blob: 350eebe770d94800acbb1b72f3b85ca611aeac94 [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * Definitions for the 'struct sk_buff' memory handlers.
3 *
4 * Authors:
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14#ifndef _LINUX_SKBUFF_H
15#define _LINUX_SKBUFF_H
16
Linus Torvalds1da177e2005-04-16 15:20:36 -070017#include <linux/kernel.h>
Vegard Nossumfe55f6d2008-08-30 12:16:35 +020018#include <linux/kmemcheck.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070019#include <linux/compiler.h>
20#include <linux/time.h>
Paul Gortmaker187f1882011-11-23 20:12:59 -050021#include <linux/bug.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070022#include <linux/cache.h>
23
Arun Sharma600634972011-07-26 16:09:06 -070024#include <linux/atomic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025#include <asm/types.h>
26#include <linux/spinlock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <linux/net.h>
Thomas Graf3fc7e8a2005-06-23 21:00:17 -070028#include <linux/textsearch.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070029#include <net/checksum.h>
Al Viroa80958f2006-12-04 20:41:19 +000030#include <linux/rcupdate.h>
Chris Leech97fc2f02006-05-23 17:55:33 -070031#include <linux/dmaengine.h>
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -070032#include <linux/hrtimer.h>
Ian Campbell131ea662011-08-19 06:25:00 +000033#include <linux/dma-mapping.h>
Michał Mirosławc8f44af2011-11-15 15:29:55 +000034#include <linux/netdev_features.h>
Eric Dumazet363ec39232014-02-26 14:02:11 -080035#include <linux/sched.h>
Jason Wang5203cd22013-03-26 23:11:21 +000036#include <net/flow_keys.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070037
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010038/* A. Checksumming of received packets by device.
39 *
40 * CHECKSUM_NONE:
41 *
42 * Device failed to checksum this packet e.g. due to lack of capabilities.
43 * The packet contains full (though not verified) checksum in packet but
44 * not in skb->csum. Thus, skb->csum is undefined in this case.
45 *
46 * CHECKSUM_UNNECESSARY:
47 *
48 * The hardware you're dealing with doesn't calculate the full checksum
49 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
50 * for specific protocols e.g. TCP/UDP/SCTP, then, for such packets it will
51 * set CHECKSUM_UNNECESSARY if their checksums are okay. skb->csum is still
52 * undefined in this case though. It is a bad option, but, unfortunately,
53 * nowadays most vendors do this. Apparently with the secret goal to sell
54 * you new devices, when you will add new protocol to your host, f.e. IPv6 8)
55 *
56 * CHECKSUM_COMPLETE:
57 *
58 * This is the most generic way. The device supplied checksum of the _whole_
59 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
60 * hardware doesn't need to parse L3/L4 headers to implement this.
61 *
62 * Note: Even if device supports only some protocols, but is able to produce
63 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
64 *
65 * CHECKSUM_PARTIAL:
66 *
67 * This is identical to the case for output below. This may occur on a packet
68 * received directly from another Linux OS, e.g., a virtualized Linux kernel
69 * on the same host. The packet can be treated in the same way as
70 * CHECKSUM_UNNECESSARY, except that on output (i.e., forwarding) the
71 * checksum must be filled in by the OS or the hardware.
72 *
73 * B. Checksumming on output.
74 *
75 * CHECKSUM_NONE:
76 *
77 * The skb was already checksummed by the protocol, or a checksum is not
78 * required.
79 *
80 * CHECKSUM_PARTIAL:
81 *
82 * The device is required to checksum the packet as seen by hard_start_xmit()
83 * from skb->csum_start up to the end, and to record/write the checksum at
84 * offset skb->csum_start + skb->csum_offset.
85 *
86 * The device must show its capabilities in dev->features, set up at device
87 * setup time, e.g. netdev_features.h:
88 *
89 * NETIF_F_HW_CSUM - It's a clever device, it's able to checksum everything.
90 * NETIF_F_IP_CSUM - Device is dumb, it's able to checksum only TCP/UDP over
91 * IPv4. Sigh. Vendors like this way for an unknown reason.
92 * Though, see comment above about CHECKSUM_UNNECESSARY. 8)
93 * NETIF_F_IPV6_CSUM - About as dumb as the last one but does IPv6 instead.
94 * NETIF_F_... - Well, you get the picture.
95 *
96 * CHECKSUM_UNNECESSARY:
97 *
98 * Normally, the device will do per protocol specific checksumming. Protocol
99 * implementations that do not want the NIC to perform the checksum
100 * calculation should use this flag in their outgoing skbs.
101 *
102 * NETIF_F_FCOE_CRC - This indicates that the device can do FCoE FC CRC
103 * offload. Correspondingly, the FCoE protocol driver
104 * stack should use CHECKSUM_UNNECESSARY.
105 *
106 * Any questions? No questions, good. --ANK
107 */
108
Herbert Xu60476372007-04-09 11:59:39 -0700109/* Don't change this without changing skb_csum_unnecessary! */
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100110#define CHECKSUM_NONE 0
111#define CHECKSUM_UNNECESSARY 1
112#define CHECKSUM_COMPLETE 2
113#define CHECKSUM_PARTIAL 3
Linus Torvalds1da177e2005-04-16 15:20:36 -0700114
115#define SKB_DATA_ALIGN(X) (((X) + (SMP_CACHE_BYTES - 1)) & \
116 ~(SMP_CACHE_BYTES - 1))
David S. Millerfc910a22007-03-25 20:27:59 -0700117#define SKB_WITH_OVERHEAD(X) \
Herbert Xudeea84b2007-10-21 16:27:46 -0700118 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
David S. Millerfc910a22007-03-25 20:27:59 -0700119#define SKB_MAX_ORDER(X, ORDER) \
120 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700121#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
122#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
123
Eric Dumazet87fb4b72011-10-13 07:28:54 +0000124/* return minimum truesize of one skb containing X bytes of data */
125#define SKB_TRUESIZE(X) ((X) + \
126 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
127 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
128
Linus Torvalds1da177e2005-04-16 15:20:36 -0700129struct net_device;
David Howells716ea3a2007-04-02 20:19:53 -0700130struct scatterlist;
Jens Axboe9c55e012007-11-06 23:30:13 -0800131struct pipe_inode_info;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700132
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700133#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700134struct nf_conntrack {
135 atomic_t use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700136};
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700137#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700138
139#ifdef CONFIG_BRIDGE_NETFILTER
140struct nf_bridge_info {
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000141 atomic_t use;
142 unsigned int mask;
143 struct net_device *physindev;
144 struct net_device *physoutdev;
145 unsigned long data[32 / sizeof(unsigned long)];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700146};
147#endif
148
Linus Torvalds1da177e2005-04-16 15:20:36 -0700149struct sk_buff_head {
150 /* These two members must be first. */
151 struct sk_buff *next;
152 struct sk_buff *prev;
153
154 __u32 qlen;
155 spinlock_t lock;
156};
157
158struct sk_buff;
159
Ian Campbell9d4dde52011-12-22 23:39:14 +0000160/* To allow 64K frame to be packed as single skb without frag_list we
161 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
162 * buffers which do not start on a page boundary.
163 *
164 * Since GRO uses frags we allocate at least 16 regardless of page
165 * size.
Anton Blancharda715dea2011-03-27 14:57:26 +0000166 */
Ian Campbell9d4dde52011-12-22 23:39:14 +0000167#if (65536/PAGE_SIZE + 1) < 16
David S. Millereec00952011-03-29 23:34:08 -0700168#define MAX_SKB_FRAGS 16UL
Anton Blancharda715dea2011-03-27 14:57:26 +0000169#else
Ian Campbell9d4dde52011-12-22 23:39:14 +0000170#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
Anton Blancharda715dea2011-03-27 14:57:26 +0000171#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700172
173typedef struct skb_frag_struct skb_frag_t;
174
175struct skb_frag_struct {
Ian Campbella8605c62011-10-19 23:01:49 +0000176 struct {
177 struct page *p;
178 } page;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000179#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
David S. Millera309bb02007-07-30 18:47:03 -0700180 __u32 page_offset;
181 __u32 size;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000182#else
183 __u16 page_offset;
184 __u16 size;
185#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700186};
187
Eric Dumazet9e903e02011-10-18 21:00:24 +0000188static inline unsigned int skb_frag_size(const skb_frag_t *frag)
189{
190 return frag->size;
191}
192
193static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
194{
195 frag->size = size;
196}
197
198static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
199{
200 frag->size += delta;
201}
202
203static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
204{
205 frag->size -= delta;
206}
207
Patrick Ohlyac45f602009-02-12 05:03:37 +0000208#define HAVE_HW_TIME_STAMP
209
210/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800211 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000212 * @hwtstamp: hardware time stamp transformed into duration
213 * since arbitrary point in time
214 * @syststamp: hwtstamp transformed to system time base
215 *
216 * Software time stamps generated by ktime_get_real() are stored in
217 * skb->tstamp. The relation between the different kinds of time
218 * stamps is as follows:
219 *
220 * syststamp and tstamp can be compared against each other in
221 * arbitrary combinations. The accuracy of a
222 * syststamp/tstamp/"syststamp from other device" comparison is
223 * limited by the accuracy of the transformation into system time
224 * base. This depends on the device driver and its underlying
225 * hardware.
226 *
227 * hwtstamps can only be compared against other hwtstamps from
228 * the same device.
229 *
230 * This structure is attached to packets as part of the
231 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
232 */
233struct skb_shared_hwtstamps {
234 ktime_t hwtstamp;
235 ktime_t syststamp;
236};
237
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000238/* Definitions for tx_flags in struct skb_shared_info */
239enum {
240 /* generate hardware time stamp */
241 SKBTX_HW_TSTAMP = 1 << 0,
242
243 /* generate software time stamp */
244 SKBTX_SW_TSTAMP = 1 << 1,
245
246 /* device driver is going to provide hardware time stamp */
247 SKBTX_IN_PROGRESS = 1 << 2,
248
Shirley Maa6686f22011-07-06 12:22:12 +0000249 /* device driver supports TX zero-copy buffers */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000250 SKBTX_DEV_ZEROCOPY = 1 << 3,
Johannes Berg6e3e9392011-11-09 10:15:42 +0100251
252 /* generate wifi status information (where possible) */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000253 SKBTX_WIFI_STATUS = 1 << 4,
Pravin B Shelarc9af6db2013-02-11 09:27:41 +0000254
255 /* This indicates at least one fragment might be overwritten
256 * (as in vmsplice(), sendfile() ...)
257 * If we need to compute a TX checksum, we'll need to copy
258 * all frags to avoid possible bad checksum
259 */
260 SKBTX_SHARED_FRAG = 1 << 5,
Shirley Maa6686f22011-07-06 12:22:12 +0000261};
262
263/*
264 * The callback notifies userspace to release buffers when skb DMA is done in
265 * lower device, the skb last reference should be 0 when calling this.
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000266 * The zerocopy_success argument is true if zero copy transmit occurred,
267 * false on data copy or out of memory error caused by data copy attempt.
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000268 * The ctx field is used to track device context.
269 * The desc field is used to track userspace buffer index.
Shirley Maa6686f22011-07-06 12:22:12 +0000270 */
271struct ubuf_info {
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000272 void (*callback)(struct ubuf_info *, bool zerocopy_success);
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000273 void *ctx;
Shirley Maa6686f22011-07-06 12:22:12 +0000274 unsigned long desc;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000275};
276
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277/* This data is invariant across clones and lives at
278 * the end of the header data, ie. at skb->end.
279 */
280struct skb_shared_info {
Ian Campbell9f42f122012-01-05 07:13:39 +0000281 unsigned char nr_frags;
282 __u8 tx_flags;
Herbert Xu79671682006-06-22 02:40:14 -0700283 unsigned short gso_size;
284 /* Warning: this field is not always filled in (UFO)! */
285 unsigned short gso_segs;
286 unsigned short gso_type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700287 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000288 struct skb_shared_hwtstamps hwtstamps;
Ian Campbell9f42f122012-01-05 07:13:39 +0000289 __be32 ip6_frag_id;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700290
291 /*
292 * Warning : all fields before dataref are cleared in __alloc_skb()
293 */
294 atomic_t dataref;
295
Johann Baudy69e3c752009-05-18 22:11:22 -0700296 /* Intermediate layers must ensure that destructor_arg
297 * remains valid until skb destructor */
298 void * destructor_arg;
Shirley Maa6686f22011-07-06 12:22:12 +0000299
Eric Dumazetfed66382010-07-22 19:09:08 +0000300 /* must be last field, see pskb_expand_head() */
301 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700302};
303
304/* We divide dataref into two halves. The higher 16 bits hold references
305 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700306 * the entire skb->data. A clone of a headerless skb holds the length of
307 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700308 *
309 * All users must obey the rule that the skb->data reference count must be
310 * greater than or equal to the payload reference count.
311 *
312 * Holding a reference to the payload part means that the user does not
313 * care about modifications to the header part of skb->data.
314 */
315#define SKB_DATAREF_SHIFT 16
316#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
317
David S. Millerd179cd12005-08-17 14:57:30 -0700318
319enum {
320 SKB_FCLONE_UNAVAILABLE,
321 SKB_FCLONE_ORIG,
322 SKB_FCLONE_CLONE,
323};
324
Herbert Xu79671682006-06-22 02:40:14 -0700325enum {
326 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700327 SKB_GSO_UDP = 1 << 1,
Herbert Xu576a30e2006-06-27 13:22:38 -0700328
329 /* This indicates the skb is from an untrusted source. */
330 SKB_GSO_DODGY = 1 << 2,
Michael Chanb0da85372006-06-29 12:30:00 -0700331
332 /* This indicates the tcp segment has CWR set. */
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700333 SKB_GSO_TCP_ECN = 1 << 3,
334
335 SKB_GSO_TCPV6 = 1 << 4,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800336
337 SKB_GSO_FCOE = 1 << 5,
Pravin B Shelar68c33162013-02-14 14:02:41 +0000338
339 SKB_GSO_GRE = 1 << 6,
Pravin B Shelar73136262013-03-07 13:21:51 +0000340
Eric Dumazetcb32f512013-10-19 11:42:57 -0700341 SKB_GSO_IPIP = 1 << 7,
Simon Horman0d89d202013-05-23 21:02:52 +0000342
Eric Dumazet61c1db72013-10-20 20:47:30 -0700343 SKB_GSO_SIT = 1 << 8,
Eric Dumazetcb32f512013-10-19 11:42:57 -0700344
Eric Dumazet61c1db72013-10-20 20:47:30 -0700345 SKB_GSO_UDP_TUNNEL = 1 << 9,
346
347 SKB_GSO_MPLS = 1 << 10,
Herbert Xu79671682006-06-22 02:40:14 -0700348};
349
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700350#if BITS_PER_LONG > 32
351#define NET_SKBUFF_DATA_USES_OFFSET 1
352#endif
353
354#ifdef NET_SKBUFF_DATA_USES_OFFSET
355typedef unsigned int sk_buff_data_t;
356#else
357typedef unsigned char *sk_buff_data_t;
358#endif
359
Eric Dumazet363ec39232014-02-26 14:02:11 -0800360/**
361 * struct skb_mstamp - multi resolution time stamps
362 * @stamp_us: timestamp in us resolution
363 * @stamp_jiffies: timestamp in jiffies
364 */
365struct skb_mstamp {
366 union {
367 u64 v64;
368 struct {
369 u32 stamp_us;
370 u32 stamp_jiffies;
371 };
372 };
373};
374
375/**
376 * skb_mstamp_get - get current timestamp
377 * @cl: place to store timestamps
378 */
379static inline void skb_mstamp_get(struct skb_mstamp *cl)
380{
381 u64 val = local_clock();
382
383 do_div(val, NSEC_PER_USEC);
384 cl->stamp_us = (u32)val;
385 cl->stamp_jiffies = (u32)jiffies;
386}
387
388/**
389 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
390 * @t1: pointer to newest sample
391 * @t0: pointer to oldest sample
392 */
393static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
394 const struct skb_mstamp *t0)
395{
396 s32 delta_us = t1->stamp_us - t0->stamp_us;
397 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
398
399 /* If delta_us is negative, this might be because interval is too big,
400 * or local_clock() drift is too big : fallback using jiffies.
401 */
402 if (delta_us <= 0 ||
403 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
404
405 delta_us = jiffies_to_usecs(delta_jiffies);
406
407 return delta_us;
408}
409
410
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411/**
412 * struct sk_buff - socket buffer
413 * @next: Next buffer in list
414 * @prev: Previous buffer in list
Eric Dumazet363ec39232014-02-26 14:02:11 -0800415 * @tstamp: Time we arrived/left
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700416 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700417 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700418 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000419 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700420 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 * @len: Length of actual data
422 * @data_len: Data length
423 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700424 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700425 * @csum: Checksum (must include start/offset pair)
426 * @csum_start: Offset from skb->head where checksumming should start
427 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700428 * @priority: Packet queueing priority
Martin Waitz67be2dd2005-05-01 08:59:26 -0700429 * @local_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700431 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700432 * @nohdr: Payload reference only, must not modify header
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700433 * @nfctinfo: Relationship of this skb to the connection
Linus Torvalds1da177e2005-04-16 15:20:36 -0700434 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700435 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700436 * @ipvs_property: skbuff is owned by ipvs
Randy Dunlap31729362008-02-18 20:52:13 -0800437 * @peeked: this packet has been seen already, so stats have been
438 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700439 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700440 * @protocol: Packet protocol from driver
441 * @destructor: Destruct function
442 * @nfct: Associated connection, if any
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800444 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445 * @tc_index: Traffic control index
446 * @tc_verd: traffic control verdict
Tom Herbert61b905d2014-03-24 15:34:47 -0700447 * @hash: the packet hash
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700448 * @queue_mapping: Queue mapping for multiqueue devices
Randy Dunlap553a5672008-04-20 10:51:01 -0700449 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700450 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Tom Herbert61b905d2014-03-24 15:34:47 -0700451 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
Changli Gao4ca24622011-08-19 07:26:44 -0700452 * ports.
Johannes Berg6e3e9392011-11-09 10:15:42 +0100453 * @wifi_acked_valid: wifi_acked was set
454 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000455 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Randy Dunlapf4b8ea72006-06-22 16:00:11 -0700456 * @dma_cookie: a cookie to one of several possible DMA operations
457 * done by skb DMA functions
Eliezer Tamir06021292013-06-10 11:39:50 +0300458 * @napi_id: id of the NAPI struct this skb came from
James Morris984bc162006-06-09 00:29:17 -0700459 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700460 * @mark: Generic packet mark
461 * @dropcount: total number of sk_receive_queue overflows
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000462 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b02008-07-14 22:49:06 -0700463 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000464 * @inner_protocol: Protocol (encapsulation)
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000465 * @inner_transport_header: Inner transport layer header (encapsulation)
466 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000467 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700468 * @transport_header: Transport layer header
469 * @network_header: Network layer header
470 * @mac_header: Link layer header
471 * @tail: Tail pointer
472 * @end: End pointer
473 * @head: Head of buffer
474 * @data: Data head pointer
475 * @truesize: Buffer size
476 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477 */
478
479struct sk_buff {
480 /* These two members must be first. */
481 struct sk_buff *next;
482 struct sk_buff *prev;
483
Eric Dumazet363ec39232014-02-26 14:02:11 -0800484 union {
485 ktime_t tstamp;
486 struct skb_mstamp skb_mstamp;
487 };
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000488
489 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700490 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700491
Linus Torvalds1da177e2005-04-16 15:20:36 -0700492 /*
493 * This is the control buffer. It is free to use for every
494 * layer. Please put your private variables there. If you
495 * want to keep them across layers you have to do a skb_clone()
496 * first. This is owned by whoever has the skb queued ATM.
497 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000498 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700499
Eric Dumazet7fee2262010-05-11 23:19:48 +0000500 unsigned long _skb_refdst;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000501#ifdef CONFIG_XFRM
502 struct sec_path *sp;
503#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700504 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700505 data_len;
506 __u16 mac_len,
507 hdr_len;
Al Viroff1dcad2006-11-20 18:07:29 -0800508 union {
509 __wsum csum;
Herbert Xu663ead32007-04-09 11:59:07 -0700510 struct {
511 __u16 csum_start;
512 __u16 csum_offset;
513 };
Al Viroff1dcad2006-11-20 18:07:29 -0800514 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700515 __u32 priority;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200516 kmemcheck_bitfield_begin(flags1);
Thomas Graf1cbb3382005-07-05 14:13:41 -0700517 __u8 local_df:1,
518 cloned:1,
519 ip_summed:2,
Harald Welte6869c4d2005-08-09 19:24:19 -0700520 nohdr:1,
521 nfctinfo:3;
David S. Millerd179cd12005-08-17 14:57:30 -0700522 __u8 pkt_type:3,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800523 fclone:2,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700524 ipvs_property:1,
Herbert Xua59322b2007-12-05 01:53:40 -0800525 peeked:1,
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700526 nf_trace:1;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200527 kmemcheck_bitfield_end(flags1);
Eric Dumazet4ab408d2010-03-01 03:09:26 +0000528 __be16 protocol;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529
530 void (*destructor)(struct sk_buff *skb);
Yasuyuki Kozakai9fb9cbb2005-11-09 16:38:16 -0800531#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700532 struct nf_conntrack *nfct;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100533#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534#ifdef CONFIG_BRIDGE_NETFILTER
535 struct nf_bridge_info *nf_bridge;
536#endif
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -0700537
Eric Dumazet8964be42009-11-20 15:35:04 -0800538 int skb_iif;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000539
Tom Herbert61b905d2014-03-24 15:34:47 -0700540 __u32 hash;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000541
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000542 __be16 vlan_proto;
Alexander Duyck4031ae62012-01-27 06:22:53 +0000543 __u16 vlan_tci;
544
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700546 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700548 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700550#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200551
Eric Dumazet0a148422011-04-20 09:27:32 +0000552 __u16 queue_mapping;
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200553 kmemcheck_bitfield_begin(flags2);
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400554#ifdef CONFIG_IPV6_NDISC_NODETYPE
Jiri Pirko8a4eb572011-03-12 03:14:39 +0000555 __u8 ndisc_nodetype:2;
YOSHIFUJI Hideakide357cc2008-03-15 23:59:18 -0400556#endif
Mel Gormanc93bdd02012-07-31 16:44:19 -0700557 __u8 pfmemalloc:1;
Tom Herbert3853b582010-11-21 13:17:29 +0000558 __u8 ooo_okay:1;
Tom Herbert61b905d2014-03-24 15:34:47 -0700559 __u8 l4_hash:1;
Johannes Berg6e3e9392011-11-09 10:15:42 +0100560 __u8 wifi_acked_valid:1;
561 __u8 wifi_acked:1;
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000562 __u8 no_fcs:1;
Eric Dumazetd3836f22012-04-27 00:33:38 +0000563 __u8 head_frag:1;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000564 /* Encapsulation protocol and NIC drivers should use
565 * this flag to indicate to each other if the skb contains
566 * encapsulated packet or not and maybe use the inner packet
567 * headers if needed
568 */
569 __u8 encapsulation:1;
Nicolas Dichtel45906722013-09-30 14:16:41 +0200570 /* 6/8 bit hole (depending on ndisc_nodetype presence) */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200571 kmemcheck_bitfield_end(flags2);
572
Cong Wange0d10952013-08-01 11:10:25 +0800573#if defined CONFIG_NET_DMA || defined CONFIG_NET_RX_BUSY_POLL
Eliezer Tamir06021292013-06-10 11:39:50 +0300574 union {
575 unsigned int napi_id;
576 dma_cookie_t dma_cookie;
577 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700578#endif
James Morris984bc162006-06-09 00:29:17 -0700579#ifdef CONFIG_NETWORK_SECMARK
580 __u32 secmark;
581#endif
Neil Horman3b885782009-10-12 13:26:31 -0700582 union {
583 __u32 mark;
584 __u32 dropcount;
Eric Dumazet16fad692013-03-14 05:40:32 +0000585 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700586 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587
Simon Horman0d89d202013-05-23 21:02:52 +0000588 __be16 inner_protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000589 __u16 inner_transport_header;
590 __u16 inner_network_header;
591 __u16 inner_mac_header;
592 __u16 transport_header;
593 __u16 network_header;
594 __u16 mac_header;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700596 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700597 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700598 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700599 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700600 unsigned int truesize;
601 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602};
603
604#ifdef __KERNEL__
605/*
606 * Handling routines are only of interest to the kernel
607 */
608#include <linux/slab.h>
609
Linus Torvalds1da177e2005-04-16 15:20:36 -0700610
Mel Gormanc93bdd02012-07-31 16:44:19 -0700611#define SKB_ALLOC_FCLONE 0x01
612#define SKB_ALLOC_RX 0x02
613
614/* Returns true if the skb was allocated from PFMEMALLOC reserves */
615static inline bool skb_pfmemalloc(const struct sk_buff *skb)
616{
617 return unlikely(skb->pfmemalloc);
618}
619
Eric Dumazet7fee2262010-05-11 23:19:48 +0000620/*
621 * skb might have a dst pointer attached, refcounted or not.
622 * _skb_refdst low order bit is set if refcount was _not_ taken
623 */
624#define SKB_DST_NOREF 1UL
625#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
626
627/**
628 * skb_dst - returns skb dst_entry
629 * @skb: buffer
630 *
631 * Returns skb dst_entry, regardless of reference taken or not.
632 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000633static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
634{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000635 /* If refdst was not refcounted, check we still are in a
636 * rcu_read_lock section
637 */
638 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
639 !rcu_read_lock_held() &&
640 !rcu_read_lock_bh_held());
641 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000642}
643
Eric Dumazet7fee2262010-05-11 23:19:48 +0000644/**
645 * skb_dst_set - sets skb dst
646 * @skb: buffer
647 * @dst: dst entry
648 *
649 * Sets skb dst, assuming a reference was taken on dst and should
650 * be released by skb_dst_drop()
651 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000652static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
653{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000654 skb->_skb_refdst = (unsigned long)dst;
655}
656
Joe Perches7965bd42013-09-26 14:48:15 -0700657void __skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst,
658 bool force);
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200659
660/**
661 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
662 * @skb: buffer
663 * @dst: dst entry
664 *
665 * Sets skb dst, assuming a reference was not taken on dst.
666 * If dst entry is cached, we do not take reference and dst_release
667 * will be avoided by refdst_drop. If dst entry is not cached, we take
668 * reference, so that last dst_release can destroy the dst immediately.
669 */
670static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
671{
672 __skb_dst_set_noref(skb, dst, false);
673}
674
675/**
676 * skb_dst_set_noref_force - sets skb dst, without taking reference
677 * @skb: buffer
678 * @dst: dst entry
679 *
680 * Sets skb dst, assuming a reference was not taken on dst.
681 * No reference is taken and no dst_release will be called. While for
682 * cached dsts deferred reclaim is a basic feature, for entries that are
683 * not cached it is caller's job to guarantee that last dst_release for
684 * provided dst happens when nobody uses it, eg. after a RCU grace period.
685 */
686static inline void skb_dst_set_noref_force(struct sk_buff *skb,
687 struct dst_entry *dst)
688{
689 __skb_dst_set_noref(skb, dst, true);
690}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000691
692/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300693 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000694 * @skb: buffer
695 */
696static inline bool skb_dst_is_noref(const struct sk_buff *skb)
697{
698 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000699}
700
Eric Dumazet511c3f92009-06-02 05:14:27 +0000701static inline struct rtable *skb_rtable(const struct sk_buff *skb)
702{
Eric Dumazetadf30902009-06-02 05:19:30 +0000703 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000704}
705
Joe Perches7965bd42013-09-26 14:48:15 -0700706void kfree_skb(struct sk_buff *skb);
707void kfree_skb_list(struct sk_buff *segs);
708void skb_tx_error(struct sk_buff *skb);
709void consume_skb(struct sk_buff *skb);
710void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000711extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000712
Joe Perches7965bd42013-09-26 14:48:15 -0700713void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
714bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
715 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000716
Joe Perches7965bd42013-09-26 14:48:15 -0700717struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
718 int node);
719struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700720static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100721 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700722{
Eric Dumazet564824b2010-10-11 19:05:25 +0000723 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700724}
725
726static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100727 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700728{
Mel Gormanc93bdd02012-07-31 16:44:19 -0700729 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700730}
731
Joe Perches7965bd42013-09-26 14:48:15 -0700732struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
Patrick McHardy0ebd0ac2013-04-17 06:46:58 +0000733static inline struct sk_buff *alloc_skb_head(gfp_t priority)
734{
735 return __alloc_skb_head(priority, -1);
736}
737
Joe Perches7965bd42013-09-26 14:48:15 -0700738struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
739int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
740struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
741struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
742struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom, gfp_t gfp_mask);
Eric Dumazet117632e2011-12-03 21:39:53 +0000743
Joe Perches7965bd42013-09-26 14:48:15 -0700744int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
745struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
746 unsigned int headroom);
747struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
748 int newtailroom, gfp_t priority);
Fan Du25a91d82014-01-18 09:54:23 +0800749int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
750 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -0700751int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
752 int len);
753int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
754int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000755#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700756
Joe Perches7965bd42013-09-26 14:48:15 -0700757int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
758 int getfrag(void *from, char *to, int offset,
759 int len, int odd, struct sk_buff *skb),
760 void *from, int length);
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700761
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800762struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700763 __u32 lower_offset;
764 __u32 upper_offset;
765 __u32 frag_idx;
766 __u32 stepped_offset;
767 struct sk_buff *root_skb;
768 struct sk_buff *cur_skb;
769 __u8 *frag_data;
770};
771
Joe Perches7965bd42013-09-26 14:48:15 -0700772void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
773 unsigned int to, struct skb_seq_state *st);
774unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
775 struct skb_seq_state *st);
776void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -0700777
Joe Perches7965bd42013-09-26 14:48:15 -0700778unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
779 unsigned int to, struct ts_config *config,
780 struct ts_state *state);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700781
Tom Herbert09323cc2013-12-15 22:16:19 -0800782/*
783 * Packet hash types specify the type of hash in skb_set_hash.
784 *
785 * Hash types refer to the protocol layer addresses which are used to
786 * construct a packet's hash. The hashes are used to differentiate or identify
787 * flows of the protocol layer for the hash type. Hash types are either
788 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
789 *
790 * Properties of hashes:
791 *
792 * 1) Two packets in different flows have different hash values
793 * 2) Two packets in the same flow should have the same hash value
794 *
795 * A hash at a higher layer is considered to be more specific. A driver should
796 * set the most specific hash possible.
797 *
798 * A driver cannot indicate a more specific hash than the layer at which a hash
799 * was computed. For instance an L3 hash cannot be set as an L4 hash.
800 *
801 * A driver may indicate a hash level which is less specific than the
802 * actual layer the hash was computed on. For instance, a hash computed
803 * at L4 may be considered an L3 hash. This should only be done if the
804 * driver can't unambiguously determine that the HW computed the hash at
805 * the higher layer. Note that the "should" in the second property above
806 * permits this.
807 */
808enum pkt_hash_types {
809 PKT_HASH_TYPE_NONE, /* Undefined type */
810 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
811 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
812 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
813};
814
815static inline void
816skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
817{
Tom Herbert61b905d2014-03-24 15:34:47 -0700818 skb->l4_hash = (type == PKT_HASH_TYPE_L4);
819 skb->hash = hash;
Tom Herbert09323cc2013-12-15 22:16:19 -0800820}
821
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800822void __skb_get_hash(struct sk_buff *skb);
823static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000824{
Tom Herbert61b905d2014-03-24 15:34:47 -0700825 if (!skb->l4_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -0800826 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000827
Tom Herbert61b905d2014-03-24 15:34:47 -0700828 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +0000829}
830
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800831static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
832{
Tom Herbert61b905d2014-03-24 15:34:47 -0700833 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -0800834}
835
Tom Herbert7539fad2013-12-15 22:12:18 -0800836static inline void skb_clear_hash(struct sk_buff *skb)
837{
Tom Herbert61b905d2014-03-24 15:34:47 -0700838 skb->hash = 0;
839 skb->l4_hash = 0;
Tom Herbert7539fad2013-12-15 22:12:18 -0800840}
841
842static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
843{
Tom Herbert61b905d2014-03-24 15:34:47 -0700844 if (!skb->l4_hash)
Tom Herbert7539fad2013-12-15 22:12:18 -0800845 skb_clear_hash(skb);
846}
847
Tom Herbert3df7a742013-12-15 22:16:29 -0800848static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
849{
Tom Herbert61b905d2014-03-24 15:34:47 -0700850 to->hash = from->hash;
851 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -0800852};
853
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700854#ifdef NET_SKBUFF_DATA_USES_OFFSET
855static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
856{
857 return skb->head + skb->end;
858}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000859
860static inline unsigned int skb_end_offset(const struct sk_buff *skb)
861{
862 return skb->end;
863}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700864#else
865static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
866{
867 return skb->end;
868}
Alexander Duyckec47ea82012-05-04 14:26:56 +0000869
870static inline unsigned int skb_end_offset(const struct sk_buff *skb)
871{
872 return skb->end - skb->head;
873}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700874#endif
875
Linus Torvalds1da177e2005-04-16 15:20:36 -0700876/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700877#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700878
Patrick Ohlyac45f602009-02-12 05:03:37 +0000879static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
880{
881 return &skb_shinfo(skb)->hwtstamps;
882}
883
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884/**
885 * skb_queue_empty - check if a queue is empty
886 * @list: queue head
887 *
888 * Returns true if the queue is empty, false otherwise.
889 */
890static inline int skb_queue_empty(const struct sk_buff_head *list)
891{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100892 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700893}
894
895/**
David S. Millerfc7ebb22008-09-23 00:34:07 -0700896 * skb_queue_is_last - check if skb is the last entry in the queue
897 * @list: queue head
898 * @skb: buffer
899 *
900 * Returns true if @skb is the last buffer on the list.
901 */
902static inline bool skb_queue_is_last(const struct sk_buff_head *list,
903 const struct sk_buff *skb)
904{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100905 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -0700906}
907
908/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800909 * skb_queue_is_first - check if skb is the first entry in the queue
910 * @list: queue head
911 * @skb: buffer
912 *
913 * Returns true if @skb is the first buffer on the list.
914 */
915static inline bool skb_queue_is_first(const struct sk_buff_head *list,
916 const struct sk_buff *skb)
917{
Daniel Borkmannfd44b932014-01-07 23:23:44 +0100918 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800919}
920
921/**
David S. Miller249c8b42008-09-23 00:44:42 -0700922 * skb_queue_next - return the next packet in the queue
923 * @list: queue head
924 * @skb: current buffer
925 *
926 * Return the next packet in @list after @skb. It is only valid to
927 * call this if skb_queue_is_last() evaluates to false.
928 */
929static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
930 const struct sk_buff *skb)
931{
932 /* This BUG_ON may seem severe, but if we just return then we
933 * are going to dereference garbage.
934 */
935 BUG_ON(skb_queue_is_last(list, skb));
936 return skb->next;
937}
938
939/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -0800940 * skb_queue_prev - return the prev packet in the queue
941 * @list: queue head
942 * @skb: current buffer
943 *
944 * Return the prev packet in @list before @skb. It is only valid to
945 * call this if skb_queue_is_first() evaluates to false.
946 */
947static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
948 const struct sk_buff *skb)
949{
950 /* This BUG_ON may seem severe, but if we just return then we
951 * are going to dereference garbage.
952 */
953 BUG_ON(skb_queue_is_first(list, skb));
954 return skb->prev;
955}
956
957/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700958 * skb_get - reference buffer
959 * @skb: buffer to reference
960 *
961 * Makes another reference to a socket buffer and returns a pointer
962 * to the buffer.
963 */
964static inline struct sk_buff *skb_get(struct sk_buff *skb)
965{
966 atomic_inc(&skb->users);
967 return skb;
968}
969
970/*
971 * If users == 1, we are the only owner and are can avoid redundant
972 * atomic change.
973 */
974
975/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700976 * skb_cloned - is the buffer a clone
977 * @skb: buffer to check
978 *
979 * Returns true if the buffer was generated with skb_clone() and is
980 * one of multiple shared copies of the buffer. Cloned buffers are
981 * shared data so must not be written to under normal circumstances.
982 */
983static inline int skb_cloned(const struct sk_buff *skb)
984{
985 return skb->cloned &&
986 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
987}
988
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +0000989static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
990{
991 might_sleep_if(pri & __GFP_WAIT);
992
993 if (skb_cloned(skb))
994 return pskb_expand_head(skb, 0, 0, pri);
995
996 return 0;
997}
998
Linus Torvalds1da177e2005-04-16 15:20:36 -0700999/**
1000 * skb_header_cloned - is the header a clone
1001 * @skb: buffer to check
1002 *
1003 * Returns true if modifying the header part of the buffer requires
1004 * the data to be copied.
1005 */
1006static inline int skb_header_cloned(const struct sk_buff *skb)
1007{
1008 int dataref;
1009
1010 if (!skb->cloned)
1011 return 0;
1012
1013 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1014 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1015 return dataref != 1;
1016}
1017
1018/**
1019 * skb_header_release - release reference to header
1020 * @skb: buffer to operate on
1021 *
1022 * Drop a reference to the header part of the buffer. This is done
1023 * by acquiring a payload reference. You must not read from the header
1024 * part of skb->data after this.
1025 */
1026static inline void skb_header_release(struct sk_buff *skb)
1027{
1028 BUG_ON(skb->nohdr);
1029 skb->nohdr = 1;
1030 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1031}
1032
1033/**
1034 * skb_shared - is the buffer shared
1035 * @skb: buffer to check
1036 *
1037 * Returns true if more than one person has a reference to this
1038 * buffer.
1039 */
1040static inline int skb_shared(const struct sk_buff *skb)
1041{
1042 return atomic_read(&skb->users) != 1;
1043}
1044
1045/**
1046 * skb_share_check - check if buffer is shared and if so clone it
1047 * @skb: buffer to check
1048 * @pri: priority for memory allocation
1049 *
1050 * If the buffer is shared the buffer is cloned and the old copy
1051 * drops a reference. A new clone with a single reference is returned.
1052 * If the buffer is not shared the original buffer is returned. When
1053 * being called from interrupt status or with spinlocks held pri must
1054 * be GFP_ATOMIC.
1055 *
1056 * NULL is returned on a memory allocation failure.
1057 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001058static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001059{
1060 might_sleep_if(pri & __GFP_WAIT);
1061 if (skb_shared(skb)) {
1062 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001063
1064 if (likely(nskb))
1065 consume_skb(skb);
1066 else
1067 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001068 skb = nskb;
1069 }
1070 return skb;
1071}
1072
1073/*
1074 * Copy shared buffers into a new sk_buff. We effectively do COW on
1075 * packets to handle cases where we have a local reader and forward
1076 * and a couple of other messy ones. The normal one is tcpdumping
1077 * a packet thats being forwarded.
1078 */
1079
1080/**
1081 * skb_unshare - make a copy of a shared buffer
1082 * @skb: buffer to check
1083 * @pri: priority for memory allocation
1084 *
1085 * If the socket buffer is a clone then this function creates a new
1086 * copy of the data, drops a reference count on the old copy and returns
1087 * the new copy with the reference count at 1. If the buffer is not a clone
1088 * the original buffer is returned. When called with a spinlock held or
1089 * from interrupt state @pri must be %GFP_ATOMIC
1090 *
1091 * %NULL is returned on a memory allocation failure.
1092 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001093static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001094 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001095{
1096 might_sleep_if(pri & __GFP_WAIT);
1097 if (skb_cloned(skb)) {
1098 struct sk_buff *nskb = skb_copy(skb, pri);
1099 kfree_skb(skb); /* Free our shared copy */
1100 skb = nskb;
1101 }
1102 return skb;
1103}
1104
1105/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001106 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001107 * @list_: list to peek at
1108 *
1109 * Peek an &sk_buff. Unlike most other operations you _MUST_
1110 * be careful with this one. A peek leaves the buffer on the
1111 * list and someone else may run off with it. You must hold
1112 * the appropriate locks or have a private queue to do this.
1113 *
1114 * Returns %NULL for an empty list or a pointer to the head element.
1115 * The reference count is not incremented and the reference is therefore
1116 * volatile. Use with caution.
1117 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001118static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001119{
Eric Dumazet18d07002012-04-30 16:31:46 +00001120 struct sk_buff *skb = list_->next;
1121
1122 if (skb == (struct sk_buff *)list_)
1123 skb = NULL;
1124 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001125}
1126
1127/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001128 * skb_peek_next - peek skb following the given one from a queue
1129 * @skb: skb to start from
1130 * @list_: list to peek at
1131 *
1132 * Returns %NULL when the end of the list is met or a pointer to the
1133 * next element. The reference count is not incremented and the
1134 * reference is therefore volatile. Use with caution.
1135 */
1136static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1137 const struct sk_buff_head *list_)
1138{
1139 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001140
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001141 if (next == (struct sk_buff *)list_)
1142 next = NULL;
1143 return next;
1144}
1145
1146/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001147 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001148 * @list_: list to peek at
1149 *
1150 * Peek an &sk_buff. Unlike most other operations you _MUST_
1151 * be careful with this one. A peek leaves the buffer on the
1152 * list and someone else may run off with it. You must hold
1153 * the appropriate locks or have a private queue to do this.
1154 *
1155 * Returns %NULL for an empty list or a pointer to the tail element.
1156 * The reference count is not incremented and the reference is therefore
1157 * volatile. Use with caution.
1158 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001159static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001160{
Eric Dumazet18d07002012-04-30 16:31:46 +00001161 struct sk_buff *skb = list_->prev;
1162
1163 if (skb == (struct sk_buff *)list_)
1164 skb = NULL;
1165 return skb;
1166
Linus Torvalds1da177e2005-04-16 15:20:36 -07001167}
1168
1169/**
1170 * skb_queue_len - get queue length
1171 * @list_: list to measure
1172 *
1173 * Return the length of an &sk_buff queue.
1174 */
1175static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1176{
1177 return list_->qlen;
1178}
1179
David S. Miller67fed452008-09-21 22:36:24 -07001180/**
1181 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1182 * @list: queue to initialize
1183 *
1184 * This initializes only the list and queue length aspects of
1185 * an sk_buff_head object. This allows to initialize the list
1186 * aspects of an sk_buff_head without reinitializing things like
1187 * the spinlock. It can also be used for on-stack sk_buff_head
1188 * objects where the spinlock is known to not be used.
1189 */
1190static inline void __skb_queue_head_init(struct sk_buff_head *list)
1191{
1192 list->prev = list->next = (struct sk_buff *)list;
1193 list->qlen = 0;
1194}
1195
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001196/*
1197 * This function creates a split out lock class for each invocation;
1198 * this is needed for now since a whole lot of users of the skb-queue
1199 * infrastructure in drivers have different locking usage (in hardirq)
1200 * than the networking core (in softirq only). In the long run either the
1201 * network layer or drivers should need annotation to consolidate the
1202 * main types of usage into 3 classes.
1203 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001204static inline void skb_queue_head_init(struct sk_buff_head *list)
1205{
1206 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001207 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001208}
1209
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001210static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1211 struct lock_class_key *class)
1212{
1213 skb_queue_head_init(list);
1214 lockdep_set_class(&list->lock, class);
1215}
1216
Linus Torvalds1da177e2005-04-16 15:20:36 -07001217/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001218 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001219 *
1220 * The "__skb_xxxx()" functions are the non-atomic ones that
1221 * can only be called with interrupts disabled.
1222 */
Joe Perches7965bd42013-09-26 14:48:15 -07001223void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1224 struct sk_buff_head *list);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001225static inline void __skb_insert(struct sk_buff *newsk,
1226 struct sk_buff *prev, struct sk_buff *next,
1227 struct sk_buff_head *list)
1228{
1229 newsk->next = next;
1230 newsk->prev = prev;
1231 next->prev = prev->next = newsk;
1232 list->qlen++;
1233}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001234
David S. Miller67fed452008-09-21 22:36:24 -07001235static inline void __skb_queue_splice(const struct sk_buff_head *list,
1236 struct sk_buff *prev,
1237 struct sk_buff *next)
1238{
1239 struct sk_buff *first = list->next;
1240 struct sk_buff *last = list->prev;
1241
1242 first->prev = prev;
1243 prev->next = first;
1244
1245 last->next = next;
1246 next->prev = last;
1247}
1248
1249/**
1250 * skb_queue_splice - join two skb lists, this is designed for stacks
1251 * @list: the new list to add
1252 * @head: the place to add it in the first list
1253 */
1254static inline void skb_queue_splice(const struct sk_buff_head *list,
1255 struct sk_buff_head *head)
1256{
1257 if (!skb_queue_empty(list)) {
1258 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001259 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001260 }
1261}
1262
1263/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001264 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001265 * @list: the new list to add
1266 * @head: the place to add it in the first list
1267 *
1268 * The list at @list is reinitialised
1269 */
1270static inline void skb_queue_splice_init(struct sk_buff_head *list,
1271 struct sk_buff_head *head)
1272{
1273 if (!skb_queue_empty(list)) {
1274 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001275 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001276 __skb_queue_head_init(list);
1277 }
1278}
1279
1280/**
1281 * skb_queue_splice_tail - join two skb lists, each list being a queue
1282 * @list: the new list to add
1283 * @head: the place to add it in the first list
1284 */
1285static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1286 struct sk_buff_head *head)
1287{
1288 if (!skb_queue_empty(list)) {
1289 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001290 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001291 }
1292}
1293
1294/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001295 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001296 * @list: the new list to add
1297 * @head: the place to add it in the first list
1298 *
1299 * Each of the lists is a queue.
1300 * The list at @list is reinitialised
1301 */
1302static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1303 struct sk_buff_head *head)
1304{
1305 if (!skb_queue_empty(list)) {
1306 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001307 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001308 __skb_queue_head_init(list);
1309 }
1310}
1311
Linus Torvalds1da177e2005-04-16 15:20:36 -07001312/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001313 * __skb_queue_after - queue a buffer at the list head
1314 * @list: list to use
1315 * @prev: place after this buffer
1316 * @newsk: buffer to queue
1317 *
1318 * Queue a buffer int the middle of a list. This function takes no locks
1319 * and you must therefore hold required locks before calling it.
1320 *
1321 * A buffer cannot be placed on two lists at the same time.
1322 */
1323static inline void __skb_queue_after(struct sk_buff_head *list,
1324 struct sk_buff *prev,
1325 struct sk_buff *newsk)
1326{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001327 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001328}
1329
Joe Perches7965bd42013-09-26 14:48:15 -07001330void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1331 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001332
Gerrit Renkerf5572852008-04-14 00:05:28 -07001333static inline void __skb_queue_before(struct sk_buff_head *list,
1334 struct sk_buff *next,
1335 struct sk_buff *newsk)
1336{
1337 __skb_insert(newsk, next->prev, next, list);
1338}
1339
Stephen Hemminger300ce172005-10-30 13:47:34 -08001340/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001341 * __skb_queue_head - queue a buffer at the list head
1342 * @list: list to use
1343 * @newsk: buffer to queue
1344 *
1345 * Queue a buffer at the start of a list. This function takes no locks
1346 * and you must therefore hold required locks before calling it.
1347 *
1348 * A buffer cannot be placed on two lists at the same time.
1349 */
Joe Perches7965bd42013-09-26 14:48:15 -07001350void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001351static inline void __skb_queue_head(struct sk_buff_head *list,
1352 struct sk_buff *newsk)
1353{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001354 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001355}
1356
1357/**
1358 * __skb_queue_tail - queue a buffer at the list tail
1359 * @list: list to use
1360 * @newsk: buffer to queue
1361 *
1362 * Queue a buffer at the end of a list. This function takes no locks
1363 * and you must therefore hold required locks before calling it.
1364 *
1365 * A buffer cannot be placed on two lists at the same time.
1366 */
Joe Perches7965bd42013-09-26 14:48:15 -07001367void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001368static inline void __skb_queue_tail(struct sk_buff_head *list,
1369 struct sk_buff *newsk)
1370{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001371 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001372}
1373
Linus Torvalds1da177e2005-04-16 15:20:36 -07001374/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001375 * remove sk_buff from list. _Must_ be called atomically, and with
1376 * the list known..
1377 */
Joe Perches7965bd42013-09-26 14:48:15 -07001378void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1380{
1381 struct sk_buff *next, *prev;
1382
1383 list->qlen--;
1384 next = skb->next;
1385 prev = skb->prev;
1386 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001387 next->prev = prev;
1388 prev->next = next;
1389}
1390
Gerrit Renkerf525c062008-04-14 00:04:12 -07001391/**
1392 * __skb_dequeue - remove from the head of the queue
1393 * @list: list to dequeue from
1394 *
1395 * Remove the head of the list. This function does not take any locks
1396 * so must be used with appropriate locks held only. The head item is
1397 * returned or %NULL if the list is empty.
1398 */
Joe Perches7965bd42013-09-26 14:48:15 -07001399struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Gerrit Renkerf525c062008-04-14 00:04:12 -07001400static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1401{
1402 struct sk_buff *skb = skb_peek(list);
1403 if (skb)
1404 __skb_unlink(skb, list);
1405 return skb;
1406}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407
1408/**
1409 * __skb_dequeue_tail - remove from the tail of the queue
1410 * @list: list to dequeue from
1411 *
1412 * Remove the tail of the list. This function does not take any locks
1413 * so must be used with appropriate locks held only. The tail item is
1414 * returned or %NULL if the list is empty.
1415 */
Joe Perches7965bd42013-09-26 14:48:15 -07001416struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001417static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1418{
1419 struct sk_buff *skb = skb_peek_tail(list);
1420 if (skb)
1421 __skb_unlink(skb, list);
1422 return skb;
1423}
1424
1425
David S. Millerbdcc0922012-03-07 20:53:36 -05001426static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001427{
1428 return skb->data_len;
1429}
1430
1431static inline unsigned int skb_headlen(const struct sk_buff *skb)
1432{
1433 return skb->len - skb->data_len;
1434}
1435
1436static inline int skb_pagelen(const struct sk_buff *skb)
1437{
1438 int i, len = 0;
1439
1440 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001441 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001442 return len + skb_headlen(skb);
1443}
1444
Ian Campbell131ea662011-08-19 06:25:00 +00001445/**
1446 * __skb_fill_page_desc - initialise a paged fragment in an skb
1447 * @skb: buffer containing fragment to be initialised
1448 * @i: paged fragment index to initialise
1449 * @page: the page to use for this fragment
1450 * @off: the offset to the data with @page
1451 * @size: the length of the data
1452 *
1453 * Initialises the @i'th fragment of @skb to point to &size bytes at
1454 * offset @off within @page.
1455 *
1456 * Does not take any additional reference on the fragment.
1457 */
1458static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1459 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001460{
1461 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1462
Mel Gormanc48a11c2012-07-31 16:44:23 -07001463 /*
1464 * Propagate page->pfmemalloc to the skb if we can. The problem is
1465 * that not all callers have unique ownership of the page. If
1466 * pfmemalloc is set, we check the mapping as a mapping implies
1467 * page->index is set (index and pfmemalloc share space).
1468 * If it's a valid mapping, we cannot use page->pfmemalloc but we
1469 * do not lose pfmemalloc information as the pages would not be
1470 * allocated using __GFP_MEMALLOC.
1471 */
Ian Campbella8605c62011-10-19 23:01:49 +00001472 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001473 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001474 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001475
1476 page = compound_head(page);
1477 if (page->pfmemalloc && !page->mapping)
1478 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00001479}
1480
1481/**
1482 * skb_fill_page_desc - initialise a paged fragment in an skb
1483 * @skb: buffer containing fragment to be initialised
1484 * @i: paged fragment index to initialise
1485 * @page: the page to use for this fragment
1486 * @off: the offset to the data with @page
1487 * @size: the length of the data
1488 *
1489 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01001490 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00001491 * addition updates @skb such that @i is the last fragment.
1492 *
1493 * Does not take any additional reference on the fragment.
1494 */
1495static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1496 struct page *page, int off, int size)
1497{
1498 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001499 skb_shinfo(skb)->nr_frags = i + 1;
1500}
1501
Joe Perches7965bd42013-09-26 14:48:15 -07001502void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1503 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001504
Jason Wangf8e617e2013-11-01 14:07:47 +08001505void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1506 unsigned int truesize);
1507
Linus Torvalds1da177e2005-04-16 15:20:36 -07001508#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001509#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001510#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1511
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001512#ifdef NET_SKBUFF_DATA_USES_OFFSET
1513static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1514{
1515 return skb->head + skb->tail;
1516}
1517
1518static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1519{
1520 skb->tail = skb->data - skb->head;
1521}
1522
1523static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1524{
1525 skb_reset_tail_pointer(skb);
1526 skb->tail += offset;
1527}
Simon Horman7cc46192013-05-28 20:34:29 +00001528
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001529#else /* NET_SKBUFF_DATA_USES_OFFSET */
1530static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1531{
1532 return skb->tail;
1533}
1534
1535static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1536{
1537 skb->tail = skb->data;
1538}
1539
1540static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1541{
1542 skb->tail = skb->data + offset;
1543}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001544
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001545#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1546
Linus Torvalds1da177e2005-04-16 15:20:36 -07001547/*
1548 * Add data to an sk_buff
1549 */
Mathias Krause0c7ddf32013-11-07 14:18:24 +01001550unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001551unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001552static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1553{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001554 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001555 SKB_LINEAR_ASSERT(skb);
1556 skb->tail += len;
1557 skb->len += len;
1558 return tmp;
1559}
1560
Joe Perches7965bd42013-09-26 14:48:15 -07001561unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001562static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1563{
1564 skb->data -= len;
1565 skb->len += len;
1566 return skb->data;
1567}
1568
Joe Perches7965bd42013-09-26 14:48:15 -07001569unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001570static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1571{
1572 skb->len -= len;
1573 BUG_ON(skb->len < skb->data_len);
1574 return skb->data += len;
1575}
1576
David S. Miller47d29642010-05-02 02:21:44 -07001577static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1578{
1579 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1580}
1581
Joe Perches7965bd42013-09-26 14:48:15 -07001582unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583
1584static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1585{
1586 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001587 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001588 return NULL;
1589 skb->len -= len;
1590 return skb->data += len;
1591}
1592
1593static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1594{
1595 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1596}
1597
1598static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1599{
1600 if (likely(len <= skb_headlen(skb)))
1601 return 1;
1602 if (unlikely(len > skb->len))
1603 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001604 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001605}
1606
1607/**
1608 * skb_headroom - bytes at buffer head
1609 * @skb: buffer to check
1610 *
1611 * Return the number of bytes of free space at the head of an &sk_buff.
1612 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001613static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001614{
1615 return skb->data - skb->head;
1616}
1617
1618/**
1619 * skb_tailroom - bytes at buffer end
1620 * @skb: buffer to check
1621 *
1622 * Return the number of bytes of free space at the tail of an sk_buff
1623 */
1624static inline int skb_tailroom(const struct sk_buff *skb)
1625{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001626 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001627}
1628
1629/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00001630 * skb_availroom - bytes at buffer end
1631 * @skb: buffer to check
1632 *
1633 * Return the number of bytes of free space at the tail of an sk_buff
1634 * allocated by sk_stream_alloc()
1635 */
1636static inline int skb_availroom(const struct sk_buff *skb)
1637{
Eric Dumazet16fad692013-03-14 05:40:32 +00001638 if (skb_is_nonlinear(skb))
1639 return 0;
1640
1641 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00001642}
1643
1644/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001645 * skb_reserve - adjust headroom
1646 * @skb: buffer to alter
1647 * @len: bytes to move
1648 *
1649 * Increase the headroom of an empty &sk_buff by reducing the tail
1650 * room. This is only allowed for an empty buffer.
1651 */
David S. Miller8243126c2006-01-17 02:54:21 -08001652static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653{
1654 skb->data += len;
1655 skb->tail += len;
1656}
1657
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001658static inline void skb_reset_inner_headers(struct sk_buff *skb)
1659{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001660 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001661 skb->inner_network_header = skb->network_header;
1662 skb->inner_transport_header = skb->transport_header;
1663}
1664
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00001665static inline void skb_reset_mac_len(struct sk_buff *skb)
1666{
1667 skb->mac_len = skb->network_header - skb->mac_header;
1668}
1669
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001670static inline unsigned char *skb_inner_transport_header(const struct sk_buff
1671 *skb)
1672{
1673 return skb->head + skb->inner_transport_header;
1674}
1675
1676static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
1677{
1678 skb->inner_transport_header = skb->data - skb->head;
1679}
1680
1681static inline void skb_set_inner_transport_header(struct sk_buff *skb,
1682 const int offset)
1683{
1684 skb_reset_inner_transport_header(skb);
1685 skb->inner_transport_header += offset;
1686}
1687
1688static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
1689{
1690 return skb->head + skb->inner_network_header;
1691}
1692
1693static inline void skb_reset_inner_network_header(struct sk_buff *skb)
1694{
1695 skb->inner_network_header = skb->data - skb->head;
1696}
1697
1698static inline void skb_set_inner_network_header(struct sk_buff *skb,
1699 const int offset)
1700{
1701 skb_reset_inner_network_header(skb);
1702 skb->inner_network_header += offset;
1703}
1704
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00001705static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
1706{
1707 return skb->head + skb->inner_mac_header;
1708}
1709
1710static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
1711{
1712 skb->inner_mac_header = skb->data - skb->head;
1713}
1714
1715static inline void skb_set_inner_mac_header(struct sk_buff *skb,
1716 const int offset)
1717{
1718 skb_reset_inner_mac_header(skb);
1719 skb->inner_mac_header += offset;
1720}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001721static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
1722{
Cong Wang35d04612013-05-29 15:16:05 +08001723 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00001724}
1725
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001726static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
1727{
1728 return skb->head + skb->transport_header;
1729}
1730
1731static inline void skb_reset_transport_header(struct sk_buff *skb)
1732{
1733 skb->transport_header = skb->data - skb->head;
1734}
1735
1736static inline void skb_set_transport_header(struct sk_buff *skb,
1737 const int offset)
1738{
1739 skb_reset_transport_header(skb);
1740 skb->transport_header += offset;
1741}
1742
1743static inline unsigned char *skb_network_header(const struct sk_buff *skb)
1744{
1745 return skb->head + skb->network_header;
1746}
1747
1748static inline void skb_reset_network_header(struct sk_buff *skb)
1749{
1750 skb->network_header = skb->data - skb->head;
1751}
1752
1753static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
1754{
1755 skb_reset_network_header(skb);
1756 skb->network_header += offset;
1757}
1758
1759static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
1760{
1761 return skb->head + skb->mac_header;
1762}
1763
1764static inline int skb_mac_header_was_set(const struct sk_buff *skb)
1765{
Cong Wang35d04612013-05-29 15:16:05 +08001766 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001767}
1768
1769static inline void skb_reset_mac_header(struct sk_buff *skb)
1770{
1771 skb->mac_header = skb->data - skb->head;
1772}
1773
1774static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
1775{
1776 skb_reset_mac_header(skb);
1777 skb->mac_header += offset;
1778}
1779
Timo Teräs0e3da5b2013-12-16 11:02:09 +02001780static inline void skb_pop_mac_header(struct sk_buff *skb)
1781{
1782 skb->mac_header = skb->network_header;
1783}
1784
Ying Xuefbbdb8f2013-03-27 16:46:06 +00001785static inline void skb_probe_transport_header(struct sk_buff *skb,
1786 const int offset_hint)
1787{
1788 struct flow_keys keys;
1789
1790 if (skb_transport_header_was_set(skb))
1791 return;
1792 else if (skb_flow_dissect(skb, &keys))
1793 skb_set_transport_header(skb, keys.thoff);
1794 else
1795 skb_set_transport_header(skb, offset_hint);
1796}
1797
Eric Dumazet03606892012-02-23 10:55:02 +00001798static inline void skb_mac_header_rebuild(struct sk_buff *skb)
1799{
1800 if (skb_mac_header_was_set(skb)) {
1801 const unsigned char *old_mac = skb_mac_header(skb);
1802
1803 skb_set_mac_header(skb, -skb->mac_len);
1804 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
1805 }
1806}
1807
Michał Mirosław04fb4512010-12-14 15:24:08 +00001808static inline int skb_checksum_start_offset(const struct sk_buff *skb)
1809{
1810 return skb->csum_start - skb_headroom(skb);
1811}
1812
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001813static inline int skb_transport_offset(const struct sk_buff *skb)
1814{
1815 return skb_transport_header(skb) - skb->data;
1816}
1817
1818static inline u32 skb_network_header_len(const struct sk_buff *skb)
1819{
1820 return skb->transport_header - skb->network_header;
1821}
1822
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001823static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
1824{
1825 return skb->inner_transport_header - skb->inner_network_header;
1826}
1827
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07001828static inline int skb_network_offset(const struct sk_buff *skb)
1829{
1830 return skb_network_header(skb) - skb->data;
1831}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03001832
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00001833static inline int skb_inner_network_offset(const struct sk_buff *skb)
1834{
1835 return skb_inner_network_header(skb) - skb->data;
1836}
1837
Changli Gaof9599ce2010-08-04 04:43:44 +00001838static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
1839{
1840 return pskb_may_pull(skb, skb_network_offset(skb) + len);
1841}
1842
Linus Torvalds1da177e2005-04-16 15:20:36 -07001843/*
1844 * CPUs often take a performance hit when accessing unaligned memory
1845 * locations. The actual performance hit varies, it can be small if the
1846 * hardware handles it or large if we have to take an exception and fix it
1847 * in software.
1848 *
1849 * Since an ethernet header is 14 bytes network drivers often end up with
1850 * the IP header at an unaligned offset. The IP header can be aligned by
1851 * shifting the start of the packet by 2 bytes. Drivers should do this
1852 * with:
1853 *
Tobias Klauser8660c122009-07-13 22:48:16 +00001854 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001855 *
1856 * The downside to this alignment of the IP header is that the DMA is now
1857 * unaligned. On some architectures the cost of an unaligned DMA is high
1858 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00001859 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07001860 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
1861 * to be overridden.
1862 */
1863#ifndef NET_IP_ALIGN
1864#define NET_IP_ALIGN 2
1865#endif
1866
Anton Blanchard025be812006-03-31 02:27:06 -08001867/*
1868 * The networking layer reserves some headroom in skb data (via
1869 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
1870 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08001871 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08001872 *
1873 * Unfortunately this headroom changes the DMA alignment of the resulting
1874 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
1875 * on some architectures. An architecture can override this value,
1876 * perhaps setting it to a cacheline in size (since that will maintain
1877 * cacheline alignment of the DMA). It must be a power of 2.
1878 *
David S. Millerd6301d32009-02-08 19:24:13 -08001879 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08001880 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07001881 *
1882 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
1883 * to reduce average number of cache lines per packet.
1884 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07001885 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08001886 */
1887#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07001888#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08001889#endif
1890
Joe Perches7965bd42013-09-26 14:48:15 -07001891int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001892
1893static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
1894{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00001895 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07001896 WARN_ON(1);
1897 return;
1898 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001899 skb->len = len;
1900 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001901}
1902
Joe Perches7965bd42013-09-26 14:48:15 -07001903void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001904
1905static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
1906{
Herbert Xu3cc0e872006-06-09 16:13:38 -07001907 if (skb->data_len)
1908 return ___pskb_trim(skb, len);
1909 __skb_trim(skb, len);
1910 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001911}
1912
1913static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
1914{
1915 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
1916}
1917
1918/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07001919 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
1920 * @skb: buffer to alter
1921 * @len: new length
1922 *
1923 * This is identical to pskb_trim except that the caller knows that
1924 * the skb is not cloned so we should never get an error due to out-
1925 * of-memory.
1926 */
1927static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
1928{
1929 int err = pskb_trim(skb, len);
1930 BUG_ON(err);
1931}
1932
1933/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001934 * skb_orphan - orphan a buffer
1935 * @skb: buffer to orphan
1936 *
1937 * If a buffer currently has an owner then we call the owner's
1938 * destructor function and make the @skb unowned. The buffer continues
1939 * to exist but is no longer charged to its former owner.
1940 */
1941static inline void skb_orphan(struct sk_buff *skb)
1942{
Eric Dumazetc34a7612013-07-30 16:11:15 -07001943 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07001944 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07001945 skb->destructor = NULL;
1946 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07001947 } else {
1948 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07001949 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07001950}
1951
1952/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00001953 * skb_orphan_frags - orphan the frags contained in a buffer
1954 * @skb: buffer to orphan frags from
1955 * @gfp_mask: allocation mask for replacement pages
1956 *
1957 * For each frag in the SKB which needs a destructor (i.e. has an
1958 * owner) create a copy of that frag and release the original
1959 * page by calling the destructor.
1960 */
1961static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
1962{
1963 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
1964 return 0;
1965 return skb_copy_ubufs(skb, gfp_mask);
1966}
1967
1968/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001969 * __skb_queue_purge - empty a list
1970 * @list: list to empty
1971 *
1972 * Delete all buffers on an &sk_buff list. Each buffer is removed from
1973 * the list and one reference dropped. This function does not take the
1974 * list lock and the caller must hold the relevant locks to use it.
1975 */
Joe Perches7965bd42013-09-26 14:48:15 -07001976void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001977static inline void __skb_queue_purge(struct sk_buff_head *list)
1978{
1979 struct sk_buff *skb;
1980 while ((skb = __skb_dequeue(list)) != NULL)
1981 kfree_skb(skb);
1982}
1983
Alexander Duycke5e67302013-02-08 10:17:15 +00001984#define NETDEV_FRAG_PAGE_MAX_ORDER get_order(32768)
1985#define NETDEV_FRAG_PAGE_MAX_SIZE (PAGE_SIZE << NETDEV_FRAG_PAGE_MAX_ORDER)
1986#define NETDEV_PAGECNT_MAX_BIAS NETDEV_FRAG_PAGE_MAX_SIZE
1987
Joe Perches7965bd42013-09-26 14:48:15 -07001988void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001989
Joe Perches7965bd42013-09-26 14:48:15 -07001990struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
1991 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07001992
1993/**
1994 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
1995 * @dev: network device to receive on
1996 * @length: length to allocate
1997 *
1998 * Allocate a new &sk_buff and assign it a usage count of one. The
1999 * buffer has unspecified headroom built in. Users should allocate
2000 * the headroom they think they need without accounting for the
2001 * built in space. The built in space is used for optimisations.
2002 *
2003 * %NULL is returned if there is no free memory. Although this function
2004 * allocates memory it can be called from an interrupt.
2005 */
2006static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002007 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002008{
2009 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2010}
2011
Eric Dumazet6f532612012-05-18 05:12:12 +00002012/* legacy helper around __netdev_alloc_skb() */
2013static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2014 gfp_t gfp_mask)
2015{
2016 return __netdev_alloc_skb(NULL, length, gfp_mask);
2017}
2018
2019/* legacy helper around netdev_alloc_skb() */
2020static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2021{
2022 return netdev_alloc_skb(NULL, length);
2023}
2024
2025
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002026static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2027 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002028{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002029 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002030
2031 if (NET_IP_ALIGN && skb)
2032 skb_reserve(skb, NET_IP_ALIGN);
2033 return skb;
2034}
2035
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002036static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2037 unsigned int length)
2038{
2039 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2040}
2041
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002042/**
2043 * __skb_alloc_pages - allocate pages for ps-rx on a skb and preserve pfmemalloc data
Mel Gorman06140022012-07-31 16:44:24 -07002044 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2045 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2046 * @order: size of the allocation
2047 *
2048 * Allocate a new page.
2049 *
2050 * %NULL is returned if there is no free memory.
2051*/
2052static inline struct page *__skb_alloc_pages(gfp_t gfp_mask,
2053 struct sk_buff *skb,
2054 unsigned int order)
2055{
2056 struct page *page;
2057
2058 gfp_mask |= __GFP_COLD;
2059
2060 if (!(gfp_mask & __GFP_NOMEMALLOC))
2061 gfp_mask |= __GFP_MEMALLOC;
2062
2063 page = alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2064 if (skb && page && page->pfmemalloc)
2065 skb->pfmemalloc = true;
2066
2067 return page;
2068}
2069
2070/**
2071 * __skb_alloc_page - allocate a page for ps-rx for a given skb and preserve pfmemalloc data
2072 * @gfp_mask: alloc_pages_node mask. Set __GFP_NOMEMALLOC if not for network packet RX
2073 * @skb: skb to set pfmemalloc on if __GFP_MEMALLOC is used
2074 *
2075 * Allocate a new page.
2076 *
2077 * %NULL is returned if there is no free memory.
2078 */
2079static inline struct page *__skb_alloc_page(gfp_t gfp_mask,
2080 struct sk_buff *skb)
2081{
2082 return __skb_alloc_pages(gfp_mask, skb, 0);
2083}
2084
2085/**
2086 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2087 * @page: The page that was allocated from skb_alloc_page
2088 * @skb: The skb that may need pfmemalloc set
2089 */
2090static inline void skb_propagate_pfmemalloc(struct page *page,
2091 struct sk_buff *skb)
2092{
2093 if (page && page->pfmemalloc)
2094 skb->pfmemalloc = true;
2095}
2096
Eric Dumazet564824b2010-10-11 19:05:25 +00002097/**
Ian Campbell131ea662011-08-19 06:25:00 +00002098 * skb_frag_page - retrieve the page refered to by a paged fragment
2099 * @frag: the paged fragment
2100 *
2101 * Returns the &struct page associated with @frag.
2102 */
2103static inline struct page *skb_frag_page(const skb_frag_t *frag)
2104{
Ian Campbella8605c62011-10-19 23:01:49 +00002105 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00002106}
2107
2108/**
2109 * __skb_frag_ref - take an addition reference on a paged fragment.
2110 * @frag: the paged fragment
2111 *
2112 * Takes an additional reference on the paged fragment @frag.
2113 */
2114static inline void __skb_frag_ref(skb_frag_t *frag)
2115{
2116 get_page(skb_frag_page(frag));
2117}
2118
2119/**
2120 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2121 * @skb: the buffer
2122 * @f: the fragment offset.
2123 *
2124 * Takes an additional reference on the @f'th paged fragment of @skb.
2125 */
2126static inline void skb_frag_ref(struct sk_buff *skb, int f)
2127{
2128 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2129}
2130
2131/**
2132 * __skb_frag_unref - release a reference on a paged fragment.
2133 * @frag: the paged fragment
2134 *
2135 * Releases a reference on the paged fragment @frag.
2136 */
2137static inline void __skb_frag_unref(skb_frag_t *frag)
2138{
2139 put_page(skb_frag_page(frag));
2140}
2141
2142/**
2143 * skb_frag_unref - release a reference on a paged fragment of an skb.
2144 * @skb: the buffer
2145 * @f: the fragment offset
2146 *
2147 * Releases a reference on the @f'th paged fragment of @skb.
2148 */
2149static inline void skb_frag_unref(struct sk_buff *skb, int f)
2150{
2151 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2152}
2153
2154/**
2155 * skb_frag_address - gets the address of the data contained in a paged fragment
2156 * @frag: the paged fragment buffer
2157 *
2158 * Returns the address of the data within @frag. The page must already
2159 * be mapped.
2160 */
2161static inline void *skb_frag_address(const skb_frag_t *frag)
2162{
2163 return page_address(skb_frag_page(frag)) + frag->page_offset;
2164}
2165
2166/**
2167 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2168 * @frag: the paged fragment buffer
2169 *
2170 * Returns the address of the data within @frag. Checks that the page
2171 * is mapped and returns %NULL otherwise.
2172 */
2173static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2174{
2175 void *ptr = page_address(skb_frag_page(frag));
2176 if (unlikely(!ptr))
2177 return NULL;
2178
2179 return ptr + frag->page_offset;
2180}
2181
2182/**
2183 * __skb_frag_set_page - sets the page contained in a paged fragment
2184 * @frag: the paged fragment
2185 * @page: the page to set
2186 *
2187 * Sets the fragment @frag to contain @page.
2188 */
2189static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2190{
Ian Campbella8605c62011-10-19 23:01:49 +00002191 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002192}
2193
2194/**
2195 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2196 * @skb: the buffer
2197 * @f: the fragment offset
2198 * @page: the page to set
2199 *
2200 * Sets the @f'th fragment of @skb to contain @page.
2201 */
2202static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2203 struct page *page)
2204{
2205 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2206}
2207
Eric Dumazet400dfd32013-10-17 16:27:07 -07002208bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2209
Ian Campbell131ea662011-08-19 06:25:00 +00002210/**
2211 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002212 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002213 * @frag: the paged fragment to map
2214 * @offset: the offset within the fragment (starting at the
2215 * fragment's own offset)
2216 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002217 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00002218 *
2219 * Maps the page associated with @frag to @device.
2220 */
2221static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2222 const skb_frag_t *frag,
2223 size_t offset, size_t size,
2224 enum dma_data_direction dir)
2225{
2226 return dma_map_page(dev, skb_frag_page(frag),
2227 frag->page_offset + offset, size, dir);
2228}
2229
Eric Dumazet117632e2011-12-03 21:39:53 +00002230static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2231 gfp_t gfp_mask)
2232{
2233 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2234}
2235
Ian Campbell131ea662011-08-19 06:25:00 +00002236/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002237 * skb_clone_writable - is the header of a clone writable
2238 * @skb: buffer to check
2239 * @len: length up to which to write
2240 *
2241 * Returns true if modifying the header part of the cloned buffer
2242 * does not requires the data to be copied.
2243 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002244static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002245{
2246 return !skb_header_cloned(skb) &&
2247 skb_headroom(skb) + len <= skb->hdr_len;
2248}
2249
Herbert Xud9cc2042007-09-16 16:21:16 -07002250static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2251 int cloned)
2252{
2253 int delta = 0;
2254
Herbert Xud9cc2042007-09-16 16:21:16 -07002255 if (headroom > skb_headroom(skb))
2256 delta = headroom - skb_headroom(skb);
2257
2258 if (delta || cloned)
2259 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2260 GFP_ATOMIC);
2261 return 0;
2262}
2263
Patrick McHardy334a8132007-06-25 04:35:20 -07002264/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002265 * skb_cow - copy header of skb when it is required
2266 * @skb: buffer to cow
2267 * @headroom: needed headroom
2268 *
2269 * If the skb passed lacks sufficient headroom or its data part
2270 * is shared, data is reallocated. If reallocation fails, an error
2271 * is returned and original skb is not changed.
2272 *
2273 * The result is skb with writable area skb->head...skb->tail
2274 * and at least @headroom of space at head.
2275 */
2276static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2277{
Herbert Xud9cc2042007-09-16 16:21:16 -07002278 return __skb_cow(skb, headroom, skb_cloned(skb));
2279}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002280
Herbert Xud9cc2042007-09-16 16:21:16 -07002281/**
2282 * skb_cow_head - skb_cow but only making the head writable
2283 * @skb: buffer to cow
2284 * @headroom: needed headroom
2285 *
2286 * This function is identical to skb_cow except that we replace the
2287 * skb_cloned check by skb_header_cloned. It should be used when
2288 * you only need to push on some header and do not need to modify
2289 * the data.
2290 */
2291static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2292{
2293 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002294}
2295
2296/**
2297 * skb_padto - pad an skbuff up to a minimal size
2298 * @skb: buffer to pad
2299 * @len: minimal length
2300 *
2301 * Pads up a buffer to ensure the trailing bytes exist and are
2302 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002303 * is untouched. Otherwise it is extended. Returns zero on
2304 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002305 */
2306
Herbert Xu5b057c62006-06-23 02:06:41 -07002307static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002308{
2309 unsigned int size = skb->len;
2310 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002311 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002312 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002313}
2314
2315static inline int skb_add_data(struct sk_buff *skb,
2316 char __user *from, int copy)
2317{
2318 const int off = skb->len;
2319
2320 if (skb->ip_summed == CHECKSUM_NONE) {
2321 int err = 0;
Al Viro50842052006-11-14 21:36:34 -08002322 __wsum csum = csum_and_copy_from_user(from, skb_put(skb, copy),
Linus Torvalds1da177e2005-04-16 15:20:36 -07002323 copy, 0, &err);
2324 if (!err) {
2325 skb->csum = csum_block_add(skb->csum, csum, off);
2326 return 0;
2327 }
2328 } else if (!copy_from_user(skb_put(skb, copy), from, copy))
2329 return 0;
2330
2331 __skb_trim(skb, off);
2332 return -EFAULT;
2333}
2334
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002335static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2336 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002337{
2338 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002339 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002340
Ian Campbellea2ab692011-08-22 23:44:58 +00002341 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002342 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002343 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002344 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002345}
2346
Herbert Xu364c6ba2006-06-09 16:10:40 -07002347static inline int __skb_linearize(struct sk_buff *skb)
2348{
2349 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2350}
2351
Linus Torvalds1da177e2005-04-16 15:20:36 -07002352/**
2353 * skb_linearize - convert paged skb to linear one
2354 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002355 *
2356 * If there is no free memory -ENOMEM is returned, otherwise zero
2357 * is returned and the old skb data released.
2358 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002359static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002360{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002361 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2362}
2363
2364/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00002365 * skb_has_shared_frag - can any frag be overwritten
2366 * @skb: buffer to test
2367 *
2368 * Return true if the skb has at least one frag that might be modified
2369 * by an external entity (as in vmsplice()/sendfile())
2370 */
2371static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2372{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00002373 return skb_is_nonlinear(skb) &&
2374 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00002375}
2376
2377/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07002378 * skb_linearize_cow - make sure skb is linear and writable
2379 * @skb: buffer to process
2380 *
2381 * If there is no free memory -ENOMEM is returned, otherwise zero
2382 * is returned and the old skb data released.
2383 */
2384static inline int skb_linearize_cow(struct sk_buff *skb)
2385{
2386 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2387 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002388}
2389
2390/**
2391 * skb_postpull_rcsum - update checksum for received skb after pull
2392 * @skb: buffer to update
2393 * @start: start of data before pull
2394 * @len: length of data pulled
2395 *
2396 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002397 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2398 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002399 */
2400
2401static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002402 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002403{
Patrick McHardy84fa7932006-08-29 16:44:56 -07002404 if (skb->ip_summed == CHECKSUM_COMPLETE)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002405 skb->csum = csum_sub(skb->csum, csum_partial(start, len, 0));
2406}
2407
Herbert Xucbb042f2006-03-20 22:43:56 -08002408unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2409
David S. Miller7ce5a272013-12-02 17:26:05 -05002410/**
2411 * pskb_trim_rcsum - trim received skb and update checksum
2412 * @skb: buffer to trim
2413 * @len: new length
2414 *
2415 * This is exactly the same as pskb_trim except that it ensures the
2416 * checksum of received packets are still valid after the operation.
2417 */
2418
2419static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2420{
2421 if (likely(len >= skb->len))
2422 return 0;
2423 if (skb->ip_summed == CHECKSUM_COMPLETE)
2424 skb->ip_summed = CHECKSUM_NONE;
2425 return __pskb_trim(skb, len);
2426}
2427
Linus Torvalds1da177e2005-04-16 15:20:36 -07002428#define skb_queue_walk(queue, skb) \
2429 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002430 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002431 skb = skb->next)
2432
James Chapman46f89142007-04-30 00:07:31 -07002433#define skb_queue_walk_safe(queue, skb, tmp) \
2434 for (skb = (queue)->next, tmp = skb->next; \
2435 skb != (struct sk_buff *)(queue); \
2436 skb = tmp, tmp = skb->next)
2437
David S. Miller1164f522008-09-23 00:49:44 -07002438#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002439 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002440 skb = skb->next)
2441
2442#define skb_queue_walk_from_safe(queue, skb, tmp) \
2443 for (tmp = skb->next; \
2444 skb != (struct sk_buff *)(queue); \
2445 skb = tmp, tmp = skb->next)
2446
Stephen Hemminger300ce172005-10-30 13:47:34 -08002447#define skb_queue_reverse_walk(queue, skb) \
2448 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002449 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08002450 skb = skb->prev)
2451
David S. Miller686a2952011-01-20 22:47:32 -08002452#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2453 for (skb = (queue)->prev, tmp = skb->prev; \
2454 skb != (struct sk_buff *)(queue); \
2455 skb = tmp, tmp = skb->prev)
2456
2457#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2458 for (tmp = skb->prev; \
2459 skb != (struct sk_buff *)(queue); \
2460 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002461
David S. Miller21dc3302010-08-23 00:13:46 -07002462static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07002463{
2464 return skb_shinfo(skb)->frag_list != NULL;
2465}
2466
2467static inline void skb_frag_list_init(struct sk_buff *skb)
2468{
2469 skb_shinfo(skb)->frag_list = NULL;
2470}
2471
2472static inline void skb_frag_add_head(struct sk_buff *skb, struct sk_buff *frag)
2473{
2474 frag->next = skb_shinfo(skb)->frag_list;
2475 skb_shinfo(skb)->frag_list = frag;
2476}
2477
2478#define skb_walk_frags(skb, iter) \
2479 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2480
Joe Perches7965bd42013-09-26 14:48:15 -07002481struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
2482 int *peeked, int *off, int *err);
2483struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
2484 int *err);
2485unsigned int datagram_poll(struct file *file, struct socket *sock,
2486 struct poll_table_struct *wait);
2487int skb_copy_datagram_iovec(const struct sk_buff *from, int offset,
2488 struct iovec *to, int size);
2489int skb_copy_and_csum_datagram_iovec(struct sk_buff *skb, int hlen,
2490 struct iovec *iov);
2491int skb_copy_datagram_from_iovec(struct sk_buff *skb, int offset,
2492 const struct iovec *from, int from_offset,
2493 int len);
2494int zerocopy_sg_from_iovec(struct sk_buff *skb, const struct iovec *frm,
2495 int offset, size_t count);
2496int skb_copy_datagram_const_iovec(const struct sk_buff *from, int offset,
2497 const struct iovec *to, int to_offset,
2498 int size);
2499void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
2500void skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb);
2501int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07002502int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
2503int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
2504__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
2505 int len, __wsum csum);
2506int skb_splice_bits(struct sk_buff *skb, unsigned int offset,
2507 struct pipe_inode_info *pipe, unsigned int len,
2508 unsigned int flags);
2509void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01002510unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00002511int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
2512 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07002513void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
2514int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
2515void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Florian Westphalde960aa2014-01-26 10:58:16 +01002516unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -07002517struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03002518
Daniel Borkmann2817a332013-10-30 11:50:51 +01002519struct skb_checksum_ops {
2520 __wsum (*update)(const void *mem, int len, __wsum wsum);
2521 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
2522};
2523
2524__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
2525 __wsum csum, const struct skb_checksum_ops *ops);
2526__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
2527 __wsum csum);
2528
Linus Torvalds1da177e2005-04-16 15:20:36 -07002529static inline void *skb_header_pointer(const struct sk_buff *skb, int offset,
2530 int len, void *buffer)
2531{
2532 int hlen = skb_headlen(skb);
2533
Patrick McHardy55820ee2005-07-05 14:08:10 -07002534 if (hlen - offset >= len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002535 return skb->data + offset;
2536
2537 if (skb_copy_bits(skb, offset, buffer, len) < 0)
2538 return NULL;
2539
2540 return buffer;
2541}
2542
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01002543/**
2544 * skb_needs_linearize - check if we need to linearize a given skb
2545 * depending on the given device features.
2546 * @skb: socket buffer to check
2547 * @features: net device features
2548 *
2549 * Returns true if either:
2550 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
2551 * 2. skb is fragmented and the device does not support SG.
2552 */
2553static inline bool skb_needs_linearize(struct sk_buff *skb,
2554 netdev_features_t features)
2555{
2556 return skb_is_nonlinear(skb) &&
2557 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
2558 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
2559}
2560
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03002561static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
2562 void *to,
2563 const unsigned int len)
2564{
2565 memcpy(to, skb->data, len);
2566}
2567
2568static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
2569 const int offset, void *to,
2570 const unsigned int len)
2571{
2572 memcpy(to, skb->data + offset, len);
2573}
2574
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03002575static inline void skb_copy_to_linear_data(struct sk_buff *skb,
2576 const void *from,
2577 const unsigned int len)
2578{
2579 memcpy(skb->data, from, len);
2580}
2581
2582static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
2583 const int offset,
2584 const void *from,
2585 const unsigned int len)
2586{
2587 memcpy(skb->data + offset, from, len);
2588}
2589
Joe Perches7965bd42013-09-26 14:48:15 -07002590void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002591
Patrick Ohlyac45f602009-02-12 05:03:37 +00002592static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
2593{
2594 return skb->tstamp;
2595}
2596
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002597/**
2598 * skb_get_timestamp - get timestamp from a skb
2599 * @skb: skb to get stamp from
2600 * @stamp: pointer to struct timeval to store stamp in
2601 *
2602 * Timestamps are stored in the skb as offsets to a base timestamp.
2603 * This function converts the offset back to a struct timeval and stores
2604 * it in stamp.
2605 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00002606static inline void skb_get_timestamp(const struct sk_buff *skb,
2607 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002608{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002609 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002610}
2611
Patrick Ohlyac45f602009-02-12 05:03:37 +00002612static inline void skb_get_timestampns(const struct sk_buff *skb,
2613 struct timespec *stamp)
2614{
2615 *stamp = ktime_to_timespec(skb->tstamp);
2616}
2617
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002618static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002619{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07002620 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002621}
2622
Stephen Hemminger164891a2007-04-23 22:26:16 -07002623static inline ktime_t net_timedelta(ktime_t t)
2624{
2625 return ktime_sub(ktime_get_real(), t);
2626}
2627
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07002628static inline ktime_t net_invalid_timestamp(void)
2629{
2630 return ktime_set(0, 0);
2631}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07002632
Richard Cochranc1f19b52010-07-17 08:49:36 +00002633#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
2634
Joe Perches7965bd42013-09-26 14:48:15 -07002635void skb_clone_tx_timestamp(struct sk_buff *skb);
2636bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00002637
2638#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
2639
2640static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
2641{
2642}
2643
2644static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
2645{
2646 return false;
2647}
2648
2649#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
2650
2651/**
2652 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
2653 *
Richard Cochranda92b192011-10-21 00:49:15 +00002654 * PHY drivers may accept clones of transmitted packets for
2655 * timestamping via their phy_driver.txtstamp method. These drivers
2656 * must call this function to return the skb back to the stack, with
2657 * or without a timestamp.
2658 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00002659 * @skb: clone of the the original outgoing packet
Richard Cochranda92b192011-10-21 00:49:15 +00002660 * @hwtstamps: hardware time stamps, may be NULL if not available
Richard Cochranc1f19b52010-07-17 08:49:36 +00002661 *
2662 */
2663void skb_complete_tx_timestamp(struct sk_buff *skb,
2664 struct skb_shared_hwtstamps *hwtstamps);
2665
Patrick Ohlyac45f602009-02-12 05:03:37 +00002666/**
2667 * skb_tstamp_tx - queue clone of skb with send time stamps
2668 * @orig_skb: the original outgoing packet
2669 * @hwtstamps: hardware time stamps, may be NULL if not available
2670 *
2671 * If the skb has a socket associated, then this function clones the
2672 * skb (thus sharing the actual data and optional structures), stores
2673 * the optional hardware time stamping information (if non NULL) or
2674 * generates a software time stamp (otherwise), then queues the clone
2675 * to the error queue of the socket. Errors are silently ignored.
2676 */
Joe Perches7965bd42013-09-26 14:48:15 -07002677void skb_tstamp_tx(struct sk_buff *orig_skb,
2678 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00002679
Richard Cochran4507a712010-07-17 08:48:28 +00002680static inline void sw_tx_timestamp(struct sk_buff *skb)
2681{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00002682 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
2683 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00002684 skb_tstamp_tx(skb, NULL);
2685}
2686
2687/**
2688 * skb_tx_timestamp() - Driver hook for transmit timestamping
2689 *
2690 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00002691 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00002692 *
David S. Miller73409f32013-12-27 13:04:33 -05002693 * Specifically, one should make absolutely sure that this function is
2694 * called before TX completion of this packet can trigger. Otherwise
2695 * the packet could potentially already be freed.
2696 *
Richard Cochran4507a712010-07-17 08:48:28 +00002697 * @skb: A socket buffer.
2698 */
2699static inline void skb_tx_timestamp(struct sk_buff *skb)
2700{
Richard Cochranc1f19b52010-07-17 08:49:36 +00002701 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00002702 sw_tx_timestamp(skb);
2703}
2704
Johannes Berg6e3e9392011-11-09 10:15:42 +01002705/**
2706 * skb_complete_wifi_ack - deliver skb with wifi status
2707 *
2708 * @skb: the original outgoing packet
2709 * @acked: ack status
2710 *
2711 */
2712void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
2713
Joe Perches7965bd42013-09-26 14:48:15 -07002714__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
2715__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002716
Herbert Xu60476372007-04-09 11:59:39 -07002717static inline int skb_csum_unnecessary(const struct sk_buff *skb)
2718{
2719 return skb->ip_summed & CHECKSUM_UNNECESSARY;
2720}
2721
Herbert Xufb286bb2005-11-10 13:01:24 -08002722/**
2723 * skb_checksum_complete - Calculate checksum of an entire packet
2724 * @skb: packet to process
2725 *
2726 * This function calculates the checksum over the entire packet plus
2727 * the value of skb->csum. The latter can be used to supply the
2728 * checksum of a pseudo header as used by TCP/UDP. It returns the
2729 * checksum.
2730 *
2731 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
2732 * this function can be used to verify that checksum on received
2733 * packets. In that case the function should return zero if the
2734 * checksum is correct. In particular, this function will return zero
2735 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
2736 * hardware has already verified the correctness of the checksum.
2737 */
Al Viro4381ca32007-07-15 21:00:11 +01002738static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08002739{
Herbert Xu60476372007-04-09 11:59:39 -07002740 return skb_csum_unnecessary(skb) ?
2741 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08002742}
2743
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002744#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07002745void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002746static inline void nf_conntrack_put(struct nf_conntrack *nfct)
2747{
2748 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07002749 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002750}
2751static inline void nf_conntrack_get(struct nf_conntrack *nfct)
2752{
2753 if (nfct)
2754 atomic_inc(&nfct->use);
2755}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002756#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002757#ifdef CONFIG_BRIDGE_NETFILTER
2758static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
2759{
2760 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
2761 kfree(nf_bridge);
2762}
2763static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
2764{
2765 if (nf_bridge)
2766 atomic_inc(&nf_bridge->use);
2767}
2768#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08002769static inline void nf_reset(struct sk_buff *skb)
2770{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002771#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08002772 nf_conntrack_put(skb->nfct);
2773 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002774#endif
Patrick McHardya193a4a2006-03-20 19:23:05 -08002775#ifdef CONFIG_BRIDGE_NETFILTER
2776 nf_bridge_put(skb->nf_bridge);
2777 skb->nf_bridge = NULL;
2778#endif
2779}
2780
Patrick McHardy124dff02013-04-05 20:42:05 +02002781static inline void nf_reset_trace(struct sk_buff *skb)
2782{
Florian Westphal478b3602014-02-15 23:48:45 +01002783#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00002784 skb->nf_trace = 0;
2785#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07002786}
2787
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002788/* Note: This doesn't put any conntrack and bridge info in dst. */
2789static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2790{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002791#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002792 dst->nfct = src->nfct;
2793 nf_conntrack_get(src->nfct);
2794 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002795#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002796#ifdef CONFIG_BRIDGE_NETFILTER
2797 dst->nf_bridge = src->nf_bridge;
2798 nf_bridge_get(src->nf_bridge);
2799#endif
Florian Westphal478b3602014-02-15 23:48:45 +01002800#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
2801 dst->nf_trace = src->nf_trace;
2802#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07002803}
2804
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002805static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
2806{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002807#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07002808 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01002809#endif
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07002810#ifdef CONFIG_BRIDGE_NETFILTER
2811 nf_bridge_put(dst->nf_bridge);
2812#endif
2813 __nf_copy(dst, src);
2814}
2815
James Morris984bc162006-06-09 00:29:17 -07002816#ifdef CONFIG_NETWORK_SECMARK
2817static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2818{
2819 to->secmark = from->secmark;
2820}
2821
2822static inline void skb_init_secmark(struct sk_buff *skb)
2823{
2824 skb->secmark = 0;
2825}
2826#else
2827static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
2828{ }
2829
2830static inline void skb_init_secmark(struct sk_buff *skb)
2831{ }
2832#endif
2833
Eric W. Biederman574f7192014-04-01 12:20:24 -07002834static inline bool skb_irq_freeable(const struct sk_buff *skb)
2835{
2836 return !skb->destructor &&
2837#if IS_ENABLED(CONFIG_XFRM)
2838 !skb->sp &&
2839#endif
2840#if IS_ENABLED(CONFIG_NF_CONNTRACK)
2841 !skb->nfct &&
2842#endif
2843 !skb->_skb_refdst &&
2844 !skb_has_frag_list(skb);
2845}
2846
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002847static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
2848{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002849 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002850}
2851
Stephen Hemminger92477442009-03-21 13:39:26 -07002852static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002853{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002854 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07002855}
2856
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002857static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
2858{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002859 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07002860}
2861
David S. Millerd5a9e242009-01-27 16:22:11 -08002862static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
2863{
2864 skb->queue_mapping = rx_queue + 1;
2865}
2866
Stephen Hemminger92477442009-03-21 13:39:26 -07002867static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002868{
2869 return skb->queue_mapping - 1;
2870}
2871
Stephen Hemminger92477442009-03-21 13:39:26 -07002872static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08002873{
Eric Dumazeta02cec22010-09-22 20:43:57 +00002874 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08002875}
2876
Joe Perches7965bd42013-09-26 14:48:15 -07002877u16 __skb_tx_hash(const struct net_device *dev, const struct sk_buff *skb,
2878 unsigned int num_tx_queues);
Stephen Hemminger92477442009-03-21 13:39:26 -07002879
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04002880static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
2881{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002882#ifdef CONFIG_XFRM
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002883 return skb->sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002884#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002885 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002886#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04002887}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07002888
Pravin B Shelar68c33162013-02-14 14:02:41 +00002889/* Keeps track of mac header offset relative to skb->head.
2890 * It is useful for TSO of Tunneling protocol. e.g. GRE.
2891 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07002892 * tunnel skb it points to outer mac header.
2893 * Keeps track of level of encapsulation of network headers.
2894 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00002895struct skb_gso_cb {
Eric Dumazet3347c962013-10-19 11:42:56 -07002896 int mac_offset;
2897 int encap_level;
Pravin B Shelar68c33162013-02-14 14:02:41 +00002898};
2899#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)(skb)->cb)
2900
2901static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
2902{
2903 return (skb_mac_header(inner_skb) - inner_skb->head) -
2904 SKB_GSO_CB(inner_skb)->mac_offset;
2905}
2906
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00002907static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
2908{
2909 int new_headroom, headroom;
2910 int ret;
2911
2912 headroom = skb_headroom(skb);
2913 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
2914 if (ret)
2915 return ret;
2916
2917 new_headroom = skb_headroom(skb);
2918 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
2919 return 0;
2920}
2921
David S. Millerbdcc0922012-03-07 20:53:36 -05002922static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07002923{
2924 return skb_shinfo(skb)->gso_size;
2925}
2926
Eric Dumazet36a8f392013-09-29 01:21:32 -07002927/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05002928static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02002929{
2930 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
2931}
2932
Joe Perches7965bd42013-09-26 14:48:15 -07002933void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07002934
2935static inline bool skb_warn_if_lro(const struct sk_buff *skb)
2936{
2937 /* LRO sets gso_size but not gso_type, whereas if GSO is really
2938 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002939 const struct skb_shared_info *shinfo = skb_shinfo(skb);
2940
Alexander Duyckb78462e2010-06-02 12:24:37 +00002941 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
2942 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07002943 __skb_warn_lro_forwarding(skb);
2944 return true;
2945 }
2946 return false;
2947}
2948
Herbert Xu35fc92a2007-03-26 23:22:20 -07002949static inline void skb_forward_csum(struct sk_buff *skb)
2950{
2951 /* Unfortunately we don't support this one. Any brave souls? */
2952 if (skb->ip_summed == CHECKSUM_COMPLETE)
2953 skb->ip_summed = CHECKSUM_NONE;
2954}
2955
Eric Dumazetbc8acf22010-09-02 13:07:41 -07002956/**
2957 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
2958 * @skb: skb to check
2959 *
2960 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
2961 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
2962 * use this helper, to document places where we make this assertion.
2963 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002964static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07002965{
2966#ifdef DEBUG
2967 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
2968#endif
2969}
2970
Rusty Russellf35d9d82008-02-04 23:49:54 -05002971bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00002972
Paul Durranted1f50c2014-01-09 10:02:46 +00002973int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
2974
Daniel Borkmannf77668d2013-03-19 06:39:30 +00002975u32 __skb_get_poff(const struct sk_buff *skb);
2976
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00002977/**
2978 * skb_head_is_locked - Determine if the skb->head is locked down
2979 * @skb: skb to check
2980 *
2981 * The head on skbs build around a head frag can be removed if they are
2982 * not cloned. This function returns true if the skb head is locked down
2983 * due to either being allocated via kmalloc, or by being a clone with
2984 * multiple references to the head.
2985 */
2986static inline bool skb_head_is_locked(const struct sk_buff *skb)
2987{
2988 return !skb->head_frag || skb_cloned(skb);
2989}
Florian Westphalfe6cc552014-02-13 23:09:12 +01002990
2991/**
2992 * skb_gso_network_seglen - Return length of individual segments of a gso packet
2993 *
2994 * @skb: GSO skb
2995 *
2996 * skb_gso_network_seglen is used to determine the real size of the
2997 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
2998 *
2999 * The MAC/L2 header is not accounted for.
3000 */
3001static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3002{
3003 unsigned int hdr_len = skb_transport_header(skb) -
3004 skb_network_header(skb);
3005 return hdr_len + skb_gso_transport_seglen(skb);
3006}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003007#endif /* __KERNEL__ */
3008#endif /* _LINUX_SKBUFF_H */