blob: b1f46a0d18e242ac1c652898cf1cd2b26997523c [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.
112 *
113 * skb->csum_level indicates the number of consecutive checksums found in
114 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
115 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
116 * and a device is able to verify the checksums for UDP (possibly zero),
117 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
118 * two. If the device were only able to verify the UDP checksum and not
119 * GRE, either because it doesn't support GRE checksum of because GRE
120 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
121 * not considered in this case).
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100122 *
123 * CHECKSUM_COMPLETE:
124 *
125 * This is the most generic way. The device supplied checksum of the _whole_
126 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
127 * hardware doesn't need to parse L3/L4 headers to implement this.
128 *
129 * Note: Even if device supports only some protocols, but is able to produce
130 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
131 *
132 * CHECKSUM_PARTIAL:
133 *
Tom Herbert6edec0e2015-02-10 16:30:28 -0800134 * A checksum is set up to be offloaded to a device as described in the
135 * output description for CHECKSUM_PARTIAL. This may occur on a packet
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100136 * received directly from another Linux OS, e.g., a virtualized Linux kernel
Tom Herbert6edec0e2015-02-10 16:30:28 -0800137 * on the same host, or it may be set in the input path in GRO or remote
138 * checksum offload. For the purposes of checksum verification, the checksum
139 * referred to by skb->csum_start + skb->csum_offset and any preceding
140 * checksums in the packet are considered verified. Any checksums in the
141 * packet that are after the checksum being offloaded are not considered to
142 * be verified.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100143 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800144 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
145 * in the skb->ip_summed for a packet. Values are:
146 *
147 * CHECKSUM_PARTIAL:
148 *
149 * The driver is required to checksum the packet as seen by hard_start_xmit()
150 * from skb->csum_start up to the end, and to record/write the checksum at
151 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
152 * csum_start and csum_offset values are valid values given the length and
153 * offset of the packet, however they should not attempt to validate that the
154 * checksum refers to a legitimate transport layer checksum-- it is the
155 * purview of the stack to validate that csum_start and csum_offset are set
156 * correctly.
157 *
158 * When the stack requests checksum offload for a packet, the driver MUST
159 * ensure that the checksum is set correctly. A driver can either offload the
160 * checksum calculation to the device, or call skb_checksum_help (in the case
161 * that the device does not support offload for a particular checksum).
162 *
163 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
164 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
165 * checksum offload capability. If a device has limited checksum capabilities
166 * (for instance can only perform NETIF_F_IP_CSUM or NETIF_F_IPV6_CSUM as
167 * described above) a helper function can be called to resolve
168 * CHECKSUM_PARTIAL. The helper functions are skb_csum_off_chk*. The helper
169 * function takes a spec argument that describes the protocol layer that is
170 * supported for checksum offload and can be called for each packet. If a
171 * packet does not match the specification for offload, skb_checksum_help
172 * 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
192 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
193 * accordingly. Note the there is no indication in the skbuff that the
194 * CHECKSUM_PARTIAL refers to an SCTP checksum, a driver that supports
195 * both IP checksum offload and SCTP CRC offload must verify which offload
196 * is configured for a packet presumably by inspecting packet headers.
197 *
198 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
199 * offloading the FCOE CRC in a packet. To perform this offload the stack
200 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
201 * accordingly. Note the there is no indication in the skbuff that the
202 * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
203 * both IP checksum offload and FCOE CRC offload must verify which offload
204 * is configured for a packet presumably by inspecting packet headers.
205 *
206 * E. Checksumming on output with GSO.
207 *
208 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
209 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
210 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
211 * part of the GSO operation is implied. If a checksum is being offloaded
212 * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
213 * are set to refer to the outermost checksum being offload (two offloaded
214 * checksums are possible with UDP encapsulation).
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100215 */
216
Herbert Xu60476372007-04-09 11:59:39 -0700217/* Don't change this without changing skb_csum_unnecessary! */
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100218#define CHECKSUM_NONE 0
219#define CHECKSUM_UNNECESSARY 1
220#define CHECKSUM_COMPLETE 2
221#define CHECKSUM_PARTIAL 3
Linus Torvalds1da177e2005-04-16 15:20:36 -0700222
Tom Herbert77cffe22014-08-27 21:26:46 -0700223/* Maximum value in skb->csum_level */
224#define SKB_MAX_CSUM_LEVEL 3
225
Tobias Klauser0bec8c82014-07-22 12:06:23 +0200226#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
David S. Millerfc910a22007-03-25 20:27:59 -0700227#define SKB_WITH_OVERHEAD(X) \
Herbert Xudeea84b2007-10-21 16:27:46 -0700228 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
David S. Millerfc910a22007-03-25 20:27:59 -0700229#define SKB_MAX_ORDER(X, ORDER) \
230 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700231#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
232#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
233
Eric Dumazet87fb4b72011-10-13 07:28:54 +0000234/* return minimum truesize of one skb containing X bytes of data */
235#define SKB_TRUESIZE(X) ((X) + \
236 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
237 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
238
Linus Torvalds1da177e2005-04-16 15:20:36 -0700239struct net_device;
David Howells716ea3a2007-04-02 20:19:53 -0700240struct scatterlist;
Jens Axboe9c55e012007-11-06 23:30:13 -0800241struct pipe_inode_info;
Herbert Xua8f820aa2014-11-07 21:22:22 +0800242struct iov_iter;
Alexander Duyckfd11a832014-12-09 19:40:49 -0800243struct napi_struct;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700245#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246struct nf_conntrack {
247 atomic_t use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248};
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700249#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +0200251#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252struct nf_bridge_info {
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000253 atomic_t use;
Florian Westphal3eaf4022015-04-02 14:31:44 +0200254 enum {
255 BRNF_PROTO_UNCHANGED,
256 BRNF_PROTO_8021Q,
257 BRNF_PROTO_PPPOE
Florian Westphal7fb48c52015-05-03 22:05:28 +0200258 } orig_proto:8;
Florian Westphal72b1e5e2015-07-23 16:21:30 +0200259 u8 pkt_otherhost:1;
260 u8 in_prerouting:1;
261 u8 bridged_dnat:1;
Bernhard Thaler411ffb42015-05-30 15:28:28 +0200262 __u16 frag_max_size;
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000263 struct net_device *physindev;
Florian Westphal63cdbc02015-09-14 17:06:27 +0200264
265 /* always valid & non-NULL from FORWARD on, for physdev match */
266 struct net_device *physoutdev;
Florian Westphal7fb48c52015-05-03 22:05:28 +0200267 union {
Florian Westphal72b1e5e2015-07-23 16:21:30 +0200268 /* prerouting: detect dnat in orig/reply direction */
Bernhard Thaler72b31f72015-05-30 15:27:40 +0200269 __be32 ipv4_daddr;
270 struct in6_addr ipv6_daddr;
Florian Westphal72b1e5e2015-07-23 16:21:30 +0200271
272 /* after prerouting + nat detected: store original source
273 * mac since neigh resolution overwrites it, only used while
274 * skb is out in neigh layer.
275 */
276 char neigh_header[8];
Bernhard Thaler72b31f72015-05-30 15:27:40 +0200277 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700278};
279#endif
280
Linus Torvalds1da177e2005-04-16 15:20:36 -0700281struct sk_buff_head {
282 /* These two members must be first. */
283 struct sk_buff *next;
284 struct sk_buff *prev;
285
286 __u32 qlen;
287 spinlock_t lock;
288};
289
290struct sk_buff;
291
Ian Campbell9d4dde52011-12-22 23:39:14 +0000292/* To allow 64K frame to be packed as single skb without frag_list we
293 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
294 * buffers which do not start on a page boundary.
295 *
296 * Since GRO uses frags we allocate at least 16 regardless of page
297 * size.
Anton Blancharda715dea2011-03-27 14:57:26 +0000298 */
Ian Campbell9d4dde52011-12-22 23:39:14 +0000299#if (65536/PAGE_SIZE + 1) < 16
David S. Millereec00952011-03-29 23:34:08 -0700300#define MAX_SKB_FRAGS 16UL
Anton Blancharda715dea2011-03-27 14:57:26 +0000301#else
Ian Campbell9d4dde52011-12-22 23:39:14 +0000302#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
Anton Blancharda715dea2011-03-27 14:57:26 +0000303#endif
Hans Westgaard Ry5f74f82e2016-02-03 09:26:57 +0100304extern int sysctl_max_skb_frags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700305
Marcelo Ricardo Leitner3953c462016-06-02 15:05:40 -0300306/* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
307 * segment using its current segmentation instead.
308 */
309#define GSO_BY_FRAGS 0xFFFF
310
Linus Torvalds1da177e2005-04-16 15:20:36 -0700311typedef struct skb_frag_struct skb_frag_t;
312
313struct skb_frag_struct {
Ian Campbella8605c62011-10-19 23:01:49 +0000314 struct {
315 struct page *p;
316 } page;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000317#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
David S. Millera309bb02007-07-30 18:47:03 -0700318 __u32 page_offset;
319 __u32 size;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000320#else
321 __u16 page_offset;
322 __u16 size;
323#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324};
325
Eric Dumazet9e903e02011-10-18 21:00:24 +0000326static inline unsigned int skb_frag_size(const skb_frag_t *frag)
327{
328 return frag->size;
329}
330
331static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
332{
333 frag->size = size;
334}
335
336static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
337{
338 frag->size += delta;
339}
340
341static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
342{
343 frag->size -= delta;
344}
345
Patrick Ohlyac45f602009-02-12 05:03:37 +0000346#define HAVE_HW_TIME_STAMP
347
348/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800349 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000350 * @hwtstamp: hardware time stamp transformed into duration
351 * since arbitrary point in time
Patrick Ohlyac45f602009-02-12 05:03:37 +0000352 *
353 * Software time stamps generated by ktime_get_real() are stored in
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400354 * skb->tstamp.
Patrick Ohlyac45f602009-02-12 05:03:37 +0000355 *
356 * hwtstamps can only be compared against other hwtstamps from
357 * the same device.
358 *
359 * This structure is attached to packets as part of the
360 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
361 */
362struct skb_shared_hwtstamps {
363 ktime_t hwtstamp;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000364};
365
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000366/* Definitions for tx_flags in struct skb_shared_info */
367enum {
368 /* generate hardware time stamp */
369 SKBTX_HW_TSTAMP = 1 << 0,
370
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400371 /* generate software time stamp when queueing packet to NIC */
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000372 SKBTX_SW_TSTAMP = 1 << 1,
373
374 /* device driver is going to provide hardware time stamp */
375 SKBTX_IN_PROGRESS = 1 << 2,
376
Shirley Maa6686f22011-07-06 12:22:12 +0000377 /* device driver supports TX zero-copy buffers */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000378 SKBTX_DEV_ZEROCOPY = 1 << 3,
Johannes Berg6e3e9392011-11-09 10:15:42 +0100379
380 /* generate wifi status information (where possible) */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000381 SKBTX_WIFI_STATUS = 1 << 4,
Pravin B Shelarc9af6db2013-02-11 09:27:41 +0000382
383 /* This indicates at least one fragment might be overwritten
384 * (as in vmsplice(), sendfile() ...)
385 * If we need to compute a TX checksum, we'll need to copy
386 * all frags to avoid possible bad checksum
387 */
388 SKBTX_SHARED_FRAG = 1 << 5,
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400389
390 /* generate software time stamp when entering packet scheduling */
391 SKBTX_SCHED_TSTAMP = 1 << 6,
Shirley Maa6686f22011-07-06 12:22:12 +0000392};
393
Willem de Bruijne1c8a602014-08-04 22:11:50 -0400394#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
Soheil Hassas Yeganeh0a2cf202016-04-27 23:39:01 -0400395 SKBTX_SCHED_TSTAMP)
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400396#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
397
Shirley Maa6686f22011-07-06 12:22:12 +0000398/*
399 * The callback notifies userspace to release buffers when skb DMA is done in
400 * lower device, the skb last reference should be 0 when calling this.
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000401 * The zerocopy_success argument is true if zero copy transmit occurred,
402 * false on data copy or out of memory error caused by data copy attempt.
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000403 * The ctx field is used to track device context.
404 * The desc field is used to track userspace buffer index.
Shirley Maa6686f22011-07-06 12:22:12 +0000405 */
406struct ubuf_info {
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000407 void (*callback)(struct ubuf_info *, bool zerocopy_success);
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000408 void *ctx;
Shirley Maa6686f22011-07-06 12:22:12 +0000409 unsigned long desc;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000410};
411
Linus Torvalds1da177e2005-04-16 15:20:36 -0700412/* This data is invariant across clones and lives at
413 * the end of the header data, ie. at skb->end.
414 */
415struct skb_shared_info {
Steffen Klassert7f564522017-04-08 20:36:24 +0200416 unsigned short _unused;
Ian Campbell9f42f122012-01-05 07:13:39 +0000417 unsigned char nr_frags;
418 __u8 tx_flags;
Herbert Xu79671682006-06-22 02:40:14 -0700419 unsigned short gso_size;
420 /* Warning: this field is not always filled in (UFO)! */
421 unsigned short gso_segs;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700422 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000423 struct skb_shared_hwtstamps hwtstamps;
Steffen Klassert7f564522017-04-08 20:36:24 +0200424 unsigned int gso_type;
Willem de Bruijn09c2d252014-08-04 22:11:47 -0400425 u32 tskey;
Ian Campbell9f42f122012-01-05 07:13:39 +0000426 __be32 ip6_frag_id;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700427
428 /*
429 * Warning : all fields before dataref are cleared in __alloc_skb()
430 */
431 atomic_t dataref;
432
Johann Baudy69e3c752009-05-18 22:11:22 -0700433 /* Intermediate layers must ensure that destructor_arg
434 * remains valid until skb destructor */
435 void * destructor_arg;
Shirley Maa6686f22011-07-06 12:22:12 +0000436
Eric Dumazetfed66382010-07-22 19:09:08 +0000437 /* must be last field, see pskb_expand_head() */
438 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439};
440
441/* We divide dataref into two halves. The higher 16 bits hold references
442 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700443 * the entire skb->data. A clone of a headerless skb holds the length of
444 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700445 *
446 * All users must obey the rule that the skb->data reference count must be
447 * greater than or equal to the payload reference count.
448 *
449 * Holding a reference to the payload part means that the user does not
450 * care about modifications to the header part of skb->data.
451 */
452#define SKB_DATAREF_SHIFT 16
453#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
454
David S. Millerd179cd12005-08-17 14:57:30 -0700455
456enum {
Vijay Subramanianc8753d52014-10-02 10:00:43 -0700457 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
458 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
459 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
David S. Millerd179cd12005-08-17 14:57:30 -0700460};
461
Herbert Xu79671682006-06-22 02:40:14 -0700462enum {
463 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700464 SKB_GSO_UDP = 1 << 1,
Herbert Xu576a30e2006-06-27 13:22:38 -0700465
466 /* This indicates the skb is from an untrusted source. */
467 SKB_GSO_DODGY = 1 << 2,
Michael Chanb0da85372006-06-29 12:30:00 -0700468
469 /* This indicates the tcp segment has CWR set. */
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700470 SKB_GSO_TCP_ECN = 1 << 3,
471
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400472 SKB_GSO_TCP_FIXEDID = 1 << 4,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800473
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400474 SKB_GSO_TCPV6 = 1 << 5,
Pravin B Shelar68c33162013-02-14 14:02:41 +0000475
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400476 SKB_GSO_FCOE = 1 << 6,
Pravin B Shelar73136262013-03-07 13:21:51 +0000477
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400478 SKB_GSO_GRE = 1 << 7,
Simon Horman0d89d202013-05-23 21:02:52 +0000479
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400480 SKB_GSO_GRE_CSUM = 1 << 8,
Eric Dumazetcb32f512013-10-19 11:42:57 -0700481
Tom Herbert7e133182016-05-18 09:06:10 -0700482 SKB_GSO_IPXIP4 = 1 << 9,
Eric Dumazet61c1db72013-10-20 20:47:30 -0700483
Tom Herbert7e133182016-05-18 09:06:10 -0700484 SKB_GSO_IPXIP6 = 1 << 10,
Tom Herbert0f4f4ff2014-06-04 17:20:16 -0700485
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400486 SKB_GSO_UDP_TUNNEL = 1 << 11,
Tom Herbert4749c092014-06-04 17:20:23 -0700487
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400488 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 12,
489
Alexander Duyck802ab552016-04-10 21:45:03 -0400490 SKB_GSO_PARTIAL = 1 << 13,
491
492 SKB_GSO_TUNNEL_REMCSUM = 1 << 14,
Marcelo Ricardo Leitner90017ac2016-06-02 15:05:43 -0300493
494 SKB_GSO_SCTP = 1 << 15,
Steffen Klassertc7ef8f02017-04-14 10:05:36 +0200495
496 SKB_GSO_ESP = 1 << 16,
Herbert Xu79671682006-06-22 02:40:14 -0700497};
498
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700499#if BITS_PER_LONG > 32
500#define NET_SKBUFF_DATA_USES_OFFSET 1
501#endif
502
503#ifdef NET_SKBUFF_DATA_USES_OFFSET
504typedef unsigned int sk_buff_data_t;
505#else
506typedef unsigned char *sk_buff_data_t;
507#endif
508
Linus Torvalds1da177e2005-04-16 15:20:36 -0700509/**
510 * struct sk_buff - socket buffer
511 * @next: Next buffer in list
512 * @prev: Previous buffer in list
Eric Dumazet363ec39232014-02-26 14:02:11 -0800513 * @tstamp: Time we arrived/left
Eric Dumazet56b17422014-11-03 08:19:53 -0800514 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700515 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700517 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000518 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700519 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700520 * @len: Length of actual data
521 * @data_len: Data length
522 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700523 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700524 * @csum: Checksum (must include start/offset pair)
525 * @csum_start: Offset from skb->head where checksumming should start
526 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700527 * @priority: Packet queueing priority
WANG Cong60ff7462014-05-04 16:39:18 -0700528 * @ignore_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700529 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700530 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700531 * @nohdr: Payload reference only, must not modify header
532 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700533 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700534 * @ipvs_property: skbuff is owned by ipvs
Willem de Bruijne7246e12017-01-07 17:06:35 -0500535 * @tc_skip_classify: do not classify packet. set by IFB device
Willem de Bruijn8dc07fd2017-01-07 17:06:37 -0500536 * @tc_at_ingress: used within tc_classify to distinguish in/egress
Willem de Bruijnbc31c902017-01-07 17:06:38 -0500537 * @tc_redirected: packet was redirected by a tc action
538 * @tc_from_ingress: if tc_redirected, tc_at_ingress at time of redirect
Randy Dunlap31729362008-02-18 20:52:13 -0800539 * @peeked: this packet has been seen already, so stats have been
540 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700541 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700542 * @protocol: Packet protocol from driver
543 * @destructor: Destruct function
Florian Westphala9e419d2017-01-23 18:21:59 +0100544 * @_nfct: Associated connection, if any (with nfctinfo bits)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700545 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800546 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700547 * @tc_index: Traffic control index
Tom Herbert61b905d2014-03-24 15:34:47 -0700548 * @hash: the packet hash
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700549 * @queue_mapping: Queue mapping for multiqueue devices
David S. Miller0b725a22014-08-25 15:51:53 -0700550 * @xmit_more: More SKBs are pending for this queue
Randy Dunlap553a5672008-04-20 10:51:01 -0700551 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700552 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Tom Herbert61b905d2014-03-24 15:34:47 -0700553 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
Changli Gao4ca24622011-08-19 07:26:44 -0700554 * ports.
Tom Herberta3b18dd2014-07-01 21:33:17 -0700555 * @sw_hash: indicates hash was computed in software stack
Johannes Berg6e3e9392011-11-09 10:15:42 +0100556 * @wifi_acked_valid: wifi_acked was set
557 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000558 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Julian Anastasov4ff06202017-02-06 23:14:12 +0200559 * @dst_pending_confirm: need to confirm neighbour
Eliezer Tamir06021292013-06-10 11:39:50 +0300560 * @napi_id: id of the NAPI struct this skb came from
James Morris984bc162006-06-09 00:29:17 -0700561 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700562 * @mark: Generic packet mark
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000563 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b2008-07-14 22:49:06 -0700564 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000565 * @inner_protocol: Protocol (encapsulation)
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000566 * @inner_transport_header: Inner transport layer header (encapsulation)
567 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000568 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700569 * @transport_header: Transport layer header
570 * @network_header: Network layer header
571 * @mac_header: Link layer header
572 * @tail: Tail pointer
573 * @end: End pointer
574 * @head: Head of buffer
575 * @data: Data head pointer
576 * @truesize: Buffer size
577 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700578 */
579
580struct sk_buff {
Eric Dumazet363ec39232014-02-26 14:02:11 -0800581 union {
Eric Dumazet56b17422014-11-03 08:19:53 -0800582 struct {
583 /* These two members must be first. */
584 struct sk_buff *next;
585 struct sk_buff *prev;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000586
Eric Dumazet56b17422014-11-03 08:19:53 -0800587 union {
588 ktime_t tstamp;
Eric Dumazet9a568de2017-05-16 14:00:14 -0700589 u64 skb_mstamp;
Eric Dumazet56b17422014-11-03 08:19:53 -0800590 };
591 };
592 struct rb_node rbnode; /* used in netem & tcp stack */
593 };
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000594 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700595
Eric Dumazetc84d9492016-12-08 11:41:55 -0800596 union {
597 struct net_device *dev;
598 /* Some protocols might use this space to store information,
599 * while device pointer would be NULL.
600 * UDP receive path is one user.
601 */
602 unsigned long dev_scratch;
603 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604 /*
605 * This is the control buffer. It is free to use for every
606 * layer. Please put your private variables there. If you
607 * want to keep them across layers you have to do a skb_clone()
608 * first. This is owned by whoever has the skb queued ATM.
609 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000610 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700611
Eric Dumazet7fee2262010-05-11 23:19:48 +0000612 unsigned long _skb_refdst;
Eric Dumazetb1937222014-09-28 22:18:47 -0700613 void (*destructor)(struct sk_buff *skb);
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000614#ifdef CONFIG_XFRM
615 struct sec_path *sp;
616#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700617#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +0100618 unsigned long _nfct;
Eric Dumazetb1937222014-09-28 22:18:47 -0700619#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +0200620#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Eric Dumazetb1937222014-09-28 22:18:47 -0700621 struct nf_bridge_info *nf_bridge;
622#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700624 data_len;
625 __u16 mac_len,
626 hdr_len;
Eric Dumazetb1937222014-09-28 22:18:47 -0700627
628 /* Following fields are _not_ copied in __copy_skb_header()
629 * Note that queue_mapping is here mostly to fill a hole.
630 */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200631 kmemcheck_bitfield_begin(flags1);
Eric Dumazetb1937222014-09-28 22:18:47 -0700632 __u16 queue_mapping;
Daniel Borkmann36bbef52016-09-20 00:26:13 +0200633
634/* if you move cloned around you also must adapt those constants */
635#ifdef __BIG_ENDIAN_BITFIELD
636#define CLONED_MASK (1 << 7)
637#else
638#define CLONED_MASK 1
639#endif
640#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
641
642 __u8 __cloned_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700643 __u8 cloned:1,
Harald Welte6869c4d2005-08-09 19:24:19 -0700644 nohdr:1,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800645 fclone:2,
Herbert Xua59322b2007-12-05 01:53:40 -0800646 peeked:1,
Eric Dumazetb1937222014-09-28 22:18:47 -0700647 head_frag:1,
Daniel Borkmann36bbef52016-09-20 00:26:13 +0200648 xmit_more:1,
649 __unused:1; /* one bit hole */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200650 kmemcheck_bitfield_end(flags1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651
Eric Dumazetb1937222014-09-28 22:18:47 -0700652 /* fields enclosed in headers_start/headers_end are copied
653 * using a single memcpy() in __copy_skb_header()
654 */
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700655 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700656 __u32 headers_start[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700657 /* public: */
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200658
659/* if you move pkt_type around you also must adapt those constants */
660#ifdef __BIG_ENDIAN_BITFIELD
661#define PKT_TYPE_MAX (7 << 5)
662#else
663#define PKT_TYPE_MAX 7
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100664#endif
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200665#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
666
667 __u8 __pkt_type_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700668 __u8 pkt_type:3;
669 __u8 pfmemalloc:1;
670 __u8 ignore_df:1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671
Eric Dumazetb1937222014-09-28 22:18:47 -0700672 __u8 nf_trace:1;
673 __u8 ip_summed:2;
674 __u8 ooo_okay:1;
675 __u8 l4_hash:1;
676 __u8 sw_hash:1;
677 __u8 wifi_acked_valid:1;
678 __u8 wifi_acked:1;
679
680 __u8 no_fcs:1;
681 /* Indicates the inner headers are valid in the skbuff. */
682 __u8 encapsulation:1;
683 __u8 encap_hdr_csum:1;
684 __u8 csum_valid:1;
685 __u8 csum_complete_sw:1;
686 __u8 csum_level:2;
687 __u8 csum_bad:1;
688
Julian Anastasov4ff06202017-02-06 23:14:12 +0200689 __u8 dst_pending_confirm:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700690#ifdef CONFIG_IPV6_NDISC_NODETYPE
691 __u8 ndisc_nodetype:2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700693 __u8 ipvs_property:1;
Tom Herbert8bce6d72014-09-29 20:22:29 -0700694 __u8 inner_protocol_type:1;
Tom Herberte585f232014-11-04 09:06:54 -0800695 __u8 remcsum_offload:1;
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200696#ifdef CONFIG_NET_SWITCHDEV
697 __u8 offload_fwd_mark:1;
698#endif
Willem de Bruijne7246e12017-01-07 17:06:35 -0500699#ifdef CONFIG_NET_CLS_ACT
700 __u8 tc_skip_classify:1;
Willem de Bruijn8dc07fd2017-01-07 17:06:37 -0500701 __u8 tc_at_ingress:1;
Willem de Bruijnbc31c902017-01-07 17:06:38 -0500702 __u8 tc_redirected:1;
703 __u8 tc_from_ingress:1;
Willem de Bruijne7246e12017-01-07 17:06:35 -0500704#endif
Alexander Duyck4031ae62012-01-27 06:22:53 +0000705
Linus Torvalds1da177e2005-04-16 15:20:36 -0700706#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700707 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700708#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200709
Eric Dumazetb1937222014-09-28 22:18:47 -0700710 union {
711 __wsum csum;
712 struct {
713 __u16 csum_start;
714 __u16 csum_offset;
715 };
716 };
717 __u32 priority;
718 int skb_iif;
719 __u32 hash;
720 __be16 vlan_proto;
721 __u16 vlan_tci;
Eric Dumazet2bd82482015-02-03 23:48:24 -0800722#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
723 union {
724 unsigned int napi_id;
725 unsigned int sender_cpu;
726 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700727#endif
James Morris984bc162006-06-09 00:29:17 -0700728#ifdef CONFIG_NETWORK_SECMARK
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200729 __u32 secmark;
James Morris984bc162006-06-09 00:29:17 -0700730#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -0700731
Neil Horman3b885782009-10-12 13:26:31 -0700732 union {
733 __u32 mark;
Eric Dumazet16fad692013-03-14 05:40:32 +0000734 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700735 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700736
Tom Herbert8bce6d72014-09-29 20:22:29 -0700737 union {
738 __be16 inner_protocol;
739 __u8 inner_ipproto;
740 };
741
Simon Horman1a37e412013-05-23 21:02:51 +0000742 __u16 inner_transport_header;
743 __u16 inner_network_header;
744 __u16 inner_mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700745
746 __be16 protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000747 __u16 transport_header;
748 __u16 network_header;
749 __u16 mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700750
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700751 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700752 __u32 headers_end[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700753 /* public: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700754
Linus Torvalds1da177e2005-04-16 15:20:36 -0700755 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700756 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700757 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700758 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700759 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700760 unsigned int truesize;
761 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762};
763
764#ifdef __KERNEL__
765/*
766 * Handling routines are only of interest to the kernel
767 */
768#include <linux/slab.h>
769
Linus Torvalds1da177e2005-04-16 15:20:36 -0700770
Mel Gormanc93bdd02012-07-31 16:44:19 -0700771#define SKB_ALLOC_FCLONE 0x01
772#define SKB_ALLOC_RX 0x02
Alexander Duyckfd11a832014-12-09 19:40:49 -0800773#define SKB_ALLOC_NAPI 0x04
Mel Gormanc93bdd02012-07-31 16:44:19 -0700774
775/* Returns true if the skb was allocated from PFMEMALLOC reserves */
776static inline bool skb_pfmemalloc(const struct sk_buff *skb)
777{
778 return unlikely(skb->pfmemalloc);
779}
780
Eric Dumazet7fee2262010-05-11 23:19:48 +0000781/*
782 * skb might have a dst pointer attached, refcounted or not.
783 * _skb_refdst low order bit is set if refcount was _not_ taken
784 */
785#define SKB_DST_NOREF 1UL
786#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
787
Florian Westphala9e419d2017-01-23 18:21:59 +0100788#define SKB_NFCT_PTRMASK ~(7UL)
Eric Dumazet7fee2262010-05-11 23:19:48 +0000789/**
790 * skb_dst - returns skb dst_entry
791 * @skb: buffer
792 *
793 * Returns skb dst_entry, regardless of reference taken or not.
794 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000795static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
796{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000797 /* If refdst was not refcounted, check we still are in a
798 * rcu_read_lock section
799 */
800 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
801 !rcu_read_lock_held() &&
802 !rcu_read_lock_bh_held());
803 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000804}
805
Eric Dumazet7fee2262010-05-11 23:19:48 +0000806/**
807 * skb_dst_set - sets skb dst
808 * @skb: buffer
809 * @dst: dst entry
810 *
811 * Sets skb dst, assuming a reference was taken on dst and should
812 * be released by skb_dst_drop()
813 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000814static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
815{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000816 skb->_skb_refdst = (unsigned long)dst;
817}
818
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200819/**
820 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
821 * @skb: buffer
822 * @dst: dst entry
823 *
824 * Sets skb dst, assuming a reference was not taken on dst.
825 * If dst entry is cached, we do not take reference and dst_release
826 * will be avoided by refdst_drop. If dst entry is not cached, we take
827 * reference, so that last dst_release can destroy the dst immediately.
828 */
829static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
830{
Hannes Frederic Sowadbfc4fb2014-12-06 19:19:42 +0100831 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
832 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200833}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000834
835/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300836 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000837 * @skb: buffer
838 */
839static inline bool skb_dst_is_noref(const struct sk_buff *skb)
840{
841 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000842}
843
Eric Dumazet511c3f92009-06-02 05:14:27 +0000844static inline struct rtable *skb_rtable(const struct sk_buff *skb)
845{
Eric Dumazetadf30902009-06-02 05:19:30 +0000846 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000847}
848
Jamal Hadi Salim8b10cab2016-07-02 06:43:14 -0400849/* For mangling skb->pkt_type from user space side from applications
850 * such as nft, tc, etc, we only allow a conservative subset of
851 * possible pkt_types to be set.
852*/
853static inline bool skb_pkt_type_ok(u32 ptype)
854{
855 return ptype <= PACKET_OTHERHOST;
856}
857
Joe Perches7965bd42013-09-26 14:48:15 -0700858void kfree_skb(struct sk_buff *skb);
859void kfree_skb_list(struct sk_buff *segs);
860void skb_tx_error(struct sk_buff *skb);
861void consume_skb(struct sk_buff *skb);
862void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000863extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000864
Joe Perches7965bd42013-09-26 14:48:15 -0700865void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
866bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
867 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000868
Joe Perches7965bd42013-09-26 14:48:15 -0700869struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
870 int node);
Eric Dumazet2ea2f622015-04-24 16:05:01 -0700871struct sk_buff *__build_skb(void *data, unsigned int frag_size);
Joe Perches7965bd42013-09-26 14:48:15 -0700872struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700873static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100874 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700875{
Eric Dumazet564824b2010-10-11 19:05:25 +0000876 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700877}
878
Eric Dumazet2e4e4412014-09-17 04:49:49 -0700879struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
880 unsigned long data_len,
881 int max_page_order,
882 int *errcode,
883 gfp_t gfp_mask);
884
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700885/* Layout of fast clones : [skb1][skb2][fclone_ref] */
886struct sk_buff_fclones {
887 struct sk_buff skb1;
888
889 struct sk_buff skb2;
890
891 atomic_t fclone_ref;
892};
893
894/**
895 * skb_fclone_busy - check if fclone is busy
Stephen Hemminger293de7d2016-10-23 09:28:29 -0700896 * @sk: socket
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700897 * @skb: buffer
898 *
Masanari Iidabda13fe2015-12-13 16:53:02 +0900899 * Returns true if skb is a fast clone, and its clone is not freed.
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700900 * Some drivers call skb_orphan() in their ndo_start_xmit(),
901 * so we also check that this didnt happen.
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700902 */
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700903static inline bool skb_fclone_busy(const struct sock *sk,
904 const struct sk_buff *skb)
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700905{
906 const struct sk_buff_fclones *fclones;
907
908 fclones = container_of(skb, struct sk_buff_fclones, skb1);
909
910 return skb->fclone == SKB_FCLONE_ORIG &&
Eric Dumazet6ffe75eb2014-12-03 17:04:39 -0800911 atomic_read(&fclones->fclone_ref) > 1 &&
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700912 fclones->skb2.sk == sk;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700913}
914
David S. Millerd179cd12005-08-17 14:57:30 -0700915static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100916 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700917{
Mel Gormanc93bdd02012-07-31 16:44:19 -0700918 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700919}
920
Joe Perches7965bd42013-09-26 14:48:15 -0700921struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
Patrick McHardy0ebd0ac2013-04-17 06:46:58 +0000922static inline struct sk_buff *alloc_skb_head(gfp_t priority)
923{
924 return __alloc_skb_head(priority, -1);
925}
926
Joe Perches7965bd42013-09-26 14:48:15 -0700927struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
928int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
929struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
930struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
Octavian Purdilabad93e92014-06-12 01:36:26 +0300931struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
932 gfp_t gfp_mask, bool fclone);
933static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
934 gfp_t gfp_mask)
935{
936 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
937}
Eric Dumazet117632e2011-12-03 21:39:53 +0000938
Joe Perches7965bd42013-09-26 14:48:15 -0700939int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
940struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
941 unsigned int headroom);
942struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
943 int newtailroom, gfp_t priority);
Fan Du25a91d82014-01-18 09:54:23 +0800944int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
945 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -0700946int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
947 int len);
948int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
949int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000950#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700951
Joe Perches7965bd42013-09-26 14:48:15 -0700952int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
953 int getfrag(void *from, char *to, int offset,
954 int len, int odd, struct sk_buff *skb),
955 void *from, int length);
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700956
Hannes Frederic Sowabe12a1f2015-05-21 16:59:58 +0200957int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
958 int offset, size_t size);
959
Eric Dumazetd94d9fe2009-11-04 09:50:58 -0800960struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -0700961 __u32 lower_offset;
962 __u32 upper_offset;
963 __u32 frag_idx;
964 __u32 stepped_offset;
965 struct sk_buff *root_skb;
966 struct sk_buff *cur_skb;
967 __u8 *frag_data;
968};
969
Joe Perches7965bd42013-09-26 14:48:15 -0700970void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
971 unsigned int to, struct skb_seq_state *st);
972unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
973 struct skb_seq_state *st);
974void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -0700975
Joe Perches7965bd42013-09-26 14:48:15 -0700976unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
Bojan Prtvar059a2442015-02-22 11:46:35 +0100977 unsigned int to, struct ts_config *config);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -0700978
Tom Herbert09323cc2013-12-15 22:16:19 -0800979/*
980 * Packet hash types specify the type of hash in skb_set_hash.
981 *
982 * Hash types refer to the protocol layer addresses which are used to
983 * construct a packet's hash. The hashes are used to differentiate or identify
984 * flows of the protocol layer for the hash type. Hash types are either
985 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
986 *
987 * Properties of hashes:
988 *
989 * 1) Two packets in different flows have different hash values
990 * 2) Two packets in the same flow should have the same hash value
991 *
992 * A hash at a higher layer is considered to be more specific. A driver should
993 * set the most specific hash possible.
994 *
995 * A driver cannot indicate a more specific hash than the layer at which a hash
996 * was computed. For instance an L3 hash cannot be set as an L4 hash.
997 *
998 * A driver may indicate a hash level which is less specific than the
999 * actual layer the hash was computed on. For instance, a hash computed
1000 * at L4 may be considered an L3 hash. This should only be done if the
1001 * driver can't unambiguously determine that the HW computed the hash at
1002 * the higher layer. Note that the "should" in the second property above
1003 * permits this.
1004 */
1005enum pkt_hash_types {
1006 PKT_HASH_TYPE_NONE, /* Undefined type */
1007 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1008 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1009 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1010};
1011
Tom Herbertbcc83832015-09-01 09:24:24 -07001012static inline void skb_clear_hash(struct sk_buff *skb)
1013{
1014 skb->hash = 0;
1015 skb->sw_hash = 0;
1016 skb->l4_hash = 0;
1017}
1018
1019static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1020{
1021 if (!skb->l4_hash)
1022 skb_clear_hash(skb);
1023}
1024
1025static inline void
1026__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1027{
1028 skb->l4_hash = is_l4;
1029 skb->sw_hash = is_sw;
1030 skb->hash = hash;
1031}
1032
Tom Herbert09323cc2013-12-15 22:16:19 -08001033static inline void
1034skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1035{
Tom Herbertbcc83832015-09-01 09:24:24 -07001036 /* Used by drivers to set hash from HW */
1037 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1038}
1039
1040static inline void
1041__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1042{
1043 __skb_set_hash(skb, hash, true, is_l4);
Tom Herbert09323cc2013-12-15 22:16:19 -08001044}
1045
Tom Herberte5276932015-09-01 09:24:23 -07001046void __skb_get_hash(struct sk_buff *skb);
Florian Westphalb9177832016-10-26 18:49:46 +02001047u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
Tom Herberte5276932015-09-01 09:24:23 -07001048u32 skb_get_poff(const struct sk_buff *skb);
1049u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1050 const struct flow_keys *keys, int hlen);
1051__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1052 void *data, int hlen_proto);
1053
1054static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1055 int thoff, u8 ip_proto)
1056{
1057 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1058}
1059
1060void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1061 const struct flow_dissector_key *key,
1062 unsigned int key_count);
1063
1064bool __skb_flow_dissect(const struct sk_buff *skb,
1065 struct flow_dissector *flow_dissector,
1066 void *target_container,
Tom Herbertcd79a232015-09-01 09:24:27 -07001067 void *data, __be16 proto, int nhoff, int hlen,
1068 unsigned int flags);
Tom Herberte5276932015-09-01 09:24:23 -07001069
1070static inline bool skb_flow_dissect(const struct sk_buff *skb,
1071 struct flow_dissector *flow_dissector,
Tom Herbertcd79a232015-09-01 09:24:27 -07001072 void *target_container, unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001073{
1074 return __skb_flow_dissect(skb, flow_dissector, target_container,
Tom Herbertcd79a232015-09-01 09:24:27 -07001075 NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001076}
1077
1078static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
Tom Herbertcd79a232015-09-01 09:24:27 -07001079 struct flow_keys *flow,
1080 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001081{
1082 memset(flow, 0, sizeof(*flow));
1083 return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
Tom Herbertcd79a232015-09-01 09:24:27 -07001084 NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001085}
1086
1087static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
1088 void *data, __be16 proto,
Tom Herbertcd79a232015-09-01 09:24:27 -07001089 int nhoff, int hlen,
1090 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001091{
1092 memset(flow, 0, sizeof(*flow));
1093 return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
Tom Herbertcd79a232015-09-01 09:24:27 -07001094 data, proto, nhoff, hlen, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001095}
1096
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001097static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001098{
Tom Herberta3b18dd2014-07-01 21:33:17 -07001099 if (!skb->l4_hash && !skb->sw_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001100 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001101
Tom Herbert61b905d2014-03-24 15:34:47 -07001102 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001103}
1104
David S. Miller20a17bf2015-09-01 21:19:17 -07001105__u32 __skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6);
Tom Herbertf70ea012015-07-31 16:52:10 -07001106
David S. Miller20a17bf2015-09-01 21:19:17 -07001107static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
Tom Herbertf70ea012015-07-31 16:52:10 -07001108{
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001109 if (!skb->l4_hash && !skb->sw_hash) {
1110 struct flow_keys keys;
Tom Herbertde4c1f82015-09-01 18:11:04 -07001111 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001112
Tom Herbertde4c1f82015-09-01 18:11:04 -07001113 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001114 }
Tom Herbertf70ea012015-07-31 16:52:10 -07001115
1116 return skb->hash;
1117}
1118
David S. Miller20a17bf2015-09-01 21:19:17 -07001119__u32 __skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl);
Tom Herbertf70ea012015-07-31 16:52:10 -07001120
David S. Miller20a17bf2015-09-01 21:19:17 -07001121static inline __u32 skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl4)
Tom Herbertf70ea012015-07-31 16:52:10 -07001122{
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001123 if (!skb->l4_hash && !skb->sw_hash) {
1124 struct flow_keys keys;
Tom Herbertde4c1f82015-09-01 18:11:04 -07001125 __u32 hash = __get_hash_from_flowi4(fl4, &keys);
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001126
Tom Herbertde4c1f82015-09-01 18:11:04 -07001127 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001128 }
Tom Herbertf70ea012015-07-31 16:52:10 -07001129
1130 return skb->hash;
1131}
1132
Tom Herbert50fb7992015-05-01 11:30:12 -07001133__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
1134
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001135static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1136{
Tom Herbert61b905d2014-03-24 15:34:47 -07001137 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001138}
1139
Tom Herbert3df7a742013-12-15 22:16:29 -08001140static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1141{
Tom Herbert61b905d2014-03-24 15:34:47 -07001142 to->hash = from->hash;
Tom Herberta3b18dd2014-07-01 21:33:17 -07001143 to->sw_hash = from->sw_hash;
Tom Herbert61b905d2014-03-24 15:34:47 -07001144 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -08001145};
1146
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001147#ifdef NET_SKBUFF_DATA_USES_OFFSET
1148static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1149{
1150 return skb->head + skb->end;
1151}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001152
1153static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1154{
1155 return skb->end;
1156}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001157#else
1158static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1159{
1160 return skb->end;
1161}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001162
1163static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1164{
1165 return skb->end - skb->head;
1166}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001167#endif
1168
Linus Torvalds1da177e2005-04-16 15:20:36 -07001169/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001170#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001171
Patrick Ohlyac45f602009-02-12 05:03:37 +00001172static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1173{
1174 return &skb_shinfo(skb)->hwtstamps;
1175}
1176
Linus Torvalds1da177e2005-04-16 15:20:36 -07001177/**
1178 * skb_queue_empty - check if a queue is empty
1179 * @list: queue head
1180 *
1181 * Returns true if the queue is empty, false otherwise.
1182 */
1183static inline int skb_queue_empty(const struct sk_buff_head *list)
1184{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001185 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001186}
1187
1188/**
David S. Millerfc7ebb22008-09-23 00:34:07 -07001189 * skb_queue_is_last - check if skb is the last entry in the queue
1190 * @list: queue head
1191 * @skb: buffer
1192 *
1193 * Returns true if @skb is the last buffer on the list.
1194 */
1195static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1196 const struct sk_buff *skb)
1197{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001198 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -07001199}
1200
1201/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001202 * skb_queue_is_first - check if skb is the first entry in the queue
1203 * @list: queue head
1204 * @skb: buffer
1205 *
1206 * Returns true if @skb is the first buffer on the list.
1207 */
1208static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1209 const struct sk_buff *skb)
1210{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001211 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001212}
1213
1214/**
David S. Miller249c8b42008-09-23 00:44:42 -07001215 * skb_queue_next - return the next packet in the queue
1216 * @list: queue head
1217 * @skb: current buffer
1218 *
1219 * Return the next packet in @list after @skb. It is only valid to
1220 * call this if skb_queue_is_last() evaluates to false.
1221 */
1222static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1223 const struct sk_buff *skb)
1224{
1225 /* This BUG_ON may seem severe, but if we just return then we
1226 * are going to dereference garbage.
1227 */
1228 BUG_ON(skb_queue_is_last(list, skb));
1229 return skb->next;
1230}
1231
1232/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001233 * skb_queue_prev - return the prev packet in the queue
1234 * @list: queue head
1235 * @skb: current buffer
1236 *
1237 * Return the prev packet in @list before @skb. It is only valid to
1238 * call this if skb_queue_is_first() evaluates to false.
1239 */
1240static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1241 const struct sk_buff *skb)
1242{
1243 /* This BUG_ON may seem severe, but if we just return then we
1244 * are going to dereference garbage.
1245 */
1246 BUG_ON(skb_queue_is_first(list, skb));
1247 return skb->prev;
1248}
1249
1250/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001251 * skb_get - reference buffer
1252 * @skb: buffer to reference
1253 *
1254 * Makes another reference to a socket buffer and returns a pointer
1255 * to the buffer.
1256 */
1257static inline struct sk_buff *skb_get(struct sk_buff *skb)
1258{
1259 atomic_inc(&skb->users);
1260 return skb;
1261}
1262
1263/*
1264 * If users == 1, we are the only owner and are can avoid redundant
1265 * atomic change.
1266 */
1267
1268/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001269 * skb_cloned - is the buffer a clone
1270 * @skb: buffer to check
1271 *
1272 * Returns true if the buffer was generated with skb_clone() and is
1273 * one of multiple shared copies of the buffer. Cloned buffers are
1274 * shared data so must not be written to under normal circumstances.
1275 */
1276static inline int skb_cloned(const struct sk_buff *skb)
1277{
1278 return skb->cloned &&
1279 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1280}
1281
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001282static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1283{
Mel Gormand0164ad2015-11-06 16:28:21 -08001284 might_sleep_if(gfpflags_allow_blocking(pri));
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001285
1286 if (skb_cloned(skb))
1287 return pskb_expand_head(skb, 0, 0, pri);
1288
1289 return 0;
1290}
1291
Linus Torvalds1da177e2005-04-16 15:20:36 -07001292/**
1293 * skb_header_cloned - is the header a clone
1294 * @skb: buffer to check
1295 *
1296 * Returns true if modifying the header part of the buffer requires
1297 * the data to be copied.
1298 */
1299static inline int skb_header_cloned(const struct sk_buff *skb)
1300{
1301 int dataref;
1302
1303 if (!skb->cloned)
1304 return 0;
1305
1306 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1307 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1308 return dataref != 1;
1309}
1310
Eric Dumazet9580bf22016-04-30 10:19:29 -07001311static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1312{
1313 might_sleep_if(gfpflags_allow_blocking(pri));
1314
1315 if (skb_header_cloned(skb))
1316 return pskb_expand_head(skb, 0, 0, pri);
1317
1318 return 0;
1319}
1320
Linus Torvalds1da177e2005-04-16 15:20:36 -07001321/**
1322 * skb_header_release - release reference to header
1323 * @skb: buffer to operate on
1324 *
1325 * Drop a reference to the header part of the buffer. This is done
1326 * by acquiring a payload reference. You must not read from the header
1327 * part of skb->data after this.
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001328 * Note : Check if you can use __skb_header_release() instead.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001329 */
1330static inline void skb_header_release(struct sk_buff *skb)
1331{
1332 BUG_ON(skb->nohdr);
1333 skb->nohdr = 1;
1334 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1335}
1336
1337/**
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001338 * __skb_header_release - release reference to header
1339 * @skb: buffer to operate on
1340 *
1341 * Variant of skb_header_release() assuming skb is private to caller.
1342 * We can avoid one atomic operation.
1343 */
1344static inline void __skb_header_release(struct sk_buff *skb)
1345{
1346 skb->nohdr = 1;
1347 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1348}
1349
1350
1351/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352 * skb_shared - is the buffer shared
1353 * @skb: buffer to check
1354 *
1355 * Returns true if more than one person has a reference to this
1356 * buffer.
1357 */
1358static inline int skb_shared(const struct sk_buff *skb)
1359{
1360 return atomic_read(&skb->users) != 1;
1361}
1362
1363/**
1364 * skb_share_check - check if buffer is shared and if so clone it
1365 * @skb: buffer to check
1366 * @pri: priority for memory allocation
1367 *
1368 * If the buffer is shared the buffer is cloned and the old copy
1369 * drops a reference. A new clone with a single reference is returned.
1370 * If the buffer is not shared the original buffer is returned. When
1371 * being called from interrupt status or with spinlocks held pri must
1372 * be GFP_ATOMIC.
1373 *
1374 * NULL is returned on a memory allocation failure.
1375 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001376static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001377{
Mel Gormand0164ad2015-11-06 16:28:21 -08001378 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001379 if (skb_shared(skb)) {
1380 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001381
1382 if (likely(nskb))
1383 consume_skb(skb);
1384 else
1385 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001386 skb = nskb;
1387 }
1388 return skb;
1389}
1390
1391/*
1392 * Copy shared buffers into a new sk_buff. We effectively do COW on
1393 * packets to handle cases where we have a local reader and forward
1394 * and a couple of other messy ones. The normal one is tcpdumping
1395 * a packet thats being forwarded.
1396 */
1397
1398/**
1399 * skb_unshare - make a copy of a shared buffer
1400 * @skb: buffer to check
1401 * @pri: priority for memory allocation
1402 *
1403 * If the socket buffer is a clone then this function creates a new
1404 * copy of the data, drops a reference count on the old copy and returns
1405 * the new copy with the reference count at 1. If the buffer is not a clone
1406 * the original buffer is returned. When called with a spinlock held or
1407 * from interrupt state @pri must be %GFP_ATOMIC
1408 *
1409 * %NULL is returned on a memory allocation failure.
1410 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001411static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001412 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001413{
Mel Gormand0164ad2015-11-06 16:28:21 -08001414 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001415 if (skb_cloned(skb)) {
1416 struct sk_buff *nskb = skb_copy(skb, pri);
Alexander Aring31eff812014-10-10 23:10:47 +02001417
1418 /* Free our shared copy */
1419 if (likely(nskb))
1420 consume_skb(skb);
1421 else
1422 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001423 skb = nskb;
1424 }
1425 return skb;
1426}
1427
1428/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001429 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001430 * @list_: list to peek at
1431 *
1432 * Peek an &sk_buff. Unlike most other operations you _MUST_
1433 * be careful with this one. A peek leaves the buffer on the
1434 * list and someone else may run off with it. You must hold
1435 * the appropriate locks or have a private queue to do this.
1436 *
1437 * Returns %NULL for an empty list or a pointer to the head element.
1438 * The reference count is not incremented and the reference is therefore
1439 * volatile. Use with caution.
1440 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001441static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001442{
Eric Dumazet18d07002012-04-30 16:31:46 +00001443 struct sk_buff *skb = list_->next;
1444
1445 if (skb == (struct sk_buff *)list_)
1446 skb = NULL;
1447 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001448}
1449
1450/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001451 * skb_peek_next - peek skb following the given one from a queue
1452 * @skb: skb to start from
1453 * @list_: list to peek at
1454 *
1455 * Returns %NULL when the end of the list is met or a pointer to the
1456 * next element. The reference count is not incremented and the
1457 * reference is therefore volatile. Use with caution.
1458 */
1459static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1460 const struct sk_buff_head *list_)
1461{
1462 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001463
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001464 if (next == (struct sk_buff *)list_)
1465 next = NULL;
1466 return next;
1467}
1468
1469/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001470 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001471 * @list_: list to peek at
1472 *
1473 * Peek an &sk_buff. Unlike most other operations you _MUST_
1474 * be careful with this one. A peek leaves the buffer on the
1475 * list and someone else may run off with it. You must hold
1476 * the appropriate locks or have a private queue to do this.
1477 *
1478 * Returns %NULL for an empty list or a pointer to the tail element.
1479 * The reference count is not incremented and the reference is therefore
1480 * volatile. Use with caution.
1481 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001482static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001483{
Eric Dumazet18d07002012-04-30 16:31:46 +00001484 struct sk_buff *skb = list_->prev;
1485
1486 if (skb == (struct sk_buff *)list_)
1487 skb = NULL;
1488 return skb;
1489
Linus Torvalds1da177e2005-04-16 15:20:36 -07001490}
1491
1492/**
1493 * skb_queue_len - get queue length
1494 * @list_: list to measure
1495 *
1496 * Return the length of an &sk_buff queue.
1497 */
1498static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1499{
1500 return list_->qlen;
1501}
1502
David S. Miller67fed452008-09-21 22:36:24 -07001503/**
1504 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1505 * @list: queue to initialize
1506 *
1507 * This initializes only the list and queue length aspects of
1508 * an sk_buff_head object. This allows to initialize the list
1509 * aspects of an sk_buff_head without reinitializing things like
1510 * the spinlock. It can also be used for on-stack sk_buff_head
1511 * objects where the spinlock is known to not be used.
1512 */
1513static inline void __skb_queue_head_init(struct sk_buff_head *list)
1514{
1515 list->prev = list->next = (struct sk_buff *)list;
1516 list->qlen = 0;
1517}
1518
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001519/*
1520 * This function creates a split out lock class for each invocation;
1521 * this is needed for now since a whole lot of users of the skb-queue
1522 * infrastructure in drivers have different locking usage (in hardirq)
1523 * than the networking core (in softirq only). In the long run either the
1524 * network layer or drivers should need annotation to consolidate the
1525 * main types of usage into 3 classes.
1526 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001527static inline void skb_queue_head_init(struct sk_buff_head *list)
1528{
1529 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001530 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531}
1532
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001533static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1534 struct lock_class_key *class)
1535{
1536 skb_queue_head_init(list);
1537 lockdep_set_class(&list->lock, class);
1538}
1539
Linus Torvalds1da177e2005-04-16 15:20:36 -07001540/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001541 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001542 *
1543 * The "__skb_xxxx()" functions are the non-atomic ones that
1544 * can only be called with interrupts disabled.
1545 */
Joe Perches7965bd42013-09-26 14:48:15 -07001546void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1547 struct sk_buff_head *list);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001548static inline void __skb_insert(struct sk_buff *newsk,
1549 struct sk_buff *prev, struct sk_buff *next,
1550 struct sk_buff_head *list)
1551{
1552 newsk->next = next;
1553 newsk->prev = prev;
1554 next->prev = prev->next = newsk;
1555 list->qlen++;
1556}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001557
David S. Miller67fed452008-09-21 22:36:24 -07001558static inline void __skb_queue_splice(const struct sk_buff_head *list,
1559 struct sk_buff *prev,
1560 struct sk_buff *next)
1561{
1562 struct sk_buff *first = list->next;
1563 struct sk_buff *last = list->prev;
1564
1565 first->prev = prev;
1566 prev->next = first;
1567
1568 last->next = next;
1569 next->prev = last;
1570}
1571
1572/**
1573 * skb_queue_splice - join two skb lists, this is designed for stacks
1574 * @list: the new list to add
1575 * @head: the place to add it in the first list
1576 */
1577static inline void skb_queue_splice(const struct sk_buff_head *list,
1578 struct sk_buff_head *head)
1579{
1580 if (!skb_queue_empty(list)) {
1581 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001582 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001583 }
1584}
1585
1586/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001587 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001588 * @list: the new list to add
1589 * @head: the place to add it in the first list
1590 *
1591 * The list at @list is reinitialised
1592 */
1593static inline void skb_queue_splice_init(struct sk_buff_head *list,
1594 struct sk_buff_head *head)
1595{
1596 if (!skb_queue_empty(list)) {
1597 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001598 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001599 __skb_queue_head_init(list);
1600 }
1601}
1602
1603/**
1604 * skb_queue_splice_tail - join two skb lists, each list being a queue
1605 * @list: the new list to add
1606 * @head: the place to add it in the first list
1607 */
1608static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1609 struct sk_buff_head *head)
1610{
1611 if (!skb_queue_empty(list)) {
1612 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001613 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001614 }
1615}
1616
1617/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001618 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001619 * @list: the new list to add
1620 * @head: the place to add it in the first list
1621 *
1622 * Each of the lists is a queue.
1623 * The list at @list is reinitialised
1624 */
1625static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1626 struct sk_buff_head *head)
1627{
1628 if (!skb_queue_empty(list)) {
1629 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001630 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001631 __skb_queue_head_init(list);
1632 }
1633}
1634
Linus Torvalds1da177e2005-04-16 15:20:36 -07001635/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001636 * __skb_queue_after - queue a buffer at the list head
1637 * @list: list to use
1638 * @prev: place after this buffer
1639 * @newsk: buffer to queue
1640 *
1641 * Queue a buffer int the middle of a list. This function takes no locks
1642 * and you must therefore hold required locks before calling it.
1643 *
1644 * A buffer cannot be placed on two lists at the same time.
1645 */
1646static inline void __skb_queue_after(struct sk_buff_head *list,
1647 struct sk_buff *prev,
1648 struct sk_buff *newsk)
1649{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001650 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001651}
1652
Joe Perches7965bd42013-09-26 14:48:15 -07001653void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1654 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001655
Gerrit Renkerf5572852008-04-14 00:05:28 -07001656static inline void __skb_queue_before(struct sk_buff_head *list,
1657 struct sk_buff *next,
1658 struct sk_buff *newsk)
1659{
1660 __skb_insert(newsk, next->prev, next, list);
1661}
1662
Stephen Hemminger300ce172005-10-30 13:47:34 -08001663/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001664 * __skb_queue_head - queue a buffer at the list head
1665 * @list: list to use
1666 * @newsk: buffer to queue
1667 *
1668 * Queue a buffer at the start of a list. This function takes no locks
1669 * and you must therefore hold required locks before calling it.
1670 *
1671 * A buffer cannot be placed on two lists at the same time.
1672 */
Joe Perches7965bd42013-09-26 14:48:15 -07001673void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001674static inline void __skb_queue_head(struct sk_buff_head *list,
1675 struct sk_buff *newsk)
1676{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001677 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001678}
1679
1680/**
1681 * __skb_queue_tail - queue a buffer at the list tail
1682 * @list: list to use
1683 * @newsk: buffer to queue
1684 *
1685 * Queue a buffer at the end of a list. This function takes no locks
1686 * and you must therefore hold required locks before calling it.
1687 *
1688 * A buffer cannot be placed on two lists at the same time.
1689 */
Joe Perches7965bd42013-09-26 14:48:15 -07001690void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001691static inline void __skb_queue_tail(struct sk_buff_head *list,
1692 struct sk_buff *newsk)
1693{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001694 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695}
1696
Linus Torvalds1da177e2005-04-16 15:20:36 -07001697/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001698 * remove sk_buff from list. _Must_ be called atomically, and with
1699 * the list known..
1700 */
Joe Perches7965bd42013-09-26 14:48:15 -07001701void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001702static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1703{
1704 struct sk_buff *next, *prev;
1705
1706 list->qlen--;
1707 next = skb->next;
1708 prev = skb->prev;
1709 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001710 next->prev = prev;
1711 prev->next = next;
1712}
1713
Gerrit Renkerf525c062008-04-14 00:04:12 -07001714/**
1715 * __skb_dequeue - remove from the head of the queue
1716 * @list: list to dequeue from
1717 *
1718 * Remove the head of the list. This function does not take any locks
1719 * so must be used with appropriate locks held only. The head item is
1720 * returned or %NULL if the list is empty.
1721 */
Joe Perches7965bd42013-09-26 14:48:15 -07001722struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Gerrit Renkerf525c062008-04-14 00:04:12 -07001723static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1724{
1725 struct sk_buff *skb = skb_peek(list);
1726 if (skb)
1727 __skb_unlink(skb, list);
1728 return skb;
1729}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001730
1731/**
1732 * __skb_dequeue_tail - remove from the tail of the queue
1733 * @list: list to dequeue from
1734 *
1735 * Remove the tail of the list. This function does not take any locks
1736 * so must be used with appropriate locks held only. The tail item is
1737 * returned or %NULL if the list is empty.
1738 */
Joe Perches7965bd42013-09-26 14:48:15 -07001739struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001740static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1741{
1742 struct sk_buff *skb = skb_peek_tail(list);
1743 if (skb)
1744 __skb_unlink(skb, list);
1745 return skb;
1746}
1747
1748
David S. Millerbdcc0922012-03-07 20:53:36 -05001749static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001750{
1751 return skb->data_len;
1752}
1753
1754static inline unsigned int skb_headlen(const struct sk_buff *skb)
1755{
1756 return skb->len - skb->data_len;
1757}
1758
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03001759static inline unsigned int skb_pagelen(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001760{
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03001761 unsigned int i, len = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001762
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03001763 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001764 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001765 return len + skb_headlen(skb);
1766}
1767
Ian Campbell131ea662011-08-19 06:25:00 +00001768/**
1769 * __skb_fill_page_desc - initialise a paged fragment in an skb
1770 * @skb: buffer containing fragment to be initialised
1771 * @i: paged fragment index to initialise
1772 * @page: the page to use for this fragment
1773 * @off: the offset to the data with @page
1774 * @size: the length of the data
1775 *
1776 * Initialises the @i'th fragment of @skb to point to &size bytes at
1777 * offset @off within @page.
1778 *
1779 * Does not take any additional reference on the fragment.
1780 */
1781static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1782 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001783{
1784 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1785
Mel Gormanc48a11c2012-07-31 16:44:23 -07001786 /*
Michal Hocko2f064f32015-08-21 14:11:51 -07001787 * Propagate page pfmemalloc to the skb if we can. The problem is
1788 * that not all callers have unique ownership of the page but rely
1789 * on page_is_pfmemalloc doing the right thing(tm).
Mel Gormanc48a11c2012-07-31 16:44:23 -07001790 */
Ian Campbella8605c62011-10-19 23:01:49 +00001791 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001792 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001793 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001794
1795 page = compound_head(page);
Michal Hocko2f064f32015-08-21 14:11:51 -07001796 if (page_is_pfmemalloc(page))
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001797 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00001798}
1799
1800/**
1801 * skb_fill_page_desc - initialise a paged fragment in an skb
1802 * @skb: buffer containing fragment to be initialised
1803 * @i: paged fragment index to initialise
1804 * @page: the page to use for this fragment
1805 * @off: the offset to the data with @page
1806 * @size: the length of the data
1807 *
1808 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01001809 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00001810 * addition updates @skb such that @i is the last fragment.
1811 *
1812 * Does not take any additional reference on the fragment.
1813 */
1814static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1815 struct page *page, int off, int size)
1816{
1817 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001818 skb_shinfo(skb)->nr_frags = i + 1;
1819}
1820
Joe Perches7965bd42013-09-26 14:48:15 -07001821void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1822 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001823
Jason Wangf8e617e2013-11-01 14:07:47 +08001824void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1825 unsigned int truesize);
1826
Linus Torvalds1da177e2005-04-16 15:20:36 -07001827#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001828#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001829#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1830
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001831#ifdef NET_SKBUFF_DATA_USES_OFFSET
1832static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1833{
1834 return skb->head + skb->tail;
1835}
1836
1837static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1838{
1839 skb->tail = skb->data - skb->head;
1840}
1841
1842static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1843{
1844 skb_reset_tail_pointer(skb);
1845 skb->tail += offset;
1846}
Simon Horman7cc46192013-05-28 20:34:29 +00001847
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001848#else /* NET_SKBUFF_DATA_USES_OFFSET */
1849static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1850{
1851 return skb->tail;
1852}
1853
1854static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1855{
1856 skb->tail = skb->data;
1857}
1858
1859static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1860{
1861 skb->tail = skb->data + offset;
1862}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001863
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001864#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1865
Linus Torvalds1da177e2005-04-16 15:20:36 -07001866/*
1867 * Add data to an sk_buff
1868 */
Mathias Krause0c7ddf32013-11-07 14:18:24 +01001869unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001870unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001871static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1872{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001873 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001874 SKB_LINEAR_ASSERT(skb);
1875 skb->tail += len;
1876 skb->len += len;
1877 return tmp;
1878}
1879
Joe Perches7965bd42013-09-26 14:48:15 -07001880unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001881static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1882{
1883 skb->data -= len;
1884 skb->len += len;
1885 return skb->data;
1886}
1887
Joe Perches7965bd42013-09-26 14:48:15 -07001888unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001889static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1890{
1891 skb->len -= len;
1892 BUG_ON(skb->len < skb->data_len);
1893 return skb->data += len;
1894}
1895
David S. Miller47d29642010-05-02 02:21:44 -07001896static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1897{
1898 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1899}
1900
Joe Perches7965bd42013-09-26 14:48:15 -07001901unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001902
1903static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1904{
1905 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001906 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001907 return NULL;
1908 skb->len -= len;
1909 return skb->data += len;
1910}
1911
1912static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1913{
1914 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1915}
1916
1917static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1918{
1919 if (likely(len <= skb_headlen(skb)))
1920 return 1;
1921 if (unlikely(len > skb->len))
1922 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001923 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001924}
1925
Eric Dumazetc8c8b122016-12-07 09:19:33 -08001926void skb_condense(struct sk_buff *skb);
1927
Linus Torvalds1da177e2005-04-16 15:20:36 -07001928/**
1929 * skb_headroom - bytes at buffer head
1930 * @skb: buffer to check
1931 *
1932 * Return the number of bytes of free space at the head of an &sk_buff.
1933 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001934static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001935{
1936 return skb->data - skb->head;
1937}
1938
1939/**
1940 * skb_tailroom - bytes at buffer end
1941 * @skb: buffer to check
1942 *
1943 * Return the number of bytes of free space at the tail of an sk_buff
1944 */
1945static inline int skb_tailroom(const struct sk_buff *skb)
1946{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001947 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001948}
1949
1950/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00001951 * skb_availroom - bytes at buffer end
1952 * @skb: buffer to check
1953 *
1954 * Return the number of bytes of free space at the tail of an sk_buff
1955 * allocated by sk_stream_alloc()
1956 */
1957static inline int skb_availroom(const struct sk_buff *skb)
1958{
Eric Dumazet16fad692013-03-14 05:40:32 +00001959 if (skb_is_nonlinear(skb))
1960 return 0;
1961
1962 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00001963}
1964
1965/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001966 * skb_reserve - adjust headroom
1967 * @skb: buffer to alter
1968 * @len: bytes to move
1969 *
1970 * Increase the headroom of an empty &sk_buff by reducing the tail
1971 * room. This is only allowed for an empty buffer.
1972 */
David S. Miller8243126c2006-01-17 02:54:21 -08001973static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001974{
1975 skb->data += len;
1976 skb->tail += len;
1977}
1978
Benjamin Poirier1837b2e2016-02-29 15:03:33 -08001979/**
1980 * skb_tailroom_reserve - adjust reserved_tailroom
1981 * @skb: buffer to alter
1982 * @mtu: maximum amount of headlen permitted
1983 * @needed_tailroom: minimum amount of reserved_tailroom
1984 *
1985 * Set reserved_tailroom so that headlen can be as large as possible but
1986 * not larger than mtu and tailroom cannot be smaller than
1987 * needed_tailroom.
1988 * The required headroom should already have been reserved before using
1989 * this function.
1990 */
1991static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
1992 unsigned int needed_tailroom)
1993{
1994 SKB_LINEAR_ASSERT(skb);
1995 if (mtu < skb_tailroom(skb) - needed_tailroom)
1996 /* use at most mtu */
1997 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
1998 else
1999 /* use up to all available space */
2000 skb->reserved_tailroom = needed_tailroom;
2001}
2002
Tom Herbert8bce6d72014-09-29 20:22:29 -07002003#define ENCAP_TYPE_ETHER 0
2004#define ENCAP_TYPE_IPPROTO 1
2005
2006static inline void skb_set_inner_protocol(struct sk_buff *skb,
2007 __be16 protocol)
2008{
2009 skb->inner_protocol = protocol;
2010 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2011}
2012
2013static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2014 __u8 ipproto)
2015{
2016 skb->inner_ipproto = ipproto;
2017 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2018}
2019
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002020static inline void skb_reset_inner_headers(struct sk_buff *skb)
2021{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002022 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002023 skb->inner_network_header = skb->network_header;
2024 skb->inner_transport_header = skb->transport_header;
2025}
2026
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00002027static inline void skb_reset_mac_len(struct sk_buff *skb)
2028{
2029 skb->mac_len = skb->network_header - skb->mac_header;
2030}
2031
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002032static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2033 *skb)
2034{
2035 return skb->head + skb->inner_transport_header;
2036}
2037
Tom Herbert55dc5a92015-12-14 11:19:40 -08002038static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2039{
2040 return skb_inner_transport_header(skb) - skb->data;
2041}
2042
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002043static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2044{
2045 skb->inner_transport_header = skb->data - skb->head;
2046}
2047
2048static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2049 const int offset)
2050{
2051 skb_reset_inner_transport_header(skb);
2052 skb->inner_transport_header += offset;
2053}
2054
2055static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2056{
2057 return skb->head + skb->inner_network_header;
2058}
2059
2060static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2061{
2062 skb->inner_network_header = skb->data - skb->head;
2063}
2064
2065static inline void skb_set_inner_network_header(struct sk_buff *skb,
2066 const int offset)
2067{
2068 skb_reset_inner_network_header(skb);
2069 skb->inner_network_header += offset;
2070}
2071
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002072static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2073{
2074 return skb->head + skb->inner_mac_header;
2075}
2076
2077static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2078{
2079 skb->inner_mac_header = skb->data - skb->head;
2080}
2081
2082static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2083 const int offset)
2084{
2085 skb_reset_inner_mac_header(skb);
2086 skb->inner_mac_header += offset;
2087}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002088static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2089{
Cong Wang35d04612013-05-29 15:16:05 +08002090 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002091}
2092
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002093static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2094{
2095 return skb->head + skb->transport_header;
2096}
2097
2098static inline void skb_reset_transport_header(struct sk_buff *skb)
2099{
2100 skb->transport_header = skb->data - skb->head;
2101}
2102
2103static inline void skb_set_transport_header(struct sk_buff *skb,
2104 const int offset)
2105{
2106 skb_reset_transport_header(skb);
2107 skb->transport_header += offset;
2108}
2109
2110static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2111{
2112 return skb->head + skb->network_header;
2113}
2114
2115static inline void skb_reset_network_header(struct sk_buff *skb)
2116{
2117 skb->network_header = skb->data - skb->head;
2118}
2119
2120static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2121{
2122 skb_reset_network_header(skb);
2123 skb->network_header += offset;
2124}
2125
2126static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2127{
2128 return skb->head + skb->mac_header;
2129}
2130
Amir Vadaiea6da4f2017-02-07 09:56:06 +02002131static inline int skb_mac_offset(const struct sk_buff *skb)
2132{
2133 return skb_mac_header(skb) - skb->data;
2134}
2135
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002136static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2137{
Cong Wang35d04612013-05-29 15:16:05 +08002138 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002139}
2140
2141static inline void skb_reset_mac_header(struct sk_buff *skb)
2142{
2143 skb->mac_header = skb->data - skb->head;
2144}
2145
2146static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2147{
2148 skb_reset_mac_header(skb);
2149 skb->mac_header += offset;
2150}
2151
Timo Teräs0e3da5b2013-12-16 11:02:09 +02002152static inline void skb_pop_mac_header(struct sk_buff *skb)
2153{
2154 skb->mac_header = skb->network_header;
2155}
2156
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002157static inline void skb_probe_transport_header(struct sk_buff *skb,
2158 const int offset_hint)
2159{
2160 struct flow_keys keys;
2161
2162 if (skb_transport_header_was_set(skb))
2163 return;
Tom Herbertcd79a232015-09-01 09:24:27 -07002164 else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
Tom Herbert42aecaa2015-06-04 09:16:39 -07002165 skb_set_transport_header(skb, keys.control.thoff);
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002166 else
2167 skb_set_transport_header(skb, offset_hint);
2168}
2169
Eric Dumazet03606892012-02-23 10:55:02 +00002170static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2171{
2172 if (skb_mac_header_was_set(skb)) {
2173 const unsigned char *old_mac = skb_mac_header(skb);
2174
2175 skb_set_mac_header(skb, -skb->mac_len);
2176 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2177 }
2178}
2179
Michał Mirosław04fb4512010-12-14 15:24:08 +00002180static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2181{
2182 return skb->csum_start - skb_headroom(skb);
2183}
2184
Alexander Duyck08b64fc2016-02-05 15:27:49 -08002185static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2186{
2187 return skb->head + skb->csum_start;
2188}
2189
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002190static inline int skb_transport_offset(const struct sk_buff *skb)
2191{
2192 return skb_transport_header(skb) - skb->data;
2193}
2194
2195static inline u32 skb_network_header_len(const struct sk_buff *skb)
2196{
2197 return skb->transport_header - skb->network_header;
2198}
2199
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002200static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2201{
2202 return skb->inner_transport_header - skb->inner_network_header;
2203}
2204
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002205static inline int skb_network_offset(const struct sk_buff *skb)
2206{
2207 return skb_network_header(skb) - skb->data;
2208}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03002209
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002210static inline int skb_inner_network_offset(const struct sk_buff *skb)
2211{
2212 return skb_inner_network_header(skb) - skb->data;
2213}
2214
Changli Gaof9599ce2010-08-04 04:43:44 +00002215static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2216{
2217 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2218}
2219
Linus Torvalds1da177e2005-04-16 15:20:36 -07002220/*
2221 * CPUs often take a performance hit when accessing unaligned memory
2222 * locations. The actual performance hit varies, it can be small if the
2223 * hardware handles it or large if we have to take an exception and fix it
2224 * in software.
2225 *
2226 * Since an ethernet header is 14 bytes network drivers often end up with
2227 * the IP header at an unaligned offset. The IP header can be aligned by
2228 * shifting the start of the packet by 2 bytes. Drivers should do this
2229 * with:
2230 *
Tobias Klauser8660c122009-07-13 22:48:16 +00002231 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002232 *
2233 * The downside to this alignment of the IP header is that the DMA is now
2234 * unaligned. On some architectures the cost of an unaligned DMA is high
2235 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00002236 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002237 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2238 * to be overridden.
2239 */
2240#ifndef NET_IP_ALIGN
2241#define NET_IP_ALIGN 2
2242#endif
2243
Anton Blanchard025be812006-03-31 02:27:06 -08002244/*
2245 * The networking layer reserves some headroom in skb data (via
2246 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2247 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08002248 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08002249 *
2250 * Unfortunately this headroom changes the DMA alignment of the resulting
2251 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2252 * on some architectures. An architecture can override this value,
2253 * perhaps setting it to a cacheline in size (since that will maintain
2254 * cacheline alignment of the DMA). It must be a power of 2.
2255 *
David S. Millerd6301d32009-02-08 19:24:13 -08002256 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08002257 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07002258 *
2259 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2260 * to reduce average number of cache lines per packet.
2261 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07002262 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08002263 */
2264#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07002265#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08002266#endif
2267
Joe Perches7965bd42013-09-26 14:48:15 -07002268int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002269
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002270static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002271{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00002272 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07002273 WARN_ON(1);
2274 return;
2275 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002276 skb->len = len;
2277 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002278}
2279
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002280static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2281{
2282 __skb_set_length(skb, len);
2283}
2284
Joe Perches7965bd42013-09-26 14:48:15 -07002285void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002286
2287static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2288{
Herbert Xu3cc0e872006-06-09 16:13:38 -07002289 if (skb->data_len)
2290 return ___pskb_trim(skb, len);
2291 __skb_trim(skb, len);
2292 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002293}
2294
2295static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2296{
2297 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2298}
2299
2300/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07002301 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2302 * @skb: buffer to alter
2303 * @len: new length
2304 *
2305 * This is identical to pskb_trim except that the caller knows that
2306 * the skb is not cloned so we should never get an error due to out-
2307 * of-memory.
2308 */
2309static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2310{
2311 int err = pskb_trim(skb, len);
2312 BUG_ON(err);
2313}
2314
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002315static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2316{
2317 unsigned int diff = len - skb->len;
2318
2319 if (skb_tailroom(skb) < diff) {
2320 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2321 GFP_ATOMIC);
2322 if (ret)
2323 return ret;
2324 }
2325 __skb_set_length(skb, len);
2326 return 0;
2327}
2328
Herbert Xue9fa4f72006-08-13 20:12:58 -07002329/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002330 * skb_orphan - orphan a buffer
2331 * @skb: buffer to orphan
2332 *
2333 * If a buffer currently has an owner then we call the owner's
2334 * destructor function and make the @skb unowned. The buffer continues
2335 * to exist but is no longer charged to its former owner.
2336 */
2337static inline void skb_orphan(struct sk_buff *skb)
2338{
Eric Dumazetc34a7612013-07-30 16:11:15 -07002339 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002340 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002341 skb->destructor = NULL;
2342 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07002343 } else {
2344 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002345 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002346}
2347
2348/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002349 * skb_orphan_frags - orphan the frags contained in a buffer
2350 * @skb: buffer to orphan frags from
2351 * @gfp_mask: allocation mask for replacement pages
2352 *
2353 * For each frag in the SKB which needs a destructor (i.e. has an
2354 * owner) create a copy of that frag and release the original
2355 * page by calling the destructor.
2356 */
2357static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2358{
2359 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2360 return 0;
2361 return skb_copy_ubufs(skb, gfp_mask);
2362}
2363
2364/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002365 * __skb_queue_purge - empty a list
2366 * @list: list to empty
2367 *
2368 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2369 * the list and one reference dropped. This function does not take the
2370 * list lock and the caller must hold the relevant locks to use it.
2371 */
Joe Perches7965bd42013-09-26 14:48:15 -07002372void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002373static inline void __skb_queue_purge(struct sk_buff_head *list)
2374{
2375 struct sk_buff *skb;
2376 while ((skb = __skb_dequeue(list)) != NULL)
2377 kfree_skb(skb);
2378}
2379
Yaogong Wang9f5afea2016-09-07 14:49:28 -07002380void skb_rbtree_purge(struct rb_root *root);
2381
Joe Perches7965bd42013-09-26 14:48:15 -07002382void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002383
Joe Perches7965bd42013-09-26 14:48:15 -07002384struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2385 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07002386
2387/**
2388 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2389 * @dev: network device to receive on
2390 * @length: length to allocate
2391 *
2392 * Allocate a new &sk_buff and assign it a usage count of one. The
2393 * buffer has unspecified headroom built in. Users should allocate
2394 * the headroom they think they need without accounting for the
2395 * built in space. The built in space is used for optimisations.
2396 *
2397 * %NULL is returned if there is no free memory. Although this function
2398 * allocates memory it can be called from an interrupt.
2399 */
2400static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002401 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002402{
2403 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2404}
2405
Eric Dumazet6f532612012-05-18 05:12:12 +00002406/* legacy helper around __netdev_alloc_skb() */
2407static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2408 gfp_t gfp_mask)
2409{
2410 return __netdev_alloc_skb(NULL, length, gfp_mask);
2411}
2412
2413/* legacy helper around netdev_alloc_skb() */
2414static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2415{
2416 return netdev_alloc_skb(NULL, length);
2417}
2418
2419
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002420static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2421 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002422{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002423 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002424
2425 if (NET_IP_ALIGN && skb)
2426 skb_reserve(skb, NET_IP_ALIGN);
2427 return skb;
2428}
2429
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002430static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2431 unsigned int length)
2432{
2433 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2434}
2435
Alexander Duyck181edb22015-05-06 21:12:03 -07002436static inline void skb_free_frag(void *addr)
2437{
Alexander Duyck8c2dd3e2017-01-10 16:58:06 -08002438 page_frag_free(addr);
Alexander Duyck181edb22015-05-06 21:12:03 -07002439}
2440
Alexander Duyckffde7322014-12-09 19:40:42 -08002441void *napi_alloc_frag(unsigned int fragsz);
Alexander Duyckfd11a832014-12-09 19:40:49 -08002442struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2443 unsigned int length, gfp_t gfp_mask);
2444static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2445 unsigned int length)
2446{
2447 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2448}
Jesper Dangaard Brouer795bb1c2016-02-08 13:14:59 +01002449void napi_consume_skb(struct sk_buff *skb, int budget);
2450
2451void __kfree_skb_flush(void);
Jesper Dangaard Brouer15fad712016-02-08 13:15:04 +01002452void __kfree_skb_defer(struct sk_buff *skb);
Alexander Duyckffde7322014-12-09 19:40:42 -08002453
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002454/**
Alexander Duyck71dfda52014-11-11 09:26:34 -08002455 * __dev_alloc_pages - allocate page for network Rx
2456 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2457 * @order: size of the allocation
2458 *
2459 * Allocate a new page.
2460 *
2461 * %NULL is returned if there is no free memory.
2462*/
2463static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2464 unsigned int order)
2465{
2466 /* This piece of code contains several assumptions.
2467 * 1. This is for device Rx, therefor a cold page is preferred.
2468 * 2. The expectation is the user wants a compound page.
2469 * 3. If requesting a order 0 page it will not be compound
2470 * due to the check to see if order has a value in prep_new_page
2471 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2472 * code in gfp_to_alloc_flags that should be enforcing this.
2473 */
2474 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2475
2476 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2477}
2478
2479static inline struct page *dev_alloc_pages(unsigned int order)
2480{
Neil Horman95829b32016-05-19 11:30:54 -04002481 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002482}
2483
2484/**
2485 * __dev_alloc_page - allocate a page for network Rx
2486 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2487 *
2488 * Allocate a new page.
2489 *
2490 * %NULL is returned if there is no free memory.
2491 */
2492static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2493{
2494 return __dev_alloc_pages(gfp_mask, 0);
2495}
2496
2497static inline struct page *dev_alloc_page(void)
2498{
Neil Horman95829b32016-05-19 11:30:54 -04002499 return dev_alloc_pages(0);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002500}
2501
2502/**
Mel Gorman06140022012-07-31 16:44:24 -07002503 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2504 * @page: The page that was allocated from skb_alloc_page
2505 * @skb: The skb that may need pfmemalloc set
2506 */
2507static inline void skb_propagate_pfmemalloc(struct page *page,
2508 struct sk_buff *skb)
2509{
Michal Hocko2f064f32015-08-21 14:11:51 -07002510 if (page_is_pfmemalloc(page))
Mel Gorman06140022012-07-31 16:44:24 -07002511 skb->pfmemalloc = true;
2512}
2513
Eric Dumazet564824b2010-10-11 19:05:25 +00002514/**
Masanari Iidae2278672014-02-18 22:54:36 +09002515 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002516 * @frag: the paged fragment
2517 *
2518 * Returns the &struct page associated with @frag.
2519 */
2520static inline struct page *skb_frag_page(const skb_frag_t *frag)
2521{
Ian Campbella8605c62011-10-19 23:01:49 +00002522 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00002523}
2524
2525/**
2526 * __skb_frag_ref - take an addition reference on a paged fragment.
2527 * @frag: the paged fragment
2528 *
2529 * Takes an additional reference on the paged fragment @frag.
2530 */
2531static inline void __skb_frag_ref(skb_frag_t *frag)
2532{
2533 get_page(skb_frag_page(frag));
2534}
2535
2536/**
2537 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2538 * @skb: the buffer
2539 * @f: the fragment offset.
2540 *
2541 * Takes an additional reference on the @f'th paged fragment of @skb.
2542 */
2543static inline void skb_frag_ref(struct sk_buff *skb, int f)
2544{
2545 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2546}
2547
2548/**
2549 * __skb_frag_unref - release a reference on a paged fragment.
2550 * @frag: the paged fragment
2551 *
2552 * Releases a reference on the paged fragment @frag.
2553 */
2554static inline void __skb_frag_unref(skb_frag_t *frag)
2555{
2556 put_page(skb_frag_page(frag));
2557}
2558
2559/**
2560 * skb_frag_unref - release a reference on a paged fragment of an skb.
2561 * @skb: the buffer
2562 * @f: the fragment offset
2563 *
2564 * Releases a reference on the @f'th paged fragment of @skb.
2565 */
2566static inline void skb_frag_unref(struct sk_buff *skb, int f)
2567{
2568 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2569}
2570
2571/**
2572 * skb_frag_address - gets the address of the data contained in a paged fragment
2573 * @frag: the paged fragment buffer
2574 *
2575 * Returns the address of the data within @frag. The page must already
2576 * be mapped.
2577 */
2578static inline void *skb_frag_address(const skb_frag_t *frag)
2579{
2580 return page_address(skb_frag_page(frag)) + frag->page_offset;
2581}
2582
2583/**
2584 * skb_frag_address_safe - 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. Checks that the page
2588 * is mapped and returns %NULL otherwise.
2589 */
2590static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2591{
2592 void *ptr = page_address(skb_frag_page(frag));
2593 if (unlikely(!ptr))
2594 return NULL;
2595
2596 return ptr + frag->page_offset;
2597}
2598
2599/**
2600 * __skb_frag_set_page - sets the page contained in a paged fragment
2601 * @frag: the paged fragment
2602 * @page: the page to set
2603 *
2604 * Sets the fragment @frag to contain @page.
2605 */
2606static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2607{
Ian Campbella8605c62011-10-19 23:01:49 +00002608 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002609}
2610
2611/**
2612 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2613 * @skb: the buffer
2614 * @f: the fragment offset
2615 * @page: the page to set
2616 *
2617 * Sets the @f'th fragment of @skb to contain @page.
2618 */
2619static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2620 struct page *page)
2621{
2622 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2623}
2624
Eric Dumazet400dfd32013-10-17 16:27:07 -07002625bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2626
Ian Campbell131ea662011-08-19 06:25:00 +00002627/**
2628 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002629 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002630 * @frag: the paged fragment to map
2631 * @offset: the offset within the fragment (starting at the
2632 * fragment's own offset)
2633 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002634 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00002635 *
2636 * Maps the page associated with @frag to @device.
2637 */
2638static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2639 const skb_frag_t *frag,
2640 size_t offset, size_t size,
2641 enum dma_data_direction dir)
2642{
2643 return dma_map_page(dev, skb_frag_page(frag),
2644 frag->page_offset + offset, size, dir);
2645}
2646
Eric Dumazet117632e2011-12-03 21:39:53 +00002647static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2648 gfp_t gfp_mask)
2649{
2650 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2651}
2652
Octavian Purdilabad93e92014-06-12 01:36:26 +03002653
2654static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2655 gfp_t gfp_mask)
2656{
2657 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2658}
2659
2660
Ian Campbell131ea662011-08-19 06:25:00 +00002661/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002662 * skb_clone_writable - is the header of a clone writable
2663 * @skb: buffer to check
2664 * @len: length up to which to write
2665 *
2666 * Returns true if modifying the header part of the cloned buffer
2667 * does not requires the data to be copied.
2668 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002669static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002670{
2671 return !skb_header_cloned(skb) &&
2672 skb_headroom(skb) + len <= skb->hdr_len;
2673}
2674
Daniel Borkmann36976492016-02-19 23:05:25 +01002675static inline int skb_try_make_writable(struct sk_buff *skb,
2676 unsigned int write_len)
2677{
2678 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
2679 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
2680}
2681
Herbert Xud9cc2042007-09-16 16:21:16 -07002682static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2683 int cloned)
2684{
2685 int delta = 0;
2686
Herbert Xud9cc2042007-09-16 16:21:16 -07002687 if (headroom > skb_headroom(skb))
2688 delta = headroom - skb_headroom(skb);
2689
2690 if (delta || cloned)
2691 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2692 GFP_ATOMIC);
2693 return 0;
2694}
2695
Patrick McHardy334a8132007-06-25 04:35:20 -07002696/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002697 * skb_cow - copy header of skb when it is required
2698 * @skb: buffer to cow
2699 * @headroom: needed headroom
2700 *
2701 * If the skb passed lacks sufficient headroom or its data part
2702 * is shared, data is reallocated. If reallocation fails, an error
2703 * is returned and original skb is not changed.
2704 *
2705 * The result is skb with writable area skb->head...skb->tail
2706 * and at least @headroom of space at head.
2707 */
2708static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2709{
Herbert Xud9cc2042007-09-16 16:21:16 -07002710 return __skb_cow(skb, headroom, skb_cloned(skb));
2711}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002712
Herbert Xud9cc2042007-09-16 16:21:16 -07002713/**
2714 * skb_cow_head - skb_cow but only making the head writable
2715 * @skb: buffer to cow
2716 * @headroom: needed headroom
2717 *
2718 * This function is identical to skb_cow except that we replace the
2719 * skb_cloned check by skb_header_cloned. It should be used when
2720 * you only need to push on some header and do not need to modify
2721 * the data.
2722 */
2723static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2724{
2725 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002726}
2727
2728/**
2729 * skb_padto - pad an skbuff up to a minimal size
2730 * @skb: buffer to pad
2731 * @len: minimal length
2732 *
2733 * Pads up a buffer to ensure the trailing bytes exist and are
2734 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002735 * is untouched. Otherwise it is extended. Returns zero on
2736 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002737 */
Herbert Xu5b057c62006-06-23 02:06:41 -07002738static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002739{
2740 unsigned int size = skb->len;
2741 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002742 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002743 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002744}
2745
Alexander Duyck9c0c1122014-12-03 08:17:33 -08002746/**
2747 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2748 * @skb: buffer to pad
2749 * @len: minimal length
2750 *
2751 * Pads up a buffer to ensure the trailing bytes exist and are
2752 * blanked. If the buffer already contains sufficient data it
2753 * is untouched. Otherwise it is extended. Returns zero on
2754 * success. The skb is freed on error.
2755 */
2756static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2757{
2758 unsigned int size = skb->len;
2759
2760 if (unlikely(size < len)) {
2761 len -= size;
2762 if (skb_pad(skb, len))
2763 return -ENOMEM;
2764 __skb_put(skb, len);
2765 }
2766 return 0;
2767}
2768
Linus Torvalds1da177e2005-04-16 15:20:36 -07002769static inline int skb_add_data(struct sk_buff *skb,
Al Viroaf2b0402014-11-27 21:44:24 -05002770 struct iov_iter *from, int copy)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002771{
2772 const int off = skb->len;
2773
2774 if (skb->ip_summed == CHECKSUM_NONE) {
Al Viroaf2b0402014-11-27 21:44:24 -05002775 __wsum csum = 0;
Al Viro15e6cb42016-11-01 22:42:45 -04002776 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
2777 &csum, from)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002778 skb->csum = csum_block_add(skb->csum, csum, off);
2779 return 0;
2780 }
Al Viro15e6cb42016-11-01 22:42:45 -04002781 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002782 return 0;
2783
2784 __skb_trim(skb, off);
2785 return -EFAULT;
2786}
2787
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002788static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2789 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002790{
2791 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002792 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002793
Ian Campbellea2ab692011-08-22 23:44:58 +00002794 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002795 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002796 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002797 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002798}
2799
Herbert Xu364c6ba2006-06-09 16:10:40 -07002800static inline int __skb_linearize(struct sk_buff *skb)
2801{
2802 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2803}
2804
Linus Torvalds1da177e2005-04-16 15:20:36 -07002805/**
2806 * skb_linearize - convert paged skb to linear one
2807 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002808 *
2809 * If there is no free memory -ENOMEM is returned, otherwise zero
2810 * is returned and the old skb data released.
2811 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002812static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002813{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002814 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2815}
2816
2817/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00002818 * skb_has_shared_frag - can any frag be overwritten
2819 * @skb: buffer to test
2820 *
2821 * Return true if the skb has at least one frag that might be modified
2822 * by an external entity (as in vmsplice()/sendfile())
2823 */
2824static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2825{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00002826 return skb_is_nonlinear(skb) &&
2827 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00002828}
2829
2830/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07002831 * skb_linearize_cow - make sure skb is linear and writable
2832 * @skb: buffer to process
2833 *
2834 * If there is no free memory -ENOMEM is returned, otherwise zero
2835 * is returned and the old skb data released.
2836 */
2837static inline int skb_linearize_cow(struct sk_buff *skb)
2838{
2839 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2840 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002841}
2842
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02002843static __always_inline void
2844__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
2845 unsigned int off)
2846{
2847 if (skb->ip_summed == CHECKSUM_COMPLETE)
2848 skb->csum = csum_block_sub(skb->csum,
2849 csum_partial(start, len, 0), off);
2850 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
2851 skb_checksum_start_offset(skb) < 0)
2852 skb->ip_summed = CHECKSUM_NONE;
2853}
2854
Linus Torvalds1da177e2005-04-16 15:20:36 -07002855/**
2856 * skb_postpull_rcsum - update checksum for received skb after pull
2857 * @skb: buffer to update
2858 * @start: start of data before pull
2859 * @len: length of data pulled
2860 *
2861 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002862 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2863 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002864 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002865static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002866 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002867{
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02002868 __skb_postpull_rcsum(skb, start, len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002869}
2870
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02002871static __always_inline void
2872__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
2873 unsigned int off)
2874{
2875 if (skb->ip_summed == CHECKSUM_COMPLETE)
2876 skb->csum = csum_block_add(skb->csum,
2877 csum_partial(start, len, 0), off);
2878}
Herbert Xucbb042f2006-03-20 22:43:56 -08002879
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02002880/**
2881 * skb_postpush_rcsum - update checksum for received skb after push
2882 * @skb: buffer to update
2883 * @start: start of data after push
2884 * @len: length of data pushed
2885 *
2886 * After doing a push on a received packet, you need to call this to
2887 * update the CHECKSUM_COMPLETE checksum.
2888 */
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01002889static inline void skb_postpush_rcsum(struct sk_buff *skb,
2890 const void *start, unsigned int len)
2891{
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02002892 __skb_postpush_rcsum(skb, start, len, 0);
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01002893}
2894
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02002895unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2896
David S. Miller7ce5a272013-12-02 17:26:05 -05002897/**
WANG Cong82a31b92016-06-30 10:15:22 -07002898 * skb_push_rcsum - push skb and update receive checksum
2899 * @skb: buffer to update
2900 * @len: length of data pulled
2901 *
2902 * This function performs an skb_push on the packet and updates
2903 * the CHECKSUM_COMPLETE checksum. It should be used on
2904 * receive path processing instead of skb_push unless you know
2905 * that the checksum difference is zero (e.g., a valid IP header)
2906 * or you are setting ip_summed to CHECKSUM_NONE.
2907 */
2908static inline unsigned char *skb_push_rcsum(struct sk_buff *skb,
2909 unsigned int len)
2910{
2911 skb_push(skb, len);
2912 skb_postpush_rcsum(skb, skb->data, len);
2913 return skb->data;
2914}
2915
2916/**
David S. Miller7ce5a272013-12-02 17:26:05 -05002917 * pskb_trim_rcsum - trim received skb and update checksum
2918 * @skb: buffer to trim
2919 * @len: new length
2920 *
2921 * This is exactly the same as pskb_trim except that it ensures the
2922 * checksum of received packets are still valid after the operation.
2923 */
2924
2925static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2926{
2927 if (likely(len >= skb->len))
2928 return 0;
2929 if (skb->ip_summed == CHECKSUM_COMPLETE)
2930 skb->ip_summed = CHECKSUM_NONE;
2931 return __pskb_trim(skb, len);
2932}
2933
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002934static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2935{
2936 if (skb->ip_summed == CHECKSUM_COMPLETE)
2937 skb->ip_summed = CHECKSUM_NONE;
2938 __skb_trim(skb, len);
2939 return 0;
2940}
2941
2942static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
2943{
2944 if (skb->ip_summed == CHECKSUM_COMPLETE)
2945 skb->ip_summed = CHECKSUM_NONE;
2946 return __skb_grow(skb, len);
2947}
2948
Linus Torvalds1da177e2005-04-16 15:20:36 -07002949#define skb_queue_walk(queue, skb) \
2950 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002951 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002952 skb = skb->next)
2953
James Chapman46f89142007-04-30 00:07:31 -07002954#define skb_queue_walk_safe(queue, skb, tmp) \
2955 for (skb = (queue)->next, tmp = skb->next; \
2956 skb != (struct sk_buff *)(queue); \
2957 skb = tmp, tmp = skb->next)
2958
David S. Miller1164f522008-09-23 00:49:44 -07002959#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002960 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002961 skb = skb->next)
2962
2963#define skb_queue_walk_from_safe(queue, skb, tmp) \
2964 for (tmp = skb->next; \
2965 skb != (struct sk_buff *)(queue); \
2966 skb = tmp, tmp = skb->next)
2967
Stephen Hemminger300ce172005-10-30 13:47:34 -08002968#define skb_queue_reverse_walk(queue, skb) \
2969 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002970 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08002971 skb = skb->prev)
2972
David S. Miller686a2952011-01-20 22:47:32 -08002973#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
2974 for (skb = (queue)->prev, tmp = skb->prev; \
2975 skb != (struct sk_buff *)(queue); \
2976 skb = tmp, tmp = skb->prev)
2977
2978#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
2979 for (tmp = skb->prev; \
2980 skb != (struct sk_buff *)(queue); \
2981 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002982
David S. Miller21dc3302010-08-23 00:13:46 -07002983static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07002984{
2985 return skb_shinfo(skb)->frag_list != NULL;
2986}
2987
2988static inline void skb_frag_list_init(struct sk_buff *skb)
2989{
2990 skb_shinfo(skb)->frag_list = NULL;
2991}
2992
David S. Milleree039872009-06-09 00:17:13 -07002993#define skb_walk_frags(skb, iter) \
2994 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
2995
Rainer Weikusatea3793e2015-12-06 21:11:34 +00002996
2997int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
2998 const struct sk_buff *skb);
Paolo Abeni65101ae2017-05-16 11:20:13 +02002999struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3000 struct sk_buff_head *queue,
3001 unsigned int flags,
3002 void (*destructor)(struct sock *sk,
3003 struct sk_buff *skb),
3004 int *peeked, int *off, int *err,
3005 struct sk_buff **last);
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003006struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
Paolo Abeni7c13f972016-11-04 11:28:59 +01003007 void (*destructor)(struct sock *sk,
3008 struct sk_buff *skb),
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003009 int *peeked, int *off, int *err,
3010 struct sk_buff **last);
Joe Perches7965bd42013-09-26 14:48:15 -07003011struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
Paolo Abeni7c13f972016-11-04 11:28:59 +01003012 void (*destructor)(struct sock *sk,
3013 struct sk_buff *skb),
Joe Perches7965bd42013-09-26 14:48:15 -07003014 int *peeked, int *off, int *err);
3015struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3016 int *err);
3017unsigned int datagram_poll(struct file *file, struct socket *sock,
3018 struct poll_table_struct *wait);
Al Viroc0371da2014-11-24 10:42:55 -05003019int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3020 struct iov_iter *to, int size);
David S. Miller51f3d022014-11-05 16:46:40 -05003021static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3022 struct msghdr *msg, int size)
3023{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003024 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
David S. Miller51f3d022014-11-05 16:46:40 -05003025}
Al Viroe5a4b0b2014-11-24 18:17:55 -05003026int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3027 struct msghdr *msg);
Al Viro3a654f92014-06-19 14:15:22 -04003028int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3029 struct iov_iter *from, int len);
Al Viro3a654f92014-06-19 14:15:22 -04003030int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
Joe Perches7965bd42013-09-26 14:48:15 -07003031void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
samanthakumar627d2d62016-04-05 12:41:16 -04003032void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3033static inline void skb_free_datagram_locked(struct sock *sk,
3034 struct sk_buff *skb)
3035{
3036 __skb_free_datagram_locked(sk, skb, 0);
3037}
Joe Perches7965bd42013-09-26 14:48:15 -07003038int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07003039int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3040int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3041__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3042 int len, __wsum csum);
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +02003043int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
Joe Perches7965bd42013-09-26 14:48:15 -07003044 struct pipe_inode_info *pipe, unsigned int len,
Al Viro25869262016-09-17 21:02:10 -04003045 unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07003046void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01003047unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00003048int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3049 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07003050void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3051int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3052void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Florian Westphalde960aa2014-01-26 10:58:16 +01003053unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
Marcelo Ricardo Leitnerae7ef812016-06-02 15:05:41 -03003054bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu);
Joe Perches7965bd42013-09-26 14:48:15 -07003055struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Vlad Yasevich0d5501c2014-08-08 14:42:13 -04003056struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
Jiri Pirkoe2195122014-11-19 14:05:01 +01003057int skb_ensure_writable(struct sk_buff *skb, int write_len);
Shmulik Ladkanibfca4c52016-09-19 19:11:09 +03003058int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
Jiri Pirko93515d52014-11-19 14:05:02 +01003059int skb_vlan_pop(struct sk_buff *skb);
3060int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
Sowmini Varadhan6fa01cc2016-04-22 18:36:35 -07003061struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3062 gfp_t gfp);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003063
Al Viro6ce8e9c2014-04-06 21:25:44 -04003064static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3065{
Al Viro3073f072017-02-17 23:13:25 -05003066 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
Al Viro6ce8e9c2014-04-06 21:25:44 -04003067}
3068
Al Viro7eab8d92014-04-06 21:51:23 -04003069static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3070{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003071 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro7eab8d92014-04-06 21:51:23 -04003072}
3073
Daniel Borkmann2817a332013-10-30 11:50:51 +01003074struct skb_checksum_ops {
3075 __wsum (*update)(const void *mem, int len, __wsum wsum);
3076 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3077};
3078
Davide Caratti96178132017-05-18 15:44:37 +02003079extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3080
Daniel Borkmann2817a332013-10-30 11:50:51 +01003081__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3082 __wsum csum, const struct skb_checksum_ops *ops);
3083__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3084 __wsum csum);
3085
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003086static inline void * __must_check
3087__skb_header_pointer(const struct sk_buff *skb, int offset,
3088 int len, void *data, int hlen, void *buffer)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003089{
Patrick McHardy55820ee2005-07-05 14:08:10 -07003090 if (hlen - offset >= len)
David S. Miller690e36e2014-08-23 12:13:41 -07003091 return data + offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003092
David S. Miller690e36e2014-08-23 12:13:41 -07003093 if (!skb ||
3094 skb_copy_bits(skb, offset, buffer, len) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003095 return NULL;
3096
3097 return buffer;
3098}
3099
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003100static inline void * __must_check
3101skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
David S. Miller690e36e2014-08-23 12:13:41 -07003102{
3103 return __skb_header_pointer(skb, offset, len, skb->data,
3104 skb_headlen(skb), buffer);
3105}
3106
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01003107/**
3108 * skb_needs_linearize - check if we need to linearize a given skb
3109 * depending on the given device features.
3110 * @skb: socket buffer to check
3111 * @features: net device features
3112 *
3113 * Returns true if either:
3114 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3115 * 2. skb is fragmented and the device does not support SG.
3116 */
3117static inline bool skb_needs_linearize(struct sk_buff *skb,
3118 netdev_features_t features)
3119{
3120 return skb_is_nonlinear(skb) &&
3121 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3122 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3123}
3124
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03003125static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3126 void *to,
3127 const unsigned int len)
3128{
3129 memcpy(to, skb->data, len);
3130}
3131
3132static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3133 const int offset, void *to,
3134 const unsigned int len)
3135{
3136 memcpy(to, skb->data + offset, len);
3137}
3138
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03003139static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3140 const void *from,
3141 const unsigned int len)
3142{
3143 memcpy(skb->data, from, len);
3144}
3145
3146static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3147 const int offset,
3148 const void *from,
3149 const unsigned int len)
3150{
3151 memcpy(skb->data + offset, from, len);
3152}
3153
Joe Perches7965bd42013-09-26 14:48:15 -07003154void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003155
Patrick Ohlyac45f602009-02-12 05:03:37 +00003156static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3157{
3158 return skb->tstamp;
3159}
3160
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003161/**
3162 * skb_get_timestamp - get timestamp from a skb
3163 * @skb: skb to get stamp from
3164 * @stamp: pointer to struct timeval to store stamp in
3165 *
3166 * Timestamps are stored in the skb as offsets to a base timestamp.
3167 * This function converts the offset back to a struct timeval and stores
3168 * it in stamp.
3169 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00003170static inline void skb_get_timestamp(const struct sk_buff *skb,
3171 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003172{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003173 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003174}
3175
Patrick Ohlyac45f602009-02-12 05:03:37 +00003176static inline void skb_get_timestampns(const struct sk_buff *skb,
3177 struct timespec *stamp)
3178{
3179 *stamp = ktime_to_timespec(skb->tstamp);
3180}
3181
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003182static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003183{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003184 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003185}
3186
Stephen Hemminger164891a2007-04-23 22:26:16 -07003187static inline ktime_t net_timedelta(ktime_t t)
3188{
3189 return ktime_sub(ktime_get_real(), t);
3190}
3191
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003192static inline ktime_t net_invalid_timestamp(void)
3193{
Thomas Gleixner8b0e1952016-12-25 12:30:41 +01003194 return 0;
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003195}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003196
Alexander Duyck62bccb82014-09-04 13:31:35 -04003197struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3198
Richard Cochranc1f19b52010-07-17 08:49:36 +00003199#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3200
Joe Perches7965bd42013-09-26 14:48:15 -07003201void skb_clone_tx_timestamp(struct sk_buff *skb);
3202bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00003203
3204#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3205
3206static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3207{
3208}
3209
3210static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3211{
3212 return false;
3213}
3214
3215#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3216
3217/**
3218 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3219 *
Richard Cochranda92b192011-10-21 00:49:15 +00003220 * PHY drivers may accept clones of transmitted packets for
3221 * timestamping via their phy_driver.txtstamp method. These drivers
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003222 * must call this function to return the skb back to the stack with a
3223 * timestamp.
Richard Cochranda92b192011-10-21 00:49:15 +00003224 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00003225 * @skb: clone of the the original outgoing packet
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003226 * @hwtstamps: hardware time stamps
Richard Cochranc1f19b52010-07-17 08:49:36 +00003227 *
3228 */
3229void skb_complete_tx_timestamp(struct sk_buff *skb,
3230 struct skb_shared_hwtstamps *hwtstamps);
3231
Willem de Bruijne7fd2882014-08-04 22:11:48 -04003232void __skb_tstamp_tx(struct sk_buff *orig_skb,
3233 struct skb_shared_hwtstamps *hwtstamps,
3234 struct sock *sk, int tstype);
3235
Patrick Ohlyac45f602009-02-12 05:03:37 +00003236/**
3237 * skb_tstamp_tx - queue clone of skb with send time stamps
3238 * @orig_skb: the original outgoing packet
3239 * @hwtstamps: hardware time stamps, may be NULL if not available
3240 *
3241 * If the skb has a socket associated, then this function clones the
3242 * skb (thus sharing the actual data and optional structures), stores
3243 * the optional hardware time stamping information (if non NULL) or
3244 * generates a software time stamp (otherwise), then queues the clone
3245 * to the error queue of the socket. Errors are silently ignored.
3246 */
Joe Perches7965bd42013-09-26 14:48:15 -07003247void skb_tstamp_tx(struct sk_buff *orig_skb,
3248 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00003249
Richard Cochran4507a712010-07-17 08:48:28 +00003250static inline void sw_tx_timestamp(struct sk_buff *skb)
3251{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00003252 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
3253 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00003254 skb_tstamp_tx(skb, NULL);
3255}
3256
3257/**
3258 * skb_tx_timestamp() - Driver hook for transmit timestamping
3259 *
3260 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00003261 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00003262 *
David S. Miller73409f32013-12-27 13:04:33 -05003263 * Specifically, one should make absolutely sure that this function is
3264 * called before TX completion of this packet can trigger. Otherwise
3265 * the packet could potentially already be freed.
3266 *
Richard Cochran4507a712010-07-17 08:48:28 +00003267 * @skb: A socket buffer.
3268 */
3269static inline void skb_tx_timestamp(struct sk_buff *skb)
3270{
Richard Cochranc1f19b52010-07-17 08:49:36 +00003271 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00003272 sw_tx_timestamp(skb);
3273}
3274
Johannes Berg6e3e9392011-11-09 10:15:42 +01003275/**
3276 * skb_complete_wifi_ack - deliver skb with wifi status
3277 *
3278 * @skb: the original outgoing packet
3279 * @acked: ack status
3280 *
3281 */
3282void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3283
Joe Perches7965bd42013-09-26 14:48:15 -07003284__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3285__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003286
Herbert Xu60476372007-04-09 11:59:39 -07003287static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3288{
Tom Herbert6edec0e2015-02-10 16:30:28 -08003289 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3290 skb->csum_valid ||
3291 (skb->ip_summed == CHECKSUM_PARTIAL &&
3292 skb_checksum_start_offset(skb) >= 0));
Herbert Xu60476372007-04-09 11:59:39 -07003293}
3294
Herbert Xufb286bb2005-11-10 13:01:24 -08003295/**
3296 * skb_checksum_complete - Calculate checksum of an entire packet
3297 * @skb: packet to process
3298 *
3299 * This function calculates the checksum over the entire packet plus
3300 * the value of skb->csum. The latter can be used to supply the
3301 * checksum of a pseudo header as used by TCP/UDP. It returns the
3302 * checksum.
3303 *
3304 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3305 * this function can be used to verify that checksum on received
3306 * packets. In that case the function should return zero if the
3307 * checksum is correct. In particular, this function will return zero
3308 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3309 * hardware has already verified the correctness of the checksum.
3310 */
Al Viro4381ca32007-07-15 21:00:11 +01003311static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08003312{
Herbert Xu60476372007-04-09 11:59:39 -07003313 return skb_csum_unnecessary(skb) ?
3314 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003315}
3316
Tom Herbert77cffe22014-08-27 21:26:46 -07003317static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3318{
3319 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3320 if (skb->csum_level == 0)
3321 skb->ip_summed = CHECKSUM_NONE;
3322 else
3323 skb->csum_level--;
3324 }
3325}
3326
3327static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3328{
3329 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3330 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3331 skb->csum_level++;
3332 } else if (skb->ip_summed == CHECKSUM_NONE) {
3333 skb->ip_summed = CHECKSUM_UNNECESSARY;
3334 skb->csum_level = 0;
3335 }
3336}
3337
Tom Herbert5a212322014-08-31 15:12:41 -07003338static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
3339{
3340 /* Mark current checksum as bad (typically called from GRO
3341 * path). In the case that ip_summed is CHECKSUM_NONE
3342 * this must be the first checksum encountered in the packet.
3343 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
3344 * checksum after the last one validated. For UDP, a zero
3345 * checksum can not be marked as bad.
3346 */
3347
3348 if (skb->ip_summed == CHECKSUM_NONE ||
3349 skb->ip_summed == CHECKSUM_UNNECESSARY)
3350 skb->csum_bad = 1;
3351}
3352
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003353/* Check if we need to perform checksum complete validation.
3354 *
3355 * Returns true if checksum complete is needed, false otherwise
3356 * (either checksum is unnecessary or zero checksum is allowed).
3357 */
3358static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3359 bool zero_okay,
3360 __sum16 check)
3361{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003362 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3363 skb->csum_valid = 1;
Tom Herbert77cffe22014-08-27 21:26:46 -07003364 __skb_decr_checksum_unnecessary(skb);
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003365 return false;
3366 }
3367
3368 return true;
3369}
3370
3371/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3372 * in checksum_init.
3373 */
3374#define CHECKSUM_BREAK 76
3375
Tom Herbert4e18b9a2015-04-20 14:10:04 -07003376/* Unset checksum-complete
3377 *
3378 * Unset checksum complete can be done when packet is being modified
3379 * (uncompressed for instance) and checksum-complete value is
3380 * invalidated.
3381 */
3382static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3383{
3384 if (skb->ip_summed == CHECKSUM_COMPLETE)
3385 skb->ip_summed = CHECKSUM_NONE;
3386}
3387
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003388/* Validate (init) checksum based on checksum complete.
3389 *
3390 * Return values:
3391 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3392 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3393 * checksum is stored in skb->csum for use in __skb_checksum_complete
3394 * non-zero: value of invalid checksum
3395 *
3396 */
3397static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3398 bool complete,
3399 __wsum psum)
3400{
3401 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3402 if (!csum_fold(csum_add(psum, skb->csum))) {
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003403 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003404 return 0;
3405 }
Tom Herbert5a212322014-08-31 15:12:41 -07003406 } else if (skb->csum_bad) {
3407 /* ip_summed == CHECKSUM_NONE in this case */
Eric Dumazetc91d4602015-05-15 05:48:07 -07003408 return (__force __sum16)1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003409 }
3410
3411 skb->csum = psum;
3412
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003413 if (complete || skb->len <= CHECKSUM_BREAK) {
3414 __sum16 csum;
3415
3416 csum = __skb_checksum_complete(skb);
3417 skb->csum_valid = !csum;
3418 return csum;
3419 }
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003420
3421 return 0;
3422}
3423
3424static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3425{
3426 return 0;
3427}
3428
3429/* Perform checksum validate (init). Note that this is a macro since we only
3430 * want to calculate the pseudo header which is an input function if necessary.
3431 * First we try to validate without any computation (checksum unnecessary) and
3432 * then calculate based on checksum complete calling the function to compute
3433 * pseudo header.
3434 *
3435 * Return values:
3436 * 0: checksum is validated or try to in skb_checksum_complete
3437 * non-zero: value of invalid checksum
3438 */
3439#define __skb_checksum_validate(skb, proto, complete, \
3440 zero_okay, check, compute_pseudo) \
3441({ \
3442 __sum16 __ret = 0; \
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003443 skb->csum_valid = 0; \
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003444 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3445 __ret = __skb_checksum_validate_complete(skb, \
3446 complete, compute_pseudo(skb, proto)); \
3447 __ret; \
3448})
3449
3450#define skb_checksum_init(skb, proto, compute_pseudo) \
3451 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3452
3453#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3454 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3455
3456#define skb_checksum_validate(skb, proto, compute_pseudo) \
3457 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3458
3459#define skb_checksum_validate_zero_check(skb, proto, check, \
3460 compute_pseudo) \
Sabrina Dubroca096a4cf2015-02-06 18:54:19 +01003461 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003462
3463#define skb_checksum_simple_validate(skb) \
3464 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3465
Tom Herbertd96535a2014-08-31 15:12:42 -07003466static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3467{
3468 return (skb->ip_summed == CHECKSUM_NONE &&
3469 skb->csum_valid && !skb->csum_bad);
3470}
3471
3472static inline void __skb_checksum_convert(struct sk_buff *skb,
3473 __sum16 check, __wsum pseudo)
3474{
3475 skb->csum = ~pseudo;
3476 skb->ip_summed = CHECKSUM_COMPLETE;
3477}
3478
3479#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3480do { \
3481 if (__skb_checksum_convert_check(skb)) \
3482 __skb_checksum_convert(skb, check, \
3483 compute_pseudo(skb, proto)); \
3484} while (0)
3485
Tom Herbert15e23962015-02-10 16:30:31 -08003486static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3487 u16 start, u16 offset)
3488{
3489 skb->ip_summed = CHECKSUM_PARTIAL;
3490 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3491 skb->csum_offset = offset - start;
3492}
3493
Tom Herbertdcdc8992015-02-02 16:07:34 -08003494/* Update skbuf and packet to reflect the remote checksum offload operation.
3495 * When called, ptr indicates the starting point for skb->csum when
3496 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3497 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3498 */
3499static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
Tom Herbert15e23962015-02-10 16:30:31 -08003500 int start, int offset, bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08003501{
3502 __wsum delta;
3503
Tom Herbert15e23962015-02-10 16:30:31 -08003504 if (!nopartial) {
3505 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3506 return;
3507 }
3508
Tom Herbertdcdc8992015-02-02 16:07:34 -08003509 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3510 __skb_checksum_complete(skb);
3511 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3512 }
3513
3514 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3515
3516 /* Adjust skb->csum since we changed the packet */
3517 skb->csum = csum_add(skb->csum, delta);
3518}
3519
Florian Westphalcb9c6832017-01-23 18:21:56 +01003520static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
3521{
3522#if IS_ENABLED(CONFIG_NF_CONNTRACK)
Florian Westphala9e419d2017-01-23 18:21:59 +01003523 return (void *)(skb->_nfct & SKB_NFCT_PTRMASK);
Florian Westphalcb9c6832017-01-23 18:21:56 +01003524#else
3525 return NULL;
3526#endif
3527}
3528
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003529#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07003530void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003531static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3532{
3533 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07003534 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003535}
3536static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3537{
3538 if (nfct)
3539 atomic_inc(&nfct->use);
3540}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003541#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003542#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003543static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3544{
3545 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3546 kfree(nf_bridge);
3547}
3548static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3549{
3550 if (nf_bridge)
3551 atomic_inc(&nf_bridge->use);
3552}
3553#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08003554static inline void nf_reset(struct sk_buff *skb)
3555{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003556#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01003557 nf_conntrack_put(skb_nfct(skb));
3558 skb->_nfct = 0;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003559#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003560#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Patrick McHardya193a4a2006-03-20 19:23:05 -08003561 nf_bridge_put(skb->nf_bridge);
3562 skb->nf_bridge = NULL;
3563#endif
3564}
3565
Patrick McHardy124dff02013-04-05 20:42:05 +02003566static inline void nf_reset_trace(struct sk_buff *skb)
3567{
Florian Westphal478b3602014-02-15 23:48:45 +01003568#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00003569 skb->nf_trace = 0;
3570#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003571}
3572
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003573/* Note: This doesn't put any conntrack and bridge info in dst. */
Eric Dumazetb1937222014-09-28 22:18:47 -07003574static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3575 bool copy)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003576{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003577#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01003578 dst->_nfct = src->_nfct;
3579 nf_conntrack_get(skb_nfct(src));
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003580#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003581#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003582 dst->nf_bridge = src->nf_bridge;
3583 nf_bridge_get(src->nf_bridge);
3584#endif
Florian Westphal478b3602014-02-15 23:48:45 +01003585#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Eric Dumazetb1937222014-09-28 22:18:47 -07003586 if (copy)
3587 dst->nf_trace = src->nf_trace;
Florian Westphal478b3602014-02-15 23:48:45 +01003588#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003589}
3590
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003591static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3592{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003593#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01003594 nf_conntrack_put(skb_nfct(dst));
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003595#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003596#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003597 nf_bridge_put(dst->nf_bridge);
3598#endif
Eric Dumazetb1937222014-09-28 22:18:47 -07003599 __nf_copy(dst, src, true);
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003600}
3601
James Morris984bc162006-06-09 00:29:17 -07003602#ifdef CONFIG_NETWORK_SECMARK
3603static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3604{
3605 to->secmark = from->secmark;
3606}
3607
3608static inline void skb_init_secmark(struct sk_buff *skb)
3609{
3610 skb->secmark = 0;
3611}
3612#else
3613static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3614{ }
3615
3616static inline void skb_init_secmark(struct sk_buff *skb)
3617{ }
3618#endif
3619
Eric W. Biederman574f7192014-04-01 12:20:24 -07003620static inline bool skb_irq_freeable(const struct sk_buff *skb)
3621{
3622 return !skb->destructor &&
3623#if IS_ENABLED(CONFIG_XFRM)
3624 !skb->sp &&
3625#endif
Florian Westphalcb9c6832017-01-23 18:21:56 +01003626 !skb_nfct(skb) &&
Eric W. Biederman574f7192014-04-01 12:20:24 -07003627 !skb->_skb_refdst &&
3628 !skb_has_frag_list(skb);
3629}
3630
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003631static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3632{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003633 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003634}
3635
Stephen Hemminger92477442009-03-21 13:39:26 -07003636static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003637{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003638 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003639}
3640
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003641static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3642{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003643 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003644}
3645
David S. Millerd5a9e242009-01-27 16:22:11 -08003646static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3647{
3648 skb->queue_mapping = rx_queue + 1;
3649}
3650
Stephen Hemminger92477442009-03-21 13:39:26 -07003651static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003652{
3653 return skb->queue_mapping - 1;
3654}
3655
Stephen Hemminger92477442009-03-21 13:39:26 -07003656static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003657{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003658 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08003659}
3660
Julian Anastasov4ff06202017-02-06 23:14:12 +02003661static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
3662{
3663 skb->dst_pending_confirm = val;
3664}
3665
3666static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
3667{
3668 return skb->dst_pending_confirm != 0;
3669}
3670
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003671static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3672{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003673#ifdef CONFIG_XFRM
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003674 return skb->sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003675#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003676 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003677#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003678}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003679
Pravin B Shelar68c33162013-02-14 14:02:41 +00003680/* Keeps track of mac header offset relative to skb->head.
3681 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3682 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07003683 * tunnel skb it points to outer mac header.
3684 * Keeps track of level of encapsulation of network headers.
3685 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00003686struct skb_gso_cb {
Alexander Duyck802ab552016-04-10 21:45:03 -04003687 union {
3688 int mac_offset;
3689 int data_offset;
3690 };
Eric Dumazet3347c962013-10-19 11:42:56 -07003691 int encap_level;
Alexander Duyck76443452016-02-05 15:27:37 -08003692 __wsum csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003693 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00003694};
Konstantin Khlebnikov9207f9d2016-01-08 15:21:46 +03003695#define SKB_SGO_CB_OFFSET 32
3696#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
Pravin B Shelar68c33162013-02-14 14:02:41 +00003697
3698static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3699{
3700 return (skb_mac_header(inner_skb) - inner_skb->head) -
3701 SKB_GSO_CB(inner_skb)->mac_offset;
3702}
3703
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00003704static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3705{
3706 int new_headroom, headroom;
3707 int ret;
3708
3709 headroom = skb_headroom(skb);
3710 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3711 if (ret)
3712 return ret;
3713
3714 new_headroom = skb_headroom(skb);
3715 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3716 return 0;
3717}
3718
Alexander Duyck08b64fc2016-02-05 15:27:49 -08003719static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
3720{
3721 /* Do not update partial checksums if remote checksum is enabled. */
3722 if (skb->remcsum_offload)
3723 return;
3724
3725 SKB_GSO_CB(skb)->csum = res;
3726 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
3727}
3728
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003729/* Compute the checksum for a gso segment. First compute the checksum value
3730 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3731 * then add in skb->csum (checksum from csum_start to end of packet).
3732 * skb->csum and csum_start are then updated to reflect the checksum of the
3733 * resultant packet starting from the transport header-- the resultant checksum
3734 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3735 * header.
3736 */
3737static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3738{
Alexander Duyck76443452016-02-05 15:27:37 -08003739 unsigned char *csum_start = skb_transport_header(skb);
3740 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
3741 __wsum partial = SKB_GSO_CB(skb)->csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003742
Alexander Duyck76443452016-02-05 15:27:37 -08003743 SKB_GSO_CB(skb)->csum = res;
3744 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003745
Alexander Duyck76443452016-02-05 15:27:37 -08003746 return csum_fold(csum_partial(csum_start, plen, partial));
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003747}
3748
David S. Millerbdcc0922012-03-07 20:53:36 -05003749static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07003750{
3751 return skb_shinfo(skb)->gso_size;
3752}
3753
Eric Dumazet36a8f392013-09-29 01:21:32 -07003754/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05003755static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02003756{
3757 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3758}
3759
Daniel Borkmann5293efe2016-08-18 01:00:39 +02003760static inline void skb_gso_reset(struct sk_buff *skb)
3761{
3762 skb_shinfo(skb)->gso_size = 0;
3763 skb_shinfo(skb)->gso_segs = 0;
3764 skb_shinfo(skb)->gso_type = 0;
3765}
3766
Joe Perches7965bd42013-09-26 14:48:15 -07003767void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07003768
3769static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3770{
3771 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3772 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003773 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3774
Alexander Duyckb78462e2010-06-02 12:24:37 +00003775 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3776 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07003777 __skb_warn_lro_forwarding(skb);
3778 return true;
3779 }
3780 return false;
3781}
3782
Herbert Xu35fc92a2007-03-26 23:22:20 -07003783static inline void skb_forward_csum(struct sk_buff *skb)
3784{
3785 /* Unfortunately we don't support this one. Any brave souls? */
3786 if (skb->ip_summed == CHECKSUM_COMPLETE)
3787 skb->ip_summed = CHECKSUM_NONE;
3788}
3789
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003790/**
3791 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3792 * @skb: skb to check
3793 *
3794 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3795 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3796 * use this helper, to document places where we make this assertion.
3797 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003798static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003799{
3800#ifdef DEBUG
3801 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3802#endif
3803}
3804
Rusty Russellf35d9d82008-02-04 23:49:54 -05003805bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00003806
Paul Durranted1f50c2014-01-09 10:02:46 +00003807int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
Linus Lüssing9afd85c2015-05-02 14:01:07 +02003808struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
3809 unsigned int transport_len,
3810 __sum16(*skb_chkf)(struct sk_buff *skb));
Paul Durranted1f50c2014-01-09 10:02:46 +00003811
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00003812/**
3813 * skb_head_is_locked - Determine if the skb->head is locked down
3814 * @skb: skb to check
3815 *
3816 * The head on skbs build around a head frag can be removed if they are
3817 * not cloned. This function returns true if the skb head is locked down
3818 * due to either being allocated via kmalloc, or by being a clone with
3819 * multiple references to the head.
3820 */
3821static inline bool skb_head_is_locked(const struct sk_buff *skb)
3822{
3823 return !skb->head_frag || skb_cloned(skb);
3824}
Florian Westphalfe6cc552014-02-13 23:09:12 +01003825
3826/**
3827 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3828 *
3829 * @skb: GSO skb
3830 *
3831 * skb_gso_network_seglen is used to determine the real size of the
3832 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3833 *
3834 * The MAC/L2 header is not accounted for.
3835 */
3836static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3837{
3838 unsigned int hdr_len = skb_transport_header(skb) -
3839 skb_network_header(skb);
3840 return hdr_len + skb_gso_transport_seglen(skb);
3841}
Thomas Grafee122c72015-07-21 10:43:58 +02003842
Edward Cree179bc672016-02-11 20:48:04 +00003843/* Local Checksum Offload.
3844 * Compute outer checksum based on the assumption that the
3845 * inner checksum will be offloaded later.
Edward Creee8ae7b02016-02-11 21:03:37 +00003846 * See Documentation/networking/checksum-offloads.txt for
3847 * explanation of how this works.
Edward Cree179bc672016-02-11 20:48:04 +00003848 * Fill in outer checksum adjustment (e.g. with sum of outer
3849 * pseudo-header) before calling.
3850 * Also ensure that inner checksum is in linear data area.
3851 */
3852static inline __wsum lco_csum(struct sk_buff *skb)
3853{
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003854 unsigned char *csum_start = skb_checksum_start(skb);
3855 unsigned char *l4_hdr = skb_transport_header(skb);
3856 __wsum partial;
Edward Cree179bc672016-02-11 20:48:04 +00003857
3858 /* Start with complement of inner checksum adjustment */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003859 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
3860 skb->csum_offset));
3861
Edward Cree179bc672016-02-11 20:48:04 +00003862 /* Add in checksum of our headers (incl. outer checksum
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003863 * adjustment filled in by caller) and return result.
Edward Cree179bc672016-02-11 20:48:04 +00003864 */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003865 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
Edward Cree179bc672016-02-11 20:48:04 +00003866}
3867
Linus Torvalds1da177e2005-04-16 15:20:36 -07003868#endif /* __KERNEL__ */
3869#endif /* _LINUX_SKBUFF_H */