blob: d460a4cbda1c8b65e5e0b2a643e29c122055df1f [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>
Eric Dumazet56b17422014-11-03 08:19:53 -080023#include <linux/rbtree.h>
David S. Miller51f3d022014-11-05 16:46:40 -050024#include <linux/socket.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025
Arun Sharma600634972011-07-26 16:09:06 -070026#include <linux/atomic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <asm/types.h>
28#include <linux/spinlock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070029#include <linux/net.h>
Thomas Graf3fc7e8a2005-06-23 21:00:17 -070030#include <linux/textsearch.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070031#include <net/checksum.h>
Al Viroa80958f2006-12-04 20:41:19 +000032#include <linux/rcupdate.h>
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -070033#include <linux/hrtimer.h>
Ian Campbell131ea662011-08-19 06:25:00 +000034#include <linux/dma-mapping.h>
Michał Mirosławc8f44af2011-11-15 15:29:55 +000035#include <linux/netdev_features.h>
Eric Dumazet363ec39232014-02-26 14:02:11 -080036#include <linux/sched.h>
Ingo Molnare6017572017-02-01 16:36:40 +010037#include <linux/sched/clock.h>
Jiri Pirko1bd758e2015-05-12 14:56:07 +020038#include <net/flow_dissector.h>
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +020039#include <linux/splice.h>
Bernhard Thaler72b31f72015-05-30 15:27:40 +020040#include <linux/in6.h>
Jamal Hadi Salim8b10cab2016-07-02 06:43:14 -040041#include <linux/if_packet.h>
Tom Herbertf70ea012015-07-31 16:52:10 -070042#include <net/flow.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070043
Tom Herbert7a6ae712015-12-14 11:19:47 -080044/* The interface for checksum offload between the stack and networking drivers
45 * is as follows...
46 *
47 * A. IP checksum related features
48 *
49 * Drivers advertise checksum offload capabilities in the features of a device.
50 * From the stack's point of view these are capabilities offered by the driver,
51 * a driver typically only advertises features that it is capable of offloading
52 * to its device.
53 *
54 * The checksum related features are:
55 *
56 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
57 * IP (one's complement) checksum for any combination
58 * of protocols or protocol layering. The checksum is
59 * computed and set in a packet per the CHECKSUM_PARTIAL
60 * interface (see below).
61 *
62 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
63 * TCP or UDP packets over IPv4. These are specifically
64 * unencapsulated packets of the form IPv4|TCP or
65 * IPv4|UDP where the Protocol field in the IPv4 header
66 * is TCP or UDP. The IPv4 header may contain IP options
67 * This feature cannot be set in features for a device
68 * with NETIF_F_HW_CSUM also set. This feature is being
69 * DEPRECATED (see below).
70 *
71 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
72 * TCP or UDP packets over IPv6. These are specifically
73 * unencapsulated packets of the form IPv6|TCP or
74 * IPv4|UDP where the Next Header field in the IPv6
75 * header is either TCP or UDP. IPv6 extension headers
76 * are not supported with this feature. This feature
77 * cannot be set in features for a device with
78 * NETIF_F_HW_CSUM also set. This feature is being
79 * DEPRECATED (see below).
80 *
81 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
82 * This flag is used only used to disable the RX checksum
83 * feature for a device. The stack will accept receive
84 * checksum indication in packets received on a device
85 * regardless of whether NETIF_F_RXCSUM is set.
86 *
87 * B. Checksumming of received packets by device. Indication of checksum
88 * verification is in set skb->ip_summed. Possible values are:
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010089 *
90 * CHECKSUM_NONE:
91 *
Tom Herbert7a6ae712015-12-14 11:19:47 -080092 * Device did not checksum this packet e.g. due to lack of capabilities.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010093 * The packet contains full (though not verified) checksum in packet but
94 * not in skb->csum. Thus, skb->csum is undefined in this case.
95 *
96 * CHECKSUM_UNNECESSARY:
97 *
98 * The hardware you're dealing with doesn't calculate the full checksum
99 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
Tom Herbert77cffe22014-08-27 21:26:46 -0700100 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
101 * if their checksums are okay. skb->csum is still undefined in this case
Tom Herbert7a6ae712015-12-14 11:19:47 -0800102 * though. A driver or device must never modify the checksum field in the
103 * packet even if checksum is verified.
Tom Herbert77cffe22014-08-27 21:26:46 -0700104 *
105 * CHECKSUM_UNNECESSARY is applicable to following protocols:
106 * TCP: IPv6 and IPv4.
107 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
108 * zero UDP checksum for either IPv4 or IPv6, the networking stack
109 * may perform further validation in this case.
110 * GRE: only if the checksum is present in the header.
111 * SCTP: indicates the CRC in SCTP header has been validated.
Davide Carattib4759dc2017-05-18 15:44:43 +0200112 * FCOE: indicates the CRC in FC frame has been validated.
Tom Herbert77cffe22014-08-27 21:26:46 -0700113 *
114 * skb->csum_level indicates the number of consecutive checksums found in
115 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
116 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
117 * and a device is able to verify the checksums for UDP (possibly zero),
118 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
119 * two. If the device were only able to verify the UDP checksum and not
120 * GRE, either because it doesn't support GRE checksum of because GRE
121 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
122 * not considered in this case).
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100123 *
124 * CHECKSUM_COMPLETE:
125 *
126 * This is the most generic way. The device supplied checksum of the _whole_
127 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
128 * hardware doesn't need to parse L3/L4 headers to implement this.
129 *
Davide Carattib4759dc2017-05-18 15:44:43 +0200130 * Notes:
131 * - Even if device supports only some protocols, but is able to produce
132 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
133 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100134 *
135 * CHECKSUM_PARTIAL:
136 *
Tom Herbert6edec0e2015-02-10 16:30:28 -0800137 * A checksum is set up to be offloaded to a device as described in the
138 * output description for CHECKSUM_PARTIAL. This may occur on a packet
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100139 * received directly from another Linux OS, e.g., a virtualized Linux kernel
Tom Herbert6edec0e2015-02-10 16:30:28 -0800140 * on the same host, or it may be set in the input path in GRO or remote
141 * checksum offload. For the purposes of checksum verification, the checksum
142 * referred to by skb->csum_start + skb->csum_offset and any preceding
143 * checksums in the packet are considered verified. Any checksums in the
144 * packet that are after the checksum being offloaded are not considered to
145 * be verified.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100146 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800147 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
148 * in the skb->ip_summed for a packet. Values are:
149 *
150 * CHECKSUM_PARTIAL:
151 *
152 * The driver is required to checksum the packet as seen by hard_start_xmit()
153 * from skb->csum_start up to the end, and to record/write the checksum at
154 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
155 * csum_start and csum_offset values are valid values given the length and
156 * offset of the packet, however they should not attempt to validate that the
157 * checksum refers to a legitimate transport layer checksum-- it is the
158 * purview of the stack to validate that csum_start and csum_offset are set
159 * correctly.
160 *
161 * When the stack requests checksum offload for a packet, the driver MUST
162 * ensure that the checksum is set correctly. A driver can either offload the
163 * checksum calculation to the device, or call skb_checksum_help (in the case
164 * that the device does not support offload for a particular checksum).
165 *
166 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
167 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
Davide Caratti43c26a12017-05-18 15:44:41 +0200168 * checksum offload capability.
169 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based
170 * on network device checksumming capabilities: if a packet does not match
171 * them, skb_checksum_help or skb_crc32c_help (depending on the value of
172 * csum_not_inet, see item D.) is called to resolve the checksum.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100173 *
174 * CHECKSUM_NONE:
175 *
176 * The skb was already checksummed by the protocol, or a checksum is not
177 * required.
178 *
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100179 * CHECKSUM_UNNECESSARY:
180 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800181 * This has the same meaning on as CHECKSUM_NONE for checksum offload on
182 * output.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100183 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800184 * CHECKSUM_COMPLETE:
185 * Not used in checksum output. If a driver observes a packet with this value
186 * set in skbuff, if should treat as CHECKSUM_NONE being set.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100187 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800188 * D. Non-IP checksum (CRC) offloads
189 *
190 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
191 * offloading the SCTP CRC in a packet. To perform this offload the stack
Davide Carattidba00302017-05-18 15:44:40 +0200192 * will set set csum_start and csum_offset accordingly, set ip_summed to
193 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
194 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
195 * A driver that supports both IP checksum offload and SCTP CRC32c offload
196 * must verify which offload is configured for a packet by testing the
197 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
198 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
Tom Herbert7a6ae712015-12-14 11:19:47 -0800199 *
200 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
201 * offloading the FCOE CRC in a packet. To perform this offload the stack
202 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
203 * accordingly. Note the there is no indication in the skbuff that the
204 * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
205 * both IP checksum offload and FCOE CRC offload must verify which offload
206 * is configured for a packet presumably by inspecting packet headers.
207 *
208 * E. Checksumming on output with GSO.
209 *
210 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
211 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
212 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
213 * part of the GSO operation is implied. If a checksum is being offloaded
214 * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
215 * are set to refer to the outermost checksum being offload (two offloaded
216 * checksums are possible with UDP encapsulation).
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100217 */
218
Herbert Xu60476372007-04-09 11:59:39 -0700219/* Don't change this without changing skb_csum_unnecessary! */
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100220#define CHECKSUM_NONE 0
221#define CHECKSUM_UNNECESSARY 1
222#define CHECKSUM_COMPLETE 2
223#define CHECKSUM_PARTIAL 3
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224
Tom Herbert77cffe22014-08-27 21:26:46 -0700225/* Maximum value in skb->csum_level */
226#define SKB_MAX_CSUM_LEVEL 3
227
Tobias Klauser0bec8c82014-07-22 12:06:23 +0200228#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
David S. Millerfc910a22007-03-25 20:27:59 -0700229#define SKB_WITH_OVERHEAD(X) \
Herbert Xudeea84b2007-10-21 16:27:46 -0700230 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
David S. Millerfc910a22007-03-25 20:27:59 -0700231#define SKB_MAX_ORDER(X, ORDER) \
232 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
234#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
235
Eric Dumazet87fb4b72011-10-13 07:28:54 +0000236/* return minimum truesize of one skb containing X bytes of data */
237#define SKB_TRUESIZE(X) ((X) + \
238 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
239 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
240
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241struct net_device;
David Howells716ea3a2007-04-02 20:19:53 -0700242struct scatterlist;
Jens Axboe9c55e012007-11-06 23:30:13 -0800243struct pipe_inode_info;
Herbert Xua8f820aa2014-11-07 21:22:22 +0800244struct iov_iter;
Alexander Duyckfd11a832014-12-09 19:40:49 -0800245struct napi_struct;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700247#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248struct nf_conntrack {
249 atomic_t use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250};
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700251#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +0200253#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254struct nf_bridge_info {
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000255 atomic_t use;
Florian Westphal3eaf4022015-04-02 14:31:44 +0200256 enum {
257 BRNF_PROTO_UNCHANGED,
258 BRNF_PROTO_8021Q,
259 BRNF_PROTO_PPPOE
Florian Westphal7fb48c52015-05-03 22:05:28 +0200260 } orig_proto:8;
Florian Westphal72b1e5e2015-07-23 16:21:30 +0200261 u8 pkt_otherhost:1;
262 u8 in_prerouting:1;
263 u8 bridged_dnat:1;
Bernhard Thaler411ffb42015-05-30 15:28:28 +0200264 __u16 frag_max_size;
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000265 struct net_device *physindev;
Florian Westphal63cdbc02015-09-14 17:06:27 +0200266
267 /* always valid & non-NULL from FORWARD on, for physdev match */
268 struct net_device *physoutdev;
Florian Westphal7fb48c52015-05-03 22:05:28 +0200269 union {
Florian Westphal72b1e5e2015-07-23 16:21:30 +0200270 /* prerouting: detect dnat in orig/reply direction */
Bernhard Thaler72b31f72015-05-30 15:27:40 +0200271 __be32 ipv4_daddr;
272 struct in6_addr ipv6_daddr;
Florian Westphal72b1e5e2015-07-23 16:21:30 +0200273
274 /* after prerouting + nat detected: store original source
275 * mac since neigh resolution overwrites it, only used while
276 * skb is out in neigh layer.
277 */
278 char neigh_header[8];
Bernhard Thaler72b31f72015-05-30 15:27:40 +0200279 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280};
281#endif
282
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283struct sk_buff_head {
284 /* These two members must be first. */
285 struct sk_buff *next;
286 struct sk_buff *prev;
287
288 __u32 qlen;
289 spinlock_t lock;
290};
291
292struct sk_buff;
293
Ian Campbell9d4dde52011-12-22 23:39:14 +0000294/* To allow 64K frame to be packed as single skb without frag_list we
295 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
296 * buffers which do not start on a page boundary.
297 *
298 * Since GRO uses frags we allocate at least 16 regardless of page
299 * size.
Anton Blancharda715dea2011-03-27 14:57:26 +0000300 */
Ian Campbell9d4dde52011-12-22 23:39:14 +0000301#if (65536/PAGE_SIZE + 1) < 16
David S. Millereec00952011-03-29 23:34:08 -0700302#define MAX_SKB_FRAGS 16UL
Anton Blancharda715dea2011-03-27 14:57:26 +0000303#else
Ian Campbell9d4dde52011-12-22 23:39:14 +0000304#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
Anton Blancharda715dea2011-03-27 14:57:26 +0000305#endif
Hans Westgaard Ry5f74f82e2016-02-03 09:26:57 +0100306extern int sysctl_max_skb_frags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307
Marcelo Ricardo Leitner3953c462016-06-02 15:05:40 -0300308/* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
309 * segment using its current segmentation instead.
310 */
311#define GSO_BY_FRAGS 0xFFFF
312
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313typedef struct skb_frag_struct skb_frag_t;
314
315struct skb_frag_struct {
Ian Campbella8605c62011-10-19 23:01:49 +0000316 struct {
317 struct page *p;
318 } page;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000319#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
David S. Millera309bb02007-07-30 18:47:03 -0700320 __u32 page_offset;
321 __u32 size;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000322#else
323 __u16 page_offset;
324 __u16 size;
325#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326};
327
Eric Dumazet9e903e02011-10-18 21:00:24 +0000328static inline unsigned int skb_frag_size(const skb_frag_t *frag)
329{
330 return frag->size;
331}
332
333static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
334{
335 frag->size = size;
336}
337
338static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
339{
340 frag->size += delta;
341}
342
343static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
344{
345 frag->size -= delta;
346}
347
Patrick Ohlyac45f602009-02-12 05:03:37 +0000348#define HAVE_HW_TIME_STAMP
349
350/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800351 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000352 * @hwtstamp: hardware time stamp transformed into duration
353 * since arbitrary point in time
Patrick Ohlyac45f602009-02-12 05:03:37 +0000354 *
355 * Software time stamps generated by ktime_get_real() are stored in
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400356 * skb->tstamp.
Patrick Ohlyac45f602009-02-12 05:03:37 +0000357 *
358 * hwtstamps can only be compared against other hwtstamps from
359 * the same device.
360 *
361 * This structure is attached to packets as part of the
362 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
363 */
364struct skb_shared_hwtstamps {
365 ktime_t hwtstamp;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000366};
367
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000368/* Definitions for tx_flags in struct skb_shared_info */
369enum {
370 /* generate hardware time stamp */
371 SKBTX_HW_TSTAMP = 1 << 0,
372
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400373 /* generate software time stamp when queueing packet to NIC */
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000374 SKBTX_SW_TSTAMP = 1 << 1,
375
376 /* device driver is going to provide hardware time stamp */
377 SKBTX_IN_PROGRESS = 1 << 2,
378
Shirley Maa6686f22011-07-06 12:22:12 +0000379 /* device driver supports TX zero-copy buffers */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000380 SKBTX_DEV_ZEROCOPY = 1 << 3,
Johannes Berg6e3e9392011-11-09 10:15:42 +0100381
382 /* generate wifi status information (where possible) */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000383 SKBTX_WIFI_STATUS = 1 << 4,
Pravin B Shelarc9af6db2013-02-11 09:27:41 +0000384
385 /* This indicates at least one fragment might be overwritten
386 * (as in vmsplice(), sendfile() ...)
387 * If we need to compute a TX checksum, we'll need to copy
388 * all frags to avoid possible bad checksum
389 */
390 SKBTX_SHARED_FRAG = 1 << 5,
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400391
392 /* generate software time stamp when entering packet scheduling */
393 SKBTX_SCHED_TSTAMP = 1 << 6,
Shirley Maa6686f22011-07-06 12:22:12 +0000394};
395
Willem de Bruijne1c8a602014-08-04 22:11:50 -0400396#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
Soheil Hassas Yeganeh0a2cf202016-04-27 23:39:01 -0400397 SKBTX_SCHED_TSTAMP)
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400398#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
399
Shirley Maa6686f22011-07-06 12:22:12 +0000400/*
401 * The callback notifies userspace to release buffers when skb DMA is done in
402 * lower device, the skb last reference should be 0 when calling this.
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000403 * The zerocopy_success argument is true if zero copy transmit occurred,
404 * false on data copy or out of memory error caused by data copy attempt.
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000405 * The ctx field is used to track device context.
406 * The desc field is used to track userspace buffer index.
Shirley Maa6686f22011-07-06 12:22:12 +0000407 */
408struct ubuf_info {
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000409 void (*callback)(struct ubuf_info *, bool zerocopy_success);
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000410 void *ctx;
Shirley Maa6686f22011-07-06 12:22:12 +0000411 unsigned long desc;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000412};
413
Linus Torvalds1da177e2005-04-16 15:20:36 -0700414/* This data is invariant across clones and lives at
415 * the end of the header data, ie. at skb->end.
416 */
417struct skb_shared_info {
Steffen Klassert7f564522017-04-08 20:36:24 +0200418 unsigned short _unused;
Ian Campbell9f42f122012-01-05 07:13:39 +0000419 unsigned char nr_frags;
420 __u8 tx_flags;
Herbert Xu79671682006-06-22 02:40:14 -0700421 unsigned short gso_size;
422 /* Warning: this field is not always filled in (UFO)! */
423 unsigned short gso_segs;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700424 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000425 struct skb_shared_hwtstamps hwtstamps;
Steffen Klassert7f564522017-04-08 20:36:24 +0200426 unsigned int gso_type;
Willem de Bruijn09c2d252014-08-04 22:11:47 -0400427 u32 tskey;
Ian Campbell9f42f122012-01-05 07:13:39 +0000428 __be32 ip6_frag_id;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700429
430 /*
431 * Warning : all fields before dataref are cleared in __alloc_skb()
432 */
433 atomic_t dataref;
434
Johann Baudy69e3c752009-05-18 22:11:22 -0700435 /* Intermediate layers must ensure that destructor_arg
436 * remains valid until skb destructor */
437 void * destructor_arg;
Shirley Maa6686f22011-07-06 12:22:12 +0000438
Eric Dumazetfed66382010-07-22 19:09:08 +0000439 /* must be last field, see pskb_expand_head() */
440 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700441};
442
443/* We divide dataref into two halves. The higher 16 bits hold references
444 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700445 * the entire skb->data. A clone of a headerless skb holds the length of
446 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 *
448 * All users must obey the rule that the skb->data reference count must be
449 * greater than or equal to the payload reference count.
450 *
451 * Holding a reference to the payload part means that the user does not
452 * care about modifications to the header part of skb->data.
453 */
454#define SKB_DATAREF_SHIFT 16
455#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
456
David S. Millerd179cd12005-08-17 14:57:30 -0700457
458enum {
Vijay Subramanianc8753d52014-10-02 10:00:43 -0700459 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
460 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
461 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
David S. Millerd179cd12005-08-17 14:57:30 -0700462};
463
Herbert Xu79671682006-06-22 02:40:14 -0700464enum {
465 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700466 SKB_GSO_UDP = 1 << 1,
Herbert Xu576a30e2006-06-27 13:22:38 -0700467
468 /* This indicates the skb is from an untrusted source. */
469 SKB_GSO_DODGY = 1 << 2,
Michael Chanb0da85372006-06-29 12:30:00 -0700470
471 /* This indicates the tcp segment has CWR set. */
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700472 SKB_GSO_TCP_ECN = 1 << 3,
473
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400474 SKB_GSO_TCP_FIXEDID = 1 << 4,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800475
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400476 SKB_GSO_TCPV6 = 1 << 5,
Pravin B Shelar68c33162013-02-14 14:02:41 +0000477
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400478 SKB_GSO_FCOE = 1 << 6,
Pravin B Shelar73136262013-03-07 13:21:51 +0000479
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400480 SKB_GSO_GRE = 1 << 7,
Simon Horman0d89d202013-05-23 21:02:52 +0000481
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400482 SKB_GSO_GRE_CSUM = 1 << 8,
Eric Dumazetcb32f512013-10-19 11:42:57 -0700483
Tom Herbert7e133182016-05-18 09:06:10 -0700484 SKB_GSO_IPXIP4 = 1 << 9,
Eric Dumazet61c1db72013-10-20 20:47:30 -0700485
Tom Herbert7e133182016-05-18 09:06:10 -0700486 SKB_GSO_IPXIP6 = 1 << 10,
Tom Herbert0f4f4ff2014-06-04 17:20:16 -0700487
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400488 SKB_GSO_UDP_TUNNEL = 1 << 11,
Tom Herbert4749c092014-06-04 17:20:23 -0700489
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400490 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 12,
491
Alexander Duyck802ab552016-04-10 21:45:03 -0400492 SKB_GSO_PARTIAL = 1 << 13,
493
494 SKB_GSO_TUNNEL_REMCSUM = 1 << 14,
Marcelo Ricardo Leitner90017ac2016-06-02 15:05:43 -0300495
496 SKB_GSO_SCTP = 1 << 15,
Steffen Klassertc7ef8f02017-04-14 10:05:36 +0200497
498 SKB_GSO_ESP = 1 << 16,
Herbert Xu79671682006-06-22 02:40:14 -0700499};
500
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700501#if BITS_PER_LONG > 32
502#define NET_SKBUFF_DATA_USES_OFFSET 1
503#endif
504
505#ifdef NET_SKBUFF_DATA_USES_OFFSET
506typedef unsigned int sk_buff_data_t;
507#else
508typedef unsigned char *sk_buff_data_t;
509#endif
510
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511/**
512 * struct sk_buff - socket buffer
513 * @next: Next buffer in list
514 * @prev: Previous buffer in list
Eric Dumazet363ec39232014-02-26 14:02:11 -0800515 * @tstamp: Time we arrived/left
Eric Dumazet56b17422014-11-03 08:19:53 -0800516 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700517 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700519 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000520 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700521 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522 * @len: Length of actual data
523 * @data_len: Data length
524 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700525 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700526 * @csum: Checksum (must include start/offset pair)
527 * @csum_start: Offset from skb->head where checksumming should start
528 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700529 * @priority: Packet queueing priority
WANG Cong60ff7462014-05-04 16:39:18 -0700530 * @ignore_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700532 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700533 * @nohdr: Payload reference only, must not modify header
534 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700535 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700536 * @ipvs_property: skbuff is owned by ipvs
Willem de Bruijne7246e12017-01-07 17:06:35 -0500537 * @tc_skip_classify: do not classify packet. set by IFB device
Willem de Bruijn8dc07fd2017-01-07 17:06:37 -0500538 * @tc_at_ingress: used within tc_classify to distinguish in/egress
Willem de Bruijnbc31c902017-01-07 17:06:38 -0500539 * @tc_redirected: packet was redirected by a tc action
540 * @tc_from_ingress: if tc_redirected, tc_at_ingress at time of redirect
Randy Dunlap31729362008-02-18 20:52:13 -0800541 * @peeked: this packet has been seen already, so stats have been
542 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700543 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700544 * @protocol: Packet protocol from driver
545 * @destructor: Destruct function
Florian Westphala9e419d2017-01-23 18:21:59 +0100546 * @_nfct: Associated connection, if any (with nfctinfo bits)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800548 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549 * @tc_index: Traffic control index
Tom Herbert61b905d2014-03-24 15:34:47 -0700550 * @hash: the packet hash
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700551 * @queue_mapping: Queue mapping for multiqueue devices
David S. Miller0b725a22014-08-25 15:51:53 -0700552 * @xmit_more: More SKBs are pending for this queue
Randy Dunlap553a5672008-04-20 10:51:01 -0700553 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700554 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Tom Herbert61b905d2014-03-24 15:34:47 -0700555 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
Changli Gao4ca24622011-08-19 07:26:44 -0700556 * ports.
Tom Herberta3b18dd2014-07-01 21:33:17 -0700557 * @sw_hash: indicates hash was computed in software stack
Johannes Berg6e3e9392011-11-09 10:15:42 +0100558 * @wifi_acked_valid: wifi_acked was set
559 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000560 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Davide Carattidba00302017-05-18 15:44:40 +0200561 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
Julian Anastasov4ff06202017-02-06 23:14:12 +0200562 * @dst_pending_confirm: need to confirm neighbour
Eliezer Tamir06021292013-06-10 11:39:50 +0300563 * @napi_id: id of the NAPI struct this skb came from
James Morris984bc162006-06-09 00:29:17 -0700564 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700565 * @mark: Generic packet mark
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000566 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b2008-07-14 22:49:06 -0700567 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000568 * @inner_protocol: Protocol (encapsulation)
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000569 * @inner_transport_header: Inner transport layer header (encapsulation)
570 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000571 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700572 * @transport_header: Transport layer header
573 * @network_header: Network layer header
574 * @mac_header: Link layer header
575 * @tail: Tail pointer
576 * @end: End pointer
577 * @head: Head of buffer
578 * @data: Data head pointer
579 * @truesize: Buffer size
580 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700581 */
582
583struct sk_buff {
Eric Dumazet363ec39232014-02-26 14:02:11 -0800584 union {
Eric Dumazet56b17422014-11-03 08:19:53 -0800585 struct {
586 /* These two members must be first. */
587 struct sk_buff *next;
588 struct sk_buff *prev;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000589
Eric Dumazet56b17422014-11-03 08:19:53 -0800590 union {
591 ktime_t tstamp;
Eric Dumazet9a568de2017-05-16 14:00:14 -0700592 u64 skb_mstamp;
Eric Dumazet56b17422014-11-03 08:19:53 -0800593 };
594 };
595 struct rb_node rbnode; /* used in netem & tcp stack */
596 };
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000597 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700598
Eric Dumazetc84d9492016-12-08 11:41:55 -0800599 union {
600 struct net_device *dev;
601 /* Some protocols might use this space to store information,
602 * while device pointer would be NULL.
603 * UDP receive path is one user.
604 */
605 unsigned long dev_scratch;
606 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607 /*
608 * This is the control buffer. It is free to use for every
609 * layer. Please put your private variables there. If you
610 * want to keep them across layers you have to do a skb_clone()
611 * first. This is owned by whoever has the skb queued ATM.
612 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000613 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700614
Eric Dumazet7fee2262010-05-11 23:19:48 +0000615 unsigned long _skb_refdst;
Eric Dumazetb1937222014-09-28 22:18:47 -0700616 void (*destructor)(struct sk_buff *skb);
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000617#ifdef CONFIG_XFRM
618 struct sec_path *sp;
619#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700620#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +0100621 unsigned long _nfct;
Eric Dumazetb1937222014-09-28 22:18:47 -0700622#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +0200623#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Eric Dumazetb1937222014-09-28 22:18:47 -0700624 struct nf_bridge_info *nf_bridge;
625#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700627 data_len;
628 __u16 mac_len,
629 hdr_len;
Eric Dumazetb1937222014-09-28 22:18:47 -0700630
631 /* Following fields are _not_ copied in __copy_skb_header()
632 * Note that queue_mapping is here mostly to fill a hole.
633 */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200634 kmemcheck_bitfield_begin(flags1);
Eric Dumazetb1937222014-09-28 22:18:47 -0700635 __u16 queue_mapping;
Daniel Borkmann36bbef52016-09-20 00:26:13 +0200636
637/* if you move cloned around you also must adapt those constants */
638#ifdef __BIG_ENDIAN_BITFIELD
639#define CLONED_MASK (1 << 7)
640#else
641#define CLONED_MASK 1
642#endif
643#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
644
645 __u8 __cloned_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700646 __u8 cloned:1,
Harald Welte6869c4d2005-08-09 19:24:19 -0700647 nohdr:1,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800648 fclone:2,
Herbert Xua59322b2007-12-05 01:53:40 -0800649 peeked:1,
Eric Dumazetb1937222014-09-28 22:18:47 -0700650 head_frag:1,
Daniel Borkmann36bbef52016-09-20 00:26:13 +0200651 xmit_more:1,
652 __unused:1; /* one bit hole */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200653 kmemcheck_bitfield_end(flags1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700654
Eric Dumazetb1937222014-09-28 22:18:47 -0700655 /* fields enclosed in headers_start/headers_end are copied
656 * using a single memcpy() in __copy_skb_header()
657 */
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700658 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700659 __u32 headers_start[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700660 /* public: */
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200661
662/* if you move pkt_type around you also must adapt those constants */
663#ifdef __BIG_ENDIAN_BITFIELD
664#define PKT_TYPE_MAX (7 << 5)
665#else
666#define PKT_TYPE_MAX 7
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100667#endif
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200668#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
669
670 __u8 __pkt_type_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700671 __u8 pkt_type:3;
672 __u8 pfmemalloc:1;
673 __u8 ignore_df:1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700674
Eric Dumazetb1937222014-09-28 22:18:47 -0700675 __u8 nf_trace:1;
676 __u8 ip_summed:2;
677 __u8 ooo_okay:1;
678 __u8 l4_hash:1;
679 __u8 sw_hash:1;
680 __u8 wifi_acked_valid:1;
681 __u8 wifi_acked:1;
682
683 __u8 no_fcs:1;
684 /* Indicates the inner headers are valid in the skbuff. */
685 __u8 encapsulation:1;
686 __u8 encap_hdr_csum:1;
687 __u8 csum_valid:1;
688 __u8 csum_complete_sw:1;
689 __u8 csum_level:2;
Davide Carattidba00302017-05-18 15:44:40 +0200690 __u8 csum_not_inet:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700691
Julian Anastasov4ff06202017-02-06 23:14:12 +0200692 __u8 dst_pending_confirm:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700693#ifdef CONFIG_IPV6_NDISC_NODETYPE
694 __u8 ndisc_nodetype:2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700695#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700696 __u8 ipvs_property:1;
Tom Herbert8bce6d72014-09-29 20:22:29 -0700697 __u8 inner_protocol_type:1;
Tom Herberte585f232014-11-04 09:06:54 -0800698 __u8 remcsum_offload:1;
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200699#ifdef CONFIG_NET_SWITCHDEV
700 __u8 offload_fwd_mark:1;
701#endif
Willem de Bruijne7246e12017-01-07 17:06:35 -0500702#ifdef CONFIG_NET_CLS_ACT
703 __u8 tc_skip_classify:1;
Willem de Bruijn8dc07fd2017-01-07 17:06:37 -0500704 __u8 tc_at_ingress:1;
Willem de Bruijnbc31c902017-01-07 17:06:38 -0500705 __u8 tc_redirected:1;
706 __u8 tc_from_ingress:1;
Willem de Bruijne7246e12017-01-07 17:06:35 -0500707#endif
Alexander Duyck4031ae62012-01-27 06:22:53 +0000708
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700710 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200712
Eric Dumazetb1937222014-09-28 22:18:47 -0700713 union {
714 __wsum csum;
715 struct {
716 __u16 csum_start;
717 __u16 csum_offset;
718 };
719 };
720 __u32 priority;
721 int skb_iif;
722 __u32 hash;
723 __be16 vlan_proto;
724 __u16 vlan_tci;
Eric Dumazet2bd82482015-02-03 23:48:24 -0800725#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
726 union {
727 unsigned int napi_id;
728 unsigned int sender_cpu;
729 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700730#endif
James Morris984bc162006-06-09 00:29:17 -0700731#ifdef CONFIG_NETWORK_SECMARK
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200732 __u32 secmark;
James Morris984bc162006-06-09 00:29:17 -0700733#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -0700734
Neil Horman3b885782009-10-12 13:26:31 -0700735 union {
736 __u32 mark;
Eric Dumazet16fad692013-03-14 05:40:32 +0000737 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700738 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700739
Tom Herbert8bce6d72014-09-29 20:22:29 -0700740 union {
741 __be16 inner_protocol;
742 __u8 inner_ipproto;
743 };
744
Simon Horman1a37e412013-05-23 21:02:51 +0000745 __u16 inner_transport_header;
746 __u16 inner_network_header;
747 __u16 inner_mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700748
749 __be16 protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000750 __u16 transport_header;
751 __u16 network_header;
752 __u16 mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700753
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700754 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700755 __u32 headers_end[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700756 /* public: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700757
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700759 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700760 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700761 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700762 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700763 unsigned int truesize;
764 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765};
766
767#ifdef __KERNEL__
768/*
769 * Handling routines are only of interest to the kernel
770 */
771#include <linux/slab.h>
772
Linus Torvalds1da177e2005-04-16 15:20:36 -0700773
Mel Gormanc93bdd02012-07-31 16:44:19 -0700774#define SKB_ALLOC_FCLONE 0x01
775#define SKB_ALLOC_RX 0x02
Alexander Duyckfd11a832014-12-09 19:40:49 -0800776#define SKB_ALLOC_NAPI 0x04
Mel Gormanc93bdd02012-07-31 16:44:19 -0700777
778/* Returns true if the skb was allocated from PFMEMALLOC reserves */
779static inline bool skb_pfmemalloc(const struct sk_buff *skb)
780{
781 return unlikely(skb->pfmemalloc);
782}
783
Eric Dumazet7fee2262010-05-11 23:19:48 +0000784/*
785 * skb might have a dst pointer attached, refcounted or not.
786 * _skb_refdst low order bit is set if refcount was _not_ taken
787 */
788#define SKB_DST_NOREF 1UL
789#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
790
Florian Westphala9e419d2017-01-23 18:21:59 +0100791#define SKB_NFCT_PTRMASK ~(7UL)
Eric Dumazet7fee2262010-05-11 23:19:48 +0000792/**
793 * skb_dst - returns skb dst_entry
794 * @skb: buffer
795 *
796 * Returns skb dst_entry, regardless of reference taken or not.
797 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000798static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
799{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000800 /* If refdst was not refcounted, check we still are in a
801 * rcu_read_lock section
802 */
803 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
804 !rcu_read_lock_held() &&
805 !rcu_read_lock_bh_held());
806 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000807}
808
Eric Dumazet7fee2262010-05-11 23:19:48 +0000809/**
810 * skb_dst_set - sets skb dst
811 * @skb: buffer
812 * @dst: dst entry
813 *
814 * Sets skb dst, assuming a reference was taken on dst and should
815 * be released by skb_dst_drop()
816 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000817static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
818{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000819 skb->_skb_refdst = (unsigned long)dst;
820}
821
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200822/**
823 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
824 * @skb: buffer
825 * @dst: dst entry
826 *
827 * Sets skb dst, assuming a reference was not taken on dst.
828 * If dst entry is cached, we do not take reference and dst_release
829 * will be avoided by refdst_drop. If dst entry is not cached, we take
830 * reference, so that last dst_release can destroy the dst immediately.
831 */
832static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
833{
Hannes Frederic Sowadbfc4fb2014-12-06 19:19:42 +0100834 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
835 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200836}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000837
838/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300839 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000840 * @skb: buffer
841 */
842static inline bool skb_dst_is_noref(const struct sk_buff *skb)
843{
844 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000845}
846
Eric Dumazet511c3f92009-06-02 05:14:27 +0000847static inline struct rtable *skb_rtable(const struct sk_buff *skb)
848{
Eric Dumazetadf30902009-06-02 05:19:30 +0000849 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000850}
851
Jamal Hadi Salim8b10cab2016-07-02 06:43:14 -0400852/* For mangling skb->pkt_type from user space side from applications
853 * such as nft, tc, etc, we only allow a conservative subset of
854 * possible pkt_types to be set.
855*/
856static inline bool skb_pkt_type_ok(u32 ptype)
857{
858 return ptype <= PACKET_OTHERHOST;
859}
860
Miroslav Lichvar90b602f2017-05-19 17:52:37 +0200861static inline unsigned int skb_napi_id(const struct sk_buff *skb)
862{
863#ifdef CONFIG_NET_RX_BUSY_POLL
864 return skb->napi_id;
865#else
866 return 0;
867#endif
868}
869
Joe Perches7965bd42013-09-26 14:48:15 -0700870void kfree_skb(struct sk_buff *skb);
871void kfree_skb_list(struct sk_buff *segs);
872void skb_tx_error(struct sk_buff *skb);
873void consume_skb(struct sk_buff *skb);
874void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000875extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000876
Joe Perches7965bd42013-09-26 14:48:15 -0700877void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
878bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
879 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000880
Joe Perches7965bd42013-09-26 14:48:15 -0700881struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
882 int node);
Eric Dumazet2ea2f622015-04-24 16:05:01 -0700883struct sk_buff *__build_skb(void *data, unsigned int frag_size);
Joe Perches7965bd42013-09-26 14:48:15 -0700884struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700885static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100886 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700887{
Eric Dumazet564824b2010-10-11 19:05:25 +0000888 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700889}
890
Eric Dumazet2e4e4412014-09-17 04:49:49 -0700891struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
892 unsigned long data_len,
893 int max_page_order,
894 int *errcode,
895 gfp_t gfp_mask);
896
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700897/* Layout of fast clones : [skb1][skb2][fclone_ref] */
898struct sk_buff_fclones {
899 struct sk_buff skb1;
900
901 struct sk_buff skb2;
902
903 atomic_t fclone_ref;
904};
905
906/**
907 * skb_fclone_busy - check if fclone is busy
Stephen Hemminger293de7d2016-10-23 09:28:29 -0700908 * @sk: socket
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700909 * @skb: buffer
910 *
Masanari Iidabda13fe2015-12-13 16:53:02 +0900911 * Returns true if skb is a fast clone, and its clone is not freed.
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700912 * Some drivers call skb_orphan() in their ndo_start_xmit(),
913 * so we also check that this didnt happen.
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700914 */
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700915static inline bool skb_fclone_busy(const struct sock *sk,
916 const struct sk_buff *skb)
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700917{
918 const struct sk_buff_fclones *fclones;
919
920 fclones = container_of(skb, struct sk_buff_fclones, skb1);
921
922 return skb->fclone == SKB_FCLONE_ORIG &&
Eric Dumazet6ffe75eb2014-12-03 17:04:39 -0800923 atomic_read(&fclones->fclone_ref) > 1 &&
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700924 fclones->skb2.sk == sk;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700925}
926
David S. Millerd179cd12005-08-17 14:57:30 -0700927static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100928 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700929{
Mel Gormanc93bdd02012-07-31 16:44:19 -0700930 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700931}
932
Joe Perches7965bd42013-09-26 14:48:15 -0700933struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
Patrick McHardy0ebd0ac2013-04-17 06:46:58 +0000934static inline struct sk_buff *alloc_skb_head(gfp_t priority)
935{
936 return __alloc_skb_head(priority, -1);
937}
938
Joe Perches7965bd42013-09-26 14:48:15 -0700939struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
940int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
941struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
942struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
Octavian Purdilabad93e92014-06-12 01:36:26 +0300943struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
944 gfp_t gfp_mask, bool fclone);
945static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
946 gfp_t gfp_mask)
947{
948 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
949}
Eric Dumazet117632e2011-12-03 21:39:53 +0000950
Joe Perches7965bd42013-09-26 14:48:15 -0700951int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
952struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
953 unsigned int headroom);
954struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
955 int newtailroom, gfp_t priority);
Jason A. Donenfeld48a1df62017-06-04 04:16:22 +0200956int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
957 int offset, int len);
958int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
959 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -0700960int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
961int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000962#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700963
Joe Perches7965bd42013-09-26 14:48:15 -0700964int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
965 int getfrag(void *from, char *to, int offset,
966 int len, int odd, struct sk_buff *skb),
967 void *from, int length);
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700968
Hannes Frederic Sowabe12a1f2015-05-21 16:59:58 +0200969int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
970 int offset, size_t size);
971
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800972struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700973 __u32 lower_offset;
974 __u32 upper_offset;
975 __u32 frag_idx;
976 __u32 stepped_offset;
977 struct sk_buff *root_skb;
978 struct sk_buff *cur_skb;
979 __u8 *frag_data;
980};
981
Joe Perches7965bd42013-09-26 14:48:15 -0700982void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
983 unsigned int to, struct skb_seq_state *st);
984unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
985 struct skb_seq_state *st);
986void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -0700987
Joe Perches7965bd42013-09-26 14:48:15 -0700988unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
Bojan Prtvar059a2442015-02-22 11:46:35 +0100989 unsigned int to, struct ts_config *config);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700990
Tom Herbert09323cc2013-12-15 22:16:19 -0800991/*
992 * Packet hash types specify the type of hash in skb_set_hash.
993 *
994 * Hash types refer to the protocol layer addresses which are used to
995 * construct a packet's hash. The hashes are used to differentiate or identify
996 * flows of the protocol layer for the hash type. Hash types are either
997 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
998 *
999 * Properties of hashes:
1000 *
1001 * 1) Two packets in different flows have different hash values
1002 * 2) Two packets in the same flow should have the same hash value
1003 *
1004 * A hash at a higher layer is considered to be more specific. A driver should
1005 * set the most specific hash possible.
1006 *
1007 * A driver cannot indicate a more specific hash than the layer at which a hash
1008 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1009 *
1010 * A driver may indicate a hash level which is less specific than the
1011 * actual layer the hash was computed on. For instance, a hash computed
1012 * at L4 may be considered an L3 hash. This should only be done if the
1013 * driver can't unambiguously determine that the HW computed the hash at
1014 * the higher layer. Note that the "should" in the second property above
1015 * permits this.
1016 */
1017enum pkt_hash_types {
1018 PKT_HASH_TYPE_NONE, /* Undefined type */
1019 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1020 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1021 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1022};
1023
Tom Herbertbcc83832015-09-01 09:24:24 -07001024static inline void skb_clear_hash(struct sk_buff *skb)
1025{
1026 skb->hash = 0;
1027 skb->sw_hash = 0;
1028 skb->l4_hash = 0;
1029}
1030
1031static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1032{
1033 if (!skb->l4_hash)
1034 skb_clear_hash(skb);
1035}
1036
1037static inline void
1038__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1039{
1040 skb->l4_hash = is_l4;
1041 skb->sw_hash = is_sw;
1042 skb->hash = hash;
1043}
1044
Tom Herbert09323cc2013-12-15 22:16:19 -08001045static inline void
1046skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1047{
Tom Herbertbcc83832015-09-01 09:24:24 -07001048 /* Used by drivers to set hash from HW */
1049 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1050}
1051
1052static inline void
1053__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1054{
1055 __skb_set_hash(skb, hash, true, is_l4);
Tom Herbert09323cc2013-12-15 22:16:19 -08001056}
1057
Tom Herberte5276932015-09-01 09:24:23 -07001058void __skb_get_hash(struct sk_buff *skb);
Florian Westphalb9177832016-10-26 18:49:46 +02001059u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
Tom Herberte5276932015-09-01 09:24:23 -07001060u32 skb_get_poff(const struct sk_buff *skb);
1061u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1062 const struct flow_keys *keys, int hlen);
1063__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1064 void *data, int hlen_proto);
1065
1066static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1067 int thoff, u8 ip_proto)
1068{
1069 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1070}
1071
1072void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1073 const struct flow_dissector_key *key,
1074 unsigned int key_count);
1075
1076bool __skb_flow_dissect(const struct sk_buff *skb,
1077 struct flow_dissector *flow_dissector,
1078 void *target_container,
Tom Herbertcd79a232015-09-01 09:24:27 -07001079 void *data, __be16 proto, int nhoff, int hlen,
1080 unsigned int flags);
Tom Herberte5276932015-09-01 09:24:23 -07001081
1082static inline bool skb_flow_dissect(const struct sk_buff *skb,
1083 struct flow_dissector *flow_dissector,
Tom Herbertcd79a232015-09-01 09:24:27 -07001084 void *target_container, unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001085{
1086 return __skb_flow_dissect(skb, flow_dissector, target_container,
Tom Herbertcd79a232015-09-01 09:24:27 -07001087 NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001088}
1089
1090static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
Tom Herbertcd79a232015-09-01 09:24:27 -07001091 struct flow_keys *flow,
1092 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001093{
1094 memset(flow, 0, sizeof(*flow));
1095 return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
Tom Herbertcd79a232015-09-01 09:24:27 -07001096 NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001097}
1098
1099static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
1100 void *data, __be16 proto,
Tom Herbertcd79a232015-09-01 09:24:27 -07001101 int nhoff, int hlen,
1102 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001103{
1104 memset(flow, 0, sizeof(*flow));
1105 return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
Tom Herbertcd79a232015-09-01 09:24:27 -07001106 data, proto, nhoff, hlen, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001107}
1108
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001109static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001110{
Tom Herberta3b18dd2014-07-01 21:33:17 -07001111 if (!skb->l4_hash && !skb->sw_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001112 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001113
Tom Herbert61b905d2014-03-24 15:34:47 -07001114 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001115}
1116
David S. Miller20a17bf2015-09-01 21:19:17 -07001117__u32 __skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6);
Tom Herbertf70ea012015-07-31 16:52:10 -07001118
David S. Miller20a17bf2015-09-01 21:19:17 -07001119static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
Tom Herbertf70ea012015-07-31 16:52:10 -07001120{
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001121 if (!skb->l4_hash && !skb->sw_hash) {
1122 struct flow_keys keys;
Tom Herbertde4c1f82015-09-01 18:11:04 -07001123 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001124
Tom Herbertde4c1f82015-09-01 18:11:04 -07001125 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001126 }
Tom Herbertf70ea012015-07-31 16:52:10 -07001127
1128 return skb->hash;
1129}
1130
David S. Miller20a17bf2015-09-01 21:19:17 -07001131__u32 __skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl);
Tom Herbertf70ea012015-07-31 16:52:10 -07001132
David S. Miller20a17bf2015-09-01 21:19:17 -07001133static inline __u32 skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl4)
Tom Herbertf70ea012015-07-31 16:52:10 -07001134{
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001135 if (!skb->l4_hash && !skb->sw_hash) {
1136 struct flow_keys keys;
Tom Herbertde4c1f82015-09-01 18:11:04 -07001137 __u32 hash = __get_hash_from_flowi4(fl4, &keys);
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001138
Tom Herbertde4c1f82015-09-01 18:11:04 -07001139 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001140 }
Tom Herbertf70ea012015-07-31 16:52:10 -07001141
1142 return skb->hash;
1143}
1144
Tom Herbert50fb7992015-05-01 11:30:12 -07001145__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
1146
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001147static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1148{
Tom Herbert61b905d2014-03-24 15:34:47 -07001149 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001150}
1151
Tom Herbert3df7a742013-12-15 22:16:29 -08001152static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1153{
Tom Herbert61b905d2014-03-24 15:34:47 -07001154 to->hash = from->hash;
Tom Herberta3b18dd2014-07-01 21:33:17 -07001155 to->sw_hash = from->sw_hash;
Tom Herbert61b905d2014-03-24 15:34:47 -07001156 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -08001157};
1158
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001159#ifdef NET_SKBUFF_DATA_USES_OFFSET
1160static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1161{
1162 return skb->head + skb->end;
1163}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001164
1165static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1166{
1167 return skb->end;
1168}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001169#else
1170static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1171{
1172 return skb->end;
1173}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001174
1175static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1176{
1177 return skb->end - skb->head;
1178}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001179#endif
1180
Linus Torvalds1da177e2005-04-16 15:20:36 -07001181/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001182#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001183
Patrick Ohlyac45f602009-02-12 05:03:37 +00001184static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1185{
1186 return &skb_shinfo(skb)->hwtstamps;
1187}
1188
Linus Torvalds1da177e2005-04-16 15:20:36 -07001189/**
1190 * skb_queue_empty - check if a queue is empty
1191 * @list: queue head
1192 *
1193 * Returns true if the queue is empty, false otherwise.
1194 */
1195static inline int skb_queue_empty(const struct sk_buff_head *list)
1196{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001197 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001198}
1199
1200/**
David S. Millerfc7ebb22008-09-23 00:34:07 -07001201 * skb_queue_is_last - check if skb is the last entry in the queue
1202 * @list: queue head
1203 * @skb: buffer
1204 *
1205 * Returns true if @skb is the last buffer on the list.
1206 */
1207static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1208 const struct sk_buff *skb)
1209{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001210 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -07001211}
1212
1213/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001214 * skb_queue_is_first - check if skb is the first entry in the queue
1215 * @list: queue head
1216 * @skb: buffer
1217 *
1218 * Returns true if @skb is the first buffer on the list.
1219 */
1220static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1221 const struct sk_buff *skb)
1222{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001223 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001224}
1225
1226/**
David S. Miller249c8b42008-09-23 00:44:42 -07001227 * skb_queue_next - return the next packet in the queue
1228 * @list: queue head
1229 * @skb: current buffer
1230 *
1231 * Return the next packet in @list after @skb. It is only valid to
1232 * call this if skb_queue_is_last() evaluates to false.
1233 */
1234static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1235 const struct sk_buff *skb)
1236{
1237 /* This BUG_ON may seem severe, but if we just return then we
1238 * are going to dereference garbage.
1239 */
1240 BUG_ON(skb_queue_is_last(list, skb));
1241 return skb->next;
1242}
1243
1244/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001245 * skb_queue_prev - return the prev packet in the queue
1246 * @list: queue head
1247 * @skb: current buffer
1248 *
1249 * Return the prev packet in @list before @skb. It is only valid to
1250 * call this if skb_queue_is_first() evaluates to false.
1251 */
1252static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1253 const struct sk_buff *skb)
1254{
1255 /* This BUG_ON may seem severe, but if we just return then we
1256 * are going to dereference garbage.
1257 */
1258 BUG_ON(skb_queue_is_first(list, skb));
1259 return skb->prev;
1260}
1261
1262/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001263 * skb_get - reference buffer
1264 * @skb: buffer to reference
1265 *
1266 * Makes another reference to a socket buffer and returns a pointer
1267 * to the buffer.
1268 */
1269static inline struct sk_buff *skb_get(struct sk_buff *skb)
1270{
1271 atomic_inc(&skb->users);
1272 return skb;
1273}
1274
1275/*
1276 * If users == 1, we are the only owner and are can avoid redundant
1277 * atomic change.
1278 */
1279
1280/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001281 * skb_cloned - is the buffer a clone
1282 * @skb: buffer to check
1283 *
1284 * Returns true if the buffer was generated with skb_clone() and is
1285 * one of multiple shared copies of the buffer. Cloned buffers are
1286 * shared data so must not be written to under normal circumstances.
1287 */
1288static inline int skb_cloned(const struct sk_buff *skb)
1289{
1290 return skb->cloned &&
1291 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1292}
1293
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001294static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1295{
Mel Gormand0164ad2015-11-06 16:28:21 -08001296 might_sleep_if(gfpflags_allow_blocking(pri));
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001297
1298 if (skb_cloned(skb))
1299 return pskb_expand_head(skb, 0, 0, pri);
1300
1301 return 0;
1302}
1303
Linus Torvalds1da177e2005-04-16 15:20:36 -07001304/**
1305 * skb_header_cloned - is the header a clone
1306 * @skb: buffer to check
1307 *
1308 * Returns true if modifying the header part of the buffer requires
1309 * the data to be copied.
1310 */
1311static inline int skb_header_cloned(const struct sk_buff *skb)
1312{
1313 int dataref;
1314
1315 if (!skb->cloned)
1316 return 0;
1317
1318 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1319 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1320 return dataref != 1;
1321}
1322
Eric Dumazet9580bf22016-04-30 10:19:29 -07001323static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1324{
1325 might_sleep_if(gfpflags_allow_blocking(pri));
1326
1327 if (skb_header_cloned(skb))
1328 return pskb_expand_head(skb, 0, 0, pri);
1329
1330 return 0;
1331}
1332
Linus Torvalds1da177e2005-04-16 15:20:36 -07001333/**
1334 * skb_header_release - release reference to header
1335 * @skb: buffer to operate on
1336 *
1337 * Drop a reference to the header part of the buffer. This is done
1338 * by acquiring a payload reference. You must not read from the header
1339 * part of skb->data after this.
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001340 * Note : Check if you can use __skb_header_release() instead.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001341 */
1342static inline void skb_header_release(struct sk_buff *skb)
1343{
1344 BUG_ON(skb->nohdr);
1345 skb->nohdr = 1;
1346 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1347}
1348
1349/**
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001350 * __skb_header_release - release reference to header
1351 * @skb: buffer to operate on
1352 *
1353 * Variant of skb_header_release() assuming skb is private to caller.
1354 * We can avoid one atomic operation.
1355 */
1356static inline void __skb_header_release(struct sk_buff *skb)
1357{
1358 skb->nohdr = 1;
1359 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1360}
1361
1362
1363/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001364 * skb_shared - is the buffer shared
1365 * @skb: buffer to check
1366 *
1367 * Returns true if more than one person has a reference to this
1368 * buffer.
1369 */
1370static inline int skb_shared(const struct sk_buff *skb)
1371{
1372 return atomic_read(&skb->users) != 1;
1373}
1374
1375/**
1376 * skb_share_check - check if buffer is shared and if so clone it
1377 * @skb: buffer to check
1378 * @pri: priority for memory allocation
1379 *
1380 * If the buffer is shared the buffer is cloned and the old copy
1381 * drops a reference. A new clone with a single reference is returned.
1382 * If the buffer is not shared the original buffer is returned. When
1383 * being called from interrupt status or with spinlocks held pri must
1384 * be GFP_ATOMIC.
1385 *
1386 * NULL is returned on a memory allocation failure.
1387 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001388static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001389{
Mel Gormand0164ad2015-11-06 16:28:21 -08001390 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001391 if (skb_shared(skb)) {
1392 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001393
1394 if (likely(nskb))
1395 consume_skb(skb);
1396 else
1397 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001398 skb = nskb;
1399 }
1400 return skb;
1401}
1402
1403/*
1404 * Copy shared buffers into a new sk_buff. We effectively do COW on
1405 * packets to handle cases where we have a local reader and forward
1406 * and a couple of other messy ones. The normal one is tcpdumping
1407 * a packet thats being forwarded.
1408 */
1409
1410/**
1411 * skb_unshare - make a copy of a shared buffer
1412 * @skb: buffer to check
1413 * @pri: priority for memory allocation
1414 *
1415 * If the socket buffer is a clone then this function creates a new
1416 * copy of the data, drops a reference count on the old copy and returns
1417 * the new copy with the reference count at 1. If the buffer is not a clone
1418 * the original buffer is returned. When called with a spinlock held or
1419 * from interrupt state @pri must be %GFP_ATOMIC
1420 *
1421 * %NULL is returned on a memory allocation failure.
1422 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001423static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001424 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001425{
Mel Gormand0164ad2015-11-06 16:28:21 -08001426 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001427 if (skb_cloned(skb)) {
1428 struct sk_buff *nskb = skb_copy(skb, pri);
Alexander Aring31eff812014-10-10 23:10:47 +02001429
1430 /* Free our shared copy */
1431 if (likely(nskb))
1432 consume_skb(skb);
1433 else
1434 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001435 skb = nskb;
1436 }
1437 return skb;
1438}
1439
1440/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001441 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001442 * @list_: list to peek at
1443 *
1444 * Peek an &sk_buff. Unlike most other operations you _MUST_
1445 * be careful with this one. A peek leaves the buffer on the
1446 * list and someone else may run off with it. You must hold
1447 * the appropriate locks or have a private queue to do this.
1448 *
1449 * Returns %NULL for an empty list or a pointer to the head element.
1450 * The reference count is not incremented and the reference is therefore
1451 * volatile. Use with caution.
1452 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001453static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001454{
Eric Dumazet18d07002012-04-30 16:31:46 +00001455 struct sk_buff *skb = list_->next;
1456
1457 if (skb == (struct sk_buff *)list_)
1458 skb = NULL;
1459 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001460}
1461
1462/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001463 * skb_peek_next - peek skb following the given one from a queue
1464 * @skb: skb to start from
1465 * @list_: list to peek at
1466 *
1467 * Returns %NULL when the end of the list is met or a pointer to the
1468 * next element. The reference count is not incremented and the
1469 * reference is therefore volatile. Use with caution.
1470 */
1471static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1472 const struct sk_buff_head *list_)
1473{
1474 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001475
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001476 if (next == (struct sk_buff *)list_)
1477 next = NULL;
1478 return next;
1479}
1480
1481/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001482 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001483 * @list_: list to peek at
1484 *
1485 * Peek an &sk_buff. Unlike most other operations you _MUST_
1486 * be careful with this one. A peek leaves the buffer on the
1487 * list and someone else may run off with it. You must hold
1488 * the appropriate locks or have a private queue to do this.
1489 *
1490 * Returns %NULL for an empty list or a pointer to the tail element.
1491 * The reference count is not incremented and the reference is therefore
1492 * volatile. Use with caution.
1493 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001494static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001495{
Eric Dumazet18d07002012-04-30 16:31:46 +00001496 struct sk_buff *skb = list_->prev;
1497
1498 if (skb == (struct sk_buff *)list_)
1499 skb = NULL;
1500 return skb;
1501
Linus Torvalds1da177e2005-04-16 15:20:36 -07001502}
1503
1504/**
1505 * skb_queue_len - get queue length
1506 * @list_: list to measure
1507 *
1508 * Return the length of an &sk_buff queue.
1509 */
1510static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1511{
1512 return list_->qlen;
1513}
1514
David S. Miller67fed452008-09-21 22:36:24 -07001515/**
1516 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1517 * @list: queue to initialize
1518 *
1519 * This initializes only the list and queue length aspects of
1520 * an sk_buff_head object. This allows to initialize the list
1521 * aspects of an sk_buff_head without reinitializing things like
1522 * the spinlock. It can also be used for on-stack sk_buff_head
1523 * objects where the spinlock is known to not be used.
1524 */
1525static inline void __skb_queue_head_init(struct sk_buff_head *list)
1526{
1527 list->prev = list->next = (struct sk_buff *)list;
1528 list->qlen = 0;
1529}
1530
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001531/*
1532 * This function creates a split out lock class for each invocation;
1533 * this is needed for now since a whole lot of users of the skb-queue
1534 * infrastructure in drivers have different locking usage (in hardirq)
1535 * than the networking core (in softirq only). In the long run either the
1536 * network layer or drivers should need annotation to consolidate the
1537 * main types of usage into 3 classes.
1538 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539static inline void skb_queue_head_init(struct sk_buff_head *list)
1540{
1541 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001542 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001543}
1544
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001545static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1546 struct lock_class_key *class)
1547{
1548 skb_queue_head_init(list);
1549 lockdep_set_class(&list->lock, class);
1550}
1551
Linus Torvalds1da177e2005-04-16 15:20:36 -07001552/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001553 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001554 *
1555 * The "__skb_xxxx()" functions are the non-atomic ones that
1556 * can only be called with interrupts disabled.
1557 */
Joe Perches7965bd42013-09-26 14:48:15 -07001558void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1559 struct sk_buff_head *list);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001560static inline void __skb_insert(struct sk_buff *newsk,
1561 struct sk_buff *prev, struct sk_buff *next,
1562 struct sk_buff_head *list)
1563{
1564 newsk->next = next;
1565 newsk->prev = prev;
1566 next->prev = prev->next = newsk;
1567 list->qlen++;
1568}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569
David S. Miller67fed452008-09-21 22:36:24 -07001570static inline void __skb_queue_splice(const struct sk_buff_head *list,
1571 struct sk_buff *prev,
1572 struct sk_buff *next)
1573{
1574 struct sk_buff *first = list->next;
1575 struct sk_buff *last = list->prev;
1576
1577 first->prev = prev;
1578 prev->next = first;
1579
1580 last->next = next;
1581 next->prev = last;
1582}
1583
1584/**
1585 * skb_queue_splice - join two skb lists, this is designed for stacks
1586 * @list: the new list to add
1587 * @head: the place to add it in the first list
1588 */
1589static inline void skb_queue_splice(const struct sk_buff_head *list,
1590 struct sk_buff_head *head)
1591{
1592 if (!skb_queue_empty(list)) {
1593 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001594 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001595 }
1596}
1597
1598/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001599 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001600 * @list: the new list to add
1601 * @head: the place to add it in the first list
1602 *
1603 * The list at @list is reinitialised
1604 */
1605static inline void skb_queue_splice_init(struct sk_buff_head *list,
1606 struct sk_buff_head *head)
1607{
1608 if (!skb_queue_empty(list)) {
1609 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001610 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001611 __skb_queue_head_init(list);
1612 }
1613}
1614
1615/**
1616 * skb_queue_splice_tail - join two skb lists, each list being a queue
1617 * @list: the new list to add
1618 * @head: the place to add it in the first list
1619 */
1620static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1621 struct sk_buff_head *head)
1622{
1623 if (!skb_queue_empty(list)) {
1624 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001625 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001626 }
1627}
1628
1629/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001630 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001631 * @list: the new list to add
1632 * @head: the place to add it in the first list
1633 *
1634 * Each of the lists is a queue.
1635 * The list at @list is reinitialised
1636 */
1637static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1638 struct sk_buff_head *head)
1639{
1640 if (!skb_queue_empty(list)) {
1641 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001642 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001643 __skb_queue_head_init(list);
1644 }
1645}
1646
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001648 * __skb_queue_after - queue a buffer at the list head
1649 * @list: list to use
1650 * @prev: place after this buffer
1651 * @newsk: buffer to queue
1652 *
1653 * Queue a buffer int the middle of a list. This function takes no locks
1654 * and you must therefore hold required locks before calling it.
1655 *
1656 * A buffer cannot be placed on two lists at the same time.
1657 */
1658static inline void __skb_queue_after(struct sk_buff_head *list,
1659 struct sk_buff *prev,
1660 struct sk_buff *newsk)
1661{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001662 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001663}
1664
Joe Perches7965bd42013-09-26 14:48:15 -07001665void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1666 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001667
Gerrit Renkerf5572852008-04-14 00:05:28 -07001668static inline void __skb_queue_before(struct sk_buff_head *list,
1669 struct sk_buff *next,
1670 struct sk_buff *newsk)
1671{
1672 __skb_insert(newsk, next->prev, next, list);
1673}
1674
Stephen Hemminger300ce172005-10-30 13:47:34 -08001675/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001676 * __skb_queue_head - queue a buffer at the list head
1677 * @list: list to use
1678 * @newsk: buffer to queue
1679 *
1680 * Queue a buffer at the start of a list. This function takes no locks
1681 * and you must therefore hold required locks before calling it.
1682 *
1683 * A buffer cannot be placed on two lists at the same time.
1684 */
Joe Perches7965bd42013-09-26 14:48:15 -07001685void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001686static inline void __skb_queue_head(struct sk_buff_head *list,
1687 struct sk_buff *newsk)
1688{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001689 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001690}
1691
1692/**
1693 * __skb_queue_tail - queue a buffer at the list tail
1694 * @list: list to use
1695 * @newsk: buffer to queue
1696 *
1697 * Queue a buffer at the end of a list. This function takes no locks
1698 * and you must therefore hold required locks before calling it.
1699 *
1700 * A buffer cannot be placed on two lists at the same time.
1701 */
Joe Perches7965bd42013-09-26 14:48:15 -07001702void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001703static inline void __skb_queue_tail(struct sk_buff_head *list,
1704 struct sk_buff *newsk)
1705{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001706 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001707}
1708
Linus Torvalds1da177e2005-04-16 15:20:36 -07001709/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710 * remove sk_buff from list. _Must_ be called atomically, and with
1711 * the list known..
1712 */
Joe Perches7965bd42013-09-26 14:48:15 -07001713void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001714static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1715{
1716 struct sk_buff *next, *prev;
1717
1718 list->qlen--;
1719 next = skb->next;
1720 prev = skb->prev;
1721 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001722 next->prev = prev;
1723 prev->next = next;
1724}
1725
Gerrit Renkerf525c062008-04-14 00:04:12 -07001726/**
1727 * __skb_dequeue - remove from the head of the queue
1728 * @list: list to dequeue from
1729 *
1730 * Remove the head of the list. This function does not take any locks
1731 * so must be used with appropriate locks held only. The head item is
1732 * returned or %NULL if the list is empty.
1733 */
Joe Perches7965bd42013-09-26 14:48:15 -07001734struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Gerrit Renkerf525c062008-04-14 00:04:12 -07001735static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1736{
1737 struct sk_buff *skb = skb_peek(list);
1738 if (skb)
1739 __skb_unlink(skb, list);
1740 return skb;
1741}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001742
1743/**
1744 * __skb_dequeue_tail - remove from the tail of the queue
1745 * @list: list to dequeue from
1746 *
1747 * Remove the tail of the list. This function does not take any locks
1748 * so must be used with appropriate locks held only. The tail item is
1749 * returned or %NULL if the list is empty.
1750 */
Joe Perches7965bd42013-09-26 14:48:15 -07001751struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001752static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1753{
1754 struct sk_buff *skb = skb_peek_tail(list);
1755 if (skb)
1756 __skb_unlink(skb, list);
1757 return skb;
1758}
1759
1760
David S. Millerbdcc0922012-03-07 20:53:36 -05001761static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001762{
1763 return skb->data_len;
1764}
1765
1766static inline unsigned int skb_headlen(const struct sk_buff *skb)
1767{
1768 return skb->len - skb->data_len;
1769}
1770
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03001771static inline unsigned int skb_pagelen(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001772{
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03001773 unsigned int i, len = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001774
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03001775 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001776 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001777 return len + skb_headlen(skb);
1778}
1779
Ian Campbell131ea662011-08-19 06:25:00 +00001780/**
1781 * __skb_fill_page_desc - initialise a paged fragment in an skb
1782 * @skb: buffer containing fragment to be initialised
1783 * @i: paged fragment index to initialise
1784 * @page: the page to use for this fragment
1785 * @off: the offset to the data with @page
1786 * @size: the length of the data
1787 *
1788 * Initialises the @i'th fragment of @skb to point to &size bytes at
1789 * offset @off within @page.
1790 *
1791 * Does not take any additional reference on the fragment.
1792 */
1793static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1794 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001795{
1796 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1797
Mel Gormanc48a11c2012-07-31 16:44:23 -07001798 /*
Michal Hocko2f064f32015-08-21 14:11:51 -07001799 * Propagate page pfmemalloc to the skb if we can. The problem is
1800 * that not all callers have unique ownership of the page but rely
1801 * on page_is_pfmemalloc doing the right thing(tm).
Mel Gormanc48a11c2012-07-31 16:44:23 -07001802 */
Ian Campbella8605c62011-10-19 23:01:49 +00001803 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001804 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001805 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001806
1807 page = compound_head(page);
Michal Hocko2f064f32015-08-21 14:11:51 -07001808 if (page_is_pfmemalloc(page))
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001809 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00001810}
1811
1812/**
1813 * skb_fill_page_desc - initialise a paged fragment in an skb
1814 * @skb: buffer containing fragment to be initialised
1815 * @i: paged fragment index to initialise
1816 * @page: the page to use for this fragment
1817 * @off: the offset to the data with @page
1818 * @size: the length of the data
1819 *
1820 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01001821 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00001822 * addition updates @skb such that @i is the last fragment.
1823 *
1824 * Does not take any additional reference on the fragment.
1825 */
1826static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1827 struct page *page, int off, int size)
1828{
1829 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001830 skb_shinfo(skb)->nr_frags = i + 1;
1831}
1832
Joe Perches7965bd42013-09-26 14:48:15 -07001833void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1834 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001835
Jason Wangf8e617e2013-11-01 14:07:47 +08001836void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1837 unsigned int truesize);
1838
Linus Torvalds1da177e2005-04-16 15:20:36 -07001839#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001840#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001841#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1842
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001843#ifdef NET_SKBUFF_DATA_USES_OFFSET
1844static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1845{
1846 return skb->head + skb->tail;
1847}
1848
1849static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1850{
1851 skb->tail = skb->data - skb->head;
1852}
1853
1854static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1855{
1856 skb_reset_tail_pointer(skb);
1857 skb->tail += offset;
1858}
Simon Horman7cc46192013-05-28 20:34:29 +00001859
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001860#else /* NET_SKBUFF_DATA_USES_OFFSET */
1861static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1862{
1863 return skb->tail;
1864}
1865
1866static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1867{
1868 skb->tail = skb->data;
1869}
1870
1871static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1872{
1873 skb->tail = skb->data + offset;
1874}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001875
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001876#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1877
Linus Torvalds1da177e2005-04-16 15:20:36 -07001878/*
1879 * Add data to an sk_buff
1880 */
Mathias Krause0c7ddf32013-11-07 14:18:24 +01001881unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001882unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001883static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1884{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001885 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001886 SKB_LINEAR_ASSERT(skb);
1887 skb->tail += len;
1888 skb->len += len;
1889 return tmp;
1890}
1891
Joe Perches7965bd42013-09-26 14:48:15 -07001892unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001893static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1894{
1895 skb->data -= len;
1896 skb->len += len;
1897 return skb->data;
1898}
1899
Joe Perches7965bd42013-09-26 14:48:15 -07001900unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001901static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1902{
1903 skb->len -= len;
1904 BUG_ON(skb->len < skb->data_len);
1905 return skb->data += len;
1906}
1907
David S. Miller47d29642010-05-02 02:21:44 -07001908static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1909{
1910 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1911}
1912
Joe Perches7965bd42013-09-26 14:48:15 -07001913unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001914
1915static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1916{
1917 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001918 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001919 return NULL;
1920 skb->len -= len;
1921 return skb->data += len;
1922}
1923
1924static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1925{
1926 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1927}
1928
1929static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1930{
1931 if (likely(len <= skb_headlen(skb)))
1932 return 1;
1933 if (unlikely(len > skb->len))
1934 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001935 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001936}
1937
Eric Dumazetc8c8b122016-12-07 09:19:33 -08001938void skb_condense(struct sk_buff *skb);
1939
Linus Torvalds1da177e2005-04-16 15:20:36 -07001940/**
1941 * skb_headroom - bytes at buffer head
1942 * @skb: buffer to check
1943 *
1944 * Return the number of bytes of free space at the head of an &sk_buff.
1945 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001946static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001947{
1948 return skb->data - skb->head;
1949}
1950
1951/**
1952 * skb_tailroom - bytes at buffer end
1953 * @skb: buffer to check
1954 *
1955 * Return the number of bytes of free space at the tail of an sk_buff
1956 */
1957static inline int skb_tailroom(const struct sk_buff *skb)
1958{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001959 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001960}
1961
1962/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00001963 * skb_availroom - bytes at buffer end
1964 * @skb: buffer to check
1965 *
1966 * Return the number of bytes of free space at the tail of an sk_buff
1967 * allocated by sk_stream_alloc()
1968 */
1969static inline int skb_availroom(const struct sk_buff *skb)
1970{
Eric Dumazet16fad692013-03-14 05:40:32 +00001971 if (skb_is_nonlinear(skb))
1972 return 0;
1973
1974 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00001975}
1976
1977/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001978 * skb_reserve - adjust headroom
1979 * @skb: buffer to alter
1980 * @len: bytes to move
1981 *
1982 * Increase the headroom of an empty &sk_buff by reducing the tail
1983 * room. This is only allowed for an empty buffer.
1984 */
David S. Miller8243126c2006-01-17 02:54:21 -08001985static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001986{
1987 skb->data += len;
1988 skb->tail += len;
1989}
1990
Benjamin Poirier1837b2e2016-02-29 15:03:33 -08001991/**
1992 * skb_tailroom_reserve - adjust reserved_tailroom
1993 * @skb: buffer to alter
1994 * @mtu: maximum amount of headlen permitted
1995 * @needed_tailroom: minimum amount of reserved_tailroom
1996 *
1997 * Set reserved_tailroom so that headlen can be as large as possible but
1998 * not larger than mtu and tailroom cannot be smaller than
1999 * needed_tailroom.
2000 * The required headroom should already have been reserved before using
2001 * this function.
2002 */
2003static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2004 unsigned int needed_tailroom)
2005{
2006 SKB_LINEAR_ASSERT(skb);
2007 if (mtu < skb_tailroom(skb) - needed_tailroom)
2008 /* use at most mtu */
2009 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2010 else
2011 /* use up to all available space */
2012 skb->reserved_tailroom = needed_tailroom;
2013}
2014
Tom Herbert8bce6d72014-09-29 20:22:29 -07002015#define ENCAP_TYPE_ETHER 0
2016#define ENCAP_TYPE_IPPROTO 1
2017
2018static inline void skb_set_inner_protocol(struct sk_buff *skb,
2019 __be16 protocol)
2020{
2021 skb->inner_protocol = protocol;
2022 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2023}
2024
2025static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2026 __u8 ipproto)
2027{
2028 skb->inner_ipproto = ipproto;
2029 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2030}
2031
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002032static inline void skb_reset_inner_headers(struct sk_buff *skb)
2033{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002034 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002035 skb->inner_network_header = skb->network_header;
2036 skb->inner_transport_header = skb->transport_header;
2037}
2038
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00002039static inline void skb_reset_mac_len(struct sk_buff *skb)
2040{
2041 skb->mac_len = skb->network_header - skb->mac_header;
2042}
2043
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002044static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2045 *skb)
2046{
2047 return skb->head + skb->inner_transport_header;
2048}
2049
Tom Herbert55dc5a92015-12-14 11:19:40 -08002050static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2051{
2052 return skb_inner_transport_header(skb) - skb->data;
2053}
2054
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002055static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2056{
2057 skb->inner_transport_header = skb->data - skb->head;
2058}
2059
2060static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2061 const int offset)
2062{
2063 skb_reset_inner_transport_header(skb);
2064 skb->inner_transport_header += offset;
2065}
2066
2067static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2068{
2069 return skb->head + skb->inner_network_header;
2070}
2071
2072static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2073{
2074 skb->inner_network_header = skb->data - skb->head;
2075}
2076
2077static inline void skb_set_inner_network_header(struct sk_buff *skb,
2078 const int offset)
2079{
2080 skb_reset_inner_network_header(skb);
2081 skb->inner_network_header += offset;
2082}
2083
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002084static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2085{
2086 return skb->head + skb->inner_mac_header;
2087}
2088
2089static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2090{
2091 skb->inner_mac_header = skb->data - skb->head;
2092}
2093
2094static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2095 const int offset)
2096{
2097 skb_reset_inner_mac_header(skb);
2098 skb->inner_mac_header += offset;
2099}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002100static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2101{
Cong Wang35d04612013-05-29 15:16:05 +08002102 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002103}
2104
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002105static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2106{
2107 return skb->head + skb->transport_header;
2108}
2109
2110static inline void skb_reset_transport_header(struct sk_buff *skb)
2111{
2112 skb->transport_header = skb->data - skb->head;
2113}
2114
2115static inline void skb_set_transport_header(struct sk_buff *skb,
2116 const int offset)
2117{
2118 skb_reset_transport_header(skb);
2119 skb->transport_header += offset;
2120}
2121
2122static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2123{
2124 return skb->head + skb->network_header;
2125}
2126
2127static inline void skb_reset_network_header(struct sk_buff *skb)
2128{
2129 skb->network_header = skb->data - skb->head;
2130}
2131
2132static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2133{
2134 skb_reset_network_header(skb);
2135 skb->network_header += offset;
2136}
2137
2138static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2139{
2140 return skb->head + skb->mac_header;
2141}
2142
Amir Vadaiea6da4f2017-02-07 09:56:06 +02002143static inline int skb_mac_offset(const struct sk_buff *skb)
2144{
2145 return skb_mac_header(skb) - skb->data;
2146}
2147
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002148static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2149{
Cong Wang35d04612013-05-29 15:16:05 +08002150 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002151}
2152
2153static inline void skb_reset_mac_header(struct sk_buff *skb)
2154{
2155 skb->mac_header = skb->data - skb->head;
2156}
2157
2158static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2159{
2160 skb_reset_mac_header(skb);
2161 skb->mac_header += offset;
2162}
2163
Timo Teräs0e3da5b2013-12-16 11:02:09 +02002164static inline void skb_pop_mac_header(struct sk_buff *skb)
2165{
2166 skb->mac_header = skb->network_header;
2167}
2168
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002169static inline void skb_probe_transport_header(struct sk_buff *skb,
2170 const int offset_hint)
2171{
2172 struct flow_keys keys;
2173
2174 if (skb_transport_header_was_set(skb))
2175 return;
Tom Herbertcd79a232015-09-01 09:24:27 -07002176 else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
Tom Herbert42aecaa2015-06-04 09:16:39 -07002177 skb_set_transport_header(skb, keys.control.thoff);
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002178 else
2179 skb_set_transport_header(skb, offset_hint);
2180}
2181
Eric Dumazet03606892012-02-23 10:55:02 +00002182static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2183{
2184 if (skb_mac_header_was_set(skb)) {
2185 const unsigned char *old_mac = skb_mac_header(skb);
2186
2187 skb_set_mac_header(skb, -skb->mac_len);
2188 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2189 }
2190}
2191
Michał Mirosław04fb4512010-12-14 15:24:08 +00002192static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2193{
2194 return skb->csum_start - skb_headroom(skb);
2195}
2196
Alexander Duyck08b64fc2016-02-05 15:27:49 -08002197static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2198{
2199 return skb->head + skb->csum_start;
2200}
2201
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002202static inline int skb_transport_offset(const struct sk_buff *skb)
2203{
2204 return skb_transport_header(skb) - skb->data;
2205}
2206
2207static inline u32 skb_network_header_len(const struct sk_buff *skb)
2208{
2209 return skb->transport_header - skb->network_header;
2210}
2211
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002212static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2213{
2214 return skb->inner_transport_header - skb->inner_network_header;
2215}
2216
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002217static inline int skb_network_offset(const struct sk_buff *skb)
2218{
2219 return skb_network_header(skb) - skb->data;
2220}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03002221
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002222static inline int skb_inner_network_offset(const struct sk_buff *skb)
2223{
2224 return skb_inner_network_header(skb) - skb->data;
2225}
2226
Changli Gaof9599ce2010-08-04 04:43:44 +00002227static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2228{
2229 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2230}
2231
Linus Torvalds1da177e2005-04-16 15:20:36 -07002232/*
2233 * CPUs often take a performance hit when accessing unaligned memory
2234 * locations. The actual performance hit varies, it can be small if the
2235 * hardware handles it or large if we have to take an exception and fix it
2236 * in software.
2237 *
2238 * Since an ethernet header is 14 bytes network drivers often end up with
2239 * the IP header at an unaligned offset. The IP header can be aligned by
2240 * shifting the start of the packet by 2 bytes. Drivers should do this
2241 * with:
2242 *
Tobias Klauser8660c122009-07-13 22:48:16 +00002243 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002244 *
2245 * The downside to this alignment of the IP header is that the DMA is now
2246 * unaligned. On some architectures the cost of an unaligned DMA is high
2247 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00002248 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002249 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2250 * to be overridden.
2251 */
2252#ifndef NET_IP_ALIGN
2253#define NET_IP_ALIGN 2
2254#endif
2255
Anton Blanchard025be812006-03-31 02:27:06 -08002256/*
2257 * The networking layer reserves some headroom in skb data (via
2258 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2259 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08002260 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08002261 *
2262 * Unfortunately this headroom changes the DMA alignment of the resulting
2263 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2264 * on some architectures. An architecture can override this value,
2265 * perhaps setting it to a cacheline in size (since that will maintain
2266 * cacheline alignment of the DMA). It must be a power of 2.
2267 *
David S. Millerd6301d32009-02-08 19:24:13 -08002268 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08002269 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07002270 *
2271 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2272 * to reduce average number of cache lines per packet.
2273 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07002274 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08002275 */
2276#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07002277#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08002278#endif
2279
Joe Perches7965bd42013-09-26 14:48:15 -07002280int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002281
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002282static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002283{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00002284 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07002285 WARN_ON(1);
2286 return;
2287 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002288 skb->len = len;
2289 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002290}
2291
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002292static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2293{
2294 __skb_set_length(skb, len);
2295}
2296
Joe Perches7965bd42013-09-26 14:48:15 -07002297void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002298
2299static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2300{
Herbert Xu3cc0e872006-06-09 16:13:38 -07002301 if (skb->data_len)
2302 return ___pskb_trim(skb, len);
2303 __skb_trim(skb, len);
2304 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002305}
2306
2307static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2308{
2309 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2310}
2311
2312/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07002313 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2314 * @skb: buffer to alter
2315 * @len: new length
2316 *
2317 * This is identical to pskb_trim except that the caller knows that
2318 * the skb is not cloned so we should never get an error due to out-
2319 * of-memory.
2320 */
2321static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2322{
2323 int err = pskb_trim(skb, len);
2324 BUG_ON(err);
2325}
2326
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002327static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2328{
2329 unsigned int diff = len - skb->len;
2330
2331 if (skb_tailroom(skb) < diff) {
2332 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2333 GFP_ATOMIC);
2334 if (ret)
2335 return ret;
2336 }
2337 __skb_set_length(skb, len);
2338 return 0;
2339}
2340
Herbert Xue9fa4f72006-08-13 20:12:58 -07002341/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002342 * skb_orphan - orphan a buffer
2343 * @skb: buffer to orphan
2344 *
2345 * If a buffer currently has an owner then we call the owner's
2346 * destructor function and make the @skb unowned. The buffer continues
2347 * to exist but is no longer charged to its former owner.
2348 */
2349static inline void skb_orphan(struct sk_buff *skb)
2350{
Eric Dumazetc34a7612013-07-30 16:11:15 -07002351 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002352 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002353 skb->destructor = NULL;
2354 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07002355 } else {
2356 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002357 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002358}
2359
2360/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002361 * skb_orphan_frags - orphan the frags contained in a buffer
2362 * @skb: buffer to orphan frags from
2363 * @gfp_mask: allocation mask for replacement pages
2364 *
2365 * For each frag in the SKB which needs a destructor (i.e. has an
2366 * owner) create a copy of that frag and release the original
2367 * page by calling the destructor.
2368 */
2369static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2370{
2371 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2372 return 0;
2373 return skb_copy_ubufs(skb, gfp_mask);
2374}
2375
2376/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002377 * __skb_queue_purge - empty a list
2378 * @list: list to empty
2379 *
2380 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2381 * the list and one reference dropped. This function does not take the
2382 * list lock and the caller must hold the relevant locks to use it.
2383 */
Joe Perches7965bd42013-09-26 14:48:15 -07002384void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002385static inline void __skb_queue_purge(struct sk_buff_head *list)
2386{
2387 struct sk_buff *skb;
2388 while ((skb = __skb_dequeue(list)) != NULL)
2389 kfree_skb(skb);
2390}
2391
Yaogong Wang9f5afea2016-09-07 14:49:28 -07002392void skb_rbtree_purge(struct rb_root *root);
2393
Joe Perches7965bd42013-09-26 14:48:15 -07002394void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002395
Joe Perches7965bd42013-09-26 14:48:15 -07002396struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2397 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07002398
2399/**
2400 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2401 * @dev: network device to receive on
2402 * @length: length to allocate
2403 *
2404 * Allocate a new &sk_buff and assign it a usage count of one. The
2405 * buffer has unspecified headroom built in. Users should allocate
2406 * the headroom they think they need without accounting for the
2407 * built in space. The built in space is used for optimisations.
2408 *
2409 * %NULL is returned if there is no free memory. Although this function
2410 * allocates memory it can be called from an interrupt.
2411 */
2412static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002413 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002414{
2415 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2416}
2417
Eric Dumazet6f532612012-05-18 05:12:12 +00002418/* legacy helper around __netdev_alloc_skb() */
2419static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2420 gfp_t gfp_mask)
2421{
2422 return __netdev_alloc_skb(NULL, length, gfp_mask);
2423}
2424
2425/* legacy helper around netdev_alloc_skb() */
2426static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2427{
2428 return netdev_alloc_skb(NULL, length);
2429}
2430
2431
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002432static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2433 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002434{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002435 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002436
2437 if (NET_IP_ALIGN && skb)
2438 skb_reserve(skb, NET_IP_ALIGN);
2439 return skb;
2440}
2441
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002442static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2443 unsigned int length)
2444{
2445 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2446}
2447
Alexander Duyck181edb22015-05-06 21:12:03 -07002448static inline void skb_free_frag(void *addr)
2449{
Alexander Duyck8c2dd3e2017-01-10 16:58:06 -08002450 page_frag_free(addr);
Alexander Duyck181edb22015-05-06 21:12:03 -07002451}
2452
Alexander Duyckffde7322014-12-09 19:40:42 -08002453void *napi_alloc_frag(unsigned int fragsz);
Alexander Duyckfd11a832014-12-09 19:40:49 -08002454struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2455 unsigned int length, gfp_t gfp_mask);
2456static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2457 unsigned int length)
2458{
2459 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2460}
Jesper Dangaard Brouer795bb1c2016-02-08 13:14:59 +01002461void napi_consume_skb(struct sk_buff *skb, int budget);
2462
2463void __kfree_skb_flush(void);
Jesper Dangaard Brouer15fad712016-02-08 13:15:04 +01002464void __kfree_skb_defer(struct sk_buff *skb);
Alexander Duyckffde7322014-12-09 19:40:42 -08002465
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002466/**
Alexander Duyck71dfda52014-11-11 09:26:34 -08002467 * __dev_alloc_pages - allocate page for network Rx
2468 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2469 * @order: size of the allocation
2470 *
2471 * Allocate a new page.
2472 *
2473 * %NULL is returned if there is no free memory.
2474*/
2475static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2476 unsigned int order)
2477{
2478 /* This piece of code contains several assumptions.
2479 * 1. This is for device Rx, therefor a cold page is preferred.
2480 * 2. The expectation is the user wants a compound page.
2481 * 3. If requesting a order 0 page it will not be compound
2482 * due to the check to see if order has a value in prep_new_page
2483 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2484 * code in gfp_to_alloc_flags that should be enforcing this.
2485 */
2486 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2487
2488 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2489}
2490
2491static inline struct page *dev_alloc_pages(unsigned int order)
2492{
Neil Horman95829b32016-05-19 11:30:54 -04002493 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002494}
2495
2496/**
2497 * __dev_alloc_page - allocate a page for network Rx
2498 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2499 *
2500 * Allocate a new page.
2501 *
2502 * %NULL is returned if there is no free memory.
2503 */
2504static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2505{
2506 return __dev_alloc_pages(gfp_mask, 0);
2507}
2508
2509static inline struct page *dev_alloc_page(void)
2510{
Neil Horman95829b32016-05-19 11:30:54 -04002511 return dev_alloc_pages(0);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002512}
2513
2514/**
Mel Gorman06140022012-07-31 16:44:24 -07002515 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2516 * @page: The page that was allocated from skb_alloc_page
2517 * @skb: The skb that may need pfmemalloc set
2518 */
2519static inline void skb_propagate_pfmemalloc(struct page *page,
2520 struct sk_buff *skb)
2521{
Michal Hocko2f064f32015-08-21 14:11:51 -07002522 if (page_is_pfmemalloc(page))
Mel Gorman06140022012-07-31 16:44:24 -07002523 skb->pfmemalloc = true;
2524}
2525
Eric Dumazet564824b2010-10-11 19:05:25 +00002526/**
Masanari Iidae2278672014-02-18 22:54:36 +09002527 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002528 * @frag: the paged fragment
2529 *
2530 * Returns the &struct page associated with @frag.
2531 */
2532static inline struct page *skb_frag_page(const skb_frag_t *frag)
2533{
Ian Campbella8605c62011-10-19 23:01:49 +00002534 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00002535}
2536
2537/**
2538 * __skb_frag_ref - take an addition reference on a paged fragment.
2539 * @frag: the paged fragment
2540 *
2541 * Takes an additional reference on the paged fragment @frag.
2542 */
2543static inline void __skb_frag_ref(skb_frag_t *frag)
2544{
2545 get_page(skb_frag_page(frag));
2546}
2547
2548/**
2549 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2550 * @skb: the buffer
2551 * @f: the fragment offset.
2552 *
2553 * Takes an additional reference on the @f'th paged fragment of @skb.
2554 */
2555static inline void skb_frag_ref(struct sk_buff *skb, int f)
2556{
2557 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2558}
2559
2560/**
2561 * __skb_frag_unref - release a reference on a paged fragment.
2562 * @frag: the paged fragment
2563 *
2564 * Releases a reference on the paged fragment @frag.
2565 */
2566static inline void __skb_frag_unref(skb_frag_t *frag)
2567{
2568 put_page(skb_frag_page(frag));
2569}
2570
2571/**
2572 * skb_frag_unref - release a reference on a paged fragment of an skb.
2573 * @skb: the buffer
2574 * @f: the fragment offset
2575 *
2576 * Releases a reference on the @f'th paged fragment of @skb.
2577 */
2578static inline void skb_frag_unref(struct sk_buff *skb, int f)
2579{
2580 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2581}
2582
2583/**
2584 * skb_frag_address - gets the address of the data contained in a paged fragment
2585 * @frag: the paged fragment buffer
2586 *
2587 * Returns the address of the data within @frag. The page must already
2588 * be mapped.
2589 */
2590static inline void *skb_frag_address(const skb_frag_t *frag)
2591{
2592 return page_address(skb_frag_page(frag)) + frag->page_offset;
2593}
2594
2595/**
2596 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2597 * @frag: the paged fragment buffer
2598 *
2599 * Returns the address of the data within @frag. Checks that the page
2600 * is mapped and returns %NULL otherwise.
2601 */
2602static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2603{
2604 void *ptr = page_address(skb_frag_page(frag));
2605 if (unlikely(!ptr))
2606 return NULL;
2607
2608 return ptr + frag->page_offset;
2609}
2610
2611/**
2612 * __skb_frag_set_page - sets the page contained in a paged fragment
2613 * @frag: the paged fragment
2614 * @page: the page to set
2615 *
2616 * Sets the fragment @frag to contain @page.
2617 */
2618static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2619{
Ian Campbella8605c62011-10-19 23:01:49 +00002620 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002621}
2622
2623/**
2624 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2625 * @skb: the buffer
2626 * @f: the fragment offset
2627 * @page: the page to set
2628 *
2629 * Sets the @f'th fragment of @skb to contain @page.
2630 */
2631static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2632 struct page *page)
2633{
2634 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2635}
2636
Eric Dumazet400dfd32013-10-17 16:27:07 -07002637bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2638
Ian Campbell131ea662011-08-19 06:25:00 +00002639/**
2640 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002641 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002642 * @frag: the paged fragment to map
2643 * @offset: the offset within the fragment (starting at the
2644 * fragment's own offset)
2645 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002646 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00002647 *
2648 * Maps the page associated with @frag to @device.
2649 */
2650static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2651 const skb_frag_t *frag,
2652 size_t offset, size_t size,
2653 enum dma_data_direction dir)
2654{
2655 return dma_map_page(dev, skb_frag_page(frag),
2656 frag->page_offset + offset, size, dir);
2657}
2658
Eric Dumazet117632e2011-12-03 21:39:53 +00002659static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2660 gfp_t gfp_mask)
2661{
2662 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2663}
2664
Octavian Purdilabad93e92014-06-12 01:36:26 +03002665
2666static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2667 gfp_t gfp_mask)
2668{
2669 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2670}
2671
2672
Ian Campbell131ea662011-08-19 06:25:00 +00002673/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002674 * skb_clone_writable - is the header of a clone writable
2675 * @skb: buffer to check
2676 * @len: length up to which to write
2677 *
2678 * Returns true if modifying the header part of the cloned buffer
2679 * does not requires the data to be copied.
2680 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002681static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002682{
2683 return !skb_header_cloned(skb) &&
2684 skb_headroom(skb) + len <= skb->hdr_len;
2685}
2686
Daniel Borkmann36976492016-02-19 23:05:25 +01002687static inline int skb_try_make_writable(struct sk_buff *skb,
2688 unsigned int write_len)
2689{
2690 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
2691 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
2692}
2693
Herbert Xud9cc2042007-09-16 16:21:16 -07002694static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2695 int cloned)
2696{
2697 int delta = 0;
2698
Herbert Xud9cc2042007-09-16 16:21:16 -07002699 if (headroom > skb_headroom(skb))
2700 delta = headroom - skb_headroom(skb);
2701
2702 if (delta || cloned)
2703 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2704 GFP_ATOMIC);
2705 return 0;
2706}
2707
Patrick McHardy334a8132007-06-25 04:35:20 -07002708/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002709 * skb_cow - copy header of skb when it is required
2710 * @skb: buffer to cow
2711 * @headroom: needed headroom
2712 *
2713 * If the skb passed lacks sufficient headroom or its data part
2714 * is shared, data is reallocated. If reallocation fails, an error
2715 * is returned and original skb is not changed.
2716 *
2717 * The result is skb with writable area skb->head...skb->tail
2718 * and at least @headroom of space at head.
2719 */
2720static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2721{
Herbert Xud9cc2042007-09-16 16:21:16 -07002722 return __skb_cow(skb, headroom, skb_cloned(skb));
2723}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002724
Herbert Xud9cc2042007-09-16 16:21:16 -07002725/**
2726 * skb_cow_head - skb_cow but only making the head writable
2727 * @skb: buffer to cow
2728 * @headroom: needed headroom
2729 *
2730 * This function is identical to skb_cow except that we replace the
2731 * skb_cloned check by skb_header_cloned. It should be used when
2732 * you only need to push on some header and do not need to modify
2733 * the data.
2734 */
2735static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2736{
2737 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002738}
2739
2740/**
2741 * skb_padto - pad an skbuff up to a minimal size
2742 * @skb: buffer to pad
2743 * @len: minimal length
2744 *
2745 * Pads up a buffer to ensure the trailing bytes exist and are
2746 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002747 * is untouched. Otherwise it is extended. Returns zero on
2748 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002749 */
Herbert Xu5b057c62006-06-23 02:06:41 -07002750static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002751{
2752 unsigned int size = skb->len;
2753 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002754 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002755 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002756}
2757
Alexander Duyck9c0c1122014-12-03 08:17:33 -08002758/**
2759 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2760 * @skb: buffer to pad
2761 * @len: minimal length
2762 *
2763 * Pads up a buffer to ensure the trailing bytes exist and are
2764 * blanked. If the buffer already contains sufficient data it
2765 * is untouched. Otherwise it is extended. Returns zero on
2766 * success. The skb is freed on error.
2767 */
2768static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2769{
2770 unsigned int size = skb->len;
2771
2772 if (unlikely(size < len)) {
2773 len -= size;
2774 if (skb_pad(skb, len))
2775 return -ENOMEM;
2776 __skb_put(skb, len);
2777 }
2778 return 0;
2779}
2780
Linus Torvalds1da177e2005-04-16 15:20:36 -07002781static inline int skb_add_data(struct sk_buff *skb,
Al Viroaf2b0402014-11-27 21:44:24 -05002782 struct iov_iter *from, int copy)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002783{
2784 const int off = skb->len;
2785
2786 if (skb->ip_summed == CHECKSUM_NONE) {
Al Viroaf2b0402014-11-27 21:44:24 -05002787 __wsum csum = 0;
Al Viro15e6cb42016-11-01 22:42:45 -04002788 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
2789 &csum, from)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002790 skb->csum = csum_block_add(skb->csum, csum, off);
2791 return 0;
2792 }
Al Viro15e6cb42016-11-01 22:42:45 -04002793 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002794 return 0;
2795
2796 __skb_trim(skb, off);
2797 return -EFAULT;
2798}
2799
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002800static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2801 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002802{
2803 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002804 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002805
Ian Campbellea2ab692011-08-22 23:44:58 +00002806 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002807 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002808 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002809 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002810}
2811
Herbert Xu364c6ba2006-06-09 16:10:40 -07002812static inline int __skb_linearize(struct sk_buff *skb)
2813{
2814 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2815}
2816
Linus Torvalds1da177e2005-04-16 15:20:36 -07002817/**
2818 * skb_linearize - convert paged skb to linear one
2819 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002820 *
2821 * If there is no free memory -ENOMEM is returned, otherwise zero
2822 * is returned and the old skb data released.
2823 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002824static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002825{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002826 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2827}
2828
2829/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00002830 * skb_has_shared_frag - can any frag be overwritten
2831 * @skb: buffer to test
2832 *
2833 * Return true if the skb has at least one frag that might be modified
2834 * by an external entity (as in vmsplice()/sendfile())
2835 */
2836static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2837{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00002838 return skb_is_nonlinear(skb) &&
2839 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00002840}
2841
2842/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07002843 * skb_linearize_cow - make sure skb is linear and writable
2844 * @skb: buffer to process
2845 *
2846 * If there is no free memory -ENOMEM is returned, otherwise zero
2847 * is returned and the old skb data released.
2848 */
2849static inline int skb_linearize_cow(struct sk_buff *skb)
2850{
2851 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2852 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002853}
2854
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02002855static __always_inline void
2856__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
2857 unsigned int off)
2858{
2859 if (skb->ip_summed == CHECKSUM_COMPLETE)
2860 skb->csum = csum_block_sub(skb->csum,
2861 csum_partial(start, len, 0), off);
2862 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
2863 skb_checksum_start_offset(skb) < 0)
2864 skb->ip_summed = CHECKSUM_NONE;
2865}
2866
Linus Torvalds1da177e2005-04-16 15:20:36 -07002867/**
2868 * skb_postpull_rcsum - update checksum for received skb after pull
2869 * @skb: buffer to update
2870 * @start: start of data before pull
2871 * @len: length of data pulled
2872 *
2873 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002874 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2875 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002876 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002877static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002878 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002879{
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02002880 __skb_postpull_rcsum(skb, start, len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002881}
2882
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02002883static __always_inline void
2884__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
2885 unsigned int off)
2886{
2887 if (skb->ip_summed == CHECKSUM_COMPLETE)
2888 skb->csum = csum_block_add(skb->csum,
2889 csum_partial(start, len, 0), off);
2890}
Herbert Xucbb042f2006-03-20 22:43:56 -08002891
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02002892/**
2893 * skb_postpush_rcsum - update checksum for received skb after push
2894 * @skb: buffer to update
2895 * @start: start of data after push
2896 * @len: length of data pushed
2897 *
2898 * After doing a push on a received packet, you need to call this to
2899 * update the CHECKSUM_COMPLETE checksum.
2900 */
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01002901static inline void skb_postpush_rcsum(struct sk_buff *skb,
2902 const void *start, unsigned int len)
2903{
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02002904 __skb_postpush_rcsum(skb, start, len, 0);
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01002905}
2906
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02002907unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2908
David S. Miller7ce5a272013-12-02 17:26:05 -05002909/**
WANG Cong82a31b92016-06-30 10:15:22 -07002910 * skb_push_rcsum - push skb and update receive checksum
2911 * @skb: buffer to update
2912 * @len: length of data pulled
2913 *
2914 * This function performs an skb_push on the packet and updates
2915 * the CHECKSUM_COMPLETE checksum. It should be used on
2916 * receive path processing instead of skb_push unless you know
2917 * that the checksum difference is zero (e.g., a valid IP header)
2918 * or you are setting ip_summed to CHECKSUM_NONE.
2919 */
2920static inline unsigned char *skb_push_rcsum(struct sk_buff *skb,
2921 unsigned int len)
2922{
2923 skb_push(skb, len);
2924 skb_postpush_rcsum(skb, skb->data, len);
2925 return skb->data;
2926}
2927
2928/**
David S. Miller7ce5a272013-12-02 17:26:05 -05002929 * pskb_trim_rcsum - trim received skb and update checksum
2930 * @skb: buffer to trim
2931 * @len: new length
2932 *
2933 * This is exactly the same as pskb_trim except that it ensures the
2934 * checksum of received packets are still valid after the operation.
2935 */
2936
2937static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2938{
2939 if (likely(len >= skb->len))
2940 return 0;
2941 if (skb->ip_summed == CHECKSUM_COMPLETE)
2942 skb->ip_summed = CHECKSUM_NONE;
2943 return __pskb_trim(skb, len);
2944}
2945
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002946static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2947{
2948 if (skb->ip_summed == CHECKSUM_COMPLETE)
2949 skb->ip_summed = CHECKSUM_NONE;
2950 __skb_trim(skb, len);
2951 return 0;
2952}
2953
2954static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
2955{
2956 if (skb->ip_summed == CHECKSUM_COMPLETE)
2957 skb->ip_summed = CHECKSUM_NONE;
2958 return __skb_grow(skb, len);
2959}
2960
Linus Torvalds1da177e2005-04-16 15:20:36 -07002961#define skb_queue_walk(queue, skb) \
2962 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002963 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002964 skb = skb->next)
2965
James Chapman46f89142007-04-30 00:07:31 -07002966#define skb_queue_walk_safe(queue, skb, tmp) \
2967 for (skb = (queue)->next, tmp = skb->next; \
2968 skb != (struct sk_buff *)(queue); \
2969 skb = tmp, tmp = skb->next)
2970
David S. Miller1164f522008-09-23 00:49:44 -07002971#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002972 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002973 skb = skb->next)
2974
2975#define skb_queue_walk_from_safe(queue, skb, tmp) \
2976 for (tmp = skb->next; \
2977 skb != (struct sk_buff *)(queue); \
2978 skb = tmp, tmp = skb->next)
2979
Stephen Hemminger300ce172005-10-30 13:47:34 -08002980#define skb_queue_reverse_walk(queue, skb) \
2981 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002982 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08002983 skb = skb->prev)
2984
David S. Miller686a2952011-01-20 22:47:32 -08002985#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2986 for (skb = (queue)->prev, tmp = skb->prev; \
2987 skb != (struct sk_buff *)(queue); \
2988 skb = tmp, tmp = skb->prev)
2989
2990#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2991 for (tmp = skb->prev; \
2992 skb != (struct sk_buff *)(queue); \
2993 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002994
David S. Miller21dc3302010-08-23 00:13:46 -07002995static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07002996{
2997 return skb_shinfo(skb)->frag_list != NULL;
2998}
2999
3000static inline void skb_frag_list_init(struct sk_buff *skb)
3001{
3002 skb_shinfo(skb)->frag_list = NULL;
3003}
3004
David S. Milleree039872009-06-09 00:17:13 -07003005#define skb_walk_frags(skb, iter) \
3006 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3007
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003008
3009int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
3010 const struct sk_buff *skb);
Paolo Abeni65101ae2017-05-16 11:20:13 +02003011struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3012 struct sk_buff_head *queue,
3013 unsigned int flags,
3014 void (*destructor)(struct sock *sk,
3015 struct sk_buff *skb),
3016 int *peeked, int *off, int *err,
3017 struct sk_buff **last);
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003018struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
Paolo Abeni7c13f972016-11-04 11:28:59 +01003019 void (*destructor)(struct sock *sk,
3020 struct sk_buff *skb),
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003021 int *peeked, int *off, int *err,
3022 struct sk_buff **last);
Joe Perches7965bd42013-09-26 14:48:15 -07003023struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
Paolo Abeni7c13f972016-11-04 11:28:59 +01003024 void (*destructor)(struct sock *sk,
3025 struct sk_buff *skb),
Joe Perches7965bd42013-09-26 14:48:15 -07003026 int *peeked, int *off, int *err);
3027struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3028 int *err);
3029unsigned int datagram_poll(struct file *file, struct socket *sock,
3030 struct poll_table_struct *wait);
Al Viroc0371da2014-11-24 10:42:55 -05003031int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3032 struct iov_iter *to, int size);
David S. Miller51f3d022014-11-05 16:46:40 -05003033static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3034 struct msghdr *msg, int size)
3035{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003036 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
David S. Miller51f3d022014-11-05 16:46:40 -05003037}
Al Viroe5a4b0b2014-11-24 18:17:55 -05003038int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3039 struct msghdr *msg);
Al Viro3a654f92014-06-19 14:15:22 -04003040int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3041 struct iov_iter *from, int len);
Al Viro3a654f92014-06-19 14:15:22 -04003042int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
Joe Perches7965bd42013-09-26 14:48:15 -07003043void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
samanthakumar627d2d62016-04-05 12:41:16 -04003044void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3045static inline void skb_free_datagram_locked(struct sock *sk,
3046 struct sk_buff *skb)
3047{
3048 __skb_free_datagram_locked(sk, skb, 0);
3049}
Joe Perches7965bd42013-09-26 14:48:15 -07003050int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07003051int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3052int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3053__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3054 int len, __wsum csum);
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +02003055int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
Joe Perches7965bd42013-09-26 14:48:15 -07003056 struct pipe_inode_info *pipe, unsigned int len,
Al Viro25869262016-09-17 21:02:10 -04003057 unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07003058void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01003059unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00003060int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3061 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07003062void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3063int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3064void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Florian Westphalde960aa2014-01-26 10:58:16 +01003065unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
Marcelo Ricardo Leitnerae7ef812016-06-02 15:05:41 -03003066bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu);
Joe Perches7965bd42013-09-26 14:48:15 -07003067struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Vlad Yasevich0d5501c2014-08-08 14:42:13 -04003068struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
Jiri Pirkoe2195122014-11-19 14:05:01 +01003069int skb_ensure_writable(struct sk_buff *skb, int write_len);
Shmulik Ladkanibfca4c52016-09-19 19:11:09 +03003070int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
Jiri Pirko93515d52014-11-19 14:05:02 +01003071int skb_vlan_pop(struct sk_buff *skb);
3072int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
Sowmini Varadhan6fa01cc2016-04-22 18:36:35 -07003073struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3074 gfp_t gfp);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003075
Al Viro6ce8e9c2014-04-06 21:25:44 -04003076static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3077{
Al Viro3073f072017-02-17 23:13:25 -05003078 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
Al Viro6ce8e9c2014-04-06 21:25:44 -04003079}
3080
Al Viro7eab8d92014-04-06 21:51:23 -04003081static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3082{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003083 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro7eab8d92014-04-06 21:51:23 -04003084}
3085
Daniel Borkmann2817a332013-10-30 11:50:51 +01003086struct skb_checksum_ops {
3087 __wsum (*update)(const void *mem, int len, __wsum wsum);
3088 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3089};
3090
Davide Caratti96178132017-05-18 15:44:37 +02003091extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3092
Daniel Borkmann2817a332013-10-30 11:50:51 +01003093__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3094 __wsum csum, const struct skb_checksum_ops *ops);
3095__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3096 __wsum csum);
3097
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003098static inline void * __must_check
3099__skb_header_pointer(const struct sk_buff *skb, int offset,
3100 int len, void *data, int hlen, void *buffer)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003101{
Patrick McHardy55820ee2005-07-05 14:08:10 -07003102 if (hlen - offset >= len)
David S. Miller690e36e2014-08-23 12:13:41 -07003103 return data + offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003104
David S. Miller690e36e2014-08-23 12:13:41 -07003105 if (!skb ||
3106 skb_copy_bits(skb, offset, buffer, len) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003107 return NULL;
3108
3109 return buffer;
3110}
3111
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003112static inline void * __must_check
3113skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
David S. Miller690e36e2014-08-23 12:13:41 -07003114{
3115 return __skb_header_pointer(skb, offset, len, skb->data,
3116 skb_headlen(skb), buffer);
3117}
3118
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01003119/**
3120 * skb_needs_linearize - check if we need to linearize a given skb
3121 * depending on the given device features.
3122 * @skb: socket buffer to check
3123 * @features: net device features
3124 *
3125 * Returns true if either:
3126 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3127 * 2. skb is fragmented and the device does not support SG.
3128 */
3129static inline bool skb_needs_linearize(struct sk_buff *skb,
3130 netdev_features_t features)
3131{
3132 return skb_is_nonlinear(skb) &&
3133 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3134 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3135}
3136
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03003137static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3138 void *to,
3139 const unsigned int len)
3140{
3141 memcpy(to, skb->data, len);
3142}
3143
3144static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3145 const int offset, void *to,
3146 const unsigned int len)
3147{
3148 memcpy(to, skb->data + offset, len);
3149}
3150
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03003151static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3152 const void *from,
3153 const unsigned int len)
3154{
3155 memcpy(skb->data, from, len);
3156}
3157
3158static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3159 const int offset,
3160 const void *from,
3161 const unsigned int len)
3162{
3163 memcpy(skb->data + offset, from, len);
3164}
3165
Joe Perches7965bd42013-09-26 14:48:15 -07003166void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003167
Patrick Ohlyac45f602009-02-12 05:03:37 +00003168static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3169{
3170 return skb->tstamp;
3171}
3172
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003173/**
3174 * skb_get_timestamp - get timestamp from a skb
3175 * @skb: skb to get stamp from
3176 * @stamp: pointer to struct timeval to store stamp in
3177 *
3178 * Timestamps are stored in the skb as offsets to a base timestamp.
3179 * This function converts the offset back to a struct timeval and stores
3180 * it in stamp.
3181 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00003182static inline void skb_get_timestamp(const struct sk_buff *skb,
3183 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003184{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003185 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003186}
3187
Patrick Ohlyac45f602009-02-12 05:03:37 +00003188static inline void skb_get_timestampns(const struct sk_buff *skb,
3189 struct timespec *stamp)
3190{
3191 *stamp = ktime_to_timespec(skb->tstamp);
3192}
3193
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003194static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003195{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003196 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003197}
3198
Stephen Hemminger164891a2007-04-23 22:26:16 -07003199static inline ktime_t net_timedelta(ktime_t t)
3200{
3201 return ktime_sub(ktime_get_real(), t);
3202}
3203
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003204static inline ktime_t net_invalid_timestamp(void)
3205{
Thomas Gleixner8b0e1952016-12-25 12:30:41 +01003206 return 0;
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003207}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003208
Alexander Duyck62bccb82014-09-04 13:31:35 -04003209struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3210
Richard Cochranc1f19b52010-07-17 08:49:36 +00003211#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3212
Joe Perches7965bd42013-09-26 14:48:15 -07003213void skb_clone_tx_timestamp(struct sk_buff *skb);
3214bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00003215
3216#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3217
3218static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3219{
3220}
3221
3222static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3223{
3224 return false;
3225}
3226
3227#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3228
3229/**
3230 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3231 *
Richard Cochranda92b192011-10-21 00:49:15 +00003232 * PHY drivers may accept clones of transmitted packets for
3233 * timestamping via their phy_driver.txtstamp method. These drivers
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003234 * must call this function to return the skb back to the stack with a
3235 * timestamp.
Richard Cochranda92b192011-10-21 00:49:15 +00003236 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00003237 * @skb: clone of the the original outgoing packet
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003238 * @hwtstamps: hardware time stamps
Richard Cochranc1f19b52010-07-17 08:49:36 +00003239 *
3240 */
3241void skb_complete_tx_timestamp(struct sk_buff *skb,
3242 struct skb_shared_hwtstamps *hwtstamps);
3243
Willem de Bruijne7fd2882014-08-04 22:11:48 -04003244void __skb_tstamp_tx(struct sk_buff *orig_skb,
3245 struct skb_shared_hwtstamps *hwtstamps,
3246 struct sock *sk, int tstype);
3247
Patrick Ohlyac45f602009-02-12 05:03:37 +00003248/**
3249 * skb_tstamp_tx - queue clone of skb with send time stamps
3250 * @orig_skb: the original outgoing packet
3251 * @hwtstamps: hardware time stamps, may be NULL if not available
3252 *
3253 * If the skb has a socket associated, then this function clones the
3254 * skb (thus sharing the actual data and optional structures), stores
3255 * the optional hardware time stamping information (if non NULL) or
3256 * generates a software time stamp (otherwise), then queues the clone
3257 * to the error queue of the socket. Errors are silently ignored.
3258 */
Joe Perches7965bd42013-09-26 14:48:15 -07003259void skb_tstamp_tx(struct sk_buff *orig_skb,
3260 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00003261
Richard Cochran4507a712010-07-17 08:48:28 +00003262/**
3263 * skb_tx_timestamp() - Driver hook for transmit timestamping
3264 *
3265 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00003266 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00003267 *
David S. Miller73409f32013-12-27 13:04:33 -05003268 * Specifically, one should make absolutely sure that this function is
3269 * called before TX completion of this packet can trigger. Otherwise
3270 * the packet could potentially already be freed.
3271 *
Richard Cochran4507a712010-07-17 08:48:28 +00003272 * @skb: A socket buffer.
3273 */
3274static inline void skb_tx_timestamp(struct sk_buff *skb)
3275{
Richard Cochranc1f19b52010-07-17 08:49:36 +00003276 skb_clone_tx_timestamp(skb);
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +02003277 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3278 skb_tstamp_tx(skb, NULL);
Richard Cochran4507a712010-07-17 08:48:28 +00003279}
3280
Johannes Berg6e3e9392011-11-09 10:15:42 +01003281/**
3282 * skb_complete_wifi_ack - deliver skb with wifi status
3283 *
3284 * @skb: the original outgoing packet
3285 * @acked: ack status
3286 *
3287 */
3288void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3289
Joe Perches7965bd42013-09-26 14:48:15 -07003290__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3291__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003292
Herbert Xu60476372007-04-09 11:59:39 -07003293static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3294{
Tom Herbert6edec0e2015-02-10 16:30:28 -08003295 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3296 skb->csum_valid ||
3297 (skb->ip_summed == CHECKSUM_PARTIAL &&
3298 skb_checksum_start_offset(skb) >= 0));
Herbert Xu60476372007-04-09 11:59:39 -07003299}
3300
Herbert Xufb286bb2005-11-10 13:01:24 -08003301/**
3302 * skb_checksum_complete - Calculate checksum of an entire packet
3303 * @skb: packet to process
3304 *
3305 * This function calculates the checksum over the entire packet plus
3306 * the value of skb->csum. The latter can be used to supply the
3307 * checksum of a pseudo header as used by TCP/UDP. It returns the
3308 * checksum.
3309 *
3310 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3311 * this function can be used to verify that checksum on received
3312 * packets. In that case the function should return zero if the
3313 * checksum is correct. In particular, this function will return zero
3314 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3315 * hardware has already verified the correctness of the checksum.
3316 */
Al Viro4381ca32007-07-15 21:00:11 +01003317static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08003318{
Herbert Xu60476372007-04-09 11:59:39 -07003319 return skb_csum_unnecessary(skb) ?
3320 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003321}
3322
Tom Herbert77cffe22014-08-27 21:26:46 -07003323static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3324{
3325 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3326 if (skb->csum_level == 0)
3327 skb->ip_summed = CHECKSUM_NONE;
3328 else
3329 skb->csum_level--;
3330 }
3331}
3332
3333static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3334{
3335 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3336 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3337 skb->csum_level++;
3338 } else if (skb->ip_summed == CHECKSUM_NONE) {
3339 skb->ip_summed = CHECKSUM_UNNECESSARY;
3340 skb->csum_level = 0;
3341 }
3342}
3343
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003344/* Check if we need to perform checksum complete validation.
3345 *
3346 * Returns true if checksum complete is needed, false otherwise
3347 * (either checksum is unnecessary or zero checksum is allowed).
3348 */
3349static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3350 bool zero_okay,
3351 __sum16 check)
3352{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003353 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3354 skb->csum_valid = 1;
Tom Herbert77cffe22014-08-27 21:26:46 -07003355 __skb_decr_checksum_unnecessary(skb);
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003356 return false;
3357 }
3358
3359 return true;
3360}
3361
3362/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3363 * in checksum_init.
3364 */
3365#define CHECKSUM_BREAK 76
3366
Tom Herbert4e18b9a2015-04-20 14:10:04 -07003367/* Unset checksum-complete
3368 *
3369 * Unset checksum complete can be done when packet is being modified
3370 * (uncompressed for instance) and checksum-complete value is
3371 * invalidated.
3372 */
3373static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3374{
3375 if (skb->ip_summed == CHECKSUM_COMPLETE)
3376 skb->ip_summed = CHECKSUM_NONE;
3377}
3378
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003379/* Validate (init) checksum based on checksum complete.
3380 *
3381 * Return values:
3382 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3383 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3384 * checksum is stored in skb->csum for use in __skb_checksum_complete
3385 * non-zero: value of invalid checksum
3386 *
3387 */
3388static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3389 bool complete,
3390 __wsum psum)
3391{
3392 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3393 if (!csum_fold(csum_add(psum, skb->csum))) {
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003394 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003395 return 0;
3396 }
3397 }
3398
3399 skb->csum = psum;
3400
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003401 if (complete || skb->len <= CHECKSUM_BREAK) {
3402 __sum16 csum;
3403
3404 csum = __skb_checksum_complete(skb);
3405 skb->csum_valid = !csum;
3406 return csum;
3407 }
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003408
3409 return 0;
3410}
3411
3412static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3413{
3414 return 0;
3415}
3416
3417/* Perform checksum validate (init). Note that this is a macro since we only
3418 * want to calculate the pseudo header which is an input function if necessary.
3419 * First we try to validate without any computation (checksum unnecessary) and
3420 * then calculate based on checksum complete calling the function to compute
3421 * pseudo header.
3422 *
3423 * Return values:
3424 * 0: checksum is validated or try to in skb_checksum_complete
3425 * non-zero: value of invalid checksum
3426 */
3427#define __skb_checksum_validate(skb, proto, complete, \
3428 zero_okay, check, compute_pseudo) \
3429({ \
3430 __sum16 __ret = 0; \
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003431 skb->csum_valid = 0; \
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003432 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3433 __ret = __skb_checksum_validate_complete(skb, \
3434 complete, compute_pseudo(skb, proto)); \
3435 __ret; \
3436})
3437
3438#define skb_checksum_init(skb, proto, compute_pseudo) \
3439 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3440
3441#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3442 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3443
3444#define skb_checksum_validate(skb, proto, compute_pseudo) \
3445 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3446
3447#define skb_checksum_validate_zero_check(skb, proto, check, \
3448 compute_pseudo) \
Sabrina Dubroca096a4cf2015-02-06 18:54:19 +01003449 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003450
3451#define skb_checksum_simple_validate(skb) \
3452 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3453
Tom Herbertd96535a2014-08-31 15:12:42 -07003454static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3455{
Davide Caratti219f1d792017-05-18 15:44:39 +02003456 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
Tom Herbertd96535a2014-08-31 15:12:42 -07003457}
3458
3459static inline void __skb_checksum_convert(struct sk_buff *skb,
3460 __sum16 check, __wsum pseudo)
3461{
3462 skb->csum = ~pseudo;
3463 skb->ip_summed = CHECKSUM_COMPLETE;
3464}
3465
3466#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3467do { \
3468 if (__skb_checksum_convert_check(skb)) \
3469 __skb_checksum_convert(skb, check, \
3470 compute_pseudo(skb, proto)); \
3471} while (0)
3472
Tom Herbert15e23962015-02-10 16:30:31 -08003473static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3474 u16 start, u16 offset)
3475{
3476 skb->ip_summed = CHECKSUM_PARTIAL;
3477 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3478 skb->csum_offset = offset - start;
3479}
3480
Tom Herbertdcdc8992015-02-02 16:07:34 -08003481/* Update skbuf and packet to reflect the remote checksum offload operation.
3482 * When called, ptr indicates the starting point for skb->csum when
3483 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3484 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3485 */
3486static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
Tom Herbert15e23962015-02-10 16:30:31 -08003487 int start, int offset, bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08003488{
3489 __wsum delta;
3490
Tom Herbert15e23962015-02-10 16:30:31 -08003491 if (!nopartial) {
3492 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3493 return;
3494 }
3495
Tom Herbertdcdc8992015-02-02 16:07:34 -08003496 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3497 __skb_checksum_complete(skb);
3498 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3499 }
3500
3501 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3502
3503 /* Adjust skb->csum since we changed the packet */
3504 skb->csum = csum_add(skb->csum, delta);
3505}
3506
Florian Westphalcb9c6832017-01-23 18:21:56 +01003507static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
3508{
3509#if IS_ENABLED(CONFIG_NF_CONNTRACK)
Florian Westphala9e419d2017-01-23 18:21:59 +01003510 return (void *)(skb->_nfct & SKB_NFCT_PTRMASK);
Florian Westphalcb9c6832017-01-23 18:21:56 +01003511#else
3512 return NULL;
3513#endif
3514}
3515
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003516#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07003517void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003518static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3519{
3520 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07003521 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003522}
3523static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3524{
3525 if (nfct)
3526 atomic_inc(&nfct->use);
3527}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003528#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003529#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003530static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3531{
3532 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3533 kfree(nf_bridge);
3534}
3535static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3536{
3537 if (nf_bridge)
3538 atomic_inc(&nf_bridge->use);
3539}
3540#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08003541static inline void nf_reset(struct sk_buff *skb)
3542{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003543#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01003544 nf_conntrack_put(skb_nfct(skb));
3545 skb->_nfct = 0;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003546#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003547#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Patrick McHardya193a4a2006-03-20 19:23:05 -08003548 nf_bridge_put(skb->nf_bridge);
3549 skb->nf_bridge = NULL;
3550#endif
3551}
3552
Patrick McHardy124dff02013-04-05 20:42:05 +02003553static inline void nf_reset_trace(struct sk_buff *skb)
3554{
Florian Westphal478b3602014-02-15 23:48:45 +01003555#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00003556 skb->nf_trace = 0;
3557#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003558}
3559
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003560/* Note: This doesn't put any conntrack and bridge info in dst. */
Eric Dumazetb1937222014-09-28 22:18:47 -07003561static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3562 bool copy)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003563{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003564#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01003565 dst->_nfct = src->_nfct;
3566 nf_conntrack_get(skb_nfct(src));
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003567#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003568#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003569 dst->nf_bridge = src->nf_bridge;
3570 nf_bridge_get(src->nf_bridge);
3571#endif
Florian Westphal478b3602014-02-15 23:48:45 +01003572#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Eric Dumazetb1937222014-09-28 22:18:47 -07003573 if (copy)
3574 dst->nf_trace = src->nf_trace;
Florian Westphal478b3602014-02-15 23:48:45 +01003575#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003576}
3577
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003578static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3579{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003580#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01003581 nf_conntrack_put(skb_nfct(dst));
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003582#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003583#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003584 nf_bridge_put(dst->nf_bridge);
3585#endif
Eric Dumazetb1937222014-09-28 22:18:47 -07003586 __nf_copy(dst, src, true);
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003587}
3588
James Morris984bc162006-06-09 00:29:17 -07003589#ifdef CONFIG_NETWORK_SECMARK
3590static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3591{
3592 to->secmark = from->secmark;
3593}
3594
3595static inline void skb_init_secmark(struct sk_buff *skb)
3596{
3597 skb->secmark = 0;
3598}
3599#else
3600static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3601{ }
3602
3603static inline void skb_init_secmark(struct sk_buff *skb)
3604{ }
3605#endif
3606
Eric W. Biederman574f7192014-04-01 12:20:24 -07003607static inline bool skb_irq_freeable(const struct sk_buff *skb)
3608{
3609 return !skb->destructor &&
3610#if IS_ENABLED(CONFIG_XFRM)
3611 !skb->sp &&
3612#endif
Florian Westphalcb9c6832017-01-23 18:21:56 +01003613 !skb_nfct(skb) &&
Eric W. Biederman574f7192014-04-01 12:20:24 -07003614 !skb->_skb_refdst &&
3615 !skb_has_frag_list(skb);
3616}
3617
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003618static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3619{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003620 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003621}
3622
Stephen Hemminger92477442009-03-21 13:39:26 -07003623static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003624{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003625 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003626}
3627
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003628static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3629{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003630 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003631}
3632
David S. Millerd5a9e242009-01-27 16:22:11 -08003633static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3634{
3635 skb->queue_mapping = rx_queue + 1;
3636}
3637
Stephen Hemminger92477442009-03-21 13:39:26 -07003638static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003639{
3640 return skb->queue_mapping - 1;
3641}
3642
Stephen Hemminger92477442009-03-21 13:39:26 -07003643static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003644{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003645 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08003646}
3647
Julian Anastasov4ff06202017-02-06 23:14:12 +02003648static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
3649{
3650 skb->dst_pending_confirm = val;
3651}
3652
3653static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
3654{
3655 return skb->dst_pending_confirm != 0;
3656}
3657
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003658static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3659{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003660#ifdef CONFIG_XFRM
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003661 return skb->sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003662#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003663 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003664#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003665}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003666
Pravin B Shelar68c33162013-02-14 14:02:41 +00003667/* Keeps track of mac header offset relative to skb->head.
3668 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3669 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07003670 * tunnel skb it points to outer mac header.
3671 * Keeps track of level of encapsulation of network headers.
3672 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00003673struct skb_gso_cb {
Alexander Duyck802ab552016-04-10 21:45:03 -04003674 union {
3675 int mac_offset;
3676 int data_offset;
3677 };
Eric Dumazet3347c962013-10-19 11:42:56 -07003678 int encap_level;
Alexander Duyck76443452016-02-05 15:27:37 -08003679 __wsum csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003680 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00003681};
Konstantin Khlebnikov9207f9d2016-01-08 15:21:46 +03003682#define SKB_SGO_CB_OFFSET 32
3683#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
Pravin B Shelar68c33162013-02-14 14:02:41 +00003684
3685static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3686{
3687 return (skb_mac_header(inner_skb) - inner_skb->head) -
3688 SKB_GSO_CB(inner_skb)->mac_offset;
3689}
3690
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00003691static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3692{
3693 int new_headroom, headroom;
3694 int ret;
3695
3696 headroom = skb_headroom(skb);
3697 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3698 if (ret)
3699 return ret;
3700
3701 new_headroom = skb_headroom(skb);
3702 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3703 return 0;
3704}
3705
Alexander Duyck08b64fc2016-02-05 15:27:49 -08003706static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
3707{
3708 /* Do not update partial checksums if remote checksum is enabled. */
3709 if (skb->remcsum_offload)
3710 return;
3711
3712 SKB_GSO_CB(skb)->csum = res;
3713 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
3714}
3715
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003716/* Compute the checksum for a gso segment. First compute the checksum value
3717 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3718 * then add in skb->csum (checksum from csum_start to end of packet).
3719 * skb->csum and csum_start are then updated to reflect the checksum of the
3720 * resultant packet starting from the transport header-- the resultant checksum
3721 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3722 * header.
3723 */
3724static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3725{
Alexander Duyck76443452016-02-05 15:27:37 -08003726 unsigned char *csum_start = skb_transport_header(skb);
3727 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
3728 __wsum partial = SKB_GSO_CB(skb)->csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003729
Alexander Duyck76443452016-02-05 15:27:37 -08003730 SKB_GSO_CB(skb)->csum = res;
3731 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003732
Alexander Duyck76443452016-02-05 15:27:37 -08003733 return csum_fold(csum_partial(csum_start, plen, partial));
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003734}
3735
David S. Millerbdcc0922012-03-07 20:53:36 -05003736static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07003737{
3738 return skb_shinfo(skb)->gso_size;
3739}
3740
Eric Dumazet36a8f392013-09-29 01:21:32 -07003741/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05003742static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02003743{
3744 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3745}
3746
Daniel Borkmann5293efe2016-08-18 01:00:39 +02003747static inline void skb_gso_reset(struct sk_buff *skb)
3748{
3749 skb_shinfo(skb)->gso_size = 0;
3750 skb_shinfo(skb)->gso_segs = 0;
3751 skb_shinfo(skb)->gso_type = 0;
3752}
3753
Joe Perches7965bd42013-09-26 14:48:15 -07003754void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07003755
3756static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3757{
3758 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3759 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003760 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3761
Alexander Duyckb78462e2010-06-02 12:24:37 +00003762 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3763 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07003764 __skb_warn_lro_forwarding(skb);
3765 return true;
3766 }
3767 return false;
3768}
3769
Herbert Xu35fc92a2007-03-26 23:22:20 -07003770static inline void skb_forward_csum(struct sk_buff *skb)
3771{
3772 /* Unfortunately we don't support this one. Any brave souls? */
3773 if (skb->ip_summed == CHECKSUM_COMPLETE)
3774 skb->ip_summed = CHECKSUM_NONE;
3775}
3776
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003777/**
3778 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3779 * @skb: skb to check
3780 *
3781 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3782 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3783 * use this helper, to document places where we make this assertion.
3784 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003785static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003786{
3787#ifdef DEBUG
3788 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3789#endif
3790}
3791
Rusty Russellf35d9d82008-02-04 23:49:54 -05003792bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00003793
Paul Durranted1f50c2014-01-09 10:02:46 +00003794int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
Linus Lüssing9afd85c2015-05-02 14:01:07 +02003795struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
3796 unsigned int transport_len,
3797 __sum16(*skb_chkf)(struct sk_buff *skb));
Paul Durranted1f50c2014-01-09 10:02:46 +00003798
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00003799/**
3800 * skb_head_is_locked - Determine if the skb->head is locked down
3801 * @skb: skb to check
3802 *
3803 * The head on skbs build around a head frag can be removed if they are
3804 * not cloned. This function returns true if the skb head is locked down
3805 * due to either being allocated via kmalloc, or by being a clone with
3806 * multiple references to the head.
3807 */
3808static inline bool skb_head_is_locked(const struct sk_buff *skb)
3809{
3810 return !skb->head_frag || skb_cloned(skb);
3811}
Florian Westphalfe6cc552014-02-13 23:09:12 +01003812
3813/**
3814 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3815 *
3816 * @skb: GSO skb
3817 *
3818 * skb_gso_network_seglen is used to determine the real size of the
3819 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3820 *
3821 * The MAC/L2 header is not accounted for.
3822 */
3823static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3824{
3825 unsigned int hdr_len = skb_transport_header(skb) -
3826 skb_network_header(skb);
3827 return hdr_len + skb_gso_transport_seglen(skb);
3828}
Thomas Grafee122c72015-07-21 10:43:58 +02003829
Edward Cree179bc672016-02-11 20:48:04 +00003830/* Local Checksum Offload.
3831 * Compute outer checksum based on the assumption that the
3832 * inner checksum will be offloaded later.
Edward Creee8ae7b02016-02-11 21:03:37 +00003833 * See Documentation/networking/checksum-offloads.txt for
3834 * explanation of how this works.
Edward Cree179bc672016-02-11 20:48:04 +00003835 * Fill in outer checksum adjustment (e.g. with sum of outer
3836 * pseudo-header) before calling.
3837 * Also ensure that inner checksum is in linear data area.
3838 */
3839static inline __wsum lco_csum(struct sk_buff *skb)
3840{
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003841 unsigned char *csum_start = skb_checksum_start(skb);
3842 unsigned char *l4_hdr = skb_transport_header(skb);
3843 __wsum partial;
Edward Cree179bc672016-02-11 20:48:04 +00003844
3845 /* Start with complement of inner checksum adjustment */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003846 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
3847 skb->csum_offset));
3848
Edward Cree179bc672016-02-11 20:48:04 +00003849 /* Add in checksum of our headers (incl. outer checksum
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003850 * adjustment filled in by caller) and return result.
Edward Cree179bc672016-02-11 20:48:04 +00003851 */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003852 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
Edward Cree179bc672016-02-11 20:48:04 +00003853}
3854
Linus Torvalds1da177e2005-04-16 15:20:36 -07003855#endif /* __KERNEL__ */
3856#endif /* _LINUX_SKBUFF_H */