148
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149struct sk_buff_head {
150 /* These two members must be first. */
151 struct sk_buff *next;
152 struct sk_buff *prev;
153
154 __u32 qlen;
155 spinlock_t lock;
156};
157
158struct sk_buff;
159
Ian Campbell9d4dde52011-12-22 23:39:14 +0000160/* To allow 64K frame to be packed as single skb without frag_list we
161 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
162 * buffers which do not start on a page boundary.
163 *
164 * Since GRO uses frags we allocate at least 16 regardless of page
165 * size.
Anton Blancharda715dea2011-03-27 14:57:26 +0000166 */
Ian Campbell9d4dde52011-12-22 23:39:14 +0000167#if (65536/PAGE_SIZE + 1) < 16
David S. Millereec00952011-03-29 23:34:08 -0700168#define MAX_SKB_FRAGS 16UL
Anton Blancharda715dea2011-03-27 14:57:26 +0000169#else
Ian Campbell9d4dde52011-12-22 23:39:14 +0000170#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
Anton Blancharda715dea2011-03-27 14:57:26 +0000171#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172
173typedef struct skb_frag_struct skb_frag_t;
174
175struct skb_frag_struct {
Ian Campbella8605c62011-10-19 23:01:49 +0000176 struct {
177 struct page *p;
178 } page;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000179#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
David S. Millera309bb02007-07-30 18:47:03 -0700180 __u32 page_offset;
181 __u32 size;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000182#else
183 __u16 page_offset;
184 __u16 size;
185#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186};
187
Eric Dumazet9e903e02011-10-18 21:00:24 +0000188static inline unsigned int skb_frag_size(const skb_frag_t *frag)
189{
190 return frag->size;
191}
192
193static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
194{
195 frag->size = size;
196}
197
198static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
199{
200 frag->size += delta;
201}
202
203static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
204{
205 frag->size -= delta;
206}
207
Patrick Ohlyac45f602009-02-12 05:03:37 +0000208#define HAVE_HW_TIME_STAMP
209
210/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800211 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000212 * @hwtstamp: hardware time stamp transformed into duration
213 * since arbitrary point in time
214 * @syststamp: hwtstamp transformed to system time base
215 *
216 * Software time stamps generated by ktime_get_real() are stored in
217 * skb->tstamp. The relation between the different kinds of time
218 * stamps is as follows:
219 *
220 * syststamp and tstamp can be compared against each other in
221 * arbitrary combinations. The accuracy of a
222 * syststamp/tstamp/"syststamp from other device" comparison is
223 * limited by the accuracy of the transformation into system time
224 * base. This depends on the device driver and its underlying
225 * hardware.
226 *
227 * hwtstamps can only be compared against other hwtstamps from
228 * the same device.
229 *
230 * This structure is attached to packets as part of the
231 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
232 */
233struct skb_shared_hwtstamps {
234 ktime_t hwtstamp;
235 ktime_t syststamp;
236};
237
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000238/* Definitions for tx_flags in struct skb_shared_info */
239enum {
240 /* generate hardware time stamp */
241 SKBTX_HW_TSTAMP = 1 << 0,
242
243 /* generate software time stamp */
244 SKBTX_SW_TSTAMP = 1 << 1,
245
246 /* device driver is going to provide hardware time stamp */
247 SKBTX_IN_PROGRESS = 1 << 2,
248
Shirley Maa6686f22011-07-06 12:22:12 +0000249 /* device driver supports TX zero-copy buffers */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000250 SKBTX_DEV_ZEROCOPY = 1 << 3,
Johannes Berg6e3e9392011-11-09 10:15:42 +0100251
252 /* generate wifi status information (where possible) */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000253 SKBTX_WIFI_STATUS = 1 << 4,
Pravin B Shelarc9af6db2013-02-11 09:27:41 +0000254
255 /* This indicates at least one fragment might be overwritten
256 * (as in vmsplice(), sendfile() ...)
257 * If we need to compute a TX checksum, we'll need to copy
258 * all frags to avoid possible bad checksum
259 */
260 SKBTX_SHARED_FRAG = 1 << 5,
Shirley Maa6686f22011-07-06 12:22:12 +0000261};
262
263/*
264 * The callback notifies userspace to release buffers when skb DMA is done in
265 * lower device, the skb last reference should be 0 when calling this.
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000266 * The zerocopy_success argument is true if zero copy transmit occurred,
267 * false on data copy or out of memory error caused by data copy attempt.
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000268 * The ctx field is used to track device context.
269 * The desc field is used to track userspace buffer index.
Shirley Maa6686f22011-07-06 12:22:12 +0000270 */
271struct ubuf_info {
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000272 void (*callback)(struct ubuf_info *, bool zerocopy_success);
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000273 void *ctx;
Shirley Maa6686f22011-07-06 12:22:12 +0000274 unsigned long desc;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000275};
276
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277/* This data is invariant across clones and lives at
278 * the end of the header data, ie. at skb->end.
279 */
280struct skb_shared_info {
Ian Campbell9f42f122012-01-05 07:13:39 +0000281 unsigned char nr_frags;
282 __u8 tx_flags;
Herbert Xu79671682006-06-22 02:40:14 -0700283 unsigned short gso_size;
284 /* Warning: this field is not always filled in (UFO)! */
285 unsigned short gso_segs;
286 unsigned short gso_type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000288 struct skb_shared_hwtstamps hwtstamps;
Ian Campbell9f42f122012-01-05 07:13:39 +0000289 __be32 ip6_frag_id;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700290
291 /*
292 * Warning : all fields before dataref are cleared in __alloc_skb()
293 */
294 atomic_t dataref;
295
Johann Baudy69e3c752009-05-18 22:11:22 -0700296 /* Intermediate layers must ensure that destructor_arg
297 * remains valid until skb destructor */
298 void * destructor_arg;
Shirley Maa6686f22011-07-06 12:22:12 +0000299
Eric Dumazetfed66382010-07-22 19:09:08 +0000300 /* must be last field, see pskb_expand_head() */
301 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302};
303
304/* We divide dataref into two halves. The higher 16 bits hold references
305 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700306 * the entire skb->data. A clone of a headerless skb holds the length of
307 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308 *
309 * All users must obey the rule that the skb->data reference count must be
310 * greater than or equal to the payload reference count.
311 *
312 * Holding a reference to the payload part means that the user does not
313 * care about modifications to the header part of skb->data.
314 */
315#define SKB_DATAREF_SHIFT 16
316#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
317
David S. Millerd179cd12005-08-17 14:57:30 -0700318
319enum {
320 SKB_FCLONE_UNAVAILABLE,
321 SKB_FCLONE_ORIG,
322 SKB_FCLONE_CLONE,
323};
324
Herbert Xu79671682006-06-22 02:40:14 -0700325enum {
326 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700327 SKB_GSO_UDP = 1 << 1,
Herbert Xu576a30e2006-06-27 13:22:38 -0700328
329 /* This indicates the skb is from an untrusted source. */
330 SKB_GSO_DODGY = 1 << 2,
Michael Chanb0da85372006-06-29 12:30:00 -0700331
332 /* This indicates the tcp segment has CWR set. */
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700333 SKB_GSO_TCP_ECN = 1 << 3,
334
335 SKB_GSO_TCPV6 = 1 << 4,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800336
337 SKB_GSO_FCOE = 1 << 5,
Pravin B Shelar68c33162013-02-14 14:02:41 +0000338
339 SKB_GSO_GRE = 1 << 6,
Pravin B Shelar73136262013-03-07 13:21:51 +0000340
Eric Dumazetcb32f512013-10-19 11:42:57 -0700341 SKB_GSO_IPIP = 1 << 7,
Simon Horman0d89d202013-05-23 21:02:52 +0000342
Eric Dumazet61c1db72013-10-20 20:47:30 -0700343 SKB_GSO_SIT = 1 << 8,
Eric Dumazetcb32f512013-10-19 11:42:57 -0700344
Eric Dumazet61c1db72013-10-20 20:47:30 -0700345 SKB_GSO_UDP_TUNNEL = 1 << 9,
346
347 SKB_GSO_MPLS = 1 << 10,
Tom Herbert0f4f4ff2014-06-04 17:20:16 -0700348
349 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
Tom Herbert4749c092014-06-04 17:20:23 -0700350
351 SKB_GSO_GRE_CSUM = 1 << 12,
Herbert Xu79671682006-06-22 02:40:14 -0700352};
353
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700354#if BITS_PER_LONG > 32
355#define NET_SKBUFF_DATA_USES_OFFSET 1
356#endif
357
358#ifdef NET_SKBUFF_DATA_USES_OFFSET
359typedef unsigned int sk_buff_data_t;
360#else
361typedef unsigned char *sk_buff_data_t;
362#endif
363
Eric Dumazet363ec39232014-02-26 14:02:11 -0800364/**
365 * struct skb_mstamp - multi resolution time stamps
366 * @stamp_us: timestamp in us resolution
367 * @stamp_jiffies: timestamp in jiffies
368 */
369struct skb_mstamp {
370 union {
371 u64 v64;
372 struct {
373 u32 stamp_us;
374 u32 stamp_jiffies;
375 };
376 };
377};
378
379/**
380 * skb_mstamp_get - get current timestamp
381 * @cl: place to store timestamps
382 */
383static inline void skb_mstamp_get(struct skb_mstamp *cl)
384{
385 u64 val = local_clock();
386
387 do_div(val, NSEC_PER_USEC);
388 cl->stamp_us = (u32)val;
389 cl->stamp_jiffies = (u32)jiffies;
390}
391
392/**
393 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
394 * @t1: pointer to newest sample
395 * @t0: pointer to oldest sample
396 */
397static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
398 const struct skb_mstamp *t0)
399{
400 s32 delta_us = t1->stamp_us - t0->stamp_us;
401 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
402
403 /* If delta_us is negative, this might be because interval is too big,
404 * or local_clock() drift is too big : fallback using jiffies.
405 */
406 if (delta_us <= 0 ||
407 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
408
409 delta_us = jiffies_to_usecs(delta_jiffies);
410
411 return delta_us;
412}
413
414
Linus Torvalds1da177e2005-04-16 15:20:36 -0700415/**
416 * struct sk_buff - socket buffer
417 * @next: Next buffer in list
418 * @prev: Previous buffer in list
Eric Dumazet363ec39232014-02-26 14:02:11 -0800419 * @tstamp: Time we arrived/left
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700420 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700422 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000423 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700424 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700425 * @len: Length of actual data
426 * @data_len: Data length
427 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700428 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700429 * @csum: Checksum (must include start/offset pair)
430 * @csum_start: Offset from skb->head where checksumming should start
431 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700432 * @priority: Packet queueing priority
WANG Cong60ff7462014-05-04 16:39:18 -0700433 * @ignore_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700435 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700436 * @nohdr: Payload reference only, must not modify header
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700437 * @nfctinfo: Relationship of this skb to the connection
Linus Torvalds1da177e2005-04-16 15:20:36 -0700438 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700439 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700440 * @ipvs_property: skbuff is owned by ipvs
Randy Dunlap31729362008-02-18 20:52:13 -0800441 * @peeked: this packet has been seen already, so stats have been
442 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700443 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700444 * @protocol: Packet protocol from driver
445 * @destructor: Destruct function
446 * @nfct: Associated connection, if any
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800448 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700449 * @tc_index: Traffic control index
450 * @tc_verd: traffic control verdict
Tom Herbert61b905d2014-03-24 15:34:47 -0700451 * @hash: the packet hash
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700452 * @queue_mapping: Queue mapping for multiqueue devices
Randy Dunlap553a5672008-04-20 10:51:01 -0700453 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700454 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Tom Herbert61b905d2014-03-24 15:34:47 -0700455 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
Changli Gao4ca24622011-08-19 07:26:44 -0700456 * ports.
Johannes Berg6e3e9392011-11-09 10:15:42 +0100457 * @wifi_acked_valid: wifi_acked was set
458 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000459 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Randy Dunlapf4b8ea72006-06-22 16:00:11 -0700460 * @dma_cookie: a cookie to one of several possible DMA operations
461 * done by skb DMA functions
Eliezer Tamir06021292013-06-10 11:39:50 +0300462 * @napi_id: id of the NAPI struct this skb came from
James Morris984bc162006-06-09 00:29:17 -0700463 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700464 * @mark: Generic packet mark
465 * @dropcount: total number of sk_receive_queue overflows
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000466 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b02008-07-14 22:49:06 -0700467 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000468 * @inner_protocol: Protocol (encapsulation)
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000469 * @inner_transport_header: Inner transport layer header (encapsulation)
470 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000471 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700472 * @transport_header: Transport layer header
473 * @network_header: Network layer header
474 * @mac_header: Link layer header
475 * @tail: Tail pointer
476 * @end: End pointer
477 * @head: Head of buffer
478 * @data: Data head pointer
479 * @truesize: Buffer size
480 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700481 */
482
483struct sk_buff {
484 /* These two members must be first. */
485 struct sk_buff *next;
486 struct sk_buff *prev;
487
Eric Dumazet363ec39232014-02-26 14:02:11 -0800488 union {
489 ktime_t tstamp;
490 struct skb_mstamp skb_mstamp;
491 };
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000492
493 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700494 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496 /*
497 * This is the control buffer. It is free to use for every
498 * layer. Please put your private variables there. If you
499 * want to keep them across layers you have to do a skb_clone()
500 * first. This is owned by whoever has the skb queued ATM.
501 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000502 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700503
Eric Dumazet7fee2262010-05-11 23:19:48 +0000504 unsigned long _skb_refdst;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000505#ifdef CONFIG_XFRM
506 struct sec_path *sp;
507#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700508 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700509 data_len;
510 __u16 mac_len,
511 hdr_len;
Al Viroff1dcad2006-11-20 18:07:29 -0800512 union {
513 __wsum csum;
Herbert Xu663ead32007-04-09 11:59:07 -0700514 struct {
515 __u16 csum_start;
516 __u16 csum_offset;
517 };
Al Viroff1dcad2006-11-20 18:07:29 -0800518 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700519 __u32 priority;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200520 kmemcheck_bitfield_begin(flags1);
WANG Cong60ff7462014-05-04 16:39:18 -0700521 __u8 ignore_df:1,
Thomas Graf1cbb3382005-07-05 14:13:41 -0700522 cloned:1,
523 ip_summed:2,
Harald Welte6869c4d2005-08-09 19:24:19 -0700524 nohdr:1,
525 nfctinfo:3;
David S. Millerd179cd12005-08-17 14:57:30 -0700526 __u8 pkt_type:3,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800527 fclone:2,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700528 ipvs_property:1,
Herbert Xua59322b2007-12-05 01:53:40 -0800529 peeked:1,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700530 nf_trace:1;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200531 kmemcheck_bitfield_end(flags1);
Eric Dumazet4ab408d2010-03-01 03:09:26 +0000532 __be16 protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533
534 void (*destructor)(struct sk_buff *skb);
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800535#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700536 struct nf_conntrack *nfct;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100537#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700538#ifdef CONFIG_BRIDGE_NETFILTER
539 struct nf_bridge_info *nf_bridge;
540#endif
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700541
Eric Dumazet8964be42009-11-20 15:35:04 -0800542 int skb_iif;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000543
Tom Herbert61b905d2014-03-24 15:34:47 -0700544 __u32 hash;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000545
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000546 __be16 vlan_proto;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000547 __u16 vlan_tci;
548
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700550 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700551#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700552 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200555
Eric Dumazet0a148422011-04-20 09:27:32 +0000556 __u16 queue_mapping;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200557 kmemcheck_bitfield_begin(flags2);
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400558#ifdef CONFIG_IPV6_NDISC_NODETYPE
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000559 __u8 ndisc_nodetype:2;
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400560#endif
Mel Gormanc93bdd02012-07-31 16:44:19 -0700561 __u8 pfmemalloc:1;
Tom Herbert3853b582010-11-21 13:17:29 +0000562 __u8 ooo_okay:1;
Tom Herbert61b905d2014-03-24 15:34:47 -0700563 __u8 l4_hash:1;
Johannes Berg6e3e9392011-11-09 10:15:42 +0100564 __u8 wifi_acked_valid:1;
565 __u8 wifi_acked:1;
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000566 __u8 no_fcs:1;
Eric Dumazetd3836f22012-04-27 00:33:38 +0000567 __u8 head_frag:1;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000568 /* Encapsulation protocol and NIC drivers should use
569 * this flag to indicate to each other if the skb contains
570 * encapsulated packet or not and maybe use the inner packet
571 * headers if needed
572 */
573 __u8 encapsulation:1;
Tom Herbert7e2b10c2014-06-04 17:20:02 -0700574 __u8 encap_hdr_csum:1;
Tom Herbert5d0c2b92014-06-10 18:54:13 -0700575 __u8 csum_valid:1;
Tom Herbert7e3cead2014-06-10 18:54:19 -0700576 __u8 csum_complete_sw:1;
577 /* 3/5 bit hole (depending on ndisc_nodetype presence) */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200578 kmemcheck_bitfield_end(flags2);
579
Cong Wange0d10952013-08-01 11:10:25 +0800580#if defined CONFIG_NET_DMA || defined CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir06021292013-06-10 11:39:50 +0300581 union {
582 unsigned int napi_id;
583 dma_cookie_t dma_cookie;
584 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700585#endif
James Morris984bc162006-06-09 00:29:17 -0700586#ifdef CONFIG_NETWORK_SECMARK
587 __u32 secmark;
588#endif
Neil Horman3b885782009-10-12 13:26:31 -0700589 union {
590 __u32 mark;
591 __u32 dropcount;
Eric Dumazet16fad692013-03-14 05:40:32 +0000592 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700593 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700594
Simon Horman0d89d202013-05-23 21:02:52 +0000595 __be16 inner_protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000596 __u16 inner_transport_header;
597 __u16 inner_network_header;
598 __u16 inner_mac_header;
599 __u16 transport_header;
600 __u16 network_header;
601 __u16 mac_header;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700603 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700604 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700605 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700606 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700607 unsigned int truesize;
608 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700609};
610
611#ifdef __KERNEL__
612/*
613 * Handling routines are only of interest to the kernel
614 */
615#include <linux/slab.h>
616
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617
Mel Gormanc93bdd02012-07-31 16:44:19 -0700618#define SKB_ALLOC_FCLONE 0x01
619#define SKB_ALLOC_RX 0x02
620
621/* Returns true if the skb was allocated from PFMEMALLOC reserves */
622static inline bool skb_pfmemalloc(const struct sk_buff *skb)
623{
624 return unlikely(skb->pfmemalloc);
625}
626
Eric Dumazet7fee2262010-05-11 23:19:48 +0000627/*
628 * skb might have a dst pointer attached, refcounted or not.
629 * _skb_refdst low order bit is set if refcount was _not_ taken
630 */
631#define SKB_DST_NOREF 1UL
632#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
633
634/**
635 * skb_dst - returns skb dst_entry
636 * @skb: buffer
637 *
638 * Returns skb dst_entry, regardless of reference taken or not.
639 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000640static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
641{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000642 /* If refdst was not refcounted, check we still are in a
643 * rcu_read_lock section
644 */
645 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
646 !rcu_read_lock_held() &&
647 !rcu_read_lock_bh_held());
648 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000649}
650
Eric Dumazet7fee2262010-05-11 23:19:48 +0000651/**
652 * skb_dst_set - sets skb dst
653 * @skb: buffer
654 * @dst: dst entry
655 *
656 * Sets skb dst, assuming a reference was taken on dst and should
657 * be released by skb_dst_drop()
658 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000659static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
660{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000661 skb->_skb_refdst = (unsigned long)dst;
662}
663
Joe Perches7965bd42013-09-26 14:48:15 -0700664void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
665 bool force);
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200666
667/**
668 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
669 * @skb: buffer
670 * @dst: dst entry
671 *
672 * Sets skb dst, assuming a reference was not taken on dst.
673 * If dst entry is cached, we do not take reference and dst_release
674 * will be avoided by refdst_drop. If dst entry is not cached, we take
675 * reference, so that last dst_release can destroy the dst immediately.
676 */
677static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
678{
679 __skb_dst_set_noref(skb, dst, false);
680}
681
682/**
683 * skb_dst_set_noref_force - sets skb dst, without taking reference
684 * @skb: buffer
685 * @dst: dst entry
686 *
687 * Sets skb dst, assuming a reference was not taken on dst.
688 * No reference is taken and no dst_release will be called. While for
689 * cached dsts deferred reclaim is a basic feature, for entries that are
690 * not cached it is caller's job to guarantee that last dst_release for
691 * provided dst happens when nobody uses it, eg. after a RCU grace period.
692 */
693static inline void skb_dst_set_noref_force(struct sk_buff *skb,
694 struct dst_entry *dst)
695{
696 __skb_dst_set_noref(skb, dst, true);
697}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000698
699/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300700 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000701 * @skb: buffer
702 */
703static inline bool skb_dst_is_noref(const struct sk_buff *skb)
704{
705 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000706}
707
Eric Dumazet511c3f92009-06-02 05:14:27 +0000708static inline struct rtable *skb_rtable(const struct sk_buff *skb)
709{
Eric Dumazetadf30902009-06-02 05:19:30 +0000710 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000711}
712
Joe Perches7965bd42013-09-26 14:48:15 -0700713void kfree_skb(struct sk_buff *skb);
714void kfree_skb_list(struct sk_buff *segs);
715void skb_tx_error(struct sk_buff *skb);
716void consume_skb(struct sk_buff *skb);
717void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000718extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000719
Joe Perches7965bd42013-09-26 14:48:15 -0700720void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
721bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
722 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000723
Joe Perches7965bd42013-09-26 14:48:15 -0700724struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
725 int node);
726struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700727static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100728 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700729{
Eric Dumazet564824b2010-10-11 19:05:25 +0000730 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700731}
732
733static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100734 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700735{
Mel Gormanc93bdd02012-07-31 16:44:19 -0700736 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700737}
738
Joe Perches7965bd42013-09-26 14:48:15 -0700739struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
Patrick McHardy0ebd0ac2013-04-17 06:46:58 +0000740static inline struct sk_buff *alloc_skb_head(gfp_t priority)
741{
742 return __alloc_skb_head(priority, -1);
743}
744
Joe Perches7965bd42013-09-26 14:48:15 -0700745struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
746int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
747struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
748struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
Octavian Purdilabad93e92014-06-12 01:36:26 +0300749struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
750 gfp_t gfp_mask, bool fclone);
751static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
752 gfp_t gfp_mask)
753{
754 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
755}
Eric Dumazet117632e2011-12-03 21:39:53 +0000756
Joe Perches7965bd42013-09-26 14:48:15 -0700757int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
758struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
759 unsigned int headroom);
760struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
761 int newtailroom, gfp_t priority);
Fan Du25a91d82014-01-18 09:54:23 +0800762int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
763 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -0700764int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
765 int len);
766int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
767int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000768#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700769
Joe Perches7965bd42013-09-26 14:48:15 -0700770int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
771 int getfrag(void *from, char *to, int offset,
772 int len, int odd, struct sk_buff *skb),
773 void *from, int length);
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700774
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800775struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700776 __u32 lower_offset;
777 __u32 upper_offset;
778 __u32 frag_idx;
779 __u32 stepped_offset;
780 struct sk_buff *root_skb;
781 struct sk_buff *cur_skb;
782 __u8 *frag_data;
783};
784
Joe Perches7965bd42013-09-26 14:48:15 -0700785void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
786 unsigned int to, struct skb_seq_state *st);
787unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
788 struct skb_seq_state *st);
789void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -0700790
Joe Perches7965bd42013-09-26 14:48:15 -0700791unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
792 unsigned int to, struct ts_config *config,
793 struct ts_state *state);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700794
Tom Herbert09323cc2013-12-15 22:16:19 -0800795/*
796 * Packet hash types specify the type of hash in skb_set_hash.
797 *
798 * Hash types refer to the protocol layer addresses which are used to
799 * construct a packet's hash. The hashes are used to differentiate or identify
800 * flows of the protocol layer for the hash type. Hash types are either
801 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
802 *
803 * Properties of hashes:
804 *
805 * 1) Two packets in different flows have different hash values
806 * 2) Two packets in the same flow should have the same hash value
807 *
808 * A hash at a higher layer is considered to be more specific. A driver should
809 * set the most specific hash possible.
810 *
811 * A driver cannot indicate a more specific hash than the layer at which a hash
812 * was computed. For instance an L3 hash cannot be set as an L4 hash.
813 *
814 * A driver may indicate a hash level which is less specific than the
815 * actual layer the hash was computed on. For instance, a hash computed
816 * at L4 may be considered an L3 hash. This should only be done if the
817 * driver can't unambiguously determine that the HW computed the hash at
818 * the higher layer. Note that the "should" in the second property above
819 * permits this.
820 */
821enum pkt_hash_types {
822 PKT_HASH_TYPE_NONE, /* Undefined type */
823 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
824 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
825 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
826};
827
828static inline void
829skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
830{
Tom Herbert61b905d2014-03-24 15:34:47 -0700831 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
832 skb->hash = hash;
Tom Herbert09323cc2013-12-15 22:16:19 -0800833}
834
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800835void __skb_get_hash(struct sk_buff *skb);
836static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000837{
Tom Herbert61b905d2014-03-24 15:34:47 -0700838 if (!skb->l4_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800839 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000840
Tom Herbert61b905d2014-03-24 15:34:47 -0700841 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000842}
843
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800844static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
845{
Tom Herbert61b905d2014-03-24 15:34:47 -0700846 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800847}
848
Tom Herbert7539fad2013-12-15 22:12:18 -0800849static inline void skb_clear_hash(struct sk_buff *skb)
850{
Tom Herbert61b905d2014-03-24 15:34:47 -0700851 skb->hash = 0;
852 skb->l4_hash = 0;
Tom Herbert7539fad2013-12-15 22:12:18 -0800853}
854
855static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
856{
Tom Herbert61b905d2014-03-24 15:34:47 -0700857 if (!skb->l4_hash)
Tom Herbert7539fad2013-12-15 22:12:18 -0800858 skb_clear_hash(skb);
859}
860
Tom Herbert3df7a742013-12-15 22:16:29 -0800861static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
862{
Tom Herbert61b905d2014-03-24 15:34:47 -0700863 to->hash = from->hash;
864 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -0800865};
866
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700867#ifdef NET_SKBUFF_DATA_USES_OFFSET
868static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
869{
870 return skb->head + skb->end;
871}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000872
873static inline unsigned int skb_end_offset(const struct sk_buff *skb)
874{
875 return skb->end;
876}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700877#else
878static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
879{
880 return skb->end;
881}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000882
883static inline unsigned int skb_end_offset(const struct sk_buff *skb)
884{
885 return skb->end - skb->head;
886}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700887#endif
888
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700890#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700891
Patrick Ohlyac45f602009-02-12 05:03:37 +0000892static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
893{
894 return &skb_shinfo(skb)->hwtstamps;
895}
896
Linus Torvalds1da177e2005-04-16 15:20:36 -0700897/**
898 * skb_queue_empty - check if a queue is empty
899 * @list: queue head
900 *
901 * Returns true if the queue is empty, false otherwise.
902 */
903static inline int skb_queue_empty(const struct sk_buff_head *list)
904{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100905 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700906}
907
908/**
David S. Millerfc7ebb22008-09-23 00:34:07 -0700909 * skb_queue_is_last - check if skb is the last entry in the queue
910 * @list: queue head
911 * @skb: buffer
912 *
913 * Returns true if @skb is the last buffer on the list.
914 */
915static inline bool skb_queue_is_last(const struct sk_buff_head *list,
916 const struct sk_buff *skb)
917{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100918 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -0700919}
920
921/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800922 * skb_queue_is_first - check if skb is the first entry in the queue
923 * @list: queue head
924 * @skb: buffer
925 *
926 * Returns true if @skb is the first buffer on the list.
927 */
928static inline bool skb_queue_is_first(const struct sk_buff_head *list,
929 const struct sk_buff *skb)
930{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100931 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800932}
933
934/**
David S. Miller249c8b42008-09-23 00:44:42 -0700935 * skb_queue_next - return the next packet in the queue
936 * @list: queue head
937 * @skb: current buffer
938 *
939 * Return the next packet in @list after @skb. It is only valid to
940 * call this if skb_queue_is_last() evaluates to false.
941 */
942static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
943 const struct sk_buff *skb)
944{
945 /* This BUG_ON may seem severe, but if we just return then we
946 * are going to dereference garbage.
947 */
948 BUG_ON(skb_queue_is_last(list, skb));
949 return skb->next;
950}
951
952/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800953 * skb_queue_prev - return the prev packet in the queue
954 * @list: queue head
955 * @skb: current buffer
956 *
957 * Return the prev packet in @list before @skb. It is only valid to
958 * call this if skb_queue_is_first() evaluates to false.
959 */
960static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
961 const struct sk_buff *skb)
962{
963 /* This BUG_ON may seem severe, but if we just return then we
964 * are going to dereference garbage.
965 */
966 BUG_ON(skb_queue_is_first(list, skb));
967 return skb->prev;
968}
969
970/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700971 * skb_get - reference buffer
972 * @skb: buffer to reference
973 *
974 * Makes another reference to a socket buffer and returns a pointer
975 * to the buffer.
976 */
977static inline struct sk_buff *skb_get(struct sk_buff *skb)
978{
979 atomic_inc(&skb->users);
980 return skb;
981}
982
983/*
984 * If users == 1, we are the only owner and are can avoid redundant
985 * atomic change.
986 */
987
988/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700989 * skb_cloned - is the buffer a clone
990 * @skb: buffer to check
991 *
992 * Returns true if the buffer was generated with skb_clone() and is
993 * one of multiple shared copies of the buffer. Cloned buffers are
994 * shared data so must not be written to under normal circumstances.
995 */
996static inline int skb_cloned(const struct sk_buff *skb)
997{
998 return skb->cloned &&
999 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1000}
1001
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001002static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1003{
1004 might_sleep_if(pri & __GFP_WAIT);
1005
1006 if (skb_cloned(skb))
1007 return pskb_expand_head(skb, 0, 0, pri);
1008
1009 return 0;
1010}
1011
Linus Torvalds1da177e2005-04-16 15:20:36 -07001012/**
1013 * skb_header_cloned - is the header a clone
1014 * @skb: buffer to check
1015 *
1016 * Returns true if modifying the header part of the buffer requires
1017 * the data to be copied.
1018 */
1019static inline int skb_header_cloned(const struct sk_buff *skb)
1020{
1021 int dataref;
1022
1023 if (!skb->cloned)
1024 return 0;
1025
1026 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1027 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1028 return dataref != 1;
1029}
1030
1031/**
1032 * skb_header_release - release reference to header
1033 * @skb: buffer to operate on
1034 *
1035 * Drop a reference to the header part of the buffer. This is done
1036 * by acquiring a payload reference. You must not read from the header
1037 * part of skb->data after this.
1038 */
1039static inline void skb_header_release(struct sk_buff *skb)
1040{
1041 BUG_ON(skb->nohdr);
1042 skb->nohdr = 1;
1043 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1044}
1045
1046/**
1047 * skb_shared - is the buffer shared
1048 * @skb: buffer to check
1049 *
1050 * Returns true if more than one person has a reference to this
1051 * buffer.
1052 */
1053static inline int skb_shared(const struct sk_buff *skb)
1054{
1055 return atomic_read(&skb->users) != 1;
1056}
1057
1058/**
1059 * skb_share_check - check if buffer is shared and if so clone it
1060 * @skb: buffer to check
1061 * @pri: priority for memory allocation
1062 *
1063 * If the buffer is shared the buffer is cloned and the old copy
1064 * drops a reference. A new clone with a single reference is returned.
1065 * If the buffer is not shared the original buffer is returned. When
1066 * being called from interrupt status or with spinlocks held pri must
1067 * be GFP_ATOMIC.
1068 *
1069 * NULL is returned on a memory allocation failure.
1070 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001071static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001072{
1073 might_sleep_if(pri & __GFP_WAIT);
1074 if (skb_shared(skb)) {
1075 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001076
1077 if (likely(nskb))
1078 consume_skb(skb);
1079 else
1080 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001081 skb = nskb;
1082 }
1083 return skb;
1084}
1085
1086/*
1087 * Copy shared buffers into a new sk_buff. We effectively do COW on
1088 * packets to handle cases where we have a local reader and forward
1089 * and a couple of other messy ones. The normal one is tcpdumping
1090 * a packet thats being forwarded.
1091 */
1092
1093/**
1094 * skb_unshare - make a copy of a shared buffer
1095 * @skb: buffer to check
1096 * @pri: priority for memory allocation
1097 *
1098 * If the socket buffer is a clone then this function creates a new
1099 * copy of the data, drops a reference count on the old copy and returns
1100 * the new copy with the reference count at 1. If the buffer is not a clone
1101 * the original buffer is returned. When called with a spinlock held or
1102 * from interrupt state @pri must be %GFP_ATOMIC
1103 *
1104 * %NULL is returned on a memory allocation failure.
1105 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001106static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001107 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001108{
1109 might_sleep_if(pri & __GFP_WAIT);
1110 if (skb_cloned(skb)) {
1111 struct sk_buff *nskb = skb_copy(skb, pri);
1112 kfree_skb(skb); /* Free our shared copy */
1113 skb = nskb;
1114 }
1115 return skb;
1116}
1117
1118/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001119 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001120 * @list_: list to peek at
1121 *
1122 * Peek an &sk_buff. Unlike most other operations you _MUST_
1123 * be careful with this one. A peek leaves the buffer on the
1124 * list and someone else may run off with it. You must hold
1125 * the appropriate locks or have a private queue to do this.
1126 *
1127 * Returns %NULL for an empty list or a pointer to the head element.
1128 * The reference count is not incremented and the reference is therefore
1129 * volatile. Use with caution.
1130 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001131static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001132{
Eric Dumazet18d07002012-04-30 16:31:46 +00001133 struct sk_buff *skb = list_->next;
1134
1135 if (skb == (struct sk_buff *)list_)
1136 skb = NULL;
1137 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001138}
1139
1140/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001141 * skb_peek_next - peek skb following the given one from a queue
1142 * @skb: skb to start from
1143 * @list_: list to peek at
1144 *
1145 * Returns %NULL when the end of the list is met or a pointer to the
1146 * next element. The reference count is not incremented and the
1147 * reference is therefore volatile. Use with caution.
1148 */
1149static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1150 const struct sk_buff_head *list_)
1151{
1152 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001153
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001154 if (next == (struct sk_buff *)list_)
1155 next = NULL;
1156 return next;
1157}
1158
1159/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001160 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001161 * @list_: list to peek at
1162 *
1163 * Peek an &sk_buff. Unlike most other operations you _MUST_
1164 * be careful with this one. A peek leaves the buffer on the
1165 * list and someone else may run off with it. You must hold
1166 * the appropriate locks or have a private queue to do this.
1167 *
1168 * Returns %NULL for an empty list or a pointer to the tail element.
1169 * The reference count is not incremented and the reference is therefore
1170 * volatile. Use with caution.
1171 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001172static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001173{
Eric Dumazet18d07002012-04-30 16:31:46 +00001174 struct sk_buff *skb = list_->prev;
1175
1176 if (skb == (struct sk_buff *)list_)
1177 skb = NULL;
1178 return skb;
1179
Linus Torvalds1da177e2005-04-16 15:20:36 -07001180}
1181
1182/**
1183 * skb_queue_len - get queue length
1184 * @list_: list to measure
1185 *
1186 * Return the length of an &sk_buff queue.
1187 */
1188static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1189{
1190 return list_->qlen;
1191}
1192
David S. Miller67fed452008-09-21 22:36:24 -07001193/**
1194 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1195 * @list: queue to initialize
1196 *
1197 * This initializes only the list and queue length aspects of
1198 * an sk_buff_head object. This allows to initialize the list
1199 * aspects of an sk_buff_head without reinitializing things like
1200 * the spinlock. It can also be used for on-stack sk_buff_head
1201 * objects where the spinlock is known to not be used.
1202 */
1203static inline void __skb_queue_head_init(struct sk_buff_head *list)
1204{
1205 list->prev = list->next = (struct sk_buff *)list;
1206 list->qlen = 0;
1207}
1208
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001209/*
1210 * This function creates a split out lock class for each invocation;
1211 * this is needed for now since a whole lot of users of the skb-queue
1212 * infrastructure in drivers have different locking usage (in hardirq)
1213 * than the networking core (in softirq only). In the long run either the
1214 * network layer or drivers should need annotation to consolidate the
1215 * main types of usage into 3 classes.
1216 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001217static inline void skb_queue_head_init(struct sk_buff_head *list)
1218{
1219 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001220 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001221}
1222
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001223static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1224 struct lock_class_key *class)
1225{
1226 skb_queue_head_init(list);
1227 lockdep_set_class(&list->lock, class);
1228}
1229
Linus Torvalds1da177e2005-04-16 15:20:36 -07001230/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001231 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001232 *
1233 * The "__skb_xxxx()" functions are the non-atomic ones that
1234 * can only be called with interrupts disabled.
1235 */
Joe Perches7965bd42013-09-26 14:48:15 -07001236void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1237 struct sk_buff_head *list);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001238static inline void __skb_insert(struct sk_buff *newsk,
1239 struct sk_buff *prev, struct sk_buff *next,
1240 struct sk_buff_head *list)
1241{
1242 newsk->next = next;
1243 newsk->prev = prev;
1244 next->prev = prev->next = newsk;
1245 list->qlen++;
1246}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001247
David S. Miller67fed452008-09-21 22:36:24 -07001248static inline void __skb_queue_splice(const struct sk_buff_head *list,
1249 struct sk_buff *prev,
1250 struct sk_buff *next)
1251{
1252 struct sk_buff *first = list->next;
1253 struct sk_buff *last = list->prev;
1254
1255 first->prev = prev;
1256 prev->next = first;
1257
1258 last->next = next;
1259 next->prev = last;
1260}
1261
1262/**
1263 * skb_queue_splice - join two skb lists, this is designed for stacks
1264 * @list: the new list to add
1265 * @head: the place to add it in the first list
1266 */
1267static inline void skb_queue_splice(const struct sk_buff_head *list,
1268 struct sk_buff_head *head)
1269{
1270 if (!skb_queue_empty(list)) {
1271 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001272 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001273 }
1274}
1275
1276/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001277 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001278 * @list: the new list to add
1279 * @head: the place to add it in the first list
1280 *
1281 * The list at @list is reinitialised
1282 */
1283static inline void skb_queue_splice_init(struct sk_buff_head *list,
1284 struct sk_buff_head *head)
1285{
1286 if (!skb_queue_empty(list)) {
1287 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001288 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001289 __skb_queue_head_init(list);
1290 }
1291}
1292
1293/**
1294 * skb_queue_splice_tail - join two skb lists, each list being a queue
1295 * @list: the new list to add
1296 * @head: the place to add it in the first list
1297 */
1298static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1299 struct sk_buff_head *head)
1300{
1301 if (!skb_queue_empty(list)) {
1302 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001303 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001304 }
1305}
1306
1307/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001308 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001309 * @list: the new list to add
1310 * @head: the place to add it in the first list
1311 *
1312 * Each of the lists is a queue.
1313 * The list at @list is reinitialised
1314 */
1315static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1316 struct sk_buff_head *head)
1317{
1318 if (!skb_queue_empty(list)) {
1319 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001320 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001321 __skb_queue_head_init(list);
1322 }
1323}
1324
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001326 * __skb_queue_after - queue a buffer at the list head
1327 * @list: list to use
1328 * @prev: place after this buffer
1329 * @newsk: buffer to queue
1330 *
1331 * Queue a buffer int the middle of a list. This function takes no locks
1332 * and you must therefore hold required locks before calling it.
1333 *
1334 * A buffer cannot be placed on two lists at the same time.
1335 */
1336static inline void __skb_queue_after(struct sk_buff_head *list,
1337 struct sk_buff *prev,
1338 struct sk_buff *newsk)
1339{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001340 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001341}
1342
Joe Perches7965bd42013-09-26 14:48:15 -07001343void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1344 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001345
Gerrit Renkerf5572852008-04-14 00:05:28 -07001346static inline void __skb_queue_before(struct sk_buff_head *list,
1347 struct sk_buff *next,
1348 struct sk_buff *newsk)
1349{
1350 __skb_insert(newsk, next->prev, next, list);
1351}
1352
Stephen Hemminger300ce172005-10-30 13:47:34 -08001353/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001354 * __skb_queue_head - queue a buffer at the list head
1355 * @list: list to use
1356 * @newsk: buffer to queue
1357 *
1358 * Queue a buffer at the start of a list. This function takes no locks
1359 * and you must therefore hold required locks before calling it.
1360 *
1361 * A buffer cannot be placed on two lists at the same time.
1362 */
Joe Perches7965bd42013-09-26 14:48:15 -07001363void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001364static inline void __skb_queue_head(struct sk_buff_head *list,
1365 struct sk_buff *newsk)
1366{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001367 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001368}
1369
1370/**
1371 * __skb_queue_tail - queue a buffer at the list tail
1372 * @list: list to use
1373 * @newsk: buffer to queue
1374 *
1375 * Queue a buffer at the end of a list. This function takes no locks
1376 * and you must therefore hold required locks before calling it.
1377 *
1378 * A buffer cannot be placed on two lists at the same time.
1379 */
Joe Perches7965bd42013-09-26 14:48:15 -07001380void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001381static inline void __skb_queue_tail(struct sk_buff_head *list,
1382 struct sk_buff *newsk)
1383{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001384 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385}
1386
Linus Torvalds1da177e2005-04-16 15:20:36 -07001387/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001388 * remove sk_buff from list. _Must_ be called atomically, and with
1389 * the list known..
1390 */
Joe Perches7965bd42013-09-26 14:48:15 -07001391void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1393{
1394 struct sk_buff *next, *prev;
1395
1396 list->qlen--;
1397 next = skb->next;
1398 prev = skb->prev;
1399 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001400 next->prev = prev;
1401 prev->next = next;
1402}
1403
Gerrit Renkerf525c062008-04-14 00:04:12 -07001404/**
1405 * __skb_dequeue - remove from the head of the queue
1406 * @list: list to dequeue from
1407 *
1408 * Remove the head of the list. This function does not take any locks
1409 * so must be used with appropriate locks held only. The head item is
1410 * returned or %NULL if the list is empty.
1411 */
Joe Perches7965bd42013-09-26 14:48:15 -07001412struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Gerrit Renkerf525c062008-04-14 00:04:12 -07001413static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1414{
1415 struct sk_buff *skb = skb_peek(list);
1416 if (skb)
1417 __skb_unlink(skb, list);
1418 return skb;
1419}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001420
1421/**
1422 * __skb_dequeue_tail - remove from the tail of the queue
1423 * @list: list to dequeue from
1424 *
1425 * Remove the tail of the list. This function does not take any locks
1426 * so must be used with appropriate locks held only. The tail item is
1427 * returned or %NULL if the list is empty.
1428 */
Joe Perches7965bd42013-09-26 14:48:15 -07001429struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001430static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1431{
1432 struct sk_buff *skb = skb_peek_tail(list);
1433 if (skb)
1434 __skb_unlink(skb, list);
1435 return skb;
1436}
1437
1438
David S. Millerbdcc0922012-03-07 20:53:36 -05001439static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001440{
1441 return skb->data_len;
1442}
1443
1444static inline unsigned int skb_headlen(const struct sk_buff *skb)
1445{
1446 return skb->len - skb->data_len;
1447}
1448
1449static inline int skb_pagelen(const struct sk_buff *skb)
1450{
1451 int i, len = 0;
1452
1453 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001454 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001455 return len + skb_headlen(skb);
1456}
1457
Ian Campbell131ea662011-08-19 06:25:00 +00001458/**
1459 * __skb_fill_page_desc - initialise a paged fragment in an skb
1460 * @skb: buffer containing fragment to be initialised
1461 * @i: paged fragment index to initialise
1462 * @page: the page to use for this fragment
1463 * @off: the offset to the data with @page
1464 * @size: the length of the data
1465 *
1466 * Initialises the @i'th fragment of @skb to point to &size bytes at
1467 * offset @off within @page.
1468 *
1469 * Does not take any additional reference on the fragment.
1470 */
1471static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1472 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001473{
1474 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1475
Mel Gormanc48a11c2012-07-31 16:44:23 -07001476 /*
1477 * Propagate page->pfmemalloc to the skb if we can. The problem is
1478 * that not all callers have unique ownership of the page. If
1479 * pfmemalloc is set, we check the mapping as a mapping implies
1480 * page->index is set (index and pfmemalloc share space).
1481 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1482 * do not lose pfmemalloc information as the pages would not be
1483 * allocated using __GFP_MEMALLOC.
1484 */
Ian Campbella8605c62011-10-19 23:01:49 +00001485 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001486 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001487 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001488
1489 page = compound_head(page);
1490 if (page->pfmemalloc && !page->mapping)
1491 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00001492}
1493
1494/**
1495 * skb_fill_page_desc - initialise a paged fragment in an skb
1496 * @skb: buffer containing fragment to be initialised
1497 * @i: paged fragment index to initialise
1498 * @page: the page to use for this fragment
1499 * @off: the offset to the data with @page
1500 * @size: the length of the data
1501 *
1502 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01001503 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00001504 * addition updates @skb such that @i is the last fragment.
1505 *
1506 * Does not take any additional reference on the fragment.
1507 */
1508static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1509 struct page *page, int off, int size)
1510{
1511 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001512 skb_shinfo(skb)->nr_frags = i + 1;
1513}
1514
Joe Perches7965bd42013-09-26 14:48:15 -07001515void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1516 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001517
Jason Wangf8e617e2013-11-01 14:07:47 +08001518void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1519 unsigned int truesize);
1520
Linus Torvalds1da177e2005-04-16 15:20:36 -07001521#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001522#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1524
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001525#ifdef NET_SKBUFF_DATA_USES_OFFSET
1526static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1527{
1528 return skb->head + skb->tail;
1529}
1530
1531static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1532{
1533 skb->tail = skb->data - skb->head;
1534}
1535
1536static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1537{
1538 skb_reset_tail_pointer(skb);
1539 skb->tail += offset;
1540}
Simon Horman7cc46192013-05-28 20:34:29 +00001541
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001542#else /* NET_SKBUFF_DATA_USES_OFFSET */
1543static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1544{
1545 return skb->tail;
1546}
1547
1548static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1549{
1550 skb->tail = skb->data;
1551}
1552
1553static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1554{
1555 skb->tail = skb->data + offset;
1556}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001557
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001558#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1559
Linus Torvalds1da177e2005-04-16 15:20:36 -07001560/*
1561 * Add data to an sk_buff
1562 */
Mathias Krause0c7ddf32013-11-07 14:18:24 +01001563unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001564unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001565static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1566{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001567 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001568 SKB_LINEAR_ASSERT(skb);
1569 skb->tail += len;
1570 skb->len += len;
1571 return tmp;
1572}
1573
Joe Perches7965bd42013-09-26 14:48:15 -07001574unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001575static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1576{
1577 skb->data -= len;
1578 skb->len += len;
1579 return skb->data;
1580}
1581
Joe Perches7965bd42013-09-26 14:48:15 -07001582unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1584{
1585 skb->len -= len;
1586 BUG_ON(skb->len < skb->data_len);
1587 return skb->data += len;
1588}
1589
David S. Miller47d29642010-05-02 02:21:44 -07001590static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1591{
1592 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1593}
1594
Joe Perches7965bd42013-09-26 14:48:15 -07001595unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001596
1597static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1598{
1599 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001600 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001601 return NULL;
1602 skb->len -= len;
1603 return skb->data += len;
1604}
1605
1606static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1607{
1608 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1609}
1610
1611static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1612{
1613 if (likely(len <= skb_headlen(skb)))
1614 return 1;
1615 if (unlikely(len > skb->len))
1616 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001617 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001618}
1619
1620/**
1621 * skb_headroom - bytes at buffer head
1622 * @skb: buffer to check
1623 *
1624 * Return the number of bytes of free space at the head of an &sk_buff.
1625 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001626static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001627{
1628 return skb->data - skb->head;
1629}
1630
1631/**
1632 * skb_tailroom - bytes at buffer end
1633 * @skb: buffer to check
1634 *
1635 * Return the number of bytes of free space at the tail of an sk_buff
1636 */
1637static inline int skb_tailroom(const struct sk_buff *skb)
1638{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001639 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001640}
1641
1642/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00001643 * skb_availroom - bytes at buffer end
1644 * @skb: buffer to check
1645 *
1646 * Return the number of bytes of free space at the tail of an sk_buff
1647 * allocated by sk_stream_alloc()
1648 */
1649static inline int skb_availroom(const struct sk_buff *skb)
1650{
Eric Dumazet16fad692013-03-14 05:40:32 +00001651 if (skb_is_nonlinear(skb))
1652 return 0;
1653
1654 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00001655}
1656
1657/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001658 * skb_reserve - adjust headroom
1659 * @skb: buffer to alter
1660 * @len: bytes to move
1661 *
1662 * Increase the headroom of an empty &sk_buff by reducing the tail
1663 * room. This is only allowed for an empty buffer.
1664 */
David S. Miller8243126c2006-01-17 02:54:21 -08001665static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001666{
1667 skb->data += len;
1668 skb->tail += len;
1669}
1670
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001671static inline void skb_reset_inner_headers(struct sk_buff *skb)
1672{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001673 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001674 skb->inner_network_header = skb->network_header;
1675 skb->inner_transport_header = skb->transport_header;
1676}
1677
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00001678static inline void skb_reset_mac_len(struct sk_buff *skb)
1679{
1680 skb->mac_len = skb->network_header - skb->mac_header;
1681}
1682
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001683static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1684 *skb)
1685{
1686 return skb->head + skb->inner_transport_header;
1687}
1688
1689static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1690{
1691 skb->inner_transport_header = skb->data - skb->head;
1692}
1693
1694static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1695 const int offset)
1696{
1697 skb_reset_inner_transport_header(skb);
1698 skb->inner_transport_header += offset;
1699}
1700
1701static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1702{
1703 return skb->head + skb->inner_network_header;
1704}
1705
1706static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1707{
1708 skb->inner_network_header = skb->data - skb->head;
1709}
1710
1711static inline void skb_set_inner_network_header(struct sk_buff *skb,
1712 const int offset)
1713{
1714 skb_reset_inner_network_header(skb);
1715 skb->inner_network_header += offset;
1716}
1717
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001718static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1719{
1720 return skb->head + skb->inner_mac_header;
1721}
1722
1723static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1724{
1725 skb->inner_mac_header = skb->data - skb->head;
1726}
1727
1728static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1729 const int offset)
1730{
1731 skb_reset_inner_mac_header(skb);
1732 skb->inner_mac_header += offset;
1733}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001734static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1735{
Cong Wang35d04612013-05-29 15:16:05 +08001736 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001737}
1738
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001739static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1740{
1741 return skb->head + skb->transport_header;
1742}
1743
1744static inline void skb_reset_transport_header(struct sk_buff *skb)
1745{
1746 skb->transport_header = skb->data - skb->head;
1747}
1748
1749static inline void skb_set_transport_header(struct sk_buff *skb,
1750 const int offset)
1751{
1752 skb_reset_transport_header(skb);
1753 skb->transport_header += offset;
1754}
1755
1756static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1757{
1758 return skb->head + skb->network_header;
1759}
1760
1761static inline void skb_reset_network_header(struct sk_buff *skb)
1762{
1763 skb->network_header = skb->data - skb->head;
1764}
1765
1766static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1767{
1768 skb_reset_network_header(skb);
1769 skb->network_header += offset;
1770}
1771
1772static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1773{
1774 return skb->head + skb->mac_header;
1775}
1776
1777static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1778{
Cong Wang35d04612013-05-29 15:16:05 +08001779 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001780}
1781
1782static inline void skb_reset_mac_header(struct sk_buff *skb)
1783{
1784 skb->mac_header = skb->data - skb->head;
1785}
1786
1787static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1788{
1789 skb_reset_mac_header(skb);
1790 skb->mac_header += offset;
1791}
1792
Timo Teräs0e3da5b2013-12-16 11:02:09 +02001793static inline void skb_pop_mac_header(struct sk_buff *skb)
1794{
1795 skb->mac_header = skb->network_header;
1796}
1797
Ying Xuefbbdb8f2013-03-27 16:46:06 +00001798static inline void skb_probe_transport_header(struct sk_buff *skb,
1799 const int offset_hint)
1800{
1801 struct flow_keys keys;
1802
1803 if (skb_transport_header_was_set(skb))
1804 return;
1805 else if (skb_flow_dissect(skb, &keys))
1806 skb_set_transport_header(skb, keys.thoff);
1807 else
1808 skb_set_transport_header(skb, offset_hint);
1809}
1810
Eric Dumazet03606892012-02-23 10:55:02 +00001811static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1812{
1813 if (skb_mac_header_was_set(skb)) {
1814 const unsigned char *old_mac = skb_mac_header(skb);
1815
1816 skb_set_mac_header(skb, -skb->mac_len);
1817 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1818 }
1819}
1820
Michał Mirosław04fb4512010-12-14 15:24:08 +00001821static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1822{
1823 return skb->csum_start - skb_headroom(skb);
1824}
1825
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001826static inline int skb_transport_offset(const struct sk_buff *skb)
1827{
1828 return skb_transport_header(skb) - skb->data;
1829}
1830
1831static inline u32 skb_network_header_len(const struct sk_buff *skb)
1832{
1833 return skb->transport_header - skb->network_header;
1834}
1835
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001836static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1837{
1838 return skb->inner_transport_header - skb->inner_network_header;
1839}
1840
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001841static inline int skb_network_offset(const struct sk_buff *skb)
1842{
1843 return skb_network_header(skb) - skb->data;
1844}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001845
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001846static inline int skb_inner_network_offset(const struct sk_buff *skb)
1847{
1848 return skb_inner_network_header(skb) - skb->data;
1849}
1850
Changli Gaof9599ce2010-08-04 04:43:44 +00001851static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1852{
1853 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1854}
1855
Linus Torvalds1da177e2005-04-16 15:20:36 -07001856/*
1857 * CPUs often take a performance hit when accessing unaligned memory
1858 * locations. The actual performance hit varies, it can be small if the
1859 * hardware handles it or large if we have to take an exception and fix it
1860 * in software.
1861 *
1862 * Since an ethernet header is 14 bytes network drivers often end up with
1863 * the IP header at an unaligned offset. The IP header can be aligned by
1864 * shifting the start of the packet by 2 bytes. Drivers should do this
1865 * with:
1866 *
Tobias Klauser8660c122009-07-13 22:48:16 +00001867 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001868 *
1869 * The downside to this alignment of the IP header is that the DMA is now
1870 * unaligned. On some architectures the cost of an unaligned DMA is high
1871 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00001872 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001873 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1874 * to be overridden.
1875 */
1876#ifndef NET_IP_ALIGN
1877#define NET_IP_ALIGN 2
1878#endif
1879
Anton Blanchard025be812006-03-31 02:27:06 -08001880/*
1881 * The networking layer reserves some headroom in skb data (via
1882 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1883 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08001884 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08001885 *
1886 * Unfortunately this headroom changes the DMA alignment of the resulting
1887 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1888 * on some architectures. An architecture can override this value,
1889 * perhaps setting it to a cacheline in size (since that will maintain
1890 * cacheline alignment of the DMA). It must be a power of 2.
1891 *
David S. Millerd6301d32009-02-08 19:24:13 -08001892 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08001893 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07001894 *
1895 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1896 * to reduce average number of cache lines per packet.
1897 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07001898 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08001899 */
1900#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07001901#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08001902#endif
1903
Joe Perches7965bd42013-09-26 14:48:15 -07001904int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001905
1906static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1907{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00001908 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07001909 WARN_ON(1);
1910 return;
1911 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001912 skb->len = len;
1913 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001914}
1915
Joe Perches7965bd42013-09-26 14:48:15 -07001916void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001917
1918static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1919{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001920 if (skb->data_len)
1921 return ___pskb_trim(skb, len);
1922 __skb_trim(skb, len);
1923 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001924}
1925
1926static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1927{
1928 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1929}
1930
1931/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07001932 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1933 * @skb: buffer to alter
1934 * @len: new length
1935 *
1936 * This is identical to pskb_trim except that the caller knows that
1937 * the skb is not cloned so we should never get an error due to out-
1938 * of-memory.
1939 */
1940static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1941{
1942 int err = pskb_trim(skb, len);
1943 BUG_ON(err);
1944}
1945
1946/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001947 * skb_orphan - orphan a buffer
1948 * @skb: buffer to orphan
1949 *
1950 * If a buffer currently has an owner then we call the owner's
1951 * destructor function and make the @skb unowned. The buffer continues
1952 * to exist but is no longer charged to its former owner.
1953 */
1954static inline void skb_orphan(struct sk_buff *skb)
1955{
Eric Dumazetc34a7612013-07-30 16:11:15 -07001956 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07001958 skb->destructor = NULL;
1959 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07001960 } else {
1961 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07001962 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001963}
1964
1965/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00001966 * skb_orphan_frags - orphan the frags contained in a buffer
1967 * @skb: buffer to orphan frags from
1968 * @gfp_mask: allocation mask for replacement pages
1969 *
1970 * For each frag in the SKB which needs a destructor (i.e. has an
1971 * owner) create a copy of that frag and release the original
1972 * page by calling the destructor.
1973 */
1974static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1975{
1976 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1977 return 0;
1978 return skb_copy_ubufs(skb, gfp_mask);
1979}
1980
1981/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001982 * __skb_queue_purge - empty a list
1983 * @list: list to empty
1984 *
1985 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1986 * the list and one reference dropped. This function does not take the
1987 * list lock and the caller must hold the relevant locks to use it.
1988 */
Joe Perches7965bd42013-09-26 14:48:15 -07001989void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001990static inline void __skb_queue_purge(struct sk_buff_head *list)
1991{
1992 struct sk_buff *skb;
1993 while ((skb = __skb_dequeue(list)) != NULL)
1994 kfree_skb(skb);
1995}
1996
Alexander Duycke5e67302013-02-08 10:17:15 +00001997#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
1998#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
1999#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
2000
Joe Perches7965bd42013-09-26 14:48:15 -07002001void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002002
Joe Perches7965bd42013-09-26 14:48:15 -07002003struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2004 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07002005
2006/**
2007 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2008 * @dev: network device to receive on
2009 * @length: length to allocate
2010 *
2011 * Allocate a new &sk_buff and assign it a usage count of one. The
2012 * buffer has unspecified headroom built in. Users should allocate
2013 * the headroom they think they need without accounting for the
2014 * built in space. The built in space is used for optimisations.
2015 *
2016 * %NULL is returned if there is no free memory. Although this function
2017 * allocates memory it can be called from an interrupt.
2018 */
2019static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002020 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002021{
2022 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2023}
2024
Eric Dumazet6f532612012-05-18 05:12:12 +00002025/* legacy helper around __netdev_alloc_skb() */
2026static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2027 gfp_t gfp_mask)
2028{
2029 return __netdev_alloc_skb(NULL, length, gfp_mask);
2030}
2031
2032/* legacy helper around netdev_alloc_skb() */
2033static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2034{
2035 return netdev_alloc_skb(NULL, length);
2036}
2037
2038
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002039static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2040 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002041{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002042 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002043
2044 if (NET_IP_ALIGN && skb)
2045 skb_reserve(skb, NET_IP_ALIGN);
2046 return skb;
2047}
2048
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002049static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2050 unsigned int length)
2051{
2052 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2053}
2054
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002055/**
2056 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
Mel Gorman06140022012-07-31 16:44:24 -07002057 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2058 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2059 * @order: size of the allocation
2060 *
2061 * Allocate a new page.
2062 *
2063 * %NULL is returned if there is no free memory.
2064*/
2065static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
2066 struct sk_buff *skb,
2067 unsigned int order)
2068{
2069 struct page *page;
2070
2071 gfp_mask |= __GFP_COLD;
2072
2073 if (!(gfp_mask & __GFP_NOMEMALLOC))
2074 gfp_mask |= __GFP_MEMALLOC;
2075
2076 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2077 if (skb && page && page->pfmemalloc)
2078 skb->pfmemalloc = true;
2079
2080 return page;
2081}
2082
2083/**
2084 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2085 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2086 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2087 *
2088 * Allocate a new page.
2089 *
2090 * %NULL is returned if there is no free memory.
2091 */
2092static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
2093 struct sk_buff *skb)
2094{
2095 return __skb_alloc_pages(gfp_mask, skb, 0);
2096}
2097
2098/**
2099 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2100 * @page: The page that was allocated from skb_alloc_page
2101 * @skb: The skb that may need pfmemalloc set
2102 */
2103static inline void skb_propagate_pfmemalloc(struct page *page,
2104 struct sk_buff *skb)
2105{
2106 if (page && page->pfmemalloc)
2107 skb->pfmemalloc = true;
2108}
2109
Eric Dumazet564824b2010-10-11 19:05:25 +00002110/**
Masanari Iidae2278672014-02-18 22:54:36 +09002111 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002112 * @frag: the paged fragment
2113 *
2114 * Returns the &struct page associated with @frag.
2115 */
2116static inline struct page *skb_frag_page(const skb_frag_t *frag)
2117{
Ian Campbella8605c62011-10-19 23:01:49 +00002118 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00002119}
2120
2121/**
2122 * __skb_frag_ref - take an addition reference on a paged fragment.
2123 * @frag: the paged fragment
2124 *
2125 * Takes an additional reference on the paged fragment @frag.
2126 */
2127static inline void __skb_frag_ref(skb_frag_t *frag)
2128{
2129 get_page(skb_frag_page(frag));
2130}
2131
2132/**
2133 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2134 * @skb: the buffer
2135 * @f: the fragment offset.
2136 *
2137 * Takes an additional reference on the @f'th paged fragment of @skb.
2138 */
2139static inline void skb_frag_ref(struct sk_buff *skb, int f)
2140{
2141 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2142}
2143
2144/**
2145 * __skb_frag_unref - release a reference on a paged fragment.
2146 * @frag: the paged fragment
2147 *
2148 * Releases a reference on the paged fragment @frag.
2149 */
2150static inline void __skb_frag_unref(skb_frag_t *frag)
2151{
2152 put_page(skb_frag_page(frag));
2153}
2154
2155/**
2156 * skb_frag_unref - release a reference on a paged fragment of an skb.
2157 * @skb: the buffer
2158 * @f: the fragment offset
2159 *
2160 * Releases a reference on the @f'th paged fragment of @skb.
2161 */
2162static inline void skb_frag_unref(struct sk_buff *skb, int f)
2163{
2164 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2165}
2166
2167/**
2168 * skb_frag_address - gets the address of the data contained in a paged fragment
2169 * @frag: the paged fragment buffer
2170 *
2171 * Returns the address of the data within @frag. The page must already
2172 * be mapped.
2173 */
2174static inline void *skb_frag_address(const skb_frag_t *frag)
2175{
2176 return page_address(skb_frag_page(frag)) + frag->page_offset;
2177}
2178
2179/**
2180 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2181 * @frag: the paged fragment buffer
2182 *
2183 * Returns the address of the data within @frag. Checks that the page
2184 * is mapped and returns %NULL otherwise.
2185 */
2186static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2187{
2188 void *ptr = page_address(skb_frag_page(frag));
2189 if (unlikely(!ptr))
2190 return NULL;
2191
2192 return ptr + frag->page_offset;
2193}
2194
2195/**
2196 * __skb_frag_set_page - sets the page contained in a paged fragment
2197 * @frag: the paged fragment
2198 * @page: the page to set
2199 *
2200 * Sets the fragment @frag to contain @page.
2201 */
2202static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2203{
Ian Campbella8605c62011-10-19 23:01:49 +00002204 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002205}
2206
2207/**
2208 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2209 * @skb: the buffer
2210 * @f: the fragment offset
2211 * @page: the page to set
2212 *
2213 * Sets the @f'th fragment of @skb to contain @page.
2214 */
2215static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2216 struct page *page)
2217{
2218 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2219}
2220
Eric Dumazet400dfd32013-10-17 16:27:07 -07002221bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2222
Ian Campbell131ea662011-08-19 06:25:00 +00002223/**
2224 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002225 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002226 * @frag: the paged fragment to map
2227 * @offset: the offset within the fragment (starting at the
2228 * fragment's own offset)
2229 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002230 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00002231 *
2232 * Maps the page associated with @frag to @device.
2233 */
2234static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2235 const skb_frag_t *frag,
2236 size_t offset, size_t size,
2237 enum dma_data_direction dir)
2238{
2239 return dma_map_page(dev, skb_frag_page(frag),
2240 frag->page_offset + offset, size, dir);
2241}
2242
Eric Dumazet117632e2011-12-03 21:39:53 +00002243static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2244 gfp_t gfp_mask)
2245{
2246 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2247}
2248
Octavian Purdilabad93e92014-06-12 01:36:26 +03002249
2250static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2251 gfp_t gfp_mask)
2252{
2253 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2254}
2255
2256
Ian Campbell131ea662011-08-19 06:25:00 +00002257/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002258 * skb_clone_writable - is the header of a clone writable
2259 * @skb: buffer to check
2260 * @len: length up to which to write
2261 *
2262 * Returns true if modifying the header part of the cloned buffer
2263 * does not requires the data to be copied.
2264 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002265static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002266{
2267 return !skb_header_cloned(skb) &&
2268 skb_headroom(skb) + len <= skb->hdr_len;
2269}
2270
Herbert Xud9cc2042007-09-16 16:21:16 -07002271static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2272 int cloned)
2273{
2274 int delta = 0;
2275
Herbert Xud9cc2042007-09-16 16:21:16 -07002276 if (headroom > skb_headroom(skb))
2277 delta = headroom - skb_headroom(skb);
2278
2279 if (delta || cloned)
2280 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2281 GFP_ATOMIC);
2282 return 0;
2283}
2284
Patrick McHardy334a8132007-06-25 04:35:20 -07002285/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002286 * skb_cow - copy header of skb when it is required
2287 * @skb: buffer to cow
2288 * @headroom: needed headroom
2289 *
2290 * If the skb passed lacks sufficient headroom or its data part
2291 * is shared, data is reallocated. If reallocation fails, an error
2292 * is returned and original skb is not changed.
2293 *
2294 * The result is skb with writable area skb->head...skb->tail
2295 * and at least @headroom of space at head.
2296 */
2297static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2298{
Herbert Xud9cc2042007-09-16 16:21:16 -07002299 return __skb_cow(skb, headroom, skb_cloned(skb));
2300}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002301
Herbert Xud9cc2042007-09-16 16:21:16 -07002302/**
2303 * skb_cow_head - skb_cow but only making the head writable
2304 * @skb: buffer to cow
2305 * @headroom: needed headroom
2306 *
2307 * This function is identical to skb_cow except that we replace the
2308 * skb_cloned check by skb_header_cloned. It should be used when
2309 * you only need to push on some header and do not need to modify
2310 * the data.
2311 */
2312static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2313{
2314 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002315}
2316
2317/**
2318 * skb_padto - pad an skbuff up to a minimal size
2319 * @skb: buffer to pad
2320 * @len: minimal length
2321 *
2322 * Pads up a buffer to ensure the trailing bytes exist and are
2323 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002324 * is untouched. Otherwise it is extended. Returns zero on
2325 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002326 */
2327
Herbert Xu5b057c62006-06-23 02:06:41 -07002328static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002329{
2330 unsigned int size = skb->len;
2331 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002332 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002333 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002334}
2335
2336static inline int skb_add_data(struct sk_buff *skb,
2337 char __user *from, int copy)
2338{
2339 const int off = skb->len;
2340
2341 if (skb->ip_summed == CHECKSUM_NONE) {
2342 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08002343 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
Linus Torvalds1da177e2005-04-16 15:20:36 -07002344 copy, 0, &err);
2345 if (!err) {
2346 skb->csum = csum_block_add(skb->csum, csum, off);
2347 return 0;
2348 }
2349 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2350 return 0;
2351
2352 __skb_trim(skb, off);
2353 return -EFAULT;
2354}
2355
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002356static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2357 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002358{
2359 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002360 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002361
Ian Campbellea2ab692011-08-22 23:44:58 +00002362 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002363 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002364 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002365 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002366}
2367
Herbert Xu364c6ba2006-06-09 16:10:40 -07002368static inline int __skb_linearize(struct sk_buff *skb)
2369{
2370 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2371}
2372
Linus Torvalds1da177e2005-04-16 15:20:36 -07002373/**
2374 * skb_linearize - convert paged skb to linear one
2375 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002376 *
2377 * If there is no free memory -ENOMEM is returned, otherwise zero
2378 * is returned and the old skb data released.
2379 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002380static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002381{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002382 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2383}
2384
2385/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00002386 * skb_has_shared_frag - can any frag be overwritten
2387 * @skb: buffer to test
2388 *
2389 * Return true if the skb has at least one frag that might be modified
2390 * by an external entity (as in vmsplice()/sendfile())
2391 */
2392static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2393{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00002394 return skb_is_nonlinear(skb) &&
2395 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00002396}
2397
2398/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07002399 * skb_linearize_cow - make sure skb is linear and writable
2400 * @skb: buffer to process
2401 *
2402 * If there is no free memory -ENOMEM is returned, otherwise zero
2403 * is returned and the old skb data released.
2404 */
2405static inline int skb_linearize_cow(struct sk_buff *skb)
2406{
2407 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2408 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002409}
2410
2411/**
2412 * skb_postpull_rcsum - update checksum for received skb after pull
2413 * @skb: buffer to update
2414 * @start: start of data before pull
2415 * @len: length of data pulled
2416 *
2417 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002418 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2419 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002420 */
2421
2422static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002423 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002424{
Patrick McHardy84fa7932006-08-29 16:44:56 -07002425 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002426 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2427}
2428
Herbert Xucbb042f2006-03-20 22:43:56 -08002429unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2430
David S. Miller7ce5a272013-12-02 17:26:05 -05002431/**
2432 * pskb_trim_rcsum - trim received skb and update checksum
2433 * @skb: buffer to trim
2434 * @len: new length
2435 *
2436 * This is exactly the same as pskb_trim except that it ensures the
2437 * checksum of received packets are still valid after the operation.
2438 */
2439
2440static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2441{
2442 if (likely(len >= skb->len))
2443 return 0;
2444 if (skb->ip_summed == CHECKSUM_COMPLETE)
2445 skb->ip_summed = CHECKSUM_NONE;
2446 return __pskb_trim(skb, len);
2447}
2448
Linus Torvalds1da177e2005-04-16 15:20:36 -07002449#define skb_queue_walk(queue, skb) \
2450 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002451 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002452 skb = skb->next)
2453
James Chapman46f89142007-04-30 00:07:31 -07002454#define skb_queue_walk_safe(queue, skb, tmp) \
2455 for (skb = (queue)->next, tmp = skb->next; \
2456 skb != (struct sk_buff *)(queue); \
2457 skb = tmp, tmp = skb->next)
2458
David S. Miller1164f522008-09-23 00:49:44 -07002459#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002460 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002461 skb = skb->next)
2462
2463#define skb_queue_walk_from_safe(queue, skb, tmp) \
2464 for (tmp = skb->next; \
2465 skb != (struct sk_buff *)(queue); \
2466 skb = tmp, tmp = skb->next)
2467
Stephen Hemminger300ce172005-10-30 13:47:34 -08002468#define skb_queue_reverse_walk(queue, skb) \
2469 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002470 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08002471 skb = skb->prev)
2472
David S. Miller686a2952011-01-20 22:47:32 -08002473#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2474 for (skb = (queue)->prev, tmp = skb->prev; \
2475 skb != (struct sk_buff *)(queue); \
2476 skb = tmp, tmp = skb->prev)
2477
2478#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2479 for (tmp = skb->prev; \
2480 skb != (struct sk_buff *)(queue); \
2481 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002482
David S. Miller21dc3302010-08-23 00:13:46 -07002483static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07002484{
2485 return skb_shinfo(skb)->frag_list != NULL;
2486}
2487
2488static inline void skb_frag_list_init(struct sk_buff *skb)
2489{
2490 skb_shinfo(skb)->frag_list = NULL;
2491}
2492
2493static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2494{
2495 frag->next = skb_shinfo(skb)->frag_list;
2496 skb_shinfo(skb)->frag_list = frag;
2497}
2498
2499#define skb_walk_frags(skb, iter) \
2500 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2501
Joe Perches7965bd42013-09-26 14:48:15 -07002502struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2503 int *peeked, int *off, int *err);
2504struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2505 int *err);
2506unsigned int datagram_poll(struct file *file, struct socket *sock,
2507 struct poll_table_struct *wait);
2508int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2509 struct iovec *to, int size);
2510int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2511 struct iovec *iov);
2512int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2513 const struct iovec *from, int from_offset,
2514 int len);
2515int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2516 int offset, size_t count);
2517int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2518 const struct iovec *to, int to_offset,
2519 int size);
2520void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2521void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2522int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07002523int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2524int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2525__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2526 int len, __wsum csum);
2527int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2528 struct pipe_inode_info *pipe, unsigned int len,
2529 unsigned int flags);
2530void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01002531unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00002532int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2533 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07002534void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2535int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2536void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Florian Westphalde960aa2014-01-26 10:58:16 +01002537unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -07002538struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002539
Daniel Borkmann2817a332013-10-30 11:50:51 +01002540struct skb_checksum_ops {
2541 __wsum (*update)(const void *mem, int len, __wsum wsum);
2542 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2543};
2544
2545__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2546 __wsum csum, const struct skb_checksum_ops *ops);
2547__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2548 __wsum csum);
2549
Linus Torvalds1da177e2005-04-16 15:20:36 -07002550static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2551 int len, void *buffer)
2552{
2553 int hlen = skb_headlen(skb);
2554
Patrick McHardy55820ee2005-07-05 14:08:10 -07002555 if (hlen - offset >= len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002556 return skb->data + offset;
2557
2558 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2559 return NULL;
2560
2561 return buffer;
2562}
2563
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01002564/**
2565 * skb_needs_linearize - check if we need to linearize a given skb
2566 * depending on the given device features.
2567 * @skb: socket buffer to check
2568 * @features: net device features
2569 *
2570 * Returns true if either:
2571 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2572 * 2. skb is fragmented and the device does not support SG.
2573 */
2574static inline bool skb_needs_linearize(struct sk_buff *skb,
2575 netdev_features_t features)
2576{
2577 return skb_is_nonlinear(skb) &&
2578 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2579 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2580}
2581
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03002582static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2583 void *to,
2584 const unsigned int len)
2585{
2586 memcpy(to, skb->data, len);
2587}
2588
2589static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2590 const int offset, void *to,
2591 const unsigned int len)
2592{
2593 memcpy(to, skb->data + offset, len);
2594}
2595
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03002596static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2597 const void *from,
2598 const unsigned int len)
2599{
2600 memcpy(skb->data, from, len);
2601}
2602
2603static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2604 const int offset,
2605 const void *from,
2606 const unsigned int len)
2607{
2608 memcpy(skb->data + offset, from, len);
2609}
2610
Joe Perches7965bd42013-09-26 14:48:15 -07002611void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002612
Patrick Ohlyac45f602009-02-12 05:03:37 +00002613static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2614{
2615 return skb->tstamp;
2616}
2617
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002618/**
2619 * skb_get_timestamp - get timestamp from a skb
2620 * @skb: skb to get stamp from
2621 * @stamp: pointer to struct timeval to store stamp in
2622 *
2623 * Timestamps are stored in the skb as offsets to a base timestamp.
2624 * This function converts the offset back to a struct timeval and stores
2625 * it in stamp.
2626 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00002627static inline void skb_get_timestamp(const struct sk_buff *skb,
2628 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002629{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002630 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002631}
2632
Patrick Ohlyac45f602009-02-12 05:03:37 +00002633static inline void skb_get_timestampns(const struct sk_buff *skb,
2634 struct timespec *stamp)
2635{
2636 *stamp = ktime_to_timespec(skb->tstamp);
2637}
2638
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002639static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002640{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002641 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002642}
2643
Stephen Hemminger164891a2007-04-23 22:26:16 -07002644static inline ktime_t net_timedelta(ktime_t t)
2645{
2646 return ktime_sub(ktime_get_real(), t);
2647}
2648
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07002649static inline ktime_t net_invalid_timestamp(void)
2650{
2651 return ktime_set(0, 0);
2652}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002653
Richard Cochranc1f19b52010-07-17 08:49:36 +00002654#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2655
Joe Perches7965bd42013-09-26 14:48:15 -07002656void skb_clone_tx_timestamp(struct sk_buff *skb);
2657bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00002658
2659#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2660
2661static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2662{
2663}
2664
2665static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2666{
2667 return false;
2668}
2669
2670#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2671
2672/**
2673 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2674 *
Richard Cochranda92b192011-10-21 00:49:15 +00002675 * PHY drivers may accept clones of transmitted packets for
2676 * timestamping via their phy_driver.txtstamp method. These drivers
2677 * must call this function to return the skb back to the stack, with
2678 * or without a timestamp.
2679 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00002680 * @skb: clone of the the original outgoing packet
Richard Cochranda92b192011-10-21 00:49:15 +00002681 * @hwtstamps: hardware time stamps, may be NULL if not available
Richard Cochranc1f19b52010-07-17 08:49:36 +00002682 *
2683 */
2684void skb_complete_tx_timestamp(struct sk_buff *skb,
2685 struct skb_shared_hwtstamps *hwtstamps);
2686
Patrick Ohlyac45f602009-02-12 05:03:37 +00002687/**
2688 * skb_tstamp_tx - queue clone of skb with send time stamps
2689 * @orig_skb: the original outgoing packet
2690 * @hwtstamps: hardware time stamps, may be NULL if not available
2691 *
2692 * If the skb has a socket associated, then this function clones the
2693 * skb (thus sharing the actual data and optional structures), stores
2694 * the optional hardware time stamping information (if non NULL) or
2695 * generates a software time stamp (otherwise), then queues the clone
2696 * to the error queue of the socket. Errors are silently ignored.
2697 */
Joe Perches7965bd42013-09-26 14:48:15 -07002698void skb_tstamp_tx(struct sk_buff *orig_skb,
2699 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00002700
Richard Cochran4507a712010-07-17 08:48:28 +00002701static inline void sw_tx_timestamp(struct sk_buff *skb)
2702{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002703 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2704 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00002705 skb_tstamp_tx(skb, NULL);
2706}
2707
2708/**
2709 * skb_tx_timestamp() - Driver hook for transmit timestamping
2710 *
2711 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00002712 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00002713 *
David S. Miller73409f32013-12-27 13:04:33 -05002714 * Specifically, one should make absolutely sure that this function is
2715 * called before TX completion of this packet can trigger. Otherwise
2716 * the packet could potentially already be freed.
2717 *
Richard Cochran4507a712010-07-17 08:48:28 +00002718 * @skb: A socket buffer.
2719 */
2720static inline void skb_tx_timestamp(struct sk_buff *skb)
2721{
Richard Cochranc1f19b52010-07-17 08:49:36 +00002722 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002723 sw_tx_timestamp(skb);
2724}
2725
Johannes Berg6e3e9392011-11-09 10:15:42 +01002726/**
2727 * skb_complete_wifi_ack - deliver skb with wifi status
2728 *
2729 * @skb: the original outgoing packet
2730 * @acked: ack status
2731 *
2732 */
2733void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2734
Joe Perches7965bd42013-09-26 14:48:15 -07002735__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2736__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002737
Herbert Xu60476372007-04-09 11:59:39 -07002738static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2739{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002740 return ((skb->ip_summed & CHECKSUM_UNNECESSARY) || skb->csum_valid);
Herbert Xu60476372007-04-09 11:59:39 -07002741}
2742
Herbert Xufb286bb2005-11-10 13:01:24 -08002743/**
2744 * skb_checksum_complete - Calculate checksum of an entire packet
2745 * @skb: packet to process
2746 *
2747 * This function calculates the checksum over the entire packet plus
2748 * the value of skb->csum. The latter can be used to supply the
2749 * checksum of a pseudo header as used by TCP/UDP. It returns the
2750 * checksum.
2751 *
2752 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2753 * this function can be used to verify that checksum on received
2754 * packets. In that case the function should return zero if the
2755 * checksum is correct. In particular, this function will return zero
2756 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2757 * hardware has already verified the correctness of the checksum.
2758 */
Al Viro4381ca32007-07-15 21:00:11 +01002759static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002760{
Herbert Xu60476372007-04-09 11:59:39 -07002761 return skb_csum_unnecessary(skb) ?
2762 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002763}
2764
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002765/* Check if we need to perform checksum complete validation.
2766 *
2767 * Returns true if checksum complete is needed, false otherwise
2768 * (either checksum is unnecessary or zero checksum is allowed).
2769 */
2770static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
2771 bool zero_okay,
2772 __sum16 check)
2773{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002774 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
2775 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002776 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))) {
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002802 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002803 return 0;
2804 }
2805 }
2806
2807 skb->csum = psum;
2808
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002809 if (complete || skb->len <= CHECKSUM_BREAK) {
2810 __sum16 csum;
2811
2812 csum = __skb_checksum_complete(skb);
2813 skb->csum_valid = !csum;
2814 return csum;
2815 }
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002816
2817 return 0;
2818}
2819
2820static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
2821{
2822 return 0;
2823}
2824
2825/* Perform checksum validate (init). Note that this is a macro since we only
2826 * want to calculate the pseudo header which is an input function if necessary.
2827 * First we try to validate without any computation (checksum unnecessary) and
2828 * then calculate based on checksum complete calling the function to compute
2829 * pseudo header.
2830 *
2831 * Return values:
2832 * 0: checksum is validated or try to in skb_checksum_complete
2833 * non-zero: value of invalid checksum
2834 */
2835#define __skb_checksum_validate(skb, proto, complete, \
2836 zero_okay, check, compute_pseudo) \
2837({ \
2838 __sum16 __ret = 0; \
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002839 skb->csum_valid = 0; \
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002840 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
2841 __ret = __skb_checksum_validate_complete(skb, \
2842 complete, compute_pseudo(skb, proto)); \
2843 __ret; \
2844})
2845
2846#define skb_checksum_init(skb, proto, compute_pseudo) \
2847 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
2848
2849#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
2850 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
2851
2852#define skb_checksum_validate(skb, proto, compute_pseudo) \
2853 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
2854
2855#define skb_checksum_validate_zero_check(skb, proto, check, \
2856 compute_pseudo) \
2857 __skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
2858
2859#define skb_checksum_simple_validate(skb) \
2860 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
2861
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002862#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07002863void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002864static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2865{
2866 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002867 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002868}
2869static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2870{
2871 if (nfct)
2872 atomic_inc(&nfct->use);
2873}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002874#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002875#ifdef CONFIG_BRIDGE_NETFILTER
2876static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2877{
2878 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2879 kfree(nf_bridge);
2880}
2881static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2882{
2883 if (nf_bridge)
2884 atomic_inc(&nf_bridge->use);
2885}
2886#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08002887static inline void nf_reset(struct sk_buff *skb)
2888{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002889#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08002890 nf_conntrack_put(skb->nfct);
2891 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002892#endif
Patrick McHardya193a4a2006-03-20 19:23:05 -08002893#ifdef CONFIG_BRIDGE_NETFILTER
2894 nf_bridge_put(skb->nf_bridge);
2895 skb->nf_bridge = NULL;
2896#endif
2897}
2898
Patrick McHardy124dff02013-04-05 20:42:05 +02002899static inline void nf_reset_trace(struct sk_buff *skb)
2900{
Florian Westphal478b3602014-02-15 23:48:45 +01002901#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00002902 skb->nf_trace = 0;
2903#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002904}
2905
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002906/* Note: This doesn't put any conntrack and bridge info in dst. */
2907static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2908{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002909#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002910 dst->nfct = src->nfct;
2911 nf_conntrack_get(src->nfct);
2912 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002913#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002914#ifdef CONFIG_BRIDGE_NETFILTER
2915 dst->nf_bridge = src->nf_bridge;
2916 nf_bridge_get(src->nf_bridge);
2917#endif
Florian Westphal478b3602014-02-15 23:48:45 +01002918#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
2919 dst->nf_trace = src->nf_trace;
2920#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002921}
2922
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002923static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2924{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002925#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002926 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002927#endif
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002928#ifdef CONFIG_BRIDGE_NETFILTER
2929 nf_bridge_put(dst->nf_bridge);
2930#endif
2931 __nf_copy(dst, src);
2932}
2933
James Morris984bc162006-06-09 00:29:17 -07002934#ifdef CONFIG_NETWORK_SECMARK
2935static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2936{
2937 to->secmark = from->secmark;
2938}
2939
2940static inline void skb_init_secmark(struct sk_buff *skb)
2941{
2942 skb->secmark = 0;
2943}
2944#else
2945static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2946{ }
2947
2948static inline void skb_init_secmark(struct sk_buff *skb)
2949{ }
2950#endif
2951
Eric W. Biederman574f7192014-04-01 12:20:24 -07002952static inline bool skb_irq_freeable(const struct sk_buff *skb)
2953{
2954 return !skb->destructor &&
2955#if IS_ENABLED(CONFIG_XFRM)
2956 !skb->sp &&
2957#endif
2958#if IS_ENABLED(CONFIG_NF_CONNTRACK)
2959 !skb->nfct &&
2960#endif
2961 !skb->_skb_refdst &&
2962 !skb_has_frag_list(skb);
2963}
2964
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002965static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2966{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002967 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002968}
2969
Stephen Hemminger92477442009-03-21 13:39:26 -07002970static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002971{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002972 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002973}
2974
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002975static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2976{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002977 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002978}
2979
David S. Millerd5a9e242009-01-27 16:22:11 -08002980static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2981{
2982 skb->queue_mapping = rx_queue + 1;
2983}
2984
Stephen Hemminger92477442009-03-21 13:39:26 -07002985static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002986{
2987 return skb->queue_mapping - 1;
2988}
2989
Stephen Hemminger92477442009-03-21 13:39:26 -07002990static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002991{
Eric Dumazeta02cec22010-09-22 20:43:57 +00002992 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08002993}
2994
Joe Perches7965bd42013-09-26 14:48:15 -07002995u16 __skb_tx_hash(const struct net_device *dev, const struct sk_buff *skb,
2996 unsigned int num_tx_queues);
Stephen Hemminger92477442009-03-21 13:39:26 -07002997
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04002998static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2999{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003000#ifdef CONFIG_XFRM
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003001 return skb->sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003002#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003003 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003004#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003005}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003006
Pravin B Shelar68c33162013-02-14 14:02:41 +00003007/* Keeps track of mac header offset relative to skb->head.
3008 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3009 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07003010 * tunnel skb it points to outer mac header.
3011 * Keeps track of level of encapsulation of network headers.
3012 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00003013struct skb_gso_cb {
Eric Dumazet3347c962013-10-19 11:42:56 -07003014 int mac_offset;
3015 int encap_level;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003016 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00003017};
3018#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3019
3020static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3021{
3022 return (skb_mac_header(inner_skb) - inner_skb->head) -
3023 SKB_GSO_CB(inner_skb)->mac_offset;
3024}
3025
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00003026static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3027{
3028 int new_headroom, headroom;
3029 int ret;
3030
3031 headroom = skb_headroom(skb);
3032 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3033 if (ret)
3034 return ret;
3035
3036 new_headroom = skb_headroom(skb);
3037 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3038 return 0;
3039}
3040
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003041/* Compute the checksum for a gso segment. First compute the checksum value
3042 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3043 * then add in skb->csum (checksum from csum_start to end of packet).
3044 * skb->csum and csum_start are then updated to reflect the checksum of the
3045 * resultant packet starting from the transport header-- the resultant checksum
3046 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3047 * header.
3048 */
3049static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3050{
3051 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3052 skb_transport_offset(skb);
3053 __u16 csum;
3054
3055 csum = csum_fold(csum_partial(skb_transport_header(skb),
3056 plen, skb->csum));
3057 skb->csum = res;
3058 SKB_GSO_CB(skb)->csum_start -= plen;
3059
3060 return csum;
3061}
3062
David S. Millerbdcc0922012-03-07 20:53:36 -05003063static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07003064{
3065 return skb_shinfo(skb)->gso_size;
3066}
3067
Eric Dumazet36a8f392013-09-29 01:21:32 -07003068/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05003069static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02003070{
3071 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3072}
3073
Joe Perches7965bd42013-09-26 14:48:15 -07003074void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07003075
3076static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3077{
3078 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3079 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003080 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3081
Alexander Duyckb78462e2010-06-02 12:24:37 +00003082 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3083 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07003084 __skb_warn_lro_forwarding(skb);
3085 return true;
3086 }
3087 return false;
3088}
3089
Herbert Xu35fc92a2007-03-26 23:22:20 -07003090static inline void skb_forward_csum(struct sk_buff *skb)
3091{
3092 /* Unfortunately we don't support this one. Any brave souls? */
3093 if (skb->ip_summed == CHECKSUM_COMPLETE)
3094 skb->ip_summed = CHECKSUM_NONE;
3095}
3096
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003097/**
3098 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3099 * @skb: skb to check
3100 *
3101 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3102 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3103 * use this helper, to document places where we make this assertion.
3104 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003105static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003106{
3107#ifdef DEBUG
3108 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3109#endif
3110}
3111
Rusty Russellf35d9d82008-02-04 23:49:54 -05003112bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00003113
Paul Durranted1f50c2014-01-09 10:02:46 +00003114int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3115
Daniel Borkmannf77668d2013-03-19 06:39:30 +00003116u32 __skb_get_poff(const struct sk_buff *skb);
3117
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00003118/**
3119 * skb_head_is_locked - Determine if the skb->head is locked down
3120 * @skb: skb to check
3121 *
3122 * The head on skbs build around a head frag can be removed if they are
3123 * not cloned. This function returns true if the skb head is locked down
3124 * due to either being allocated via kmalloc, or by being a clone with
3125 * multiple references to the head.
3126 */
3127static inline bool skb_head_is_locked(const struct sk_buff *skb)
3128{
3129 return !skb->head_frag || skb_cloned(skb);
3130}
Florian Westphalfe6cc552014-02-13 23:09:12 +01003131
3132/**
3133 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3134 *
3135 * @skb: GSO skb
3136 *
3137 * skb_gso_network_seglen is used to determine the real size of the
3138 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3139 *
3140 * The MAC/L2 header is not accounted for.
3141 */
3142static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3143{
3144 unsigned int hdr_len = skb_transport_header(skb) -
3145 skb_network_header(skb);
3146 return hdr_len + skb_gso_transport_seglen(skb);
3147}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003148#endif /* __KERNEL__ */
3149#endif /* _LINUX_SKBUFF_H */