blob: e13ed90be7c2ef19d19f200b26c2649177d9eb95 [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
Tom Herbert4b282522014-06-14 23:23:52 -0700341 SKB_GSO_GRE_CSUM = 1 << 7,
Simon Horman0d89d202013-05-23 21:02:52 +0000342
Tom Herbert4b282522014-06-14 23:23:52 -0700343 SKB_GSO_IPIP = 1 << 8,
Eric Dumazetcb32f512013-10-19 11:42:57 -0700344
Tom Herbert4b282522014-06-14 23:23:52 -0700345 SKB_GSO_SIT = 1 << 9,
Eric Dumazet61c1db72013-10-20 20:47:30 -0700346
Tom Herbert4b282522014-06-14 23:23:52 -0700347 SKB_GSO_UDP_TUNNEL = 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
Tom Herbert4b282522014-06-14 23:23:52 -0700351 SKB_GSO_MPLS = 1 << 12,
352
Herbert Xu79671682006-06-22 02:40:14 -0700353};
354
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700355#if BITS_PER_LONG > 32
356#define NET_SKBUFF_DATA_USES_OFFSET 1
357#endif
358
359#ifdef NET_SKBUFF_DATA_USES_OFFSET
360typedef unsigned int sk_buff_data_t;
361#else
362typedef unsigned char *sk_buff_data_t;
363#endif
364
Eric Dumazet363ec39232014-02-26 14:02:11 -0800365/**
366 * struct skb_mstamp - multi resolution time stamps
367 * @stamp_us: timestamp in us resolution
368 * @stamp_jiffies: timestamp in jiffies
369 */
370struct skb_mstamp {
371 union {
372 u64 v64;
373 struct {
374 u32 stamp_us;
375 u32 stamp_jiffies;
376 };
377 };
378};
379
380/**
381 * skb_mstamp_get - get current timestamp
382 * @cl: place to store timestamps
383 */
384static inline void skb_mstamp_get(struct skb_mstamp *cl)
385{
386 u64 val = local_clock();
387
388 do_div(val, NSEC_PER_USEC);
389 cl->stamp_us = (u32)val;
390 cl->stamp_jiffies = (u32)jiffies;
391}
392
393/**
394 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
395 * @t1: pointer to newest sample
396 * @t0: pointer to oldest sample
397 */
398static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
399 const struct skb_mstamp *t0)
400{
401 s32 delta_us = t1->stamp_us - t0->stamp_us;
402 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
403
404 /* If delta_us is negative, this might be because interval is too big,
405 * or local_clock() drift is too big : fallback using jiffies.
406 */
407 if (delta_us <= 0 ||
408 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
409
410 delta_us = jiffies_to_usecs(delta_jiffies);
411
412 return delta_us;
413}
414
415
Linus Torvalds1da177e2005-04-16 15:20:36 -0700416/**
417 * struct sk_buff - socket buffer
418 * @next: Next buffer in list
419 * @prev: Previous buffer in list
Eric Dumazet363ec39232014-02-26 14:02:11 -0800420 * @tstamp: Time we arrived/left
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700421 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700423 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000424 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700425 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700426 * @len: Length of actual data
427 * @data_len: Data length
428 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700429 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700430 * @csum: Checksum (must include start/offset pair)
431 * @csum_start: Offset from skb->head where checksumming should start
432 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700433 * @priority: Packet queueing priority
WANG Cong60ff7462014-05-04 16:39:18 -0700434 * @ignore_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700435 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700436 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437 * @nohdr: Payload reference only, must not modify header
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700438 * @nfctinfo: Relationship of this skb to the connection
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700440 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700441 * @ipvs_property: skbuff is owned by ipvs
Randy Dunlap31729362008-02-18 20:52:13 -0800442 * @peeked: this packet has been seen already, so stats have been
443 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700444 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700445 * @protocol: Packet protocol from driver
446 * @destructor: Destruct function
447 * @nfct: Associated connection, if any
Linus Torvalds1da177e2005-04-16 15:20:36 -0700448 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800449 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450 * @tc_index: Traffic control index
451 * @tc_verd: traffic control verdict
Tom Herbert61b905d2014-03-24 15:34:47 -0700452 * @hash: the packet hash
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700453 * @queue_mapping: Queue mapping for multiqueue devices
Randy Dunlap553a5672008-04-20 10:51:01 -0700454 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700455 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Tom Herbert61b905d2014-03-24 15:34:47 -0700456 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
Changli Gao4ca24622011-08-19 07:26:44 -0700457 * ports.
Johannes Berg6e3e9392011-11-09 10:15:42 +0100458 * @wifi_acked_valid: wifi_acked was set
459 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000460 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Randy Dunlapf4b8ea72006-06-22 16:00:11 -0700461 * @dma_cookie: a cookie to one of several possible DMA operations
462 * done by skb DMA functions
Eliezer Tamir06021292013-06-10 11:39:50 +0300463 * @napi_id: id of the NAPI struct this skb came from
James Morris984bc162006-06-09 00:29:17 -0700464 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700465 * @mark: Generic packet mark
466 * @dropcount: total number of sk_receive_queue overflows
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000467 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b02008-07-14 22:49:06 -0700468 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000469 * @inner_protocol: Protocol (encapsulation)
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000470 * @inner_transport_header: Inner transport layer header (encapsulation)
471 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000472 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700473 * @transport_header: Transport layer header
474 * @network_header: Network layer header
475 * @mac_header: Link layer header
476 * @tail: Tail pointer
477 * @end: End pointer
478 * @head: Head of buffer
479 * @data: Data head pointer
480 * @truesize: Buffer size
481 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700482 */
483
484struct sk_buff {
485 /* These two members must be first. */
486 struct sk_buff *next;
487 struct sk_buff *prev;
488
Eric Dumazet363ec39232014-02-26 14:02:11 -0800489 union {
490 ktime_t tstamp;
491 struct skb_mstamp skb_mstamp;
492 };
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000493
494 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700495 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700496
Linus Torvalds1da177e2005-04-16 15:20:36 -0700497 /*
498 * This is the control buffer. It is free to use for every
499 * layer. Please put your private variables there. If you
500 * want to keep them across layers you have to do a skb_clone()
501 * first. This is owned by whoever has the skb queued ATM.
502 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000503 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504
Eric Dumazet7fee2262010-05-11 23:19:48 +0000505 unsigned long _skb_refdst;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000506#ifdef CONFIG_XFRM
507 struct sec_path *sp;
508#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700509 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700510 data_len;
511 __u16 mac_len,
512 hdr_len;
Al Viroff1dcad2006-11-20 18:07:29 -0800513 union {
514 __wsum csum;
Herbert Xu663ead32007-04-09 11:59:07 -0700515 struct {
516 __u16 csum_start;
517 __u16 csum_offset;
518 };
Al Viroff1dcad2006-11-20 18:07:29 -0800519 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520 __u32 priority;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200521 kmemcheck_bitfield_begin(flags1);
WANG Cong60ff7462014-05-04 16:39:18 -0700522 __u8 ignore_df:1,
Thomas Graf1cbb3382005-07-05 14:13:41 -0700523 cloned:1,
524 ip_summed:2,
Harald Welte6869c4d2005-08-09 19:24:19 -0700525 nohdr:1,
526 nfctinfo:3;
David S. Millerd179cd12005-08-17 14:57:30 -0700527 __u8 pkt_type:3,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800528 fclone:2,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700529 ipvs_property:1,
Herbert Xua59322b2007-12-05 01:53:40 -0800530 peeked:1,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700531 nf_trace:1;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200532 kmemcheck_bitfield_end(flags1);
Eric Dumazet4ab408d2010-03-01 03:09:26 +0000533 __be16 protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534
535 void (*destructor)(struct sk_buff *skb);
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800536#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700537 struct nf_conntrack *nfct;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100538#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700539#ifdef CONFIG_BRIDGE_NETFILTER
540 struct nf_bridge_info *nf_bridge;
541#endif
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700542
Eric Dumazet8964be42009-11-20 15:35:04 -0800543 int skb_iif;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000544
Tom Herbert61b905d2014-03-24 15:34:47 -0700545 __u32 hash;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000546
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000547 __be16 vlan_proto;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000548 __u16 vlan_tci;
549
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700551 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700552#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700553 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700554#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700555#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200556
Eric Dumazet0a148422011-04-20 09:27:32 +0000557 __u16 queue_mapping;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200558 kmemcheck_bitfield_begin(flags2);
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400559#ifdef CONFIG_IPV6_NDISC_NODETYPE
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000560 __u8 ndisc_nodetype:2;
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400561#endif
Mel Gormanc93bdd02012-07-31 16:44:19 -0700562 __u8 pfmemalloc:1;
Tom Herbert3853b582010-11-21 13:17:29 +0000563 __u8 ooo_okay:1;
Tom Herbert61b905d2014-03-24 15:34:47 -0700564 __u8 l4_hash:1;
Johannes Berg6e3e9392011-11-09 10:15:42 +0100565 __u8 wifi_acked_valid:1;
566 __u8 wifi_acked:1;
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000567 __u8 no_fcs:1;
Eric Dumazetd3836f22012-04-27 00:33:38 +0000568 __u8 head_frag:1;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000569 /* Encapsulation protocol and NIC drivers should use
570 * this flag to indicate to each other if the skb contains
571 * encapsulated packet or not and maybe use the inner packet
572 * headers if needed
573 */
574 __u8 encapsulation:1;
Tom Herbert7e2b10c2014-06-04 17:20:02 -0700575 __u8 encap_hdr_csum:1;
Tom Herbert5d0c2b92014-06-10 18:54:13 -0700576 __u8 csum_valid:1;
Tom Herbert7e3cead2014-06-10 18:54:19 -0700577 __u8 csum_complete_sw:1;
578 /* 3/5 bit hole (depending on ndisc_nodetype presence) */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200579 kmemcheck_bitfield_end(flags2);
580
Cong Wange0d10952013-08-01 11:10:25 +0800581#if defined CONFIG_NET_DMA || defined CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir06021292013-06-10 11:39:50 +0300582 union {
583 unsigned int napi_id;
584 dma_cookie_t dma_cookie;
585 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700586#endif
James Morris984bc162006-06-09 00:29:17 -0700587#ifdef CONFIG_NETWORK_SECMARK
588 __u32 secmark;
589#endif
Neil Horman3b885782009-10-12 13:26:31 -0700590 union {
591 __u32 mark;
592 __u32 dropcount;
Eric Dumazet16fad692013-03-14 05:40:32 +0000593 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700594 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595
Simon Horman0d89d202013-05-23 21:02:52 +0000596 __be16 inner_protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000597 __u16 inner_transport_header;
598 __u16 inner_network_header;
599 __u16 inner_mac_header;
600 __u16 transport_header;
601 __u16 network_header;
602 __u16 mac_header;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700604 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700605 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700606 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700607 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700608 unsigned int truesize;
609 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700610};
611
612#ifdef __KERNEL__
613/*
614 * Handling routines are only of interest to the kernel
615 */
616#include <linux/slab.h>
617
Linus Torvalds1da177e2005-04-16 15:20:36 -0700618
Mel Gormanc93bdd02012-07-31 16:44:19 -0700619#define SKB_ALLOC_FCLONE 0x01
620#define SKB_ALLOC_RX 0x02
621
622/* Returns true if the skb was allocated from PFMEMALLOC reserves */
623static inline bool skb_pfmemalloc(const struct sk_buff *skb)
624{
625 return unlikely(skb->pfmemalloc);
626}
627
Eric Dumazet7fee2262010-05-11 23:19:48 +0000628/*
629 * skb might have a dst pointer attached, refcounted or not.
630 * _skb_refdst low order bit is set if refcount was _not_ taken
631 */
632#define SKB_DST_NOREF 1UL
633#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
634
635/**
636 * skb_dst - returns skb dst_entry
637 * @skb: buffer
638 *
639 * Returns skb dst_entry, regardless of reference taken or not.
640 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000641static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
642{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000643 /* If refdst was not refcounted, check we still are in a
644 * rcu_read_lock section
645 */
646 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
647 !rcu_read_lock_held() &&
648 !rcu_read_lock_bh_held());
649 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000650}
651
Eric Dumazet7fee2262010-05-11 23:19:48 +0000652/**
653 * skb_dst_set - sets skb dst
654 * @skb: buffer
655 * @dst: dst entry
656 *
657 * Sets skb dst, assuming a reference was taken on dst and should
658 * be released by skb_dst_drop()
659 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000660static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
661{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000662 skb->_skb_refdst = (unsigned long)dst;
663}
664
Joe Perches7965bd42013-09-26 14:48:15 -0700665void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
666 bool force);
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200667
668/**
669 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
670 * @skb: buffer
671 * @dst: dst entry
672 *
673 * Sets skb dst, assuming a reference was not taken on dst.
674 * If dst entry is cached, we do not take reference and dst_release
675 * will be avoided by refdst_drop. If dst entry is not cached, we take
676 * reference, so that last dst_release can destroy the dst immediately.
677 */
678static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
679{
680 __skb_dst_set_noref(skb, dst, false);
681}
682
683/**
684 * skb_dst_set_noref_force - sets skb dst, without taking reference
685 * @skb: buffer
686 * @dst: dst entry
687 *
688 * Sets skb dst, assuming a reference was not taken on dst.
689 * No reference is taken and no dst_release will be called. While for
690 * cached dsts deferred reclaim is a basic feature, for entries that are
691 * not cached it is caller's job to guarantee that last dst_release for
692 * provided dst happens when nobody uses it, eg. after a RCU grace period.
693 */
694static inline void skb_dst_set_noref_force(struct sk_buff *skb,
695 struct dst_entry *dst)
696{
697 __skb_dst_set_noref(skb, dst, true);
698}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000699
700/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300701 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000702 * @skb: buffer
703 */
704static inline bool skb_dst_is_noref(const struct sk_buff *skb)
705{
706 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000707}
708
Eric Dumazet511c3f92009-06-02 05:14:27 +0000709static inline struct rtable *skb_rtable(const struct sk_buff *skb)
710{
Eric Dumazetadf30902009-06-02 05:19:30 +0000711 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000712}
713
Joe Perches7965bd42013-09-26 14:48:15 -0700714void kfree_skb(struct sk_buff *skb);
715void kfree_skb_list(struct sk_buff *segs);
716void skb_tx_error(struct sk_buff *skb);
717void consume_skb(struct sk_buff *skb);
718void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000719extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000720
Joe Perches7965bd42013-09-26 14:48:15 -0700721void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
722bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
723 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000724
Joe Perches7965bd42013-09-26 14:48:15 -0700725struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
726 int node);
727struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700728static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100729 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700730{
Eric Dumazet564824b2010-10-11 19:05:25 +0000731 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700732}
733
734static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100735 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700736{
Mel Gormanc93bdd02012-07-31 16:44:19 -0700737 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700738}
739
Joe Perches7965bd42013-09-26 14:48:15 -0700740struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
Patrick McHardy0ebd0ac2013-04-17 06:46:58 +0000741static inline struct sk_buff *alloc_skb_head(gfp_t priority)
742{
743 return __alloc_skb_head(priority, -1);
744}
745
Joe Perches7965bd42013-09-26 14:48:15 -0700746struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
747int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
748struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
749struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
Octavian Purdilabad93e92014-06-12 01:36:26 +0300750struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
751 gfp_t gfp_mask, bool fclone);
752static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
753 gfp_t gfp_mask)
754{
755 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
756}
Eric Dumazet117632e2011-12-03 21:39:53 +0000757
Joe Perches7965bd42013-09-26 14:48:15 -0700758int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
759struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
760 unsigned int headroom);
761struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
762 int newtailroom, gfp_t priority);
Fan Du25a91d82014-01-18 09:54:23 +0800763int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
764 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -0700765int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
766 int len);
767int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
768int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000769#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770
Joe Perches7965bd42013-09-26 14:48:15 -0700771int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
772 int getfrag(void *from, char *to, int offset,
773 int len, int odd, struct sk_buff *skb),
774 void *from, int length);
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700775
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800776struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700777 __u32 lower_offset;
778 __u32 upper_offset;
779 __u32 frag_idx;
780 __u32 stepped_offset;
781 struct sk_buff *root_skb;
782 struct sk_buff *cur_skb;
783 __u8 *frag_data;
784};
785
Joe Perches7965bd42013-09-26 14:48:15 -0700786void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
787 unsigned int to, struct skb_seq_state *st);
788unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
789 struct skb_seq_state *st);
790void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -0700791
Joe Perches7965bd42013-09-26 14:48:15 -0700792unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
793 unsigned int to, struct ts_config *config,
794 struct ts_state *state);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700795
Tom Herbert09323cc2013-12-15 22:16:19 -0800796/*
797 * Packet hash types specify the type of hash in skb_set_hash.
798 *
799 * Hash types refer to the protocol layer addresses which are used to
800 * construct a packet's hash. The hashes are used to differentiate or identify
801 * flows of the protocol layer for the hash type. Hash types are either
802 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
803 *
804 * Properties of hashes:
805 *
806 * 1) Two packets in different flows have different hash values
807 * 2) Two packets in the same flow should have the same hash value
808 *
809 * A hash at a higher layer is considered to be more specific. A driver should
810 * set the most specific hash possible.
811 *
812 * A driver cannot indicate a more specific hash than the layer at which a hash
813 * was computed. For instance an L3 hash cannot be set as an L4 hash.
814 *
815 * A driver may indicate a hash level which is less specific than the
816 * actual layer the hash was computed on. For instance, a hash computed
817 * at L4 may be considered an L3 hash. This should only be done if the
818 * driver can't unambiguously determine that the HW computed the hash at
819 * the higher layer. Note that the "should" in the second property above
820 * permits this.
821 */
822enum pkt_hash_types {
823 PKT_HASH_TYPE_NONE, /* Undefined type */
824 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
825 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
826 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
827};
828
829static inline void
830skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
831{
Tom Herbert61b905d2014-03-24 15:34:47 -0700832 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
833 skb->hash = hash;
Tom Herbert09323cc2013-12-15 22:16:19 -0800834}
835
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800836void __skb_get_hash(struct sk_buff *skb);
837static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000838{
Tom Herbert61b905d2014-03-24 15:34:47 -0700839 if (!skb->l4_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800840 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000841
Tom Herbert61b905d2014-03-24 15:34:47 -0700842 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000843}
844
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800845static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
846{
Tom Herbert61b905d2014-03-24 15:34:47 -0700847 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800848}
849
Tom Herbert7539fad2013-12-15 22:12:18 -0800850static inline void skb_clear_hash(struct sk_buff *skb)
851{
Tom Herbert61b905d2014-03-24 15:34:47 -0700852 skb->hash = 0;
853 skb->l4_hash = 0;
Tom Herbert7539fad2013-12-15 22:12:18 -0800854}
855
856static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
857{
Tom Herbert61b905d2014-03-24 15:34:47 -0700858 if (!skb->l4_hash)
Tom Herbert7539fad2013-12-15 22:12:18 -0800859 skb_clear_hash(skb);
860}
861
Tom Herbert3df7a742013-12-15 22:16:29 -0800862static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
863{
Tom Herbert61b905d2014-03-24 15:34:47 -0700864 to->hash = from->hash;
865 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -0800866};
867
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700868#ifdef NET_SKBUFF_DATA_USES_OFFSET
869static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
870{
871 return skb->head + skb->end;
872}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000873
874static inline unsigned int skb_end_offset(const struct sk_buff *skb)
875{
876 return skb->end;
877}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700878#else
879static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
880{
881 return skb->end;
882}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000883
884static inline unsigned int skb_end_offset(const struct sk_buff *skb)
885{
886 return skb->end - skb->head;
887}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700888#endif
889
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700891#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700892
Patrick Ohlyac45f602009-02-12 05:03:37 +0000893static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
894{
895 return &skb_shinfo(skb)->hwtstamps;
896}
897
Linus Torvalds1da177e2005-04-16 15:20:36 -0700898/**
899 * skb_queue_empty - check if a queue is empty
900 * @list: queue head
901 *
902 * Returns true if the queue is empty, false otherwise.
903 */
904static inline int skb_queue_empty(const struct sk_buff_head *list)
905{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100906 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700907}
908
909/**
David S. Millerfc7ebb22008-09-23 00:34:07 -0700910 * skb_queue_is_last - check if skb is the last entry in the queue
911 * @list: queue head
912 * @skb: buffer
913 *
914 * Returns true if @skb is the last buffer on the list.
915 */
916static inline bool skb_queue_is_last(const struct sk_buff_head *list,
917 const struct sk_buff *skb)
918{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100919 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -0700920}
921
922/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800923 * skb_queue_is_first - check if skb is the first entry in the queue
924 * @list: queue head
925 * @skb: buffer
926 *
927 * Returns true if @skb is the first buffer on the list.
928 */
929static inline bool skb_queue_is_first(const struct sk_buff_head *list,
930 const struct sk_buff *skb)
931{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100932 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800933}
934
935/**
David S. Miller249c8b42008-09-23 00:44:42 -0700936 * skb_queue_next - return the next packet in the queue
937 * @list: queue head
938 * @skb: current buffer
939 *
940 * Return the next packet in @list after @skb. It is only valid to
941 * call this if skb_queue_is_last() evaluates to false.
942 */
943static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
944 const struct sk_buff *skb)
945{
946 /* This BUG_ON may seem severe, but if we just return then we
947 * are going to dereference garbage.
948 */
949 BUG_ON(skb_queue_is_last(list, skb));
950 return skb->next;
951}
952
953/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800954 * skb_queue_prev - return the prev packet in the queue
955 * @list: queue head
956 * @skb: current buffer
957 *
958 * Return the prev packet in @list before @skb. It is only valid to
959 * call this if skb_queue_is_first() evaluates to false.
960 */
961static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
962 const struct sk_buff *skb)
963{
964 /* This BUG_ON may seem severe, but if we just return then we
965 * are going to dereference garbage.
966 */
967 BUG_ON(skb_queue_is_first(list, skb));
968 return skb->prev;
969}
970
971/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700972 * skb_get - reference buffer
973 * @skb: buffer to reference
974 *
975 * Makes another reference to a socket buffer and returns a pointer
976 * to the buffer.
977 */
978static inline struct sk_buff *skb_get(struct sk_buff *skb)
979{
980 atomic_inc(&skb->users);
981 return skb;
982}
983
984/*
985 * If users == 1, we are the only owner and are can avoid redundant
986 * atomic change.
987 */
988
989/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700990 * skb_cloned - is the buffer a clone
991 * @skb: buffer to check
992 *
993 * Returns true if the buffer was generated with skb_clone() and is
994 * one of multiple shared copies of the buffer. Cloned buffers are
995 * shared data so must not be written to under normal circumstances.
996 */
997static inline int skb_cloned(const struct sk_buff *skb)
998{
999 return skb->cloned &&
1000 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1001}
1002
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001003static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1004{
1005 might_sleep_if(pri & __GFP_WAIT);
1006
1007 if (skb_cloned(skb))
1008 return pskb_expand_head(skb, 0, 0, pri);
1009
1010 return 0;
1011}
1012
Linus Torvalds1da177e2005-04-16 15:20:36 -07001013/**
1014 * skb_header_cloned - is the header a clone
1015 * @skb: buffer to check
1016 *
1017 * Returns true if modifying the header part of the buffer requires
1018 * the data to be copied.
1019 */
1020static inline int skb_header_cloned(const struct sk_buff *skb)
1021{
1022 int dataref;
1023
1024 if (!skb->cloned)
1025 return 0;
1026
1027 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1028 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1029 return dataref != 1;
1030}
1031
1032/**
1033 * skb_header_release - release reference to header
1034 * @skb: buffer to operate on
1035 *
1036 * Drop a reference to the header part of the buffer. This is done
1037 * by acquiring a payload reference. You must not read from the header
1038 * part of skb->data after this.
1039 */
1040static inline void skb_header_release(struct sk_buff *skb)
1041{
1042 BUG_ON(skb->nohdr);
1043 skb->nohdr = 1;
1044 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1045}
1046
1047/**
1048 * skb_shared - is the buffer shared
1049 * @skb: buffer to check
1050 *
1051 * Returns true if more than one person has a reference to this
1052 * buffer.
1053 */
1054static inline int skb_shared(const struct sk_buff *skb)
1055{
1056 return atomic_read(&skb->users) != 1;
1057}
1058
1059/**
1060 * skb_share_check - check if buffer is shared and if so clone it
1061 * @skb: buffer to check
1062 * @pri: priority for memory allocation
1063 *
1064 * If the buffer is shared the buffer is cloned and the old copy
1065 * drops a reference. A new clone with a single reference is returned.
1066 * If the buffer is not shared the original buffer is returned. When
1067 * being called from interrupt status or with spinlocks held pri must
1068 * be GFP_ATOMIC.
1069 *
1070 * NULL is returned on a memory allocation failure.
1071 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001072static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001073{
1074 might_sleep_if(pri & __GFP_WAIT);
1075 if (skb_shared(skb)) {
1076 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001077
1078 if (likely(nskb))
1079 consume_skb(skb);
1080 else
1081 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001082 skb = nskb;
1083 }
1084 return skb;
1085}
1086
1087/*
1088 * Copy shared buffers into a new sk_buff. We effectively do COW on
1089 * packets to handle cases where we have a local reader and forward
1090 * and a couple of other messy ones. The normal one is tcpdumping
1091 * a packet thats being forwarded.
1092 */
1093
1094/**
1095 * skb_unshare - make a copy of a shared buffer
1096 * @skb: buffer to check
1097 * @pri: priority for memory allocation
1098 *
1099 * If the socket buffer is a clone then this function creates a new
1100 * copy of the data, drops a reference count on the old copy and returns
1101 * the new copy with the reference count at 1. If the buffer is not a clone
1102 * the original buffer is returned. When called with a spinlock held or
1103 * from interrupt state @pri must be %GFP_ATOMIC
1104 *
1105 * %NULL is returned on a memory allocation failure.
1106 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001107static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001108 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001109{
1110 might_sleep_if(pri & __GFP_WAIT);
1111 if (skb_cloned(skb)) {
1112 struct sk_buff *nskb = skb_copy(skb, pri);
1113 kfree_skb(skb); /* Free our shared copy */
1114 skb = nskb;
1115 }
1116 return skb;
1117}
1118
1119/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001120 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001121 * @list_: list to peek at
1122 *
1123 * Peek an &sk_buff. Unlike most other operations you _MUST_
1124 * be careful with this one. A peek leaves the buffer on the
1125 * list and someone else may run off with it. You must hold
1126 * the appropriate locks or have a private queue to do this.
1127 *
1128 * Returns %NULL for an empty list or a pointer to the head element.
1129 * The reference count is not incremented and the reference is therefore
1130 * volatile. Use with caution.
1131 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001132static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001133{
Eric Dumazet18d07002012-04-30 16:31:46 +00001134 struct sk_buff *skb = list_->next;
1135
1136 if (skb == (struct sk_buff *)list_)
1137 skb = NULL;
1138 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001139}
1140
1141/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001142 * skb_peek_next - peek skb following the given one from a queue
1143 * @skb: skb to start from
1144 * @list_: list to peek at
1145 *
1146 * Returns %NULL when the end of the list is met or a pointer to the
1147 * next element. The reference count is not incremented and the
1148 * reference is therefore volatile. Use with caution.
1149 */
1150static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1151 const struct sk_buff_head *list_)
1152{
1153 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001154
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001155 if (next == (struct sk_buff *)list_)
1156 next = NULL;
1157 return next;
1158}
1159
1160/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001161 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001162 * @list_: list to peek at
1163 *
1164 * Peek an &sk_buff. Unlike most other operations you _MUST_
1165 * be careful with this one. A peek leaves the buffer on the
1166 * list and someone else may run off with it. You must hold
1167 * the appropriate locks or have a private queue to do this.
1168 *
1169 * Returns %NULL for an empty list or a pointer to the tail element.
1170 * The reference count is not incremented and the reference is therefore
1171 * volatile. Use with caution.
1172 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001173static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001174{
Eric Dumazet18d07002012-04-30 16:31:46 +00001175 struct sk_buff *skb = list_->prev;
1176
1177 if (skb == (struct sk_buff *)list_)
1178 skb = NULL;
1179 return skb;
1180
Linus Torvalds1da177e2005-04-16 15:20:36 -07001181}
1182
1183/**
1184 * skb_queue_len - get queue length
1185 * @list_: list to measure
1186 *
1187 * Return the length of an &sk_buff queue.
1188 */
1189static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1190{
1191 return list_->qlen;
1192}
1193
David S. Miller67fed452008-09-21 22:36:24 -07001194/**
1195 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1196 * @list: queue to initialize
1197 *
1198 * This initializes only the list and queue length aspects of
1199 * an sk_buff_head object. This allows to initialize the list
1200 * aspects of an sk_buff_head without reinitializing things like
1201 * the spinlock. It can also be used for on-stack sk_buff_head
1202 * objects where the spinlock is known to not be used.
1203 */
1204static inline void __skb_queue_head_init(struct sk_buff_head *list)
1205{
1206 list->prev = list->next = (struct sk_buff *)list;
1207 list->qlen = 0;
1208}
1209
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001210/*
1211 * This function creates a split out lock class for each invocation;
1212 * this is needed for now since a whole lot of users of the skb-queue
1213 * infrastructure in drivers have different locking usage (in hardirq)
1214 * than the networking core (in softirq only). In the long run either the
1215 * network layer or drivers should need annotation to consolidate the
1216 * main types of usage into 3 classes.
1217 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001218static inline void skb_queue_head_init(struct sk_buff_head *list)
1219{
1220 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001221 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001222}
1223
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001224static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1225 struct lock_class_key *class)
1226{
1227 skb_queue_head_init(list);
1228 lockdep_set_class(&list->lock, class);
1229}
1230
Linus Torvalds1da177e2005-04-16 15:20:36 -07001231/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001232 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001233 *
1234 * The "__skb_xxxx()" functions are the non-atomic ones that
1235 * can only be called with interrupts disabled.
1236 */
Joe Perches7965bd42013-09-26 14:48:15 -07001237void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1238 struct sk_buff_head *list);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001239static inline void __skb_insert(struct sk_buff *newsk,
1240 struct sk_buff *prev, struct sk_buff *next,
1241 struct sk_buff_head *list)
1242{
1243 newsk->next = next;
1244 newsk->prev = prev;
1245 next->prev = prev->next = newsk;
1246 list->qlen++;
1247}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001248
David S. Miller67fed452008-09-21 22:36:24 -07001249static inline void __skb_queue_splice(const struct sk_buff_head *list,
1250 struct sk_buff *prev,
1251 struct sk_buff *next)
1252{
1253 struct sk_buff *first = list->next;
1254 struct sk_buff *last = list->prev;
1255
1256 first->prev = prev;
1257 prev->next = first;
1258
1259 last->next = next;
1260 next->prev = last;
1261}
1262
1263/**
1264 * skb_queue_splice - join two skb lists, this is designed for stacks
1265 * @list: the new list to add
1266 * @head: the place to add it in the first list
1267 */
1268static inline void skb_queue_splice(const struct sk_buff_head *list,
1269 struct sk_buff_head *head)
1270{
1271 if (!skb_queue_empty(list)) {
1272 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001273 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001274 }
1275}
1276
1277/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001278 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001279 * @list: the new list to add
1280 * @head: the place to add it in the first list
1281 *
1282 * The list at @list is reinitialised
1283 */
1284static inline void skb_queue_splice_init(struct sk_buff_head *list,
1285 struct sk_buff_head *head)
1286{
1287 if (!skb_queue_empty(list)) {
1288 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001289 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001290 __skb_queue_head_init(list);
1291 }
1292}
1293
1294/**
1295 * skb_queue_splice_tail - join two skb lists, each list being a queue
1296 * @list: the new list to add
1297 * @head: the place to add it in the first list
1298 */
1299static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1300 struct sk_buff_head *head)
1301{
1302 if (!skb_queue_empty(list)) {
1303 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001304 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001305 }
1306}
1307
1308/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001309 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001310 * @list: the new list to add
1311 * @head: the place to add it in the first list
1312 *
1313 * Each of the lists is a queue.
1314 * The list at @list is reinitialised
1315 */
1316static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1317 struct sk_buff_head *head)
1318{
1319 if (!skb_queue_empty(list)) {
1320 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001321 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001322 __skb_queue_head_init(list);
1323 }
1324}
1325
Linus Torvalds1da177e2005-04-16 15:20:36 -07001326/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001327 * __skb_queue_after - queue a buffer at the list head
1328 * @list: list to use
1329 * @prev: place after this buffer
1330 * @newsk: buffer to queue
1331 *
1332 * Queue a buffer int the middle of a list. This function takes no locks
1333 * and you must therefore hold required locks before calling it.
1334 *
1335 * A buffer cannot be placed on two lists at the same time.
1336 */
1337static inline void __skb_queue_after(struct sk_buff_head *list,
1338 struct sk_buff *prev,
1339 struct sk_buff *newsk)
1340{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001341 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001342}
1343
Joe Perches7965bd42013-09-26 14:48:15 -07001344void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1345 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001346
Gerrit Renkerf5572852008-04-14 00:05:28 -07001347static inline void __skb_queue_before(struct sk_buff_head *list,
1348 struct sk_buff *next,
1349 struct sk_buff *newsk)
1350{
1351 __skb_insert(newsk, next->prev, next, list);
1352}
1353
Stephen Hemminger300ce172005-10-30 13:47:34 -08001354/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001355 * __skb_queue_head - queue a buffer at the list head
1356 * @list: list to use
1357 * @newsk: buffer to queue
1358 *
1359 * Queue a buffer at the start of a list. This function takes no locks
1360 * and you must therefore hold required locks before calling it.
1361 *
1362 * A buffer cannot be placed on two lists at the same time.
1363 */
Joe Perches7965bd42013-09-26 14:48:15 -07001364void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001365static inline void __skb_queue_head(struct sk_buff_head *list,
1366 struct sk_buff *newsk)
1367{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001368 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001369}
1370
1371/**
1372 * __skb_queue_tail - queue a buffer at the list tail
1373 * @list: list to use
1374 * @newsk: buffer to queue
1375 *
1376 * Queue a buffer at the end of a list. This function takes no locks
1377 * and you must therefore hold required locks before calling it.
1378 *
1379 * A buffer cannot be placed on two lists at the same time.
1380 */
Joe Perches7965bd42013-09-26 14:48:15 -07001381void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001382static inline void __skb_queue_tail(struct sk_buff_head *list,
1383 struct sk_buff *newsk)
1384{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001385 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001386}
1387
Linus Torvalds1da177e2005-04-16 15:20:36 -07001388/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001389 * remove sk_buff from list. _Must_ be called atomically, and with
1390 * the list known..
1391 */
Joe Perches7965bd42013-09-26 14:48:15 -07001392void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001393static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1394{
1395 struct sk_buff *next, *prev;
1396
1397 list->qlen--;
1398 next = skb->next;
1399 prev = skb->prev;
1400 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001401 next->prev = prev;
1402 prev->next = next;
1403}
1404
Gerrit Renkerf525c062008-04-14 00:04:12 -07001405/**
1406 * __skb_dequeue - remove from the head of the queue
1407 * @list: list to dequeue from
1408 *
1409 * Remove the head of the list. This function does not take any locks
1410 * so must be used with appropriate locks held only. The head item is
1411 * returned or %NULL if the list is empty.
1412 */
Joe Perches7965bd42013-09-26 14:48:15 -07001413struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Gerrit Renkerf525c062008-04-14 00:04:12 -07001414static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1415{
1416 struct sk_buff *skb = skb_peek(list);
1417 if (skb)
1418 __skb_unlink(skb, list);
1419 return skb;
1420}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421
1422/**
1423 * __skb_dequeue_tail - remove from the tail of the queue
1424 * @list: list to dequeue from
1425 *
1426 * Remove the tail of the list. This function does not take any locks
1427 * so must be used with appropriate locks held only. The tail item is
1428 * returned or %NULL if the list is empty.
1429 */
Joe Perches7965bd42013-09-26 14:48:15 -07001430struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001431static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1432{
1433 struct sk_buff *skb = skb_peek_tail(list);
1434 if (skb)
1435 __skb_unlink(skb, list);
1436 return skb;
1437}
1438
1439
David S. Millerbdcc0922012-03-07 20:53:36 -05001440static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001441{
1442 return skb->data_len;
1443}
1444
1445static inline unsigned int skb_headlen(const struct sk_buff *skb)
1446{
1447 return skb->len - skb->data_len;
1448}
1449
1450static inline int skb_pagelen(const struct sk_buff *skb)
1451{
1452 int i, len = 0;
1453
1454 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001455 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001456 return len + skb_headlen(skb);
1457}
1458
Ian Campbell131ea662011-08-19 06:25:00 +00001459/**
1460 * __skb_fill_page_desc - initialise a paged fragment in an skb
1461 * @skb: buffer containing fragment to be initialised
1462 * @i: paged fragment index to initialise
1463 * @page: the page to use for this fragment
1464 * @off: the offset to the data with @page
1465 * @size: the length of the data
1466 *
1467 * Initialises the @i'th fragment of @skb to point to &size bytes at
1468 * offset @off within @page.
1469 *
1470 * Does not take any additional reference on the fragment.
1471 */
1472static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1473 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001474{
1475 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1476
Mel Gormanc48a11c2012-07-31 16:44:23 -07001477 /*
1478 * Propagate page->pfmemalloc to the skb if we can. The problem is
1479 * that not all callers have unique ownership of the page. If
1480 * pfmemalloc is set, we check the mapping as a mapping implies
1481 * page->index is set (index and pfmemalloc share space).
1482 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1483 * do not lose pfmemalloc information as the pages would not be
1484 * allocated using __GFP_MEMALLOC.
1485 */
Ian Campbella8605c62011-10-19 23:01:49 +00001486 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001487 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001488 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001489
1490 page = compound_head(page);
1491 if (page->pfmemalloc && !page->mapping)
1492 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00001493}
1494
1495/**
1496 * skb_fill_page_desc - initialise a paged fragment in an skb
1497 * @skb: buffer containing fragment to be initialised
1498 * @i: paged fragment index to initialise
1499 * @page: the page to use for this fragment
1500 * @off: the offset to the data with @page
1501 * @size: the length of the data
1502 *
1503 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01001504 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00001505 * addition updates @skb such that @i is the last fragment.
1506 *
1507 * Does not take any additional reference on the fragment.
1508 */
1509static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1510 struct page *page, int off, int size)
1511{
1512 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001513 skb_shinfo(skb)->nr_frags = i + 1;
1514}
1515
Joe Perches7965bd42013-09-26 14:48:15 -07001516void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1517 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001518
Jason Wangf8e617e2013-11-01 14:07:47 +08001519void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1520 unsigned int truesize);
1521
Linus Torvalds1da177e2005-04-16 15:20:36 -07001522#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001523#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001524#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1525
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001526#ifdef NET_SKBUFF_DATA_USES_OFFSET
1527static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1528{
1529 return skb->head + skb->tail;
1530}
1531
1532static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1533{
1534 skb->tail = skb->data - skb->head;
1535}
1536
1537static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1538{
1539 skb_reset_tail_pointer(skb);
1540 skb->tail += offset;
1541}
Simon Horman7cc46192013-05-28 20:34:29 +00001542
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001543#else /* NET_SKBUFF_DATA_USES_OFFSET */
1544static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1545{
1546 return skb->tail;
1547}
1548
1549static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1550{
1551 skb->tail = skb->data;
1552}
1553
1554static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1555{
1556 skb->tail = skb->data + offset;
1557}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001558
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001559#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1560
Linus Torvalds1da177e2005-04-16 15:20:36 -07001561/*
1562 * Add data to an sk_buff
1563 */
Mathias Krause0c7ddf32013-11-07 14:18:24 +01001564unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001565unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001566static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1567{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001568 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569 SKB_LINEAR_ASSERT(skb);
1570 skb->tail += len;
1571 skb->len += len;
1572 return tmp;
1573}
1574
Joe Perches7965bd42013-09-26 14:48:15 -07001575unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001576static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1577{
1578 skb->data -= len;
1579 skb->len += len;
1580 return skb->data;
1581}
1582
Joe Perches7965bd42013-09-26 14:48:15 -07001583unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001584static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1585{
1586 skb->len -= len;
1587 BUG_ON(skb->len < skb->data_len);
1588 return skb->data += len;
1589}
1590
David S. Miller47d29642010-05-02 02:21:44 -07001591static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1592{
1593 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1594}
1595
Joe Perches7965bd42013-09-26 14:48:15 -07001596unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597
1598static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1599{
1600 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001601 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001602 return NULL;
1603 skb->len -= len;
1604 return skb->data += len;
1605}
1606
1607static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1608{
1609 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1610}
1611
1612static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1613{
1614 if (likely(len <= skb_headlen(skb)))
1615 return 1;
1616 if (unlikely(len > skb->len))
1617 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001618 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001619}
1620
1621/**
1622 * skb_headroom - bytes at buffer head
1623 * @skb: buffer to check
1624 *
1625 * Return the number of bytes of free space at the head of an &sk_buff.
1626 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001627static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001628{
1629 return skb->data - skb->head;
1630}
1631
1632/**
1633 * skb_tailroom - bytes at buffer end
1634 * @skb: buffer to check
1635 *
1636 * Return the number of bytes of free space at the tail of an sk_buff
1637 */
1638static inline int skb_tailroom(const struct sk_buff *skb)
1639{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001640 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001641}
1642
1643/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00001644 * skb_availroom - bytes at buffer end
1645 * @skb: buffer to check
1646 *
1647 * Return the number of bytes of free space at the tail of an sk_buff
1648 * allocated by sk_stream_alloc()
1649 */
1650static inline int skb_availroom(const struct sk_buff *skb)
1651{
Eric Dumazet16fad692013-03-14 05:40:32 +00001652 if (skb_is_nonlinear(skb))
1653 return 0;
1654
1655 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00001656}
1657
1658/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001659 * skb_reserve - adjust headroom
1660 * @skb: buffer to alter
1661 * @len: bytes to move
1662 *
1663 * Increase the headroom of an empty &sk_buff by reducing the tail
1664 * room. This is only allowed for an empty buffer.
1665 */
David S. Miller8243126c2006-01-17 02:54:21 -08001666static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001667{
1668 skb->data += len;
1669 skb->tail += len;
1670}
1671
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001672static inline void skb_reset_inner_headers(struct sk_buff *skb)
1673{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001674 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001675 skb->inner_network_header = skb->network_header;
1676 skb->inner_transport_header = skb->transport_header;
1677}
1678
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00001679static inline void skb_reset_mac_len(struct sk_buff *skb)
1680{
1681 skb->mac_len = skb->network_header - skb->mac_header;
1682}
1683
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001684static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1685 *skb)
1686{
1687 return skb->head + skb->inner_transport_header;
1688}
1689
1690static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1691{
1692 skb->inner_transport_header = skb->data - skb->head;
1693}
1694
1695static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1696 const int offset)
1697{
1698 skb_reset_inner_transport_header(skb);
1699 skb->inner_transport_header += offset;
1700}
1701
1702static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1703{
1704 return skb->head + skb->inner_network_header;
1705}
1706
1707static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1708{
1709 skb->inner_network_header = skb->data - skb->head;
1710}
1711
1712static inline void skb_set_inner_network_header(struct sk_buff *skb,
1713 const int offset)
1714{
1715 skb_reset_inner_network_header(skb);
1716 skb->inner_network_header += offset;
1717}
1718
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001719static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1720{
1721 return skb->head + skb->inner_mac_header;
1722}
1723
1724static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1725{
1726 skb->inner_mac_header = skb->data - skb->head;
1727}
1728
1729static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1730 const int offset)
1731{
1732 skb_reset_inner_mac_header(skb);
1733 skb->inner_mac_header += offset;
1734}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001735static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1736{
Cong Wang35d04612013-05-29 15:16:05 +08001737 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001738}
1739
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001740static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1741{
1742 return skb->head + skb->transport_header;
1743}
1744
1745static inline void skb_reset_transport_header(struct sk_buff *skb)
1746{
1747 skb->transport_header = skb->data - skb->head;
1748}
1749
1750static inline void skb_set_transport_header(struct sk_buff *skb,
1751 const int offset)
1752{
1753 skb_reset_transport_header(skb);
1754 skb->transport_header += offset;
1755}
1756
1757static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1758{
1759 return skb->head + skb->network_header;
1760}
1761
1762static inline void skb_reset_network_header(struct sk_buff *skb)
1763{
1764 skb->network_header = skb->data - skb->head;
1765}
1766
1767static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1768{
1769 skb_reset_network_header(skb);
1770 skb->network_header += offset;
1771}
1772
1773static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1774{
1775 return skb->head + skb->mac_header;
1776}
1777
1778static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1779{
Cong Wang35d04612013-05-29 15:16:05 +08001780 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001781}
1782
1783static inline void skb_reset_mac_header(struct sk_buff *skb)
1784{
1785 skb->mac_header = skb->data - skb->head;
1786}
1787
1788static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1789{
1790 skb_reset_mac_header(skb);
1791 skb->mac_header += offset;
1792}
1793
Timo Teräs0e3da5b2013-12-16 11:02:09 +02001794static inline void skb_pop_mac_header(struct sk_buff *skb)
1795{
1796 skb->mac_header = skb->network_header;
1797}
1798
Ying Xuefbbdb8f2013-03-27 16:46:06 +00001799static inline void skb_probe_transport_header(struct sk_buff *skb,
1800 const int offset_hint)
1801{
1802 struct flow_keys keys;
1803
1804 if (skb_transport_header_was_set(skb))
1805 return;
1806 else if (skb_flow_dissect(skb, &keys))
1807 skb_set_transport_header(skb, keys.thoff);
1808 else
1809 skb_set_transport_header(skb, offset_hint);
1810}
1811
Eric Dumazet03606892012-02-23 10:55:02 +00001812static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1813{
1814 if (skb_mac_header_was_set(skb)) {
1815 const unsigned char *old_mac = skb_mac_header(skb);
1816
1817 skb_set_mac_header(skb, -skb->mac_len);
1818 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1819 }
1820}
1821
Michał Mirosław04fb4512010-12-14 15:24:08 +00001822static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1823{
1824 return skb->csum_start - skb_headroom(skb);
1825}
1826
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001827static inline int skb_transport_offset(const struct sk_buff *skb)
1828{
1829 return skb_transport_header(skb) - skb->data;
1830}
1831
1832static inline u32 skb_network_header_len(const struct sk_buff *skb)
1833{
1834 return skb->transport_header - skb->network_header;
1835}
1836
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001837static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1838{
1839 return skb->inner_transport_header - skb->inner_network_header;
1840}
1841
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001842static inline int skb_network_offset(const struct sk_buff *skb)
1843{
1844 return skb_network_header(skb) - skb->data;
1845}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001846
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001847static inline int skb_inner_network_offset(const struct sk_buff *skb)
1848{
1849 return skb_inner_network_header(skb) - skb->data;
1850}
1851
Changli Gaof9599ce2010-08-04 04:43:44 +00001852static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1853{
1854 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1855}
1856
Linus Torvalds1da177e2005-04-16 15:20:36 -07001857/*
1858 * CPUs often take a performance hit when accessing unaligned memory
1859 * locations. The actual performance hit varies, it can be small if the
1860 * hardware handles it or large if we have to take an exception and fix it
1861 * in software.
1862 *
1863 * Since an ethernet header is 14 bytes network drivers often end up with
1864 * the IP header at an unaligned offset. The IP header can be aligned by
1865 * shifting the start of the packet by 2 bytes. Drivers should do this
1866 * with:
1867 *
Tobias Klauser8660c122009-07-13 22:48:16 +00001868 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001869 *
1870 * The downside to this alignment of the IP header is that the DMA is now
1871 * unaligned. On some architectures the cost of an unaligned DMA is high
1872 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00001873 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001874 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1875 * to be overridden.
1876 */
1877#ifndef NET_IP_ALIGN
1878#define NET_IP_ALIGN 2
1879#endif
1880
Anton Blanchard025be812006-03-31 02:27:06 -08001881/*
1882 * The networking layer reserves some headroom in skb data (via
1883 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1884 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08001885 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08001886 *
1887 * Unfortunately this headroom changes the DMA alignment of the resulting
1888 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1889 * on some architectures. An architecture can override this value,
1890 * perhaps setting it to a cacheline in size (since that will maintain
1891 * cacheline alignment of the DMA). It must be a power of 2.
1892 *
David S. Millerd6301d32009-02-08 19:24:13 -08001893 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08001894 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07001895 *
1896 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1897 * to reduce average number of cache lines per packet.
1898 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07001899 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08001900 */
1901#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07001902#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08001903#endif
1904
Joe Perches7965bd42013-09-26 14:48:15 -07001905int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001906
1907static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1908{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00001909 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07001910 WARN_ON(1);
1911 return;
1912 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001913 skb->len = len;
1914 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001915}
1916
Joe Perches7965bd42013-09-26 14:48:15 -07001917void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001918
1919static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1920{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001921 if (skb->data_len)
1922 return ___pskb_trim(skb, len);
1923 __skb_trim(skb, len);
1924 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001925}
1926
1927static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1928{
1929 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1930}
1931
1932/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07001933 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1934 * @skb: buffer to alter
1935 * @len: new length
1936 *
1937 * This is identical to pskb_trim except that the caller knows that
1938 * the skb is not cloned so we should never get an error due to out-
1939 * of-memory.
1940 */
1941static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1942{
1943 int err = pskb_trim(skb, len);
1944 BUG_ON(err);
1945}
1946
1947/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001948 * skb_orphan - orphan a buffer
1949 * @skb: buffer to orphan
1950 *
1951 * If a buffer currently has an owner then we call the owner's
1952 * destructor function and make the @skb unowned. The buffer continues
1953 * to exist but is no longer charged to its former owner.
1954 */
1955static inline void skb_orphan(struct sk_buff *skb)
1956{
Eric Dumazetc34a7612013-07-30 16:11:15 -07001957 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001958 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07001959 skb->destructor = NULL;
1960 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07001961 } else {
1962 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07001963 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001964}
1965
1966/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00001967 * skb_orphan_frags - orphan the frags contained in a buffer
1968 * @skb: buffer to orphan frags from
1969 * @gfp_mask: allocation mask for replacement pages
1970 *
1971 * For each frag in the SKB which needs a destructor (i.e. has an
1972 * owner) create a copy of that frag and release the original
1973 * page by calling the destructor.
1974 */
1975static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1976{
1977 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1978 return 0;
1979 return skb_copy_ubufs(skb, gfp_mask);
1980}
1981
1982/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001983 * __skb_queue_purge - empty a list
1984 * @list: list to empty
1985 *
1986 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1987 * the list and one reference dropped. This function does not take the
1988 * list lock and the caller must hold the relevant locks to use it.
1989 */
Joe Perches7965bd42013-09-26 14:48:15 -07001990void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001991static inline void __skb_queue_purge(struct sk_buff_head *list)
1992{
1993 struct sk_buff *skb;
1994 while ((skb = __skb_dequeue(list)) != NULL)
1995 kfree_skb(skb);
1996}
1997
Alexander Duycke5e67302013-02-08 10:17:15 +00001998#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
1999#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
2000#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
2001
Joe Perches7965bd42013-09-26 14:48:15 -07002002void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002003
Joe Perches7965bd42013-09-26 14:48:15 -07002004struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2005 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07002006
2007/**
2008 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2009 * @dev: network device to receive on
2010 * @length: length to allocate
2011 *
2012 * Allocate a new &sk_buff and assign it a usage count of one. The
2013 * buffer has unspecified headroom built in. Users should allocate
2014 * the headroom they think they need without accounting for the
2015 * built in space. The built in space is used for optimisations.
2016 *
2017 * %NULL is returned if there is no free memory. Although this function
2018 * allocates memory it can be called from an interrupt.
2019 */
2020static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002021 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002022{
2023 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2024}
2025
Eric Dumazet6f532612012-05-18 05:12:12 +00002026/* legacy helper around __netdev_alloc_skb() */
2027static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2028 gfp_t gfp_mask)
2029{
2030 return __netdev_alloc_skb(NULL, length, gfp_mask);
2031}
2032
2033/* legacy helper around netdev_alloc_skb() */
2034static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2035{
2036 return netdev_alloc_skb(NULL, length);
2037}
2038
2039
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002040static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2041 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002042{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002043 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002044
2045 if (NET_IP_ALIGN && skb)
2046 skb_reserve(skb, NET_IP_ALIGN);
2047 return skb;
2048}
2049
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002050static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2051 unsigned int length)
2052{
2053 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2054}
2055
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002056/**
2057 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
Mel Gorman06140022012-07-31 16:44:24 -07002058 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2059 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2060 * @order: size of the allocation
2061 *
2062 * Allocate a new page.
2063 *
2064 * %NULL is returned if there is no free memory.
2065*/
2066static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
2067 struct sk_buff *skb,
2068 unsigned int order)
2069{
2070 struct page *page;
2071
2072 gfp_mask |= __GFP_COLD;
2073
2074 if (!(gfp_mask & __GFP_NOMEMALLOC))
2075 gfp_mask |= __GFP_MEMALLOC;
2076
2077 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2078 if (skb && page && page->pfmemalloc)
2079 skb->pfmemalloc = true;
2080
2081 return page;
2082}
2083
2084/**
2085 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2086 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2087 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2088 *
2089 * Allocate a new page.
2090 *
2091 * %NULL is returned if there is no free memory.
2092 */
2093static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
2094 struct sk_buff *skb)
2095{
2096 return __skb_alloc_pages(gfp_mask, skb, 0);
2097}
2098
2099/**
2100 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2101 * @page: The page that was allocated from skb_alloc_page
2102 * @skb: The skb that may need pfmemalloc set
2103 */
2104static inline void skb_propagate_pfmemalloc(struct page *page,
2105 struct sk_buff *skb)
2106{
2107 if (page && page->pfmemalloc)
2108 skb->pfmemalloc = true;
2109}
2110
Eric Dumazet564824b2010-10-11 19:05:25 +00002111/**
Masanari Iidae2278672014-02-18 22:54:36 +09002112 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002113 * @frag: the paged fragment
2114 *
2115 * Returns the &struct page associated with @frag.
2116 */
2117static inline struct page *skb_frag_page(const skb_frag_t *frag)
2118{
Ian Campbella8605c62011-10-19 23:01:49 +00002119 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00002120}
2121
2122/**
2123 * __skb_frag_ref - take an addition reference on a paged fragment.
2124 * @frag: the paged fragment
2125 *
2126 * Takes an additional reference on the paged fragment @frag.
2127 */
2128static inline void __skb_frag_ref(skb_frag_t *frag)
2129{
2130 get_page(skb_frag_page(frag));
2131}
2132
2133/**
2134 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2135 * @skb: the buffer
2136 * @f: the fragment offset.
2137 *
2138 * Takes an additional reference on the @f'th paged fragment of @skb.
2139 */
2140static inline void skb_frag_ref(struct sk_buff *skb, int f)
2141{
2142 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2143}
2144
2145/**
2146 * __skb_frag_unref - release a reference on a paged fragment.
2147 * @frag: the paged fragment
2148 *
2149 * Releases a reference on the paged fragment @frag.
2150 */
2151static inline void __skb_frag_unref(skb_frag_t *frag)
2152{
2153 put_page(skb_frag_page(frag));
2154}
2155
2156/**
2157 * skb_frag_unref - release a reference on a paged fragment of an skb.
2158 * @skb: the buffer
2159 * @f: the fragment offset
2160 *
2161 * Releases a reference on the @f'th paged fragment of @skb.
2162 */
2163static inline void skb_frag_unref(struct sk_buff *skb, int f)
2164{
2165 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2166}
2167
2168/**
2169 * skb_frag_address - gets the address of the data contained in a paged fragment
2170 * @frag: the paged fragment buffer
2171 *
2172 * Returns the address of the data within @frag. The page must already
2173 * be mapped.
2174 */
2175static inline void *skb_frag_address(const skb_frag_t *frag)
2176{
2177 return page_address(skb_frag_page(frag)) + frag->page_offset;
2178}
2179
2180/**
2181 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2182 * @frag: the paged fragment buffer
2183 *
2184 * Returns the address of the data within @frag. Checks that the page
2185 * is mapped and returns %NULL otherwise.
2186 */
2187static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2188{
2189 void *ptr = page_address(skb_frag_page(frag));
2190 if (unlikely(!ptr))
2191 return NULL;
2192
2193 return ptr + frag->page_offset;
2194}
2195
2196/**
2197 * __skb_frag_set_page - sets the page contained in a paged fragment
2198 * @frag: the paged fragment
2199 * @page: the page to set
2200 *
2201 * Sets the fragment @frag to contain @page.
2202 */
2203static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2204{
Ian Campbella8605c62011-10-19 23:01:49 +00002205 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002206}
2207
2208/**
2209 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2210 * @skb: the buffer
2211 * @f: the fragment offset
2212 * @page: the page to set
2213 *
2214 * Sets the @f'th fragment of @skb to contain @page.
2215 */
2216static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2217 struct page *page)
2218{
2219 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2220}
2221
Eric Dumazet400dfd32013-10-17 16:27:07 -07002222bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2223
Ian Campbell131ea662011-08-19 06:25:00 +00002224/**
2225 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002226 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002227 * @frag: the paged fragment to map
2228 * @offset: the offset within the fragment (starting at the
2229 * fragment's own offset)
2230 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002231 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00002232 *
2233 * Maps the page associated with @frag to @device.
2234 */
2235static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2236 const skb_frag_t *frag,
2237 size_t offset, size_t size,
2238 enum dma_data_direction dir)
2239{
2240 return dma_map_page(dev, skb_frag_page(frag),
2241 frag->page_offset + offset, size, dir);
2242}
2243
Eric Dumazet117632e2011-12-03 21:39:53 +00002244static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2245 gfp_t gfp_mask)
2246{
2247 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2248}
2249
Octavian Purdilabad93e92014-06-12 01:36:26 +03002250
2251static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2252 gfp_t gfp_mask)
2253{
2254 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2255}
2256
2257
Ian Campbell131ea662011-08-19 06:25:00 +00002258/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002259 * skb_clone_writable - is the header of a clone writable
2260 * @skb: buffer to check
2261 * @len: length up to which to write
2262 *
2263 * Returns true if modifying the header part of the cloned buffer
2264 * does not requires the data to be copied.
2265 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002266static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002267{
2268 return !skb_header_cloned(skb) &&
2269 skb_headroom(skb) + len <= skb->hdr_len;
2270}
2271
Herbert Xud9cc2042007-09-16 16:21:16 -07002272static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2273 int cloned)
2274{
2275 int delta = 0;
2276
Herbert Xud9cc2042007-09-16 16:21:16 -07002277 if (headroom > skb_headroom(skb))
2278 delta = headroom - skb_headroom(skb);
2279
2280 if (delta || cloned)
2281 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2282 GFP_ATOMIC);
2283 return 0;
2284}
2285
Patrick McHardy334a8132007-06-25 04:35:20 -07002286/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002287 * skb_cow - copy header of skb when it is required
2288 * @skb: buffer to cow
2289 * @headroom: needed headroom
2290 *
2291 * If the skb passed lacks sufficient headroom or its data part
2292 * is shared, data is reallocated. If reallocation fails, an error
2293 * is returned and original skb is not changed.
2294 *
2295 * The result is skb with writable area skb->head...skb->tail
2296 * and at least @headroom of space at head.
2297 */
2298static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2299{
Herbert Xud9cc2042007-09-16 16:21:16 -07002300 return __skb_cow(skb, headroom, skb_cloned(skb));
2301}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002302
Herbert Xud9cc2042007-09-16 16:21:16 -07002303/**
2304 * skb_cow_head - skb_cow but only making the head writable
2305 * @skb: buffer to cow
2306 * @headroom: needed headroom
2307 *
2308 * This function is identical to skb_cow except that we replace the
2309 * skb_cloned check by skb_header_cloned. It should be used when
2310 * you only need to push on some header and do not need to modify
2311 * the data.
2312 */
2313static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2314{
2315 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002316}
2317
2318/**
2319 * skb_padto - pad an skbuff up to a minimal size
2320 * @skb: buffer to pad
2321 * @len: minimal length
2322 *
2323 * Pads up a buffer to ensure the trailing bytes exist and are
2324 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002325 * is untouched. Otherwise it is extended. Returns zero on
2326 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002327 */
2328
Herbert Xu5b057c62006-06-23 02:06:41 -07002329static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002330{
2331 unsigned int size = skb->len;
2332 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002333 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002334 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002335}
2336
2337static inline int skb_add_data(struct sk_buff *skb,
2338 char __user *from, int copy)
2339{
2340 const int off = skb->len;
2341
2342 if (skb->ip_summed == CHECKSUM_NONE) {
2343 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08002344 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
Linus Torvalds1da177e2005-04-16 15:20:36 -07002345 copy, 0, &err);
2346 if (!err) {
2347 skb->csum = csum_block_add(skb->csum, csum, off);
2348 return 0;
2349 }
2350 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2351 return 0;
2352
2353 __skb_trim(skb, off);
2354 return -EFAULT;
2355}
2356
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002357static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2358 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002359{
2360 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002361 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002362
Ian Campbellea2ab692011-08-22 23:44:58 +00002363 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002364 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002365 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002366 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002367}
2368
Herbert Xu364c6ba2006-06-09 16:10:40 -07002369static inline int __skb_linearize(struct sk_buff *skb)
2370{
2371 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2372}
2373
Linus Torvalds1da177e2005-04-16 15:20:36 -07002374/**
2375 * skb_linearize - convert paged skb to linear one
2376 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002377 *
2378 * If there is no free memory -ENOMEM is returned, otherwise zero
2379 * is returned and the old skb data released.
2380 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002381static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002382{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002383 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2384}
2385
2386/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00002387 * skb_has_shared_frag - can any frag be overwritten
2388 * @skb: buffer to test
2389 *
2390 * Return true if the skb has at least one frag that might be modified
2391 * by an external entity (as in vmsplice()/sendfile())
2392 */
2393static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2394{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00002395 return skb_is_nonlinear(skb) &&
2396 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00002397}
2398
2399/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07002400 * skb_linearize_cow - make sure skb is linear and writable
2401 * @skb: buffer to process
2402 *
2403 * If there is no free memory -ENOMEM is returned, otherwise zero
2404 * is returned and the old skb data released.
2405 */
2406static inline int skb_linearize_cow(struct sk_buff *skb)
2407{
2408 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2409 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002410}
2411
2412/**
2413 * skb_postpull_rcsum - update checksum for received skb after pull
2414 * @skb: buffer to update
2415 * @start: start of data before pull
2416 * @len: length of data pulled
2417 *
2418 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002419 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2420 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002421 */
2422
2423static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002424 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002425{
Patrick McHardy84fa7932006-08-29 16:44:56 -07002426 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002427 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2428}
2429
Herbert Xucbb042f2006-03-20 22:43:56 -08002430unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2431
David S. Miller7ce5a272013-12-02 17:26:05 -05002432/**
2433 * pskb_trim_rcsum - trim received skb and update checksum
2434 * @skb: buffer to trim
2435 * @len: new length
2436 *
2437 * This is exactly the same as pskb_trim except that it ensures the
2438 * checksum of received packets are still valid after the operation.
2439 */
2440
2441static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2442{
2443 if (likely(len >= skb->len))
2444 return 0;
2445 if (skb->ip_summed == CHECKSUM_COMPLETE)
2446 skb->ip_summed = CHECKSUM_NONE;
2447 return __pskb_trim(skb, len);
2448}
2449
Linus Torvalds1da177e2005-04-16 15:20:36 -07002450#define skb_queue_walk(queue, skb) \
2451 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002452 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002453 skb = skb->next)
2454
James Chapman46f89142007-04-30 00:07:31 -07002455#define skb_queue_walk_safe(queue, skb, tmp) \
2456 for (skb = (queue)->next, tmp = skb->next; \
2457 skb != (struct sk_buff *)(queue); \
2458 skb = tmp, tmp = skb->next)
2459
David S. Miller1164f522008-09-23 00:49:44 -07002460#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002461 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002462 skb = skb->next)
2463
2464#define skb_queue_walk_from_safe(queue, skb, tmp) \
2465 for (tmp = skb->next; \
2466 skb != (struct sk_buff *)(queue); \
2467 skb = tmp, tmp = skb->next)
2468
Stephen Hemminger300ce172005-10-30 13:47:34 -08002469#define skb_queue_reverse_walk(queue, skb) \
2470 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002471 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08002472 skb = skb->prev)
2473
David S. Miller686a2952011-01-20 22:47:32 -08002474#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2475 for (skb = (queue)->prev, tmp = skb->prev; \
2476 skb != (struct sk_buff *)(queue); \
2477 skb = tmp, tmp = skb->prev)
2478
2479#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2480 for (tmp = skb->prev; \
2481 skb != (struct sk_buff *)(queue); \
2482 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002483
David S. Miller21dc3302010-08-23 00:13:46 -07002484static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07002485{
2486 return skb_shinfo(skb)->frag_list != NULL;
2487}
2488
2489static inline void skb_frag_list_init(struct sk_buff *skb)
2490{
2491 skb_shinfo(skb)->frag_list = NULL;
2492}
2493
2494static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2495{
2496 frag->next = skb_shinfo(skb)->frag_list;
2497 skb_shinfo(skb)->frag_list = frag;
2498}
2499
2500#define skb_walk_frags(skb, iter) \
2501 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2502
Joe Perches7965bd42013-09-26 14:48:15 -07002503struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2504 int *peeked, int *off, int *err);
2505struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2506 int *err);
2507unsigned int datagram_poll(struct file *file, struct socket *sock,
2508 struct poll_table_struct *wait);
2509int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2510 struct iovec *to, int size);
2511int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2512 struct iovec *iov);
2513int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2514 const struct iovec *from, int from_offset,
2515 int len);
2516int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2517 int offset, size_t count);
2518int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2519 const struct iovec *to, int to_offset,
2520 int size);
2521void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2522void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2523int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07002524int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2525int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2526__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2527 int len, __wsum csum);
2528int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2529 struct pipe_inode_info *pipe, unsigned int len,
2530 unsigned int flags);
2531void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01002532unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00002533int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2534 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07002535void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2536int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2537void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Florian Westphalde960aa2014-01-26 10:58:16 +01002538unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -07002539struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002540
Daniel Borkmann2817a332013-10-30 11:50:51 +01002541struct skb_checksum_ops {
2542 __wsum (*update)(const void *mem, int len, __wsum wsum);
2543 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2544};
2545
2546__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2547 __wsum csum, const struct skb_checksum_ops *ops);
2548__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2549 __wsum csum);
2550
Linus Torvalds1da177e2005-04-16 15:20:36 -07002551static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2552 int len, void *buffer)
2553{
2554 int hlen = skb_headlen(skb);
2555
Patrick McHardy55820ee2005-07-05 14:08:10 -07002556 if (hlen - offset >= len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002557 return skb->data + offset;
2558
2559 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2560 return NULL;
2561
2562 return buffer;
2563}
2564
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01002565/**
2566 * skb_needs_linearize - check if we need to linearize a given skb
2567 * depending on the given device features.
2568 * @skb: socket buffer to check
2569 * @features: net device features
2570 *
2571 * Returns true if either:
2572 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2573 * 2. skb is fragmented and the device does not support SG.
2574 */
2575static inline bool skb_needs_linearize(struct sk_buff *skb,
2576 netdev_features_t features)
2577{
2578 return skb_is_nonlinear(skb) &&
2579 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2580 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2581}
2582
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03002583static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2584 void *to,
2585 const unsigned int len)
2586{
2587 memcpy(to, skb->data, len);
2588}
2589
2590static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2591 const int offset, void *to,
2592 const unsigned int len)
2593{
2594 memcpy(to, skb->data + offset, len);
2595}
2596
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03002597static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2598 const void *from,
2599 const unsigned int len)
2600{
2601 memcpy(skb->data, from, len);
2602}
2603
2604static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2605 const int offset,
2606 const void *from,
2607 const unsigned int len)
2608{
2609 memcpy(skb->data + offset, from, len);
2610}
2611
Joe Perches7965bd42013-09-26 14:48:15 -07002612void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002613
Patrick Ohlyac45f602009-02-12 05:03:37 +00002614static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2615{
2616 return skb->tstamp;
2617}
2618
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002619/**
2620 * skb_get_timestamp - get timestamp from a skb
2621 * @skb: skb to get stamp from
2622 * @stamp: pointer to struct timeval to store stamp in
2623 *
2624 * Timestamps are stored in the skb as offsets to a base timestamp.
2625 * This function converts the offset back to a struct timeval and stores
2626 * it in stamp.
2627 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00002628static inline void skb_get_timestamp(const struct sk_buff *skb,
2629 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002630{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002631 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002632}
2633
Patrick Ohlyac45f602009-02-12 05:03:37 +00002634static inline void skb_get_timestampns(const struct sk_buff *skb,
2635 struct timespec *stamp)
2636{
2637 *stamp = ktime_to_timespec(skb->tstamp);
2638}
2639
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002640static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002641{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002642 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002643}
2644
Stephen Hemminger164891a2007-04-23 22:26:16 -07002645static inline ktime_t net_timedelta(ktime_t t)
2646{
2647 return ktime_sub(ktime_get_real(), t);
2648}
2649
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07002650static inline ktime_t net_invalid_timestamp(void)
2651{
2652 return ktime_set(0, 0);
2653}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002654
Richard Cochranc1f19b52010-07-17 08:49:36 +00002655#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2656
Joe Perches7965bd42013-09-26 14:48:15 -07002657void skb_clone_tx_timestamp(struct sk_buff *skb);
2658bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00002659
2660#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2661
2662static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2663{
2664}
2665
2666static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2667{
2668 return false;
2669}
2670
2671#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2672
2673/**
2674 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2675 *
Richard Cochranda92b192011-10-21 00:49:15 +00002676 * PHY drivers may accept clones of transmitted packets for
2677 * timestamping via their phy_driver.txtstamp method. These drivers
2678 * must call this function to return the skb back to the stack, with
2679 * or without a timestamp.
2680 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00002681 * @skb: clone of the the original outgoing packet
Richard Cochranda92b192011-10-21 00:49:15 +00002682 * @hwtstamps: hardware time stamps, may be NULL if not available
Richard Cochranc1f19b52010-07-17 08:49:36 +00002683 *
2684 */
2685void skb_complete_tx_timestamp(struct sk_buff *skb,
2686 struct skb_shared_hwtstamps *hwtstamps);
2687
Patrick Ohlyac45f602009-02-12 05:03:37 +00002688/**
2689 * skb_tstamp_tx - queue clone of skb with send time stamps
2690 * @orig_skb: the original outgoing packet
2691 * @hwtstamps: hardware time stamps, may be NULL if not available
2692 *
2693 * If the skb has a socket associated, then this function clones the
2694 * skb (thus sharing the actual data and optional structures), stores
2695 * the optional hardware time stamping information (if non NULL) or
2696 * generates a software time stamp (otherwise), then queues the clone
2697 * to the error queue of the socket. Errors are silently ignored.
2698 */
Joe Perches7965bd42013-09-26 14:48:15 -07002699void skb_tstamp_tx(struct sk_buff *orig_skb,
2700 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00002701
Richard Cochran4507a712010-07-17 08:48:28 +00002702static inline void sw_tx_timestamp(struct sk_buff *skb)
2703{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002704 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2705 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00002706 skb_tstamp_tx(skb, NULL);
2707}
2708
2709/**
2710 * skb_tx_timestamp() - Driver hook for transmit timestamping
2711 *
2712 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00002713 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00002714 *
David S. Miller73409f32013-12-27 13:04:33 -05002715 * Specifically, one should make absolutely sure that this function is
2716 * called before TX completion of this packet can trigger. Otherwise
2717 * the packet could potentially already be freed.
2718 *
Richard Cochran4507a712010-07-17 08:48:28 +00002719 * @skb: A socket buffer.
2720 */
2721static inline void skb_tx_timestamp(struct sk_buff *skb)
2722{
Richard Cochranc1f19b52010-07-17 08:49:36 +00002723 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002724 sw_tx_timestamp(skb);
2725}
2726
Johannes Berg6e3e9392011-11-09 10:15:42 +01002727/**
2728 * skb_complete_wifi_ack - deliver skb with wifi status
2729 *
2730 * @skb: the original outgoing packet
2731 * @acked: ack status
2732 *
2733 */
2734void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2735
Joe Perches7965bd42013-09-26 14:48:15 -07002736__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2737__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002738
Herbert Xu60476372007-04-09 11:59:39 -07002739static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2740{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002741 return ((skb->ip_summed & CHECKSUM_UNNECESSARY) || skb->csum_valid);
Herbert Xu60476372007-04-09 11:59:39 -07002742}
2743
Herbert Xufb286bb2005-11-10 13:01:24 -08002744/**
2745 * skb_checksum_complete - Calculate checksum of an entire packet
2746 * @skb: packet to process
2747 *
2748 * This function calculates the checksum over the entire packet plus
2749 * the value of skb->csum. The latter can be used to supply the
2750 * checksum of a pseudo header as used by TCP/UDP. It returns the
2751 * checksum.
2752 *
2753 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2754 * this function can be used to verify that checksum on received
2755 * packets. In that case the function should return zero if the
2756 * checksum is correct. In particular, this function will return zero
2757 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2758 * hardware has already verified the correctness of the checksum.
2759 */
Al Viro4381ca32007-07-15 21:00:11 +01002760static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002761{
Herbert Xu60476372007-04-09 11:59:39 -07002762 return skb_csum_unnecessary(skb) ?
2763 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002764}
2765
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002766/* Check if we need to perform checksum complete validation.
2767 *
2768 * Returns true if checksum complete is needed, false otherwise
2769 * (either checksum is unnecessary or zero checksum is allowed).
2770 */
2771static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
2772 bool zero_okay,
2773 __sum16 check)
2774{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002775 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
2776 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002777 return false;
2778 }
2779
2780 return true;
2781}
2782
2783/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
2784 * in checksum_init.
2785 */
2786#define CHECKSUM_BREAK 76
2787
2788/* Validate (init) checksum based on checksum complete.
2789 *
2790 * Return values:
2791 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
2792 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
2793 * checksum is stored in skb->csum for use in __skb_checksum_complete
2794 * non-zero: value of invalid checksum
2795 *
2796 */
2797static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
2798 bool complete,
2799 __wsum psum)
2800{
2801 if (skb->ip_summed == CHECKSUM_COMPLETE) {
2802 if (!csum_fold(csum_add(psum, skb->csum))) {
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002803 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002804 return 0;
2805 }
2806 }
2807
2808 skb->csum = psum;
2809
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002810 if (complete || skb->len <= CHECKSUM_BREAK) {
2811 __sum16 csum;
2812
2813 csum = __skb_checksum_complete(skb);
2814 skb->csum_valid = !csum;
2815 return csum;
2816 }
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002817
2818 return 0;
2819}
2820
2821static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
2822{
2823 return 0;
2824}
2825
2826/* Perform checksum validate (init). Note that this is a macro since we only
2827 * want to calculate the pseudo header which is an input function if necessary.
2828 * First we try to validate without any computation (checksum unnecessary) and
2829 * then calculate based on checksum complete calling the function to compute
2830 * pseudo header.
2831 *
2832 * Return values:
2833 * 0: checksum is validated or try to in skb_checksum_complete
2834 * non-zero: value of invalid checksum
2835 */
2836#define __skb_checksum_validate(skb, proto, complete, \
2837 zero_okay, check, compute_pseudo) \
2838({ \
2839 __sum16 __ret = 0; \
Tom Herbert5d0c2b92014-06-10 18:54:13 -07002840 skb->csum_valid = 0; \
Tom Herbert76ba0aa2014-05-02 16:29:18 -07002841 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
2842 __ret = __skb_checksum_validate_complete(skb, \
2843 complete, compute_pseudo(skb, proto)); \
2844 __ret; \
2845})
2846
2847#define skb_checksum_init(skb, proto, compute_pseudo) \
2848 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
2849
2850#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
2851 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
2852
2853#define skb_checksum_validate(skb, proto, compute_pseudo) \
2854 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
2855
2856#define skb_checksum_validate_zero_check(skb, proto, check, \
2857 compute_pseudo) \
2858 __skb_checksum_validate_(skb, proto, true, true, check, compute_pseudo)
2859
2860#define skb_checksum_simple_validate(skb) \
2861 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
2862
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002863#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07002864void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002865static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2866{
2867 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002868 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002869}
2870static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2871{
2872 if (nfct)
2873 atomic_inc(&nfct->use);
2874}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002875#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002876#ifdef CONFIG_BRIDGE_NETFILTER
2877static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2878{
2879 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2880 kfree(nf_bridge);
2881}
2882static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2883{
2884 if (nf_bridge)
2885 atomic_inc(&nf_bridge->use);
2886}
2887#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08002888static inline void nf_reset(struct sk_buff *skb)
2889{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002890#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08002891 nf_conntrack_put(skb->nfct);
2892 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002893#endif
Patrick McHardya193a4a2006-03-20 19:23:05 -08002894#ifdef CONFIG_BRIDGE_NETFILTER
2895 nf_bridge_put(skb->nf_bridge);
2896 skb->nf_bridge = NULL;
2897#endif
2898}
2899
Patrick McHardy124dff02013-04-05 20:42:05 +02002900static inline void nf_reset_trace(struct sk_buff *skb)
2901{
Florian Westphal478b3602014-02-15 23:48:45 +01002902#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00002903 skb->nf_trace = 0;
2904#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002905}
2906
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002907/* Note: This doesn't put any conntrack and bridge info in dst. */
2908static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2909{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002910#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002911 dst->nfct = src->nfct;
2912 nf_conntrack_get(src->nfct);
2913 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002914#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002915#ifdef CONFIG_BRIDGE_NETFILTER
2916 dst->nf_bridge = src->nf_bridge;
2917 nf_bridge_get(src->nf_bridge);
2918#endif
Florian Westphal478b3602014-02-15 23:48:45 +01002919#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
2920 dst->nf_trace = src->nf_trace;
2921#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002922}
2923
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002924static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2925{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002926#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002927 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002928#endif
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002929#ifdef CONFIG_BRIDGE_NETFILTER
2930 nf_bridge_put(dst->nf_bridge);
2931#endif
2932 __nf_copy(dst, src);
2933}
2934
James Morris984bc162006-06-09 00:29:17 -07002935#ifdef CONFIG_NETWORK_SECMARK
2936static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2937{
2938 to->secmark = from->secmark;
2939}
2940
2941static inline void skb_init_secmark(struct sk_buff *skb)
2942{
2943 skb->secmark = 0;
2944}
2945#else
2946static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2947{ }
2948
2949static inline void skb_init_secmark(struct sk_buff *skb)
2950{ }
2951#endif
2952
Eric W. Biederman574f7192014-04-01 12:20:24 -07002953static inline bool skb_irq_freeable(const struct sk_buff *skb)
2954{
2955 return !skb->destructor &&
2956#if IS_ENABLED(CONFIG_XFRM)
2957 !skb->sp &&
2958#endif
2959#if IS_ENABLED(CONFIG_NF_CONNTRACK)
2960 !skb->nfct &&
2961#endif
2962 !skb->_skb_refdst &&
2963 !skb_has_frag_list(skb);
2964}
2965
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002966static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2967{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002968 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002969}
2970
Stephen Hemminger92477442009-03-21 13:39:26 -07002971static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002972{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002973 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002974}
2975
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002976static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2977{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002978 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002979}
2980
David S. Millerd5a9e242009-01-27 16:22:11 -08002981static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2982{
2983 skb->queue_mapping = rx_queue + 1;
2984}
2985
Stephen Hemminger92477442009-03-21 13:39:26 -07002986static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002987{
2988 return skb->queue_mapping - 1;
2989}
2990
Stephen Hemminger92477442009-03-21 13:39:26 -07002991static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002992{
Eric Dumazeta02cec22010-09-22 20:43:57 +00002993 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08002994}
2995
Joe Perches7965bd42013-09-26 14:48:15 -07002996u16 __skb_tx_hash(const struct net_device *dev, const struct sk_buff *skb,
2997 unsigned int num_tx_queues);
Stephen Hemminger92477442009-03-21 13:39:26 -07002998
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04002999static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3000{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003001#ifdef CONFIG_XFRM
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003002 return skb->sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003003#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003004 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003005#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003006}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003007
Pravin B Shelar68c33162013-02-14 14:02:41 +00003008/* Keeps track of mac header offset relative to skb->head.
3009 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3010 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07003011 * tunnel skb it points to outer mac header.
3012 * Keeps track of level of encapsulation of network headers.
3013 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00003014struct skb_gso_cb {
Eric Dumazet3347c962013-10-19 11:42:56 -07003015 int mac_offset;
3016 int encap_level;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003017 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00003018};
3019#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
3020
3021static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3022{
3023 return (skb_mac_header(inner_skb) - inner_skb->head) -
3024 SKB_GSO_CB(inner_skb)->mac_offset;
3025}
3026
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00003027static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3028{
3029 int new_headroom, headroom;
3030 int ret;
3031
3032 headroom = skb_headroom(skb);
3033 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3034 if (ret)
3035 return ret;
3036
3037 new_headroom = skb_headroom(skb);
3038 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3039 return 0;
3040}
3041
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003042/* Compute the checksum for a gso segment. First compute the checksum value
3043 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3044 * then add in skb->csum (checksum from csum_start to end of packet).
3045 * skb->csum and csum_start are then updated to reflect the checksum of the
3046 * resultant packet starting from the transport header-- the resultant checksum
3047 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3048 * header.
3049 */
3050static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3051{
3052 int plen = SKB_GSO_CB(skb)->csum_start - skb_headroom(skb) -
3053 skb_transport_offset(skb);
3054 __u16 csum;
3055
3056 csum = csum_fold(csum_partial(skb_transport_header(skb),
3057 plen, skb->csum));
3058 skb->csum = res;
3059 SKB_GSO_CB(skb)->csum_start -= plen;
3060
3061 return csum;
3062}
3063
David S. Millerbdcc0922012-03-07 20:53:36 -05003064static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07003065{
3066 return skb_shinfo(skb)->gso_size;
3067}
3068
Eric Dumazet36a8f392013-09-29 01:21:32 -07003069/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05003070static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02003071{
3072 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3073}
3074
Joe Perches7965bd42013-09-26 14:48:15 -07003075void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07003076
3077static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3078{
3079 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3080 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003081 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3082
Alexander Duyckb78462e2010-06-02 12:24:37 +00003083 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3084 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07003085 __skb_warn_lro_forwarding(skb);
3086 return true;
3087 }
3088 return false;
3089}
3090
Herbert Xu35fc92a2007-03-26 23:22:20 -07003091static inline void skb_forward_csum(struct sk_buff *skb)
3092{
3093 /* Unfortunately we don't support this one. Any brave souls? */
3094 if (skb->ip_summed == CHECKSUM_COMPLETE)
3095 skb->ip_summed = CHECKSUM_NONE;
3096}
3097
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003098/**
3099 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3100 * @skb: skb to check
3101 *
3102 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3103 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3104 * use this helper, to document places where we make this assertion.
3105 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003106static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003107{
3108#ifdef DEBUG
3109 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3110#endif
3111}
3112
Rusty Russellf35d9d82008-02-04 23:49:54 -05003113bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00003114
Paul Durranted1f50c2014-01-09 10:02:46 +00003115int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
3116
Daniel Borkmannf77668d2013-03-19 06:39:30 +00003117u32 __skb_get_poff(const struct sk_buff *skb);
3118
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00003119/**
3120 * skb_head_is_locked - Determine if the skb->head is locked down
3121 * @skb: skb to check
3122 *
3123 * The head on skbs build around a head frag can be removed if they are
3124 * not cloned. This function returns true if the skb head is locked down
3125 * due to either being allocated via kmalloc, or by being a clone with
3126 * multiple references to the head.
3127 */
3128static inline bool skb_head_is_locked(const struct sk_buff *skb)
3129{
3130 return !skb->head_frag || skb_cloned(skb);
3131}
Florian Westphalfe6cc552014-02-13 23:09:12 +01003132
3133/**
3134 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3135 *
3136 * @skb: GSO skb
3137 *
3138 * skb_gso_network_seglen is used to determine the real size of the
3139 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3140 *
3141 * The MAC/L2 header is not accounted for.
3142 */
3143static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3144{
3145 unsigned int hdr_len = skb_transport_header(skb) -
3146 skb_network_header(skb);
3147 return hdr_len + skb_gso_transport_seglen(skb);
3148}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003149#endif /* __KERNEL__ */
3150#endif /* _LINUX_SKBUFF_H */