blob: d92056b2da44de91e4a13e4f00b1ea06906e6fee [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
Eric Dumazet363ec39232014-02-26 14:02:11 -0800509/**
510 * struct skb_mstamp - multi resolution time stamps
511 * @stamp_us: timestamp in us resolution
512 * @stamp_jiffies: timestamp in jiffies
513 */
514struct skb_mstamp {
515 union {
516 u64 v64;
517 struct {
518 u32 stamp_us;
519 u32 stamp_jiffies;
520 };
521 };
522};
523
524/**
525 * skb_mstamp_get - get current timestamp
526 * @cl: place to store timestamps
527 */
528static inline void skb_mstamp_get(struct skb_mstamp *cl)
529{
530 u64 val = local_clock();
531
532 do_div(val, NSEC_PER_USEC);
533 cl->stamp_us = (u32)val;
534 cl->stamp_jiffies = (u32)jiffies;
535}
536
537/**
538 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
539 * @t1: pointer to newest sample
540 * @t0: pointer to oldest sample
541 */
542static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
543 const struct skb_mstamp *t0)
544{
545 s32 delta_us = t1->stamp_us - t0->stamp_us;
546 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
547
548 /* If delta_us is negative, this might be because interval is too big,
549 * or local_clock() drift is too big : fallback using jiffies.
550 */
551 if (delta_us <= 0 ||
552 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
553
554 delta_us = jiffies_to_usecs(delta_jiffies);
555
556 return delta_us;
557}
558
Yuchung Cheng625a5e12015-10-16 21:57:45 -0700559static inline bool skb_mstamp_after(const struct skb_mstamp *t1,
560 const struct skb_mstamp *t0)
561{
562 s32 diff = t1->stamp_jiffies - t0->stamp_jiffies;
563
564 if (!diff)
565 diff = t1->stamp_us - t0->stamp_us;
566 return diff > 0;
567}
Eric Dumazet363ec39232014-02-26 14:02:11 -0800568
Linus Torvalds1da177e2005-04-16 15:20:36 -0700569/**
570 * struct sk_buff - socket buffer
571 * @next: Next buffer in list
572 * @prev: Previous buffer in list
Eric Dumazet363ec39232014-02-26 14:02:11 -0800573 * @tstamp: Time we arrived/left
Eric Dumazet56b17422014-11-03 08:19:53 -0800574 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700575 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700577 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000578 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700579 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700580 * @len: Length of actual data
581 * @data_len: Data length
582 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700583 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700584 * @csum: Checksum (must include start/offset pair)
585 * @csum_start: Offset from skb->head where checksumming should start
586 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700587 * @priority: Packet queueing priority
WANG Cong60ff7462014-05-04 16:39:18 -0700588 * @ignore_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700590 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700591 * @nohdr: Payload reference only, must not modify header
592 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700593 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700594 * @ipvs_property: skbuff is owned by ipvs
Willem de Bruijne7246e12017-01-07 17:06:35 -0500595 * @tc_skip_classify: do not classify packet. set by IFB device
Willem de Bruijn8dc07fd2017-01-07 17:06:37 -0500596 * @tc_at_ingress: used within tc_classify to distinguish in/egress
Willem de Bruijnbc31c902017-01-07 17:06:38 -0500597 * @tc_redirected: packet was redirected by a tc action
598 * @tc_from_ingress: if tc_redirected, tc_at_ingress at time of redirect
Randy Dunlap31729362008-02-18 20:52:13 -0800599 * @peeked: this packet has been seen already, so stats have been
600 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700601 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700602 * @protocol: Packet protocol from driver
603 * @destructor: Destruct function
Florian Westphala9e419d2017-01-23 18:21:59 +0100604 * @_nfct: Associated connection, if any (with nfctinfo bits)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700605 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800606 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700607 * @tc_index: Traffic control index
Tom Herbert61b905d2014-03-24 15:34:47 -0700608 * @hash: the packet hash
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700609 * @queue_mapping: Queue mapping for multiqueue devices
David S. Miller0b725a22014-08-25 15:51:53 -0700610 * @xmit_more: More SKBs are pending for this queue
Randy Dunlap553a5672008-04-20 10:51:01 -0700611 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700612 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Tom Herbert61b905d2014-03-24 15:34:47 -0700613 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
Changli Gao4ca24622011-08-19 07:26:44 -0700614 * ports.
Tom Herberta3b18dd2014-07-01 21:33:17 -0700615 * @sw_hash: indicates hash was computed in software stack
Johannes Berg6e3e9392011-11-09 10:15:42 +0100616 * @wifi_acked_valid: wifi_acked was set
617 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000618 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Julian Anastasov4ff06202017-02-06 23:14:12 +0200619 * @dst_pending_confirm: need to confirm neighbour
Eliezer Tamir06021292013-06-10 11:39:50 +0300620 * @napi_id: id of the NAPI struct this skb came from
James Morris984bc162006-06-09 00:29:17 -0700621 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700622 * @mark: Generic packet mark
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000623 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b2008-07-14 22:49:06 -0700624 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000625 * @inner_protocol: Protocol (encapsulation)
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000626 * @inner_transport_header: Inner transport layer header (encapsulation)
627 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000628 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700629 * @transport_header: Transport layer header
630 * @network_header: Network layer header
631 * @mac_header: Link layer header
632 * @tail: Tail pointer
633 * @end: End pointer
634 * @head: Head of buffer
635 * @data: Data head pointer
636 * @truesize: Buffer size
637 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638 */
639
640struct sk_buff {
Eric Dumazet363ec39232014-02-26 14:02:11 -0800641 union {
Eric Dumazet56b17422014-11-03 08:19:53 -0800642 struct {
643 /* These two members must be first. */
644 struct sk_buff *next;
645 struct sk_buff *prev;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000646
Eric Dumazet56b17422014-11-03 08:19:53 -0800647 union {
648 ktime_t tstamp;
649 struct skb_mstamp skb_mstamp;
650 };
651 };
652 struct rb_node rbnode; /* used in netem & tcp stack */
653 };
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000654 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655
Eric Dumazetc84d9492016-12-08 11:41:55 -0800656 union {
657 struct net_device *dev;
658 /* Some protocols might use this space to store information,
659 * while device pointer would be NULL.
660 * UDP receive path is one user.
661 */
662 unsigned long dev_scratch;
663 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700664 /*
665 * This is the control buffer. It is free to use for every
666 * layer. Please put your private variables there. If you
667 * want to keep them across layers you have to do a skb_clone()
668 * first. This is owned by whoever has the skb queued ATM.
669 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000670 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700671
Eric Dumazet7fee2262010-05-11 23:19:48 +0000672 unsigned long _skb_refdst;
Eric Dumazetb1937222014-09-28 22:18:47 -0700673 void (*destructor)(struct sk_buff *skb);
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000674#ifdef CONFIG_XFRM
675 struct sec_path *sp;
676#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700677#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +0100678 unsigned long _nfct;
Eric Dumazetb1937222014-09-28 22:18:47 -0700679#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +0200680#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Eric Dumazetb1937222014-09-28 22:18:47 -0700681 struct nf_bridge_info *nf_bridge;
682#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700683 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700684 data_len;
685 __u16 mac_len,
686 hdr_len;
Eric Dumazetb1937222014-09-28 22:18:47 -0700687
688 /* Following fields are _not_ copied in __copy_skb_header()
689 * Note that queue_mapping is here mostly to fill a hole.
690 */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200691 kmemcheck_bitfield_begin(flags1);
Eric Dumazetb1937222014-09-28 22:18:47 -0700692 __u16 queue_mapping;
Daniel Borkmann36bbef52016-09-20 00:26:13 +0200693
694/* if you move cloned around you also must adapt those constants */
695#ifdef __BIG_ENDIAN_BITFIELD
696#define CLONED_MASK (1 << 7)
697#else
698#define CLONED_MASK 1
699#endif
700#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
701
702 __u8 __cloned_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700703 __u8 cloned:1,
Harald Welte6869c4d2005-08-09 19:24:19 -0700704 nohdr:1,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800705 fclone:2,
Herbert Xua59322b2007-12-05 01:53:40 -0800706 peeked:1,
Eric Dumazetb1937222014-09-28 22:18:47 -0700707 head_frag:1,
Daniel Borkmann36bbef52016-09-20 00:26:13 +0200708 xmit_more:1,
709 __unused:1; /* one bit hole */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200710 kmemcheck_bitfield_end(flags1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711
Eric Dumazetb1937222014-09-28 22:18:47 -0700712 /* fields enclosed in headers_start/headers_end are copied
713 * using a single memcpy() in __copy_skb_header()
714 */
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700715 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700716 __u32 headers_start[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700717 /* public: */
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200718
719/* if you move pkt_type around you also must adapt those constants */
720#ifdef __BIG_ENDIAN_BITFIELD
721#define PKT_TYPE_MAX (7 << 5)
722#else
723#define PKT_TYPE_MAX 7
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100724#endif
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200725#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
726
727 __u8 __pkt_type_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700728 __u8 pkt_type:3;
729 __u8 pfmemalloc:1;
730 __u8 ignore_df:1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700731
Eric Dumazetb1937222014-09-28 22:18:47 -0700732 __u8 nf_trace:1;
733 __u8 ip_summed:2;
734 __u8 ooo_okay:1;
735 __u8 l4_hash:1;
736 __u8 sw_hash:1;
737 __u8 wifi_acked_valid:1;
738 __u8 wifi_acked:1;
739
740 __u8 no_fcs:1;
741 /* Indicates the inner headers are valid in the skbuff. */
742 __u8 encapsulation:1;
743 __u8 encap_hdr_csum:1;
744 __u8 csum_valid:1;
745 __u8 csum_complete_sw:1;
746 __u8 csum_level:2;
747 __u8 csum_bad:1;
748
Julian Anastasov4ff06202017-02-06 23:14:12 +0200749 __u8 dst_pending_confirm:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700750#ifdef CONFIG_IPV6_NDISC_NODETYPE
751 __u8 ndisc_nodetype:2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700753 __u8 ipvs_property:1;
Tom Herbert8bce6d72014-09-29 20:22:29 -0700754 __u8 inner_protocol_type:1;
Tom Herberte585f232014-11-04 09:06:54 -0800755 __u8 remcsum_offload:1;
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200756#ifdef CONFIG_NET_SWITCHDEV
757 __u8 offload_fwd_mark:1;
758#endif
Willem de Bruijne7246e12017-01-07 17:06:35 -0500759#ifdef CONFIG_NET_CLS_ACT
760 __u8 tc_skip_classify:1;
Willem de Bruijn8dc07fd2017-01-07 17:06:37 -0500761 __u8 tc_at_ingress:1;
Willem de Bruijnbc31c902017-01-07 17:06:38 -0500762 __u8 tc_redirected:1;
763 __u8 tc_from_ingress:1;
Willem de Bruijne7246e12017-01-07 17:06:35 -0500764#endif
Alexander Duyck4031ae62012-01-27 06:22:53 +0000765
Linus Torvalds1da177e2005-04-16 15:20:36 -0700766#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700767 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700768#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200769
Eric Dumazetb1937222014-09-28 22:18:47 -0700770 union {
771 __wsum csum;
772 struct {
773 __u16 csum_start;
774 __u16 csum_offset;
775 };
776 };
777 __u32 priority;
778 int skb_iif;
779 __u32 hash;
780 __be16 vlan_proto;
781 __u16 vlan_tci;
Eric Dumazet2bd82482015-02-03 23:48:24 -0800782#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
783 union {
784 unsigned int napi_id;
785 unsigned int sender_cpu;
786 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700787#endif
James Morris984bc162006-06-09 00:29:17 -0700788#ifdef CONFIG_NETWORK_SECMARK
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200789 __u32 secmark;
James Morris984bc162006-06-09 00:29:17 -0700790#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -0700791
Neil Horman3b885782009-10-12 13:26:31 -0700792 union {
793 __u32 mark;
Eric Dumazet16fad692013-03-14 05:40:32 +0000794 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700795 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796
Tom Herbert8bce6d72014-09-29 20:22:29 -0700797 union {
798 __be16 inner_protocol;
799 __u8 inner_ipproto;
800 };
801
Simon Horman1a37e412013-05-23 21:02:51 +0000802 __u16 inner_transport_header;
803 __u16 inner_network_header;
804 __u16 inner_mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700805
806 __be16 protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000807 __u16 transport_header;
808 __u16 network_header;
809 __u16 mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700810
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700811 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700812 __u32 headers_end[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700813 /* public: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700814
Linus Torvalds1da177e2005-04-16 15:20:36 -0700815 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700816 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700817 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700819 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700820 unsigned int truesize;
821 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700822};
823
824#ifdef __KERNEL__
825/*
826 * Handling routines are only of interest to the kernel
827 */
828#include <linux/slab.h>
829
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830
Mel Gormanc93bdd02012-07-31 16:44:19 -0700831#define SKB_ALLOC_FCLONE 0x01
832#define SKB_ALLOC_RX 0x02
Alexander Duyckfd11a832014-12-09 19:40:49 -0800833#define SKB_ALLOC_NAPI 0x04
Mel Gormanc93bdd02012-07-31 16:44:19 -0700834
835/* Returns true if the skb was allocated from PFMEMALLOC reserves */
836static inline bool skb_pfmemalloc(const struct sk_buff *skb)
837{
838 return unlikely(skb->pfmemalloc);
839}
840
Eric Dumazet7fee2262010-05-11 23:19:48 +0000841/*
842 * skb might have a dst pointer attached, refcounted or not.
843 * _skb_refdst low order bit is set if refcount was _not_ taken
844 */
845#define SKB_DST_NOREF 1UL
846#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
847
Florian Westphala9e419d2017-01-23 18:21:59 +0100848#define SKB_NFCT_PTRMASK ~(7UL)
Eric Dumazet7fee2262010-05-11 23:19:48 +0000849/**
850 * skb_dst - returns skb dst_entry
851 * @skb: buffer
852 *
853 * Returns skb dst_entry, regardless of reference taken or not.
854 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000855static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
856{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000857 /* If refdst was not refcounted, check we still are in a
858 * rcu_read_lock section
859 */
860 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
861 !rcu_read_lock_held() &&
862 !rcu_read_lock_bh_held());
863 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000864}
865
Eric Dumazet7fee2262010-05-11 23:19:48 +0000866/**
867 * skb_dst_set - sets skb dst
868 * @skb: buffer
869 * @dst: dst entry
870 *
871 * Sets skb dst, assuming a reference was taken on dst and should
872 * be released by skb_dst_drop()
873 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000874static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
875{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000876 skb->_skb_refdst = (unsigned long)dst;
877}
878
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200879/**
880 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
881 * @skb: buffer
882 * @dst: dst entry
883 *
884 * Sets skb dst, assuming a reference was not taken on dst.
885 * If dst entry is cached, we do not take reference and dst_release
886 * will be avoided by refdst_drop. If dst entry is not cached, we take
887 * reference, so that last dst_release can destroy the dst immediately.
888 */
889static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
890{
Hannes Frederic Sowadbfc4fb2014-12-06 19:19:42 +0100891 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
892 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200893}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000894
895/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300896 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000897 * @skb: buffer
898 */
899static inline bool skb_dst_is_noref(const struct sk_buff *skb)
900{
901 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000902}
903
Eric Dumazet511c3f92009-06-02 05:14:27 +0000904static inline struct rtable *skb_rtable(const struct sk_buff *skb)
905{
Eric Dumazetadf30902009-06-02 05:19:30 +0000906 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000907}
908
Jamal Hadi Salim8b10cab2016-07-02 06:43:14 -0400909/* For mangling skb->pkt_type from user space side from applications
910 * such as nft, tc, etc, we only allow a conservative subset of
911 * possible pkt_types to be set.
912*/
913static inline bool skb_pkt_type_ok(u32 ptype)
914{
915 return ptype <= PACKET_OTHERHOST;
916}
917
Joe Perches7965bd42013-09-26 14:48:15 -0700918void kfree_skb(struct sk_buff *skb);
919void kfree_skb_list(struct sk_buff *segs);
920void skb_tx_error(struct sk_buff *skb);
921void consume_skb(struct sk_buff *skb);
922void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000923extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000924
Joe Perches7965bd42013-09-26 14:48:15 -0700925void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
926bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
927 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000928
Joe Perches7965bd42013-09-26 14:48:15 -0700929struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
930 int node);
Eric Dumazet2ea2f622015-04-24 16:05:01 -0700931struct sk_buff *__build_skb(void *data, unsigned int frag_size);
Joe Perches7965bd42013-09-26 14:48:15 -0700932struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700933static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100934 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700935{
Eric Dumazet564824b2010-10-11 19:05:25 +0000936 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700937}
938
Eric Dumazet2e4e4412014-09-17 04:49:49 -0700939struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
940 unsigned long data_len,
941 int max_page_order,
942 int *errcode,
943 gfp_t gfp_mask);
944
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700945/* Layout of fast clones : [skb1][skb2][fclone_ref] */
946struct sk_buff_fclones {
947 struct sk_buff skb1;
948
949 struct sk_buff skb2;
950
951 atomic_t fclone_ref;
952};
953
954/**
955 * skb_fclone_busy - check if fclone is busy
Stephen Hemminger293de7d2016-10-23 09:28:29 -0700956 * @sk: socket
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700957 * @skb: buffer
958 *
Masanari Iidabda13fe2015-12-13 16:53:02 +0900959 * Returns true if skb is a fast clone, and its clone is not freed.
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700960 * Some drivers call skb_orphan() in their ndo_start_xmit(),
961 * so we also check that this didnt happen.
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700962 */
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700963static inline bool skb_fclone_busy(const struct sock *sk,
964 const struct sk_buff *skb)
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700965{
966 const struct sk_buff_fclones *fclones;
967
968 fclones = container_of(skb, struct sk_buff_fclones, skb1);
969
970 return skb->fclone == SKB_FCLONE_ORIG &&
Eric Dumazet6ffe75eb2014-12-03 17:04:39 -0800971 atomic_read(&fclones->fclone_ref) > 1 &&
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700972 fclones->skb2.sk == sk;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700973}
974
David S. Millerd179cd12005-08-17 14:57:30 -0700975static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100976 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700977{
Mel Gormanc93bdd02012-07-31 16:44:19 -0700978 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700979}
980
Joe Perches7965bd42013-09-26 14:48:15 -0700981struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
Patrick McHardy0ebd0ac2013-04-17 06:46:58 +0000982static inline struct sk_buff *alloc_skb_head(gfp_t priority)
983{
984 return __alloc_skb_head(priority, -1);
985}
986
Joe Perches7965bd42013-09-26 14:48:15 -0700987struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
988int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
989struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
990struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
Octavian Purdilabad93e92014-06-12 01:36:26 +0300991struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
992 gfp_t gfp_mask, bool fclone);
993static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
994 gfp_t gfp_mask)
995{
996 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
997}
Eric Dumazet117632e2011-12-03 21:39:53 +0000998
Joe Perches7965bd42013-09-26 14:48:15 -0700999int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1000struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1001 unsigned int headroom);
1002struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1003 int newtailroom, gfp_t priority);
Fan Du25a91d82014-01-18 09:54:23 +08001004int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1005 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001006int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
1007 int len);
1008int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
1009int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +00001010#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001011
Joe Perches7965bd42013-09-26 14:48:15 -07001012int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
1013 int getfrag(void *from, char *to, int offset,
1014 int len, int odd, struct sk_buff *skb),
1015 void *from, int length);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001016
Hannes Frederic Sowabe12a1f2015-05-21 16:59:58 +02001017int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1018 int offset, size_t size);
1019
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08001020struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -07001021 __u32 lower_offset;
1022 __u32 upper_offset;
1023 __u32 frag_idx;
1024 __u32 stepped_offset;
1025 struct sk_buff *root_skb;
1026 struct sk_buff *cur_skb;
1027 __u8 *frag_data;
1028};
1029
Joe Perches7965bd42013-09-26 14:48:15 -07001030void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1031 unsigned int to, struct skb_seq_state *st);
1032unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1033 struct skb_seq_state *st);
1034void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -07001035
Joe Perches7965bd42013-09-26 14:48:15 -07001036unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
Bojan Prtvar059a2442015-02-22 11:46:35 +01001037 unsigned int to, struct ts_config *config);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -07001038
Tom Herbert09323cc2013-12-15 22:16:19 -08001039/*
1040 * Packet hash types specify the type of hash in skb_set_hash.
1041 *
1042 * Hash types refer to the protocol layer addresses which are used to
1043 * construct a packet's hash. The hashes are used to differentiate or identify
1044 * flows of the protocol layer for the hash type. Hash types are either
1045 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1046 *
1047 * Properties of hashes:
1048 *
1049 * 1) Two packets in different flows have different hash values
1050 * 2) Two packets in the same flow should have the same hash value
1051 *
1052 * A hash at a higher layer is considered to be more specific. A driver should
1053 * set the most specific hash possible.
1054 *
1055 * A driver cannot indicate a more specific hash than the layer at which a hash
1056 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1057 *
1058 * A driver may indicate a hash level which is less specific than the
1059 * actual layer the hash was computed on. For instance, a hash computed
1060 * at L4 may be considered an L3 hash. This should only be done if the
1061 * driver can't unambiguously determine that the HW computed the hash at
1062 * the higher layer. Note that the "should" in the second property above
1063 * permits this.
1064 */
1065enum pkt_hash_types {
1066 PKT_HASH_TYPE_NONE, /* Undefined type */
1067 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1068 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1069 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1070};
1071
Tom Herbertbcc83832015-09-01 09:24:24 -07001072static inline void skb_clear_hash(struct sk_buff *skb)
1073{
1074 skb->hash = 0;
1075 skb->sw_hash = 0;
1076 skb->l4_hash = 0;
1077}
1078
1079static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1080{
1081 if (!skb->l4_hash)
1082 skb_clear_hash(skb);
1083}
1084
1085static inline void
1086__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1087{
1088 skb->l4_hash = is_l4;
1089 skb->sw_hash = is_sw;
1090 skb->hash = hash;
1091}
1092
Tom Herbert09323cc2013-12-15 22:16:19 -08001093static inline void
1094skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1095{
Tom Herbertbcc83832015-09-01 09:24:24 -07001096 /* Used by drivers to set hash from HW */
1097 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1098}
1099
1100static inline void
1101__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1102{
1103 __skb_set_hash(skb, hash, true, is_l4);
Tom Herbert09323cc2013-12-15 22:16:19 -08001104}
1105
Tom Herberte5276932015-09-01 09:24:23 -07001106void __skb_get_hash(struct sk_buff *skb);
Florian Westphalb9177832016-10-26 18:49:46 +02001107u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
Tom Herberte5276932015-09-01 09:24:23 -07001108u32 skb_get_poff(const struct sk_buff *skb);
1109u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1110 const struct flow_keys *keys, int hlen);
1111__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1112 void *data, int hlen_proto);
1113
1114static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1115 int thoff, u8 ip_proto)
1116{
1117 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1118}
1119
1120void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1121 const struct flow_dissector_key *key,
1122 unsigned int key_count);
1123
1124bool __skb_flow_dissect(const struct sk_buff *skb,
1125 struct flow_dissector *flow_dissector,
1126 void *target_container,
Tom Herbertcd79a232015-09-01 09:24:27 -07001127 void *data, __be16 proto, int nhoff, int hlen,
1128 unsigned int flags);
Tom Herberte5276932015-09-01 09:24:23 -07001129
1130static inline bool skb_flow_dissect(const struct sk_buff *skb,
1131 struct flow_dissector *flow_dissector,
Tom Herbertcd79a232015-09-01 09:24:27 -07001132 void *target_container, unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001133{
1134 return __skb_flow_dissect(skb, flow_dissector, target_container,
Tom Herbertcd79a232015-09-01 09:24:27 -07001135 NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001136}
1137
1138static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
Tom Herbertcd79a232015-09-01 09:24:27 -07001139 struct flow_keys *flow,
1140 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001141{
1142 memset(flow, 0, sizeof(*flow));
1143 return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
Tom Herbertcd79a232015-09-01 09:24:27 -07001144 NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001145}
1146
1147static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
1148 void *data, __be16 proto,
Tom Herbertcd79a232015-09-01 09:24:27 -07001149 int nhoff, int hlen,
1150 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001151{
1152 memset(flow, 0, sizeof(*flow));
1153 return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
Tom Herbertcd79a232015-09-01 09:24:27 -07001154 data, proto, nhoff, hlen, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001155}
1156
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001157static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001158{
Tom Herberta3b18dd2014-07-01 21:33:17 -07001159 if (!skb->l4_hash && !skb->sw_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001160 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001161
Tom Herbert61b905d2014-03-24 15:34:47 -07001162 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001163}
1164
David S. Miller20a17bf2015-09-01 21:19:17 -07001165__u32 __skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6);
Tom Herbertf70ea012015-07-31 16:52:10 -07001166
David S. Miller20a17bf2015-09-01 21:19:17 -07001167static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
Tom Herbertf70ea012015-07-31 16:52:10 -07001168{
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001169 if (!skb->l4_hash && !skb->sw_hash) {
1170 struct flow_keys keys;
Tom Herbertde4c1f82015-09-01 18:11:04 -07001171 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001172
Tom Herbertde4c1f82015-09-01 18:11:04 -07001173 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001174 }
Tom Herbertf70ea012015-07-31 16:52:10 -07001175
1176 return skb->hash;
1177}
1178
David S. Miller20a17bf2015-09-01 21:19:17 -07001179__u32 __skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl);
Tom Herbertf70ea012015-07-31 16:52:10 -07001180
David S. Miller20a17bf2015-09-01 21:19:17 -07001181static inline __u32 skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl4)
Tom Herbertf70ea012015-07-31 16:52:10 -07001182{
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001183 if (!skb->l4_hash && !skb->sw_hash) {
1184 struct flow_keys keys;
Tom Herbertde4c1f82015-09-01 18:11:04 -07001185 __u32 hash = __get_hash_from_flowi4(fl4, &keys);
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001186
Tom Herbertde4c1f82015-09-01 18:11:04 -07001187 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001188 }
Tom Herbertf70ea012015-07-31 16:52:10 -07001189
1190 return skb->hash;
1191}
1192
Tom Herbert50fb7992015-05-01 11:30:12 -07001193__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
1194
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001195static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1196{
Tom Herbert61b905d2014-03-24 15:34:47 -07001197 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001198}
1199
Tom Herbert3df7a742013-12-15 22:16:29 -08001200static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1201{
Tom Herbert61b905d2014-03-24 15:34:47 -07001202 to->hash = from->hash;
Tom Herberta3b18dd2014-07-01 21:33:17 -07001203 to->sw_hash = from->sw_hash;
Tom Herbert61b905d2014-03-24 15:34:47 -07001204 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -08001205};
1206
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001207#ifdef NET_SKBUFF_DATA_USES_OFFSET
1208static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1209{
1210 return skb->head + skb->end;
1211}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001212
1213static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1214{
1215 return skb->end;
1216}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001217#else
1218static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1219{
1220 return skb->end;
1221}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001222
1223static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1224{
1225 return skb->end - skb->head;
1226}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001227#endif
1228
Linus Torvalds1da177e2005-04-16 15:20:36 -07001229/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001230#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001231
Patrick Ohlyac45f602009-02-12 05:03:37 +00001232static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1233{
1234 return &skb_shinfo(skb)->hwtstamps;
1235}
1236
Linus Torvalds1da177e2005-04-16 15:20:36 -07001237/**
1238 * skb_queue_empty - check if a queue is empty
1239 * @list: queue head
1240 *
1241 * Returns true if the queue is empty, false otherwise.
1242 */
1243static inline int skb_queue_empty(const struct sk_buff_head *list)
1244{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001245 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001246}
1247
1248/**
David S. Millerfc7ebb22008-09-23 00:34:07 -07001249 * skb_queue_is_last - check if skb is the last entry in the queue
1250 * @list: queue head
1251 * @skb: buffer
1252 *
1253 * Returns true if @skb is the last buffer on the list.
1254 */
1255static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1256 const struct sk_buff *skb)
1257{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001258 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -07001259}
1260
1261/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001262 * skb_queue_is_first - check if skb is the first entry in the queue
1263 * @list: queue head
1264 * @skb: buffer
1265 *
1266 * Returns true if @skb is the first buffer on the list.
1267 */
1268static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1269 const struct sk_buff *skb)
1270{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001271 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001272}
1273
1274/**
David S. Miller249c8b42008-09-23 00:44:42 -07001275 * skb_queue_next - return the next packet in the queue
1276 * @list: queue head
1277 * @skb: current buffer
1278 *
1279 * Return the next packet in @list after @skb. It is only valid to
1280 * call this if skb_queue_is_last() evaluates to false.
1281 */
1282static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1283 const struct sk_buff *skb)
1284{
1285 /* This BUG_ON may seem severe, but if we just return then we
1286 * are going to dereference garbage.
1287 */
1288 BUG_ON(skb_queue_is_last(list, skb));
1289 return skb->next;
1290}
1291
1292/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001293 * skb_queue_prev - return the prev packet in the queue
1294 * @list: queue head
1295 * @skb: current buffer
1296 *
1297 * Return the prev packet in @list before @skb. It is only valid to
1298 * call this if skb_queue_is_first() evaluates to false.
1299 */
1300static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1301 const struct sk_buff *skb)
1302{
1303 /* This BUG_ON may seem severe, but if we just return then we
1304 * are going to dereference garbage.
1305 */
1306 BUG_ON(skb_queue_is_first(list, skb));
1307 return skb->prev;
1308}
1309
1310/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001311 * skb_get - reference buffer
1312 * @skb: buffer to reference
1313 *
1314 * Makes another reference to a socket buffer and returns a pointer
1315 * to the buffer.
1316 */
1317static inline struct sk_buff *skb_get(struct sk_buff *skb)
1318{
1319 atomic_inc(&skb->users);
1320 return skb;
1321}
1322
1323/*
1324 * If users == 1, we are the only owner and are can avoid redundant
1325 * atomic change.
1326 */
1327
1328/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001329 * skb_cloned - is the buffer a clone
1330 * @skb: buffer to check
1331 *
1332 * Returns true if the buffer was generated with skb_clone() and is
1333 * one of multiple shared copies of the buffer. Cloned buffers are
1334 * shared data so must not be written to under normal circumstances.
1335 */
1336static inline int skb_cloned(const struct sk_buff *skb)
1337{
1338 return skb->cloned &&
1339 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1340}
1341
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001342static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1343{
Mel Gormand0164ad2015-11-06 16:28:21 -08001344 might_sleep_if(gfpflags_allow_blocking(pri));
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001345
1346 if (skb_cloned(skb))
1347 return pskb_expand_head(skb, 0, 0, pri);
1348
1349 return 0;
1350}
1351
Linus Torvalds1da177e2005-04-16 15:20:36 -07001352/**
1353 * skb_header_cloned - is the header a clone
1354 * @skb: buffer to check
1355 *
1356 * Returns true if modifying the header part of the buffer requires
1357 * the data to be copied.
1358 */
1359static inline int skb_header_cloned(const struct sk_buff *skb)
1360{
1361 int dataref;
1362
1363 if (!skb->cloned)
1364 return 0;
1365
1366 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1367 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1368 return dataref != 1;
1369}
1370
Eric Dumazet9580bf22016-04-30 10:19:29 -07001371static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1372{
1373 might_sleep_if(gfpflags_allow_blocking(pri));
1374
1375 if (skb_header_cloned(skb))
1376 return pskb_expand_head(skb, 0, 0, pri);
1377
1378 return 0;
1379}
1380
Linus Torvalds1da177e2005-04-16 15:20:36 -07001381/**
1382 * skb_header_release - release reference to header
1383 * @skb: buffer to operate on
1384 *
1385 * Drop a reference to the header part of the buffer. This is done
1386 * by acquiring a payload reference. You must not read from the header
1387 * part of skb->data after this.
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001388 * Note : Check if you can use __skb_header_release() instead.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001389 */
1390static inline void skb_header_release(struct sk_buff *skb)
1391{
1392 BUG_ON(skb->nohdr);
1393 skb->nohdr = 1;
1394 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1395}
1396
1397/**
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001398 * __skb_header_release - release reference to header
1399 * @skb: buffer to operate on
1400 *
1401 * Variant of skb_header_release() assuming skb is private to caller.
1402 * We can avoid one atomic operation.
1403 */
1404static inline void __skb_header_release(struct sk_buff *skb)
1405{
1406 skb->nohdr = 1;
1407 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1408}
1409
1410
1411/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001412 * skb_shared - is the buffer shared
1413 * @skb: buffer to check
1414 *
1415 * Returns true if more than one person has a reference to this
1416 * buffer.
1417 */
1418static inline int skb_shared(const struct sk_buff *skb)
1419{
1420 return atomic_read(&skb->users) != 1;
1421}
1422
1423/**
1424 * skb_share_check - check if buffer is shared and if so clone it
1425 * @skb: buffer to check
1426 * @pri: priority for memory allocation
1427 *
1428 * If the buffer is shared the buffer is cloned and the old copy
1429 * drops a reference. A new clone with a single reference is returned.
1430 * If the buffer is not shared the original buffer is returned. When
1431 * being called from interrupt status or with spinlocks held pri must
1432 * be GFP_ATOMIC.
1433 *
1434 * NULL is returned on a memory allocation failure.
1435 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001436static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001437{
Mel Gormand0164ad2015-11-06 16:28:21 -08001438 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001439 if (skb_shared(skb)) {
1440 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001441
1442 if (likely(nskb))
1443 consume_skb(skb);
1444 else
1445 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001446 skb = nskb;
1447 }
1448 return skb;
1449}
1450
1451/*
1452 * Copy shared buffers into a new sk_buff. We effectively do COW on
1453 * packets to handle cases where we have a local reader and forward
1454 * and a couple of other messy ones. The normal one is tcpdumping
1455 * a packet thats being forwarded.
1456 */
1457
1458/**
1459 * skb_unshare - make a copy of a shared buffer
1460 * @skb: buffer to check
1461 * @pri: priority for memory allocation
1462 *
1463 * If the socket buffer is a clone then this function creates a new
1464 * copy of the data, drops a reference count on the old copy and returns
1465 * the new copy with the reference count at 1. If the buffer is not a clone
1466 * the original buffer is returned. When called with a spinlock held or
1467 * from interrupt state @pri must be %GFP_ATOMIC
1468 *
1469 * %NULL is returned on a memory allocation failure.
1470 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001471static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001472 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001473{
Mel Gormand0164ad2015-11-06 16:28:21 -08001474 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001475 if (skb_cloned(skb)) {
1476 struct sk_buff *nskb = skb_copy(skb, pri);
Alexander Aring31eff812014-10-10 23:10:47 +02001477
1478 /* Free our shared copy */
1479 if (likely(nskb))
1480 consume_skb(skb);
1481 else
1482 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001483 skb = nskb;
1484 }
1485 return skb;
1486}
1487
1488/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001489 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001490 * @list_: list to peek at
1491 *
1492 * Peek an &sk_buff. Unlike most other operations you _MUST_
1493 * be careful with this one. A peek leaves the buffer on the
1494 * list and someone else may run off with it. You must hold
1495 * the appropriate locks or have a private queue to do this.
1496 *
1497 * Returns %NULL for an empty list or a pointer to the head element.
1498 * The reference count is not incremented and the reference is therefore
1499 * volatile. Use with caution.
1500 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001501static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001502{
Eric Dumazet18d07002012-04-30 16:31:46 +00001503 struct sk_buff *skb = list_->next;
1504
1505 if (skb == (struct sk_buff *)list_)
1506 skb = NULL;
1507 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001508}
1509
1510/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001511 * skb_peek_next - peek skb following the given one from a queue
1512 * @skb: skb to start from
1513 * @list_: list to peek at
1514 *
1515 * Returns %NULL when the end of the list is met or a pointer to the
1516 * next element. The reference count is not incremented and the
1517 * reference is therefore volatile. Use with caution.
1518 */
1519static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1520 const struct sk_buff_head *list_)
1521{
1522 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001523
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001524 if (next == (struct sk_buff *)list_)
1525 next = NULL;
1526 return next;
1527}
1528
1529/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001530 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531 * @list_: list to peek at
1532 *
1533 * Peek an &sk_buff. Unlike most other operations you _MUST_
1534 * be careful with this one. A peek leaves the buffer on the
1535 * list and someone else may run off with it. You must hold
1536 * the appropriate locks or have a private queue to do this.
1537 *
1538 * Returns %NULL for an empty list or a pointer to the tail element.
1539 * The reference count is not incremented and the reference is therefore
1540 * volatile. Use with caution.
1541 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001542static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001543{
Eric Dumazet18d07002012-04-30 16:31:46 +00001544 struct sk_buff *skb = list_->prev;
1545
1546 if (skb == (struct sk_buff *)list_)
1547 skb = NULL;
1548 return skb;
1549
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550}
1551
1552/**
1553 * skb_queue_len - get queue length
1554 * @list_: list to measure
1555 *
1556 * Return the length of an &sk_buff queue.
1557 */
1558static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1559{
1560 return list_->qlen;
1561}
1562
David S. Miller67fed452008-09-21 22:36:24 -07001563/**
1564 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1565 * @list: queue to initialize
1566 *
1567 * This initializes only the list and queue length aspects of
1568 * an sk_buff_head object. This allows to initialize the list
1569 * aspects of an sk_buff_head without reinitializing things like
1570 * the spinlock. It can also be used for on-stack sk_buff_head
1571 * objects where the spinlock is known to not be used.
1572 */
1573static inline void __skb_queue_head_init(struct sk_buff_head *list)
1574{
1575 list->prev = list->next = (struct sk_buff *)list;
1576 list->qlen = 0;
1577}
1578
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001579/*
1580 * This function creates a split out lock class for each invocation;
1581 * this is needed for now since a whole lot of users of the skb-queue
1582 * infrastructure in drivers have different locking usage (in hardirq)
1583 * than the networking core (in softirq only). In the long run either the
1584 * network layer or drivers should need annotation to consolidate the
1585 * main types of usage into 3 classes.
1586 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001587static inline void skb_queue_head_init(struct sk_buff_head *list)
1588{
1589 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001590 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001591}
1592
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001593static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1594 struct lock_class_key *class)
1595{
1596 skb_queue_head_init(list);
1597 lockdep_set_class(&list->lock, class);
1598}
1599
Linus Torvalds1da177e2005-04-16 15:20:36 -07001600/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001601 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001602 *
1603 * The "__skb_xxxx()" functions are the non-atomic ones that
1604 * can only be called with interrupts disabled.
1605 */
Joe Perches7965bd42013-09-26 14:48:15 -07001606void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1607 struct sk_buff_head *list);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001608static inline void __skb_insert(struct sk_buff *newsk,
1609 struct sk_buff *prev, struct sk_buff *next,
1610 struct sk_buff_head *list)
1611{
1612 newsk->next = next;
1613 newsk->prev = prev;
1614 next->prev = prev->next = newsk;
1615 list->qlen++;
1616}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001617
David S. Miller67fed452008-09-21 22:36:24 -07001618static inline void __skb_queue_splice(const struct sk_buff_head *list,
1619 struct sk_buff *prev,
1620 struct sk_buff *next)
1621{
1622 struct sk_buff *first = list->next;
1623 struct sk_buff *last = list->prev;
1624
1625 first->prev = prev;
1626 prev->next = first;
1627
1628 last->next = next;
1629 next->prev = last;
1630}
1631
1632/**
1633 * skb_queue_splice - join two skb lists, this is designed for stacks
1634 * @list: the new list to add
1635 * @head: the place to add it in the first list
1636 */
1637static inline void skb_queue_splice(const struct sk_buff_head *list,
1638 struct sk_buff_head *head)
1639{
1640 if (!skb_queue_empty(list)) {
1641 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001642 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001643 }
1644}
1645
1646/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001647 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001648 * @list: the new list to add
1649 * @head: the place to add it in the first list
1650 *
1651 * The list at @list is reinitialised
1652 */
1653static inline void skb_queue_splice_init(struct sk_buff_head *list,
1654 struct sk_buff_head *head)
1655{
1656 if (!skb_queue_empty(list)) {
1657 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001658 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001659 __skb_queue_head_init(list);
1660 }
1661}
1662
1663/**
1664 * skb_queue_splice_tail - join two skb lists, each list being a queue
1665 * @list: the new list to add
1666 * @head: the place to add it in the first list
1667 */
1668static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1669 struct sk_buff_head *head)
1670{
1671 if (!skb_queue_empty(list)) {
1672 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001673 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001674 }
1675}
1676
1677/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001678 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001679 * @list: the new list to add
1680 * @head: the place to add it in the first list
1681 *
1682 * Each of the lists is a queue.
1683 * The list at @list is reinitialised
1684 */
1685static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1686 struct sk_buff_head *head)
1687{
1688 if (!skb_queue_empty(list)) {
1689 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001690 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001691 __skb_queue_head_init(list);
1692 }
1693}
1694
Linus Torvalds1da177e2005-04-16 15:20:36 -07001695/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001696 * __skb_queue_after - queue a buffer at the list head
1697 * @list: list to use
1698 * @prev: place after this buffer
1699 * @newsk: buffer to queue
1700 *
1701 * Queue a buffer int the middle of a list. This function takes no locks
1702 * and you must therefore hold required locks before calling it.
1703 *
1704 * A buffer cannot be placed on two lists at the same time.
1705 */
1706static inline void __skb_queue_after(struct sk_buff_head *list,
1707 struct sk_buff *prev,
1708 struct sk_buff *newsk)
1709{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001710 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001711}
1712
Joe Perches7965bd42013-09-26 14:48:15 -07001713void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1714 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001715
Gerrit Renkerf5572852008-04-14 00:05:28 -07001716static inline void __skb_queue_before(struct sk_buff_head *list,
1717 struct sk_buff *next,
1718 struct sk_buff *newsk)
1719{
1720 __skb_insert(newsk, next->prev, next, list);
1721}
1722
Stephen Hemminger300ce172005-10-30 13:47:34 -08001723/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724 * __skb_queue_head - queue a buffer at the list head
1725 * @list: list to use
1726 * @newsk: buffer to queue
1727 *
1728 * Queue a buffer at the start of a list. This function takes no locks
1729 * and you must therefore hold required locks before calling it.
1730 *
1731 * A buffer cannot be placed on two lists at the same time.
1732 */
Joe Perches7965bd42013-09-26 14:48:15 -07001733void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001734static inline void __skb_queue_head(struct sk_buff_head *list,
1735 struct sk_buff *newsk)
1736{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001737 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001738}
1739
1740/**
1741 * __skb_queue_tail - queue a buffer at the list tail
1742 * @list: list to use
1743 * @newsk: buffer to queue
1744 *
1745 * Queue a buffer at the end of a list. This function takes no locks
1746 * and you must therefore hold required locks before calling it.
1747 *
1748 * A buffer cannot be placed on two lists at the same time.
1749 */
Joe Perches7965bd42013-09-26 14:48:15 -07001750void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751static inline void __skb_queue_tail(struct sk_buff_head *list,
1752 struct sk_buff *newsk)
1753{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001754 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001755}
1756
Linus Torvalds1da177e2005-04-16 15:20:36 -07001757/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001758 * remove sk_buff from list. _Must_ be called atomically, and with
1759 * the list known..
1760 */
Joe Perches7965bd42013-09-26 14:48:15 -07001761void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001762static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1763{
1764 struct sk_buff *next, *prev;
1765
1766 list->qlen--;
1767 next = skb->next;
1768 prev = skb->prev;
1769 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001770 next->prev = prev;
1771 prev->next = next;
1772}
1773
Gerrit Renkerf525c062008-04-14 00:04:12 -07001774/**
1775 * __skb_dequeue - remove from the head of the queue
1776 * @list: list to dequeue from
1777 *
1778 * Remove the head of the list. This function does not take any locks
1779 * so must be used with appropriate locks held only. The head item is
1780 * returned or %NULL if the list is empty.
1781 */
Joe Perches7965bd42013-09-26 14:48:15 -07001782struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Gerrit Renkerf525c062008-04-14 00:04:12 -07001783static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1784{
1785 struct sk_buff *skb = skb_peek(list);
1786 if (skb)
1787 __skb_unlink(skb, list);
1788 return skb;
1789}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001790
1791/**
1792 * __skb_dequeue_tail - remove from the tail of the queue
1793 * @list: list to dequeue from
1794 *
1795 * Remove the tail of the list. This function does not take any locks
1796 * so must be used with appropriate locks held only. The tail item is
1797 * returned or %NULL if the list is empty.
1798 */
Joe Perches7965bd42013-09-26 14:48:15 -07001799struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001800static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1801{
1802 struct sk_buff *skb = skb_peek_tail(list);
1803 if (skb)
1804 __skb_unlink(skb, list);
1805 return skb;
1806}
1807
1808
David S. Millerbdcc0922012-03-07 20:53:36 -05001809static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001810{
1811 return skb->data_len;
1812}
1813
1814static inline unsigned int skb_headlen(const struct sk_buff *skb)
1815{
1816 return skb->len - skb->data_len;
1817}
1818
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03001819static inline unsigned int skb_pagelen(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001820{
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03001821 unsigned int i, len = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001822
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03001823 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001824 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001825 return len + skb_headlen(skb);
1826}
1827
Ian Campbell131ea662011-08-19 06:25:00 +00001828/**
1829 * __skb_fill_page_desc - initialise a paged fragment in an skb
1830 * @skb: buffer containing fragment to be initialised
1831 * @i: paged fragment index to initialise
1832 * @page: the page to use for this fragment
1833 * @off: the offset to the data with @page
1834 * @size: the length of the data
1835 *
1836 * Initialises the @i'th fragment of @skb to point to &size bytes at
1837 * offset @off within @page.
1838 *
1839 * Does not take any additional reference on the fragment.
1840 */
1841static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1842 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001843{
1844 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1845
Mel Gormanc48a11c2012-07-31 16:44:23 -07001846 /*
Michal Hocko2f064f32015-08-21 14:11:51 -07001847 * Propagate page pfmemalloc to the skb if we can. The problem is
1848 * that not all callers have unique ownership of the page but rely
1849 * on page_is_pfmemalloc doing the right thing(tm).
Mel Gormanc48a11c2012-07-31 16:44:23 -07001850 */
Ian Campbella8605c62011-10-19 23:01:49 +00001851 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001852 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001853 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001854
1855 page = compound_head(page);
Michal Hocko2f064f32015-08-21 14:11:51 -07001856 if (page_is_pfmemalloc(page))
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001857 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00001858}
1859
1860/**
1861 * skb_fill_page_desc - initialise a paged fragment in an skb
1862 * @skb: buffer containing fragment to be initialised
1863 * @i: paged fragment index to initialise
1864 * @page: the page to use for this fragment
1865 * @off: the offset to the data with @page
1866 * @size: the length of the data
1867 *
1868 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01001869 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00001870 * addition updates @skb such that @i is the last fragment.
1871 *
1872 * Does not take any additional reference on the fragment.
1873 */
1874static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1875 struct page *page, int off, int size)
1876{
1877 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001878 skb_shinfo(skb)->nr_frags = i + 1;
1879}
1880
Joe Perches7965bd42013-09-26 14:48:15 -07001881void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1882 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001883
Jason Wangf8e617e2013-11-01 14:07:47 +08001884void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1885 unsigned int truesize);
1886
Linus Torvalds1da177e2005-04-16 15:20:36 -07001887#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001888#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001889#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1890
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001891#ifdef NET_SKBUFF_DATA_USES_OFFSET
1892static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1893{
1894 return skb->head + skb->tail;
1895}
1896
1897static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1898{
1899 skb->tail = skb->data - skb->head;
1900}
1901
1902static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1903{
1904 skb_reset_tail_pointer(skb);
1905 skb->tail += offset;
1906}
Simon Horman7cc46192013-05-28 20:34:29 +00001907
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001908#else /* NET_SKBUFF_DATA_USES_OFFSET */
1909static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1910{
1911 return skb->tail;
1912}
1913
1914static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1915{
1916 skb->tail = skb->data;
1917}
1918
1919static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1920{
1921 skb->tail = skb->data + offset;
1922}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001923
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001924#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1925
Linus Torvalds1da177e2005-04-16 15:20:36 -07001926/*
1927 * Add data to an sk_buff
1928 */
Mathias Krause0c7ddf32013-11-07 14:18:24 +01001929unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001930unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001931static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1932{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001933 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001934 SKB_LINEAR_ASSERT(skb);
1935 skb->tail += len;
1936 skb->len += len;
1937 return tmp;
1938}
1939
Johannes Berge45a79d2017-05-24 09:07:47 +02001940static inline unsigned char *skb_put_zero(struct sk_buff *skb, unsigned int len)
1941{
1942 unsigned char *tmp = skb_put(skb, len);
1943
1944 memset(tmp, 0, len);
1945
1946 return tmp;
1947}
1948
Joe Perches7965bd42013-09-26 14:48:15 -07001949unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001950static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1951{
1952 skb->data -= len;
1953 skb->len += len;
1954 return skb->data;
1955}
1956
Joe Perches7965bd42013-09-26 14:48:15 -07001957unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001958static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1959{
1960 skb->len -= len;
1961 BUG_ON(skb->len < skb->data_len);
1962 return skb->data += len;
1963}
1964
David S. Miller47d29642010-05-02 02:21:44 -07001965static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1966{
1967 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1968}
1969
Joe Perches7965bd42013-09-26 14:48:15 -07001970unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001971
1972static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1973{
1974 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001975 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001976 return NULL;
1977 skb->len -= len;
1978 return skb->data += len;
1979}
1980
1981static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1982{
1983 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1984}
1985
1986static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1987{
1988 if (likely(len <= skb_headlen(skb)))
1989 return 1;
1990 if (unlikely(len > skb->len))
1991 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001992 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001993}
1994
Eric Dumazetc8c8b122016-12-07 09:19:33 -08001995void skb_condense(struct sk_buff *skb);
1996
Linus Torvalds1da177e2005-04-16 15:20:36 -07001997/**
1998 * skb_headroom - bytes at buffer head
1999 * @skb: buffer to check
2000 *
2001 * Return the number of bytes of free space at the head of an &sk_buff.
2002 */
Chuck Leverc2636b42007-10-23 21:07:32 -07002003static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002004{
2005 return skb->data - skb->head;
2006}
2007
2008/**
2009 * skb_tailroom - bytes at buffer end
2010 * @skb: buffer to check
2011 *
2012 * Return the number of bytes of free space at the tail of an sk_buff
2013 */
2014static inline int skb_tailroom(const struct sk_buff *skb)
2015{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07002016 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002017}
2018
2019/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00002020 * skb_availroom - bytes at buffer end
2021 * @skb: buffer to check
2022 *
2023 * Return the number of bytes of free space at the tail of an sk_buff
2024 * allocated by sk_stream_alloc()
2025 */
2026static inline int skb_availroom(const struct sk_buff *skb)
2027{
Eric Dumazet16fad692013-03-14 05:40:32 +00002028 if (skb_is_nonlinear(skb))
2029 return 0;
2030
2031 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00002032}
2033
2034/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002035 * skb_reserve - adjust headroom
2036 * @skb: buffer to alter
2037 * @len: bytes to move
2038 *
2039 * Increase the headroom of an empty &sk_buff by reducing the tail
2040 * room. This is only allowed for an empty buffer.
2041 */
David S. Miller8243126c2006-01-17 02:54:21 -08002042static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002043{
2044 skb->data += len;
2045 skb->tail += len;
2046}
2047
Benjamin Poirier1837b2e2016-02-29 15:03:33 -08002048/**
2049 * skb_tailroom_reserve - adjust reserved_tailroom
2050 * @skb: buffer to alter
2051 * @mtu: maximum amount of headlen permitted
2052 * @needed_tailroom: minimum amount of reserved_tailroom
2053 *
2054 * Set reserved_tailroom so that headlen can be as large as possible but
2055 * not larger than mtu and tailroom cannot be smaller than
2056 * needed_tailroom.
2057 * The required headroom should already have been reserved before using
2058 * this function.
2059 */
2060static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2061 unsigned int needed_tailroom)
2062{
2063 SKB_LINEAR_ASSERT(skb);
2064 if (mtu < skb_tailroom(skb) - needed_tailroom)
2065 /* use at most mtu */
2066 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2067 else
2068 /* use up to all available space */
2069 skb->reserved_tailroom = needed_tailroom;
2070}
2071
Tom Herbert8bce6d72014-09-29 20:22:29 -07002072#define ENCAP_TYPE_ETHER 0
2073#define ENCAP_TYPE_IPPROTO 1
2074
2075static inline void skb_set_inner_protocol(struct sk_buff *skb,
2076 __be16 protocol)
2077{
2078 skb->inner_protocol = protocol;
2079 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2080}
2081
2082static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2083 __u8 ipproto)
2084{
2085 skb->inner_ipproto = ipproto;
2086 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2087}
2088
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002089static inline void skb_reset_inner_headers(struct sk_buff *skb)
2090{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002091 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002092 skb->inner_network_header = skb->network_header;
2093 skb->inner_transport_header = skb->transport_header;
2094}
2095
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00002096static inline void skb_reset_mac_len(struct sk_buff *skb)
2097{
2098 skb->mac_len = skb->network_header - skb->mac_header;
2099}
2100
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002101static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2102 *skb)
2103{
2104 return skb->head + skb->inner_transport_header;
2105}
2106
Tom Herbert55dc5a92015-12-14 11:19:40 -08002107static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2108{
2109 return skb_inner_transport_header(skb) - skb->data;
2110}
2111
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002112static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2113{
2114 skb->inner_transport_header = skb->data - skb->head;
2115}
2116
2117static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2118 const int offset)
2119{
2120 skb_reset_inner_transport_header(skb);
2121 skb->inner_transport_header += offset;
2122}
2123
2124static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2125{
2126 return skb->head + skb->inner_network_header;
2127}
2128
2129static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2130{
2131 skb->inner_network_header = skb->data - skb->head;
2132}
2133
2134static inline void skb_set_inner_network_header(struct sk_buff *skb,
2135 const int offset)
2136{
2137 skb_reset_inner_network_header(skb);
2138 skb->inner_network_header += offset;
2139}
2140
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002141static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2142{
2143 return skb->head + skb->inner_mac_header;
2144}
2145
2146static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2147{
2148 skb->inner_mac_header = skb->data - skb->head;
2149}
2150
2151static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2152 const int offset)
2153{
2154 skb_reset_inner_mac_header(skb);
2155 skb->inner_mac_header += offset;
2156}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002157static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2158{
Cong Wang35d04612013-05-29 15:16:05 +08002159 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002160}
2161
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002162static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2163{
2164 return skb->head + skb->transport_header;
2165}
2166
2167static inline void skb_reset_transport_header(struct sk_buff *skb)
2168{
2169 skb->transport_header = skb->data - skb->head;
2170}
2171
2172static inline void skb_set_transport_header(struct sk_buff *skb,
2173 const int offset)
2174{
2175 skb_reset_transport_header(skb);
2176 skb->transport_header += offset;
2177}
2178
2179static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2180{
2181 return skb->head + skb->network_header;
2182}
2183
2184static inline void skb_reset_network_header(struct sk_buff *skb)
2185{
2186 skb->network_header = skb->data - skb->head;
2187}
2188
2189static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2190{
2191 skb_reset_network_header(skb);
2192 skb->network_header += offset;
2193}
2194
2195static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2196{
2197 return skb->head + skb->mac_header;
2198}
2199
Amir Vadaiea6da4f2017-02-07 09:56:06 +02002200static inline int skb_mac_offset(const struct sk_buff *skb)
2201{
2202 return skb_mac_header(skb) - skb->data;
2203}
2204
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002205static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2206{
Cong Wang35d04612013-05-29 15:16:05 +08002207 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002208}
2209
2210static inline void skb_reset_mac_header(struct sk_buff *skb)
2211{
2212 skb->mac_header = skb->data - skb->head;
2213}
2214
2215static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2216{
2217 skb_reset_mac_header(skb);
2218 skb->mac_header += offset;
2219}
2220
Timo Teräs0e3da5b2013-12-16 11:02:09 +02002221static inline void skb_pop_mac_header(struct sk_buff *skb)
2222{
2223 skb->mac_header = skb->network_header;
2224}
2225
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002226static inline void skb_probe_transport_header(struct sk_buff *skb,
2227 const int offset_hint)
2228{
2229 struct flow_keys keys;
2230
2231 if (skb_transport_header_was_set(skb))
2232 return;
Tom Herbertcd79a232015-09-01 09:24:27 -07002233 else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
Tom Herbert42aecaa2015-06-04 09:16:39 -07002234 skb_set_transport_header(skb, keys.control.thoff);
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002235 else
2236 skb_set_transport_header(skb, offset_hint);
2237}
2238
Eric Dumazet03606892012-02-23 10:55:02 +00002239static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2240{
2241 if (skb_mac_header_was_set(skb)) {
2242 const unsigned char *old_mac = skb_mac_header(skb);
2243
2244 skb_set_mac_header(skb, -skb->mac_len);
2245 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2246 }
2247}
2248
Michał Mirosław04fb4512010-12-14 15:24:08 +00002249static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2250{
2251 return skb->csum_start - skb_headroom(skb);
2252}
2253
Alexander Duyck08b64fc2016-02-05 15:27:49 -08002254static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2255{
2256 return skb->head + skb->csum_start;
2257}
2258
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002259static inline int skb_transport_offset(const struct sk_buff *skb)
2260{
2261 return skb_transport_header(skb) - skb->data;
2262}
2263
2264static inline u32 skb_network_header_len(const struct sk_buff *skb)
2265{
2266 return skb->transport_header - skb->network_header;
2267}
2268
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002269static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2270{
2271 return skb->inner_transport_header - skb->inner_network_header;
2272}
2273
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002274static inline int skb_network_offset(const struct sk_buff *skb)
2275{
2276 return skb_network_header(skb) - skb->data;
2277}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03002278
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002279static inline int skb_inner_network_offset(const struct sk_buff *skb)
2280{
2281 return skb_inner_network_header(skb) - skb->data;
2282}
2283
Changli Gaof9599ce2010-08-04 04:43:44 +00002284static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2285{
2286 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2287}
2288
Linus Torvalds1da177e2005-04-16 15:20:36 -07002289/*
2290 * CPUs often take a performance hit when accessing unaligned memory
2291 * locations. The actual performance hit varies, it can be small if the
2292 * hardware handles it or large if we have to take an exception and fix it
2293 * in software.
2294 *
2295 * Since an ethernet header is 14 bytes network drivers often end up with
2296 * the IP header at an unaligned offset. The IP header can be aligned by
2297 * shifting the start of the packet by 2 bytes. Drivers should do this
2298 * with:
2299 *
Tobias Klauser8660c122009-07-13 22:48:16 +00002300 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002301 *
2302 * The downside to this alignment of the IP header is that the DMA is now
2303 * unaligned. On some architectures the cost of an unaligned DMA is high
2304 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00002305 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002306 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2307 * to be overridden.
2308 */
2309#ifndef NET_IP_ALIGN
2310#define NET_IP_ALIGN 2
2311#endif
2312
Anton Blanchard025be812006-03-31 02:27:06 -08002313/*
2314 * The networking layer reserves some headroom in skb data (via
2315 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2316 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08002317 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08002318 *
2319 * Unfortunately this headroom changes the DMA alignment of the resulting
2320 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2321 * on some architectures. An architecture can override this value,
2322 * perhaps setting it to a cacheline in size (since that will maintain
2323 * cacheline alignment of the DMA). It must be a power of 2.
2324 *
David S. Millerd6301d32009-02-08 19:24:13 -08002325 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08002326 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07002327 *
2328 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2329 * to reduce average number of cache lines per packet.
2330 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07002331 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08002332 */
2333#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07002334#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08002335#endif
2336
Joe Perches7965bd42013-09-26 14:48:15 -07002337int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002338
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002339static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002340{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00002341 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07002342 WARN_ON(1);
2343 return;
2344 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002345 skb->len = len;
2346 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002347}
2348
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002349static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2350{
2351 __skb_set_length(skb, len);
2352}
2353
Joe Perches7965bd42013-09-26 14:48:15 -07002354void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002355
2356static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2357{
Herbert Xu3cc0e872006-06-09 16:13:38 -07002358 if (skb->data_len)
2359 return ___pskb_trim(skb, len);
2360 __skb_trim(skb, len);
2361 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002362}
2363
2364static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2365{
2366 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2367}
2368
2369/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07002370 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2371 * @skb: buffer to alter
2372 * @len: new length
2373 *
2374 * This is identical to pskb_trim except that the caller knows that
2375 * the skb is not cloned so we should never get an error due to out-
2376 * of-memory.
2377 */
2378static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2379{
2380 int err = pskb_trim(skb, len);
2381 BUG_ON(err);
2382}
2383
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002384static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2385{
2386 unsigned int diff = len - skb->len;
2387
2388 if (skb_tailroom(skb) < diff) {
2389 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2390 GFP_ATOMIC);
2391 if (ret)
2392 return ret;
2393 }
2394 __skb_set_length(skb, len);
2395 return 0;
2396}
2397
Herbert Xue9fa4f72006-08-13 20:12:58 -07002398/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002399 * skb_orphan - orphan a buffer
2400 * @skb: buffer to orphan
2401 *
2402 * If a buffer currently has an owner then we call the owner's
2403 * destructor function and make the @skb unowned. The buffer continues
2404 * to exist but is no longer charged to its former owner.
2405 */
2406static inline void skb_orphan(struct sk_buff *skb)
2407{
Eric Dumazetc34a7612013-07-30 16:11:15 -07002408 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002409 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002410 skb->destructor = NULL;
2411 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07002412 } else {
2413 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002414 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002415}
2416
2417/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002418 * skb_orphan_frags - orphan the frags contained in a buffer
2419 * @skb: buffer to orphan frags from
2420 * @gfp_mask: allocation mask for replacement pages
2421 *
2422 * For each frag in the SKB which needs a destructor (i.e. has an
2423 * owner) create a copy of that frag and release the original
2424 * page by calling the destructor.
2425 */
2426static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2427{
2428 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2429 return 0;
2430 return skb_copy_ubufs(skb, gfp_mask);
2431}
2432
2433/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002434 * __skb_queue_purge - empty a list
2435 * @list: list to empty
2436 *
2437 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2438 * the list and one reference dropped. This function does not take the
2439 * list lock and the caller must hold the relevant locks to use it.
2440 */
Joe Perches7965bd42013-09-26 14:48:15 -07002441void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002442static inline void __skb_queue_purge(struct sk_buff_head *list)
2443{
2444 struct sk_buff *skb;
2445 while ((skb = __skb_dequeue(list)) != NULL)
2446 kfree_skb(skb);
2447}
2448
Yaogong Wang9f5afea2016-09-07 14:49:28 -07002449void skb_rbtree_purge(struct rb_root *root);
2450
Joe Perches7965bd42013-09-26 14:48:15 -07002451void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002452
Joe Perches7965bd42013-09-26 14:48:15 -07002453struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2454 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07002455
2456/**
2457 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2458 * @dev: network device to receive on
2459 * @length: length to allocate
2460 *
2461 * Allocate a new &sk_buff and assign it a usage count of one. The
2462 * buffer has unspecified headroom built in. Users should allocate
2463 * the headroom they think they need without accounting for the
2464 * built in space. The built in space is used for optimisations.
2465 *
2466 * %NULL is returned if there is no free memory. Although this function
2467 * allocates memory it can be called from an interrupt.
2468 */
2469static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002470 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002471{
2472 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2473}
2474
Eric Dumazet6f532612012-05-18 05:12:12 +00002475/* legacy helper around __netdev_alloc_skb() */
2476static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2477 gfp_t gfp_mask)
2478{
2479 return __netdev_alloc_skb(NULL, length, gfp_mask);
2480}
2481
2482/* legacy helper around netdev_alloc_skb() */
2483static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2484{
2485 return netdev_alloc_skb(NULL, length);
2486}
2487
2488
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002489static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2490 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002491{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002492 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002493
2494 if (NET_IP_ALIGN && skb)
2495 skb_reserve(skb, NET_IP_ALIGN);
2496 return skb;
2497}
2498
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002499static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2500 unsigned int length)
2501{
2502 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2503}
2504
Alexander Duyck181edb22015-05-06 21:12:03 -07002505static inline void skb_free_frag(void *addr)
2506{
Alexander Duyck8c2dd3e2017-01-10 16:58:06 -08002507 page_frag_free(addr);
Alexander Duyck181edb22015-05-06 21:12:03 -07002508}
2509
Alexander Duyckffde7322014-12-09 19:40:42 -08002510void *napi_alloc_frag(unsigned int fragsz);
Alexander Duyckfd11a832014-12-09 19:40:49 -08002511struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2512 unsigned int length, gfp_t gfp_mask);
2513static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2514 unsigned int length)
2515{
2516 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2517}
Jesper Dangaard Brouer795bb1c2016-02-08 13:14:59 +01002518void napi_consume_skb(struct sk_buff *skb, int budget);
2519
2520void __kfree_skb_flush(void);
Jesper Dangaard Brouer15fad712016-02-08 13:15:04 +01002521void __kfree_skb_defer(struct sk_buff *skb);
Alexander Duyckffde7322014-12-09 19:40:42 -08002522
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002523/**
Alexander Duyck71dfda52014-11-11 09:26:34 -08002524 * __dev_alloc_pages - allocate page for network Rx
2525 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2526 * @order: size of the allocation
2527 *
2528 * Allocate a new page.
2529 *
2530 * %NULL is returned if there is no free memory.
2531*/
2532static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2533 unsigned int order)
2534{
2535 /* This piece of code contains several assumptions.
2536 * 1. This is for device Rx, therefor a cold page is preferred.
2537 * 2. The expectation is the user wants a compound page.
2538 * 3. If requesting a order 0 page it will not be compound
2539 * due to the check to see if order has a value in prep_new_page
2540 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2541 * code in gfp_to_alloc_flags that should be enforcing this.
2542 */
2543 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2544
2545 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2546}
2547
2548static inline struct page *dev_alloc_pages(unsigned int order)
2549{
Neil Horman95829b32016-05-19 11:30:54 -04002550 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002551}
2552
2553/**
2554 * __dev_alloc_page - allocate a page for network Rx
2555 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2556 *
2557 * Allocate a new page.
2558 *
2559 * %NULL is returned if there is no free memory.
2560 */
2561static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2562{
2563 return __dev_alloc_pages(gfp_mask, 0);
2564}
2565
2566static inline struct page *dev_alloc_page(void)
2567{
Neil Horman95829b32016-05-19 11:30:54 -04002568 return dev_alloc_pages(0);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002569}
2570
2571/**
Mel Gorman06140022012-07-31 16:44:24 -07002572 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2573 * @page: The page that was allocated from skb_alloc_page
2574 * @skb: The skb that may need pfmemalloc set
2575 */
2576static inline void skb_propagate_pfmemalloc(struct page *page,
2577 struct sk_buff *skb)
2578{
Michal Hocko2f064f32015-08-21 14:11:51 -07002579 if (page_is_pfmemalloc(page))
Mel Gorman06140022012-07-31 16:44:24 -07002580 skb->pfmemalloc = true;
2581}
2582
Eric Dumazet564824b2010-10-11 19:05:25 +00002583/**
Masanari Iidae2278672014-02-18 22:54:36 +09002584 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002585 * @frag: the paged fragment
2586 *
2587 * Returns the &struct page associated with @frag.
2588 */
2589static inline struct page *skb_frag_page(const skb_frag_t *frag)
2590{
Ian Campbella8605c62011-10-19 23:01:49 +00002591 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00002592}
2593
2594/**
2595 * __skb_frag_ref - take an addition reference on a paged fragment.
2596 * @frag: the paged fragment
2597 *
2598 * Takes an additional reference on the paged fragment @frag.
2599 */
2600static inline void __skb_frag_ref(skb_frag_t *frag)
2601{
2602 get_page(skb_frag_page(frag));
2603}
2604
2605/**
2606 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2607 * @skb: the buffer
2608 * @f: the fragment offset.
2609 *
2610 * Takes an additional reference on the @f'th paged fragment of @skb.
2611 */
2612static inline void skb_frag_ref(struct sk_buff *skb, int f)
2613{
2614 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2615}
2616
2617/**
2618 * __skb_frag_unref - release a reference on a paged fragment.
2619 * @frag: the paged fragment
2620 *
2621 * Releases a reference on the paged fragment @frag.
2622 */
2623static inline void __skb_frag_unref(skb_frag_t *frag)
2624{
2625 put_page(skb_frag_page(frag));
2626}
2627
2628/**
2629 * skb_frag_unref - release a reference on a paged fragment of an skb.
2630 * @skb: the buffer
2631 * @f: the fragment offset
2632 *
2633 * Releases a reference on the @f'th paged fragment of @skb.
2634 */
2635static inline void skb_frag_unref(struct sk_buff *skb, int f)
2636{
2637 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2638}
2639
2640/**
2641 * skb_frag_address - gets the address of the data contained in a paged fragment
2642 * @frag: the paged fragment buffer
2643 *
2644 * Returns the address of the data within @frag. The page must already
2645 * be mapped.
2646 */
2647static inline void *skb_frag_address(const skb_frag_t *frag)
2648{
2649 return page_address(skb_frag_page(frag)) + frag->page_offset;
2650}
2651
2652/**
2653 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2654 * @frag: the paged fragment buffer
2655 *
2656 * Returns the address of the data within @frag. Checks that the page
2657 * is mapped and returns %NULL otherwise.
2658 */
2659static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2660{
2661 void *ptr = page_address(skb_frag_page(frag));
2662 if (unlikely(!ptr))
2663 return NULL;
2664
2665 return ptr + frag->page_offset;
2666}
2667
2668/**
2669 * __skb_frag_set_page - sets the page contained in a paged fragment
2670 * @frag: the paged fragment
2671 * @page: the page to set
2672 *
2673 * Sets the fragment @frag to contain @page.
2674 */
2675static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2676{
Ian Campbella8605c62011-10-19 23:01:49 +00002677 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002678}
2679
2680/**
2681 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2682 * @skb: the buffer
2683 * @f: the fragment offset
2684 * @page: the page to set
2685 *
2686 * Sets the @f'th fragment of @skb to contain @page.
2687 */
2688static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2689 struct page *page)
2690{
2691 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2692}
2693
Eric Dumazet400dfd32013-10-17 16:27:07 -07002694bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2695
Ian Campbell131ea662011-08-19 06:25:00 +00002696/**
2697 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002698 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002699 * @frag: the paged fragment to map
2700 * @offset: the offset within the fragment (starting at the
2701 * fragment's own offset)
2702 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002703 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00002704 *
2705 * Maps the page associated with @frag to @device.
2706 */
2707static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2708 const skb_frag_t *frag,
2709 size_t offset, size_t size,
2710 enum dma_data_direction dir)
2711{
2712 return dma_map_page(dev, skb_frag_page(frag),
2713 frag->page_offset + offset, size, dir);
2714}
2715
Eric Dumazet117632e2011-12-03 21:39:53 +00002716static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2717 gfp_t gfp_mask)
2718{
2719 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2720}
2721
Octavian Purdilabad93e92014-06-12 01:36:26 +03002722
2723static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2724 gfp_t gfp_mask)
2725{
2726 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2727}
2728
2729
Ian Campbell131ea662011-08-19 06:25:00 +00002730/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002731 * skb_clone_writable - is the header of a clone writable
2732 * @skb: buffer to check
2733 * @len: length up to which to write
2734 *
2735 * Returns true if modifying the header part of the cloned buffer
2736 * does not requires the data to be copied.
2737 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002738static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002739{
2740 return !skb_header_cloned(skb) &&
2741 skb_headroom(skb) + len <= skb->hdr_len;
2742}
2743
Daniel Borkmann36976492016-02-19 23:05:25 +01002744static inline int skb_try_make_writable(struct sk_buff *skb,
2745 unsigned int write_len)
2746{
2747 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
2748 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
2749}
2750
Herbert Xud9cc2042007-09-16 16:21:16 -07002751static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2752 int cloned)
2753{
2754 int delta = 0;
2755
Herbert Xud9cc2042007-09-16 16:21:16 -07002756 if (headroom > skb_headroom(skb))
2757 delta = headroom - skb_headroom(skb);
2758
2759 if (delta || cloned)
2760 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2761 GFP_ATOMIC);
2762 return 0;
2763}
2764
Patrick McHardy334a8132007-06-25 04:35:20 -07002765/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002766 * skb_cow - copy header of skb when it is required
2767 * @skb: buffer to cow
2768 * @headroom: needed headroom
2769 *
2770 * If the skb passed lacks sufficient headroom or its data part
2771 * is shared, data is reallocated. If reallocation fails, an error
2772 * is returned and original skb is not changed.
2773 *
2774 * The result is skb with writable area skb->head...skb->tail
2775 * and at least @headroom of space at head.
2776 */
2777static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2778{
Herbert Xud9cc2042007-09-16 16:21:16 -07002779 return __skb_cow(skb, headroom, skb_cloned(skb));
2780}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002781
Herbert Xud9cc2042007-09-16 16:21:16 -07002782/**
2783 * skb_cow_head - skb_cow but only making the head writable
2784 * @skb: buffer to cow
2785 * @headroom: needed headroom
2786 *
2787 * This function is identical to skb_cow except that we replace the
2788 * skb_cloned check by skb_header_cloned. It should be used when
2789 * you only need to push on some header and do not need to modify
2790 * the data.
2791 */
2792static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2793{
2794 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002795}
2796
2797/**
2798 * skb_padto - pad an skbuff up to a minimal size
2799 * @skb: buffer to pad
2800 * @len: minimal length
2801 *
2802 * Pads up a buffer to ensure the trailing bytes exist and are
2803 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002804 * is untouched. Otherwise it is extended. Returns zero on
2805 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002806 */
Herbert Xu5b057c62006-06-23 02:06:41 -07002807static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002808{
2809 unsigned int size = skb->len;
2810 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002811 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002812 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002813}
2814
Alexander Duyck9c0c1122014-12-03 08:17:33 -08002815/**
2816 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2817 * @skb: buffer to pad
2818 * @len: minimal length
2819 *
2820 * Pads up a buffer to ensure the trailing bytes exist and are
2821 * blanked. If the buffer already contains sufficient data it
2822 * is untouched. Otherwise it is extended. Returns zero on
2823 * success. The skb is freed on error.
2824 */
2825static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2826{
2827 unsigned int size = skb->len;
2828
2829 if (unlikely(size < len)) {
2830 len -= size;
2831 if (skb_pad(skb, len))
2832 return -ENOMEM;
2833 __skb_put(skb, len);
2834 }
2835 return 0;
2836}
2837
Linus Torvalds1da177e2005-04-16 15:20:36 -07002838static inline int skb_add_data(struct sk_buff *skb,
Al Viroaf2b0402014-11-27 21:44:24 -05002839 struct iov_iter *from, int copy)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002840{
2841 const int off = skb->len;
2842
2843 if (skb->ip_summed == CHECKSUM_NONE) {
Al Viroaf2b0402014-11-27 21:44:24 -05002844 __wsum csum = 0;
Al Viro15e6cb42016-11-01 22:42:45 -04002845 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
2846 &csum, from)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002847 skb->csum = csum_block_add(skb->csum, csum, off);
2848 return 0;
2849 }
Al Viro15e6cb42016-11-01 22:42:45 -04002850 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002851 return 0;
2852
2853 __skb_trim(skb, off);
2854 return -EFAULT;
2855}
2856
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002857static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2858 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002859{
2860 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002861 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002862
Ian Campbellea2ab692011-08-22 23:44:58 +00002863 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002864 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002865 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002866 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002867}
2868
Herbert Xu364c6ba2006-06-09 16:10:40 -07002869static inline int __skb_linearize(struct sk_buff *skb)
2870{
2871 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2872}
2873
Linus Torvalds1da177e2005-04-16 15:20:36 -07002874/**
2875 * skb_linearize - convert paged skb to linear one
2876 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002877 *
2878 * If there is no free memory -ENOMEM is returned, otherwise zero
2879 * is returned and the old skb data released.
2880 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002881static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002882{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002883 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2884}
2885
2886/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00002887 * skb_has_shared_frag - can any frag be overwritten
2888 * @skb: buffer to test
2889 *
2890 * Return true if the skb has at least one frag that might be modified
2891 * by an external entity (as in vmsplice()/sendfile())
2892 */
2893static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2894{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00002895 return skb_is_nonlinear(skb) &&
2896 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00002897}
2898
2899/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07002900 * skb_linearize_cow - make sure skb is linear and writable
2901 * @skb: buffer to process
2902 *
2903 * If there is no free memory -ENOMEM is returned, otherwise zero
2904 * is returned and the old skb data released.
2905 */
2906static inline int skb_linearize_cow(struct sk_buff *skb)
2907{
2908 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2909 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002910}
2911
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02002912static __always_inline void
2913__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
2914 unsigned int off)
2915{
2916 if (skb->ip_summed == CHECKSUM_COMPLETE)
2917 skb->csum = csum_block_sub(skb->csum,
2918 csum_partial(start, len, 0), off);
2919 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
2920 skb_checksum_start_offset(skb) < 0)
2921 skb->ip_summed = CHECKSUM_NONE;
2922}
2923
Linus Torvalds1da177e2005-04-16 15:20:36 -07002924/**
2925 * skb_postpull_rcsum - update checksum for received skb after pull
2926 * @skb: buffer to update
2927 * @start: start of data before pull
2928 * @len: length of data pulled
2929 *
2930 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002931 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2932 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002933 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002934static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f92006-03-20 22:43:56 -08002935 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002936{
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02002937 __skb_postpull_rcsum(skb, start, len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002938}
2939
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02002940static __always_inline void
2941__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
2942 unsigned int off)
2943{
2944 if (skb->ip_summed == CHECKSUM_COMPLETE)
2945 skb->csum = csum_block_add(skb->csum,
2946 csum_partial(start, len, 0), off);
2947}
Herbert Xucbb042f92006-03-20 22:43:56 -08002948
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02002949/**
2950 * skb_postpush_rcsum - update checksum for received skb after push
2951 * @skb: buffer to update
2952 * @start: start of data after push
2953 * @len: length of data pushed
2954 *
2955 * After doing a push on a received packet, you need to call this to
2956 * update the CHECKSUM_COMPLETE checksum.
2957 */
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01002958static inline void skb_postpush_rcsum(struct sk_buff *skb,
2959 const void *start, unsigned int len)
2960{
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02002961 __skb_postpush_rcsum(skb, start, len, 0);
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01002962}
2963
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02002964unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2965
David S. Miller7ce5a272013-12-02 17:26:05 -05002966/**
WANG Cong82a31b92016-06-30 10:15:22 -07002967 * skb_push_rcsum - push skb and update receive checksum
2968 * @skb: buffer to update
2969 * @len: length of data pulled
2970 *
2971 * This function performs an skb_push on the packet and updates
2972 * the CHECKSUM_COMPLETE checksum. It should be used on
2973 * receive path processing instead of skb_push unless you know
2974 * that the checksum difference is zero (e.g., a valid IP header)
2975 * or you are setting ip_summed to CHECKSUM_NONE.
2976 */
2977static inline unsigned char *skb_push_rcsum(struct sk_buff *skb,
2978 unsigned int len)
2979{
2980 skb_push(skb, len);
2981 skb_postpush_rcsum(skb, skb->data, len);
2982 return skb->data;
2983}
2984
2985/**
David S. Miller7ce5a272013-12-02 17:26:05 -05002986 * pskb_trim_rcsum - trim received skb and update checksum
2987 * @skb: buffer to trim
2988 * @len: new length
2989 *
2990 * This is exactly the same as pskb_trim except that it ensures the
2991 * checksum of received packets are still valid after the operation.
2992 */
2993
2994static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2995{
2996 if (likely(len >= skb->len))
2997 return 0;
2998 if (skb->ip_summed == CHECKSUM_COMPLETE)
2999 skb->ip_summed = CHECKSUM_NONE;
3000 return __pskb_trim(skb, len);
3001}
3002
Daniel Borkmann5293efe2016-08-18 01:00:39 +02003003static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3004{
3005 if (skb->ip_summed == CHECKSUM_COMPLETE)
3006 skb->ip_summed = CHECKSUM_NONE;
3007 __skb_trim(skb, len);
3008 return 0;
3009}
3010
3011static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3012{
3013 if (skb->ip_summed == CHECKSUM_COMPLETE)
3014 skb->ip_summed = CHECKSUM_NONE;
3015 return __skb_grow(skb, len);
3016}
3017
Linus Torvalds1da177e2005-04-16 15:20:36 -07003018#define skb_queue_walk(queue, skb) \
3019 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003020 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07003021 skb = skb->next)
3022
James Chapman46f89142007-04-30 00:07:31 -07003023#define skb_queue_walk_safe(queue, skb, tmp) \
3024 for (skb = (queue)->next, tmp = skb->next; \
3025 skb != (struct sk_buff *)(queue); \
3026 skb = tmp, tmp = skb->next)
3027
David S. Miller1164f522008-09-23 00:49:44 -07003028#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003029 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07003030 skb = skb->next)
3031
3032#define skb_queue_walk_from_safe(queue, skb, tmp) \
3033 for (tmp = skb->next; \
3034 skb != (struct sk_buff *)(queue); \
3035 skb = tmp, tmp = skb->next)
3036
Stephen Hemminger300ce172005-10-30 13:47:34 -08003037#define skb_queue_reverse_walk(queue, skb) \
3038 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003039 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08003040 skb = skb->prev)
3041
David S. Miller686a2952011-01-20 22:47:32 -08003042#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3043 for (skb = (queue)->prev, tmp = skb->prev; \
3044 skb != (struct sk_buff *)(queue); \
3045 skb = tmp, tmp = skb->prev)
3046
3047#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3048 for (tmp = skb->prev; \
3049 skb != (struct sk_buff *)(queue); \
3050 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003051
David S. Miller21dc3302010-08-23 00:13:46 -07003052static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07003053{
3054 return skb_shinfo(skb)->frag_list != NULL;
3055}
3056
3057static inline void skb_frag_list_init(struct sk_buff *skb)
3058{
3059 skb_shinfo(skb)->frag_list = NULL;
3060}
3061
David S. Milleree039872009-06-09 00:17:13 -07003062#define skb_walk_frags(skb, iter) \
3063 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3064
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003065
3066int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
3067 const struct sk_buff *skb);
Paolo Abeni65101ae2017-05-16 11:20:13 +02003068struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3069 struct sk_buff_head *queue,
3070 unsigned int flags,
3071 void (*destructor)(struct sock *sk,
3072 struct sk_buff *skb),
3073 int *peeked, int *off, int *err,
3074 struct sk_buff **last);
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003075struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
Paolo Abeni7c13f972016-11-04 11:28:59 +01003076 void (*destructor)(struct sock *sk,
3077 struct sk_buff *skb),
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003078 int *peeked, int *off, int *err,
3079 struct sk_buff **last);
Joe Perches7965bd42013-09-26 14:48:15 -07003080struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
Paolo Abeni7c13f972016-11-04 11:28:59 +01003081 void (*destructor)(struct sock *sk,
3082 struct sk_buff *skb),
Joe Perches7965bd42013-09-26 14:48:15 -07003083 int *peeked, int *off, int *err);
3084struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3085 int *err);
3086unsigned int datagram_poll(struct file *file, struct socket *sock,
3087 struct poll_table_struct *wait);
Al Viroc0371da2014-11-24 10:42:55 -05003088int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3089 struct iov_iter *to, int size);
David S. Miller51f3d022014-11-05 16:46:40 -05003090static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3091 struct msghdr *msg, int size)
3092{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003093 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
David S. Miller51f3d022014-11-05 16:46:40 -05003094}
Al Viroe5a4b0b2014-11-24 18:17:55 -05003095int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3096 struct msghdr *msg);
Al Viro3a654f92014-06-19 14:15:22 -04003097int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3098 struct iov_iter *from, int len);
Al Viro3a654f92014-06-19 14:15:22 -04003099int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
Joe Perches7965bd42013-09-26 14:48:15 -07003100void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
samanthakumar627d2d62016-04-05 12:41:16 -04003101void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3102static inline void skb_free_datagram_locked(struct sock *sk,
3103 struct sk_buff *skb)
3104{
3105 __skb_free_datagram_locked(sk, skb, 0);
3106}
Joe Perches7965bd42013-09-26 14:48:15 -07003107int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07003108int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3109int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3110__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3111 int len, __wsum csum);
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +02003112int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
Joe Perches7965bd42013-09-26 14:48:15 -07003113 struct pipe_inode_info *pipe, unsigned int len,
Al Viro25869262016-09-17 21:02:10 -04003114 unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07003115void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01003116unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00003117int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3118 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07003119void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3120int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3121void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Florian Westphalde960aa2014-01-26 10:58:16 +01003122unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
Marcelo Ricardo Leitnerae7ef812016-06-02 15:05:41 -03003123bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu);
Joe Perches7965bd42013-09-26 14:48:15 -07003124struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Vlad Yasevich0d5501c2014-08-08 14:42:13 -04003125struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
Jiri Pirkoe2195122014-11-19 14:05:01 +01003126int skb_ensure_writable(struct sk_buff *skb, int write_len);
Shmulik Ladkanibfca4c52016-09-19 19:11:09 +03003127int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
Jiri Pirko93515d52014-11-19 14:05:02 +01003128int skb_vlan_pop(struct sk_buff *skb);
3129int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
Sowmini Varadhan6fa01cc2016-04-22 18:36:35 -07003130struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3131 gfp_t gfp);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003132
Al Viro6ce8e9c2014-04-06 21:25:44 -04003133static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3134{
Al Viro3073f072017-02-17 23:13:25 -05003135 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
Al Viro6ce8e9c2014-04-06 21:25:44 -04003136}
3137
Al Viro7eab8d92014-04-06 21:51:23 -04003138static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3139{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003140 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro7eab8d92014-04-06 21:51:23 -04003141}
3142
Daniel Borkmann2817a332013-10-30 11:50:51 +01003143struct skb_checksum_ops {
3144 __wsum (*update)(const void *mem, int len, __wsum wsum);
3145 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3146};
3147
3148__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3149 __wsum csum, const struct skb_checksum_ops *ops);
3150__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3151 __wsum csum);
3152
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003153static inline void * __must_check
3154__skb_header_pointer(const struct sk_buff *skb, int offset,
3155 int len, void *data, int hlen, void *buffer)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003156{
Patrick McHardy55820ee2005-07-05 14:08:10 -07003157 if (hlen - offset >= len)
David S. Miller690e36e2014-08-23 12:13:41 -07003158 return data + offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003159
David S. Miller690e36e2014-08-23 12:13:41 -07003160 if (!skb ||
3161 skb_copy_bits(skb, offset, buffer, len) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003162 return NULL;
3163
3164 return buffer;
3165}
3166
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003167static inline void * __must_check
3168skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
David S. Miller690e36e2014-08-23 12:13:41 -07003169{
3170 return __skb_header_pointer(skb, offset, len, skb->data,
3171 skb_headlen(skb), buffer);
3172}
3173
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01003174/**
3175 * skb_needs_linearize - check if we need to linearize a given skb
3176 * depending on the given device features.
3177 * @skb: socket buffer to check
3178 * @features: net device features
3179 *
3180 * Returns true if either:
3181 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3182 * 2. skb is fragmented and the device does not support SG.
3183 */
3184static inline bool skb_needs_linearize(struct sk_buff *skb,
3185 netdev_features_t features)
3186{
3187 return skb_is_nonlinear(skb) &&
3188 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3189 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3190}
3191
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03003192static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3193 void *to,
3194 const unsigned int len)
3195{
3196 memcpy(to, skb->data, len);
3197}
3198
3199static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3200 const int offset, void *to,
3201 const unsigned int len)
3202{
3203 memcpy(to, skb->data + offset, len);
3204}
3205
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03003206static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3207 const void *from,
3208 const unsigned int len)
3209{
3210 memcpy(skb->data, from, len);
3211}
3212
3213static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3214 const int offset,
3215 const void *from,
3216 const unsigned int len)
3217{
3218 memcpy(skb->data + offset, from, len);
3219}
3220
Joe Perches7965bd42013-09-26 14:48:15 -07003221void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003222
Patrick Ohlyac45f602009-02-12 05:03:37 +00003223static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3224{
3225 return skb->tstamp;
3226}
3227
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003228/**
3229 * skb_get_timestamp - get timestamp from a skb
3230 * @skb: skb to get stamp from
3231 * @stamp: pointer to struct timeval to store stamp in
3232 *
3233 * Timestamps are stored in the skb as offsets to a base timestamp.
3234 * This function converts the offset back to a struct timeval and stores
3235 * it in stamp.
3236 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00003237static inline void skb_get_timestamp(const struct sk_buff *skb,
3238 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003239{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003240 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003241}
3242
Patrick Ohlyac45f602009-02-12 05:03:37 +00003243static inline void skb_get_timestampns(const struct sk_buff *skb,
3244 struct timespec *stamp)
3245{
3246 *stamp = ktime_to_timespec(skb->tstamp);
3247}
3248
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003249static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003250{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003251 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003252}
3253
Stephen Hemminger164891a2007-04-23 22:26:16 -07003254static inline ktime_t net_timedelta(ktime_t t)
3255{
3256 return ktime_sub(ktime_get_real(), t);
3257}
3258
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003259static inline ktime_t net_invalid_timestamp(void)
3260{
Thomas Gleixner8b0e1952016-12-25 12:30:41 +01003261 return 0;
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003262}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003263
Alexander Duyck62bccb82014-09-04 13:31:35 -04003264struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3265
Richard Cochranc1f19b52010-07-17 08:49:36 +00003266#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3267
Joe Perches7965bd42013-09-26 14:48:15 -07003268void skb_clone_tx_timestamp(struct sk_buff *skb);
3269bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00003270
3271#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3272
3273static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3274{
3275}
3276
3277static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3278{
3279 return false;
3280}
3281
3282#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3283
3284/**
3285 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3286 *
Richard Cochranda92b192011-10-21 00:49:15 +00003287 * PHY drivers may accept clones of transmitted packets for
3288 * timestamping via their phy_driver.txtstamp method. These drivers
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003289 * must call this function to return the skb back to the stack with a
3290 * timestamp.
Richard Cochranda92b192011-10-21 00:49:15 +00003291 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00003292 * @skb: clone of the the original outgoing packet
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003293 * @hwtstamps: hardware time stamps
Richard Cochranc1f19b52010-07-17 08:49:36 +00003294 *
3295 */
3296void skb_complete_tx_timestamp(struct sk_buff *skb,
3297 struct skb_shared_hwtstamps *hwtstamps);
3298
Willem de Bruijne7fd2882014-08-04 22:11:48 -04003299void __skb_tstamp_tx(struct sk_buff *orig_skb,
3300 struct skb_shared_hwtstamps *hwtstamps,
3301 struct sock *sk, int tstype);
3302
Patrick Ohlyac45f602009-02-12 05:03:37 +00003303/**
3304 * skb_tstamp_tx - queue clone of skb with send time stamps
3305 * @orig_skb: the original outgoing packet
3306 * @hwtstamps: hardware time stamps, may be NULL if not available
3307 *
3308 * If the skb has a socket associated, then this function clones the
3309 * skb (thus sharing the actual data and optional structures), stores
3310 * the optional hardware time stamping information (if non NULL) or
3311 * generates a software time stamp (otherwise), then queues the clone
3312 * to the error queue of the socket. Errors are silently ignored.
3313 */
Joe Perches7965bd42013-09-26 14:48:15 -07003314void skb_tstamp_tx(struct sk_buff *orig_skb,
3315 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00003316
Richard Cochran4507a712010-07-17 08:48:28 +00003317static inline void sw_tx_timestamp(struct sk_buff *skb)
3318{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00003319 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
3320 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00003321 skb_tstamp_tx(skb, NULL);
3322}
3323
3324/**
3325 * skb_tx_timestamp() - Driver hook for transmit timestamping
3326 *
3327 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00003328 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00003329 *
David S. Miller73409f32013-12-27 13:04:33 -05003330 * Specifically, one should make absolutely sure that this function is
3331 * called before TX completion of this packet can trigger. Otherwise
3332 * the packet could potentially already be freed.
3333 *
Richard Cochran4507a712010-07-17 08:48:28 +00003334 * @skb: A socket buffer.
3335 */
3336static inline void skb_tx_timestamp(struct sk_buff *skb)
3337{
Richard Cochranc1f19b52010-07-17 08:49:36 +00003338 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00003339 sw_tx_timestamp(skb);
3340}
3341
Johannes Berg6e3e9392011-11-09 10:15:42 +01003342/**
3343 * skb_complete_wifi_ack - deliver skb with wifi status
3344 *
3345 * @skb: the original outgoing packet
3346 * @acked: ack status
3347 *
3348 */
3349void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3350
Joe Perches7965bd42013-09-26 14:48:15 -07003351__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3352__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003353
Herbert Xu60476372007-04-09 11:59:39 -07003354static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3355{
Tom Herbert6edec0e2015-02-10 16:30:28 -08003356 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3357 skb->csum_valid ||
3358 (skb->ip_summed == CHECKSUM_PARTIAL &&
3359 skb_checksum_start_offset(skb) >= 0));
Herbert Xu60476372007-04-09 11:59:39 -07003360}
3361
Herbert Xufb286bb2005-11-10 13:01:24 -08003362/**
3363 * skb_checksum_complete - Calculate checksum of an entire packet
3364 * @skb: packet to process
3365 *
3366 * This function calculates the checksum over the entire packet plus
3367 * the value of skb->csum. The latter can be used to supply the
3368 * checksum of a pseudo header as used by TCP/UDP. It returns the
3369 * checksum.
3370 *
3371 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3372 * this function can be used to verify that checksum on received
3373 * packets. In that case the function should return zero if the
3374 * checksum is correct. In particular, this function will return zero
3375 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3376 * hardware has already verified the correctness of the checksum.
3377 */
Al Viro4381ca32007-07-15 21:00:11 +01003378static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08003379{
Herbert Xu60476372007-04-09 11:59:39 -07003380 return skb_csum_unnecessary(skb) ?
3381 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003382}
3383
Tom Herbert77cffe22014-08-27 21:26:46 -07003384static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3385{
3386 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3387 if (skb->csum_level == 0)
3388 skb->ip_summed = CHECKSUM_NONE;
3389 else
3390 skb->csum_level--;
3391 }
3392}
3393
3394static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3395{
3396 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3397 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3398 skb->csum_level++;
3399 } else if (skb->ip_summed == CHECKSUM_NONE) {
3400 skb->ip_summed = CHECKSUM_UNNECESSARY;
3401 skb->csum_level = 0;
3402 }
3403}
3404
Tom Herbert5a212322014-08-31 15:12:41 -07003405static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
3406{
3407 /* Mark current checksum as bad (typically called from GRO
3408 * path). In the case that ip_summed is CHECKSUM_NONE
3409 * this must be the first checksum encountered in the packet.
3410 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
3411 * checksum after the last one validated. For UDP, a zero
3412 * checksum can not be marked as bad.
3413 */
3414
3415 if (skb->ip_summed == CHECKSUM_NONE ||
3416 skb->ip_summed == CHECKSUM_UNNECESSARY)
3417 skb->csum_bad = 1;
3418}
3419
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003420/* Check if we need to perform checksum complete validation.
3421 *
3422 * Returns true if checksum complete is needed, false otherwise
3423 * (either checksum is unnecessary or zero checksum is allowed).
3424 */
3425static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3426 bool zero_okay,
3427 __sum16 check)
3428{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003429 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3430 skb->csum_valid = 1;
Tom Herbert77cffe22014-08-27 21:26:46 -07003431 __skb_decr_checksum_unnecessary(skb);
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003432 return false;
3433 }
3434
3435 return true;
3436}
3437
3438/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3439 * in checksum_init.
3440 */
3441#define CHECKSUM_BREAK 76
3442
Tom Herbert4e18b9a2015-04-20 14:10:04 -07003443/* Unset checksum-complete
3444 *
3445 * Unset checksum complete can be done when packet is being modified
3446 * (uncompressed for instance) and checksum-complete value is
3447 * invalidated.
3448 */
3449static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3450{
3451 if (skb->ip_summed == CHECKSUM_COMPLETE)
3452 skb->ip_summed = CHECKSUM_NONE;
3453}
3454
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003455/* Validate (init) checksum based on checksum complete.
3456 *
3457 * Return values:
3458 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3459 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3460 * checksum is stored in skb->csum for use in __skb_checksum_complete
3461 * non-zero: value of invalid checksum
3462 *
3463 */
3464static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3465 bool complete,
3466 __wsum psum)
3467{
3468 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3469 if (!csum_fold(csum_add(psum, skb->csum))) {
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003470 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003471 return 0;
3472 }
Tom Herbert5a212322014-08-31 15:12:41 -07003473 } else if (skb->csum_bad) {
3474 /* ip_summed == CHECKSUM_NONE in this case */
Eric Dumazetc91d4602015-05-15 05:48:07 -07003475 return (__force __sum16)1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003476 }
3477
3478 skb->csum = psum;
3479
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003480 if (complete || skb->len <= CHECKSUM_BREAK) {
3481 __sum16 csum;
3482
3483 csum = __skb_checksum_complete(skb);
3484 skb->csum_valid = !csum;
3485 return csum;
3486 }
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003487
3488 return 0;
3489}
3490
3491static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3492{
3493 return 0;
3494}
3495
3496/* Perform checksum validate (init). Note that this is a macro since we only
3497 * want to calculate the pseudo header which is an input function if necessary.
3498 * First we try to validate without any computation (checksum unnecessary) and
3499 * then calculate based on checksum complete calling the function to compute
3500 * pseudo header.
3501 *
3502 * Return values:
3503 * 0: checksum is validated or try to in skb_checksum_complete
3504 * non-zero: value of invalid checksum
3505 */
3506#define __skb_checksum_validate(skb, proto, complete, \
3507 zero_okay, check, compute_pseudo) \
3508({ \
3509 __sum16 __ret = 0; \
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003510 skb->csum_valid = 0; \
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003511 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3512 __ret = __skb_checksum_validate_complete(skb, \
3513 complete, compute_pseudo(skb, proto)); \
3514 __ret; \
3515})
3516
3517#define skb_checksum_init(skb, proto, compute_pseudo) \
3518 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3519
3520#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3521 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3522
3523#define skb_checksum_validate(skb, proto, compute_pseudo) \
3524 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3525
3526#define skb_checksum_validate_zero_check(skb, proto, check, \
3527 compute_pseudo) \
Sabrina Dubroca096a4cf2015-02-06 18:54:19 +01003528 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003529
3530#define skb_checksum_simple_validate(skb) \
3531 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3532
Tom Herbertd96535a2014-08-31 15:12:42 -07003533static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3534{
3535 return (skb->ip_summed == CHECKSUM_NONE &&
3536 skb->csum_valid && !skb->csum_bad);
3537}
3538
3539static inline void __skb_checksum_convert(struct sk_buff *skb,
3540 __sum16 check, __wsum pseudo)
3541{
3542 skb->csum = ~pseudo;
3543 skb->ip_summed = CHECKSUM_COMPLETE;
3544}
3545
3546#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3547do { \
3548 if (__skb_checksum_convert_check(skb)) \
3549 __skb_checksum_convert(skb, check, \
3550 compute_pseudo(skb, proto)); \
3551} while (0)
3552
Tom Herbert15e23962015-02-10 16:30:31 -08003553static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3554 u16 start, u16 offset)
3555{
3556 skb->ip_summed = CHECKSUM_PARTIAL;
3557 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3558 skb->csum_offset = offset - start;
3559}
3560
Tom Herbertdcdc8992015-02-02 16:07:34 -08003561/* Update skbuf and packet to reflect the remote checksum offload operation.
3562 * When called, ptr indicates the starting point for skb->csum when
3563 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3564 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3565 */
3566static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
Tom Herbert15e23962015-02-10 16:30:31 -08003567 int start, int offset, bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08003568{
3569 __wsum delta;
3570
Tom Herbert15e23962015-02-10 16:30:31 -08003571 if (!nopartial) {
3572 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3573 return;
3574 }
3575
Tom Herbertdcdc8992015-02-02 16:07:34 -08003576 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3577 __skb_checksum_complete(skb);
3578 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3579 }
3580
3581 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3582
3583 /* Adjust skb->csum since we changed the packet */
3584 skb->csum = csum_add(skb->csum, delta);
3585}
3586
Florian Westphalcb9c6832017-01-23 18:21:56 +01003587static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
3588{
3589#if IS_ENABLED(CONFIG_NF_CONNTRACK)
Florian Westphala9e419d2017-01-23 18:21:59 +01003590 return (void *)(skb->_nfct & SKB_NFCT_PTRMASK);
Florian Westphalcb9c6832017-01-23 18:21:56 +01003591#else
3592 return NULL;
3593#endif
3594}
3595
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003596#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07003597void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003598static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3599{
3600 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07003601 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003602}
3603static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3604{
3605 if (nfct)
3606 atomic_inc(&nfct->use);
3607}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003608#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003609#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003610static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3611{
3612 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3613 kfree(nf_bridge);
3614}
3615static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3616{
3617 if (nf_bridge)
3618 atomic_inc(&nf_bridge->use);
3619}
3620#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08003621static inline void nf_reset(struct sk_buff *skb)
3622{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003623#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01003624 nf_conntrack_put(skb_nfct(skb));
3625 skb->_nfct = 0;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003626#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003627#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Patrick McHardya193a4a2006-03-20 19:23:05 -08003628 nf_bridge_put(skb->nf_bridge);
3629 skb->nf_bridge = NULL;
3630#endif
3631}
3632
Patrick McHardy124dff02013-04-05 20:42:05 +02003633static inline void nf_reset_trace(struct sk_buff *skb)
3634{
Florian Westphal478b3602014-02-15 23:48:45 +01003635#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00003636 skb->nf_trace = 0;
3637#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003638}
3639
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003640/* Note: This doesn't put any conntrack and bridge info in dst. */
Eric Dumazetb1937222014-09-28 22:18:47 -07003641static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3642 bool copy)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003643{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003644#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01003645 dst->_nfct = src->_nfct;
3646 nf_conntrack_get(skb_nfct(src));
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003647#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003648#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003649 dst->nf_bridge = src->nf_bridge;
3650 nf_bridge_get(src->nf_bridge);
3651#endif
Florian Westphal478b3602014-02-15 23:48:45 +01003652#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Eric Dumazetb1937222014-09-28 22:18:47 -07003653 if (copy)
3654 dst->nf_trace = src->nf_trace;
Florian Westphal478b3602014-02-15 23:48:45 +01003655#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003656}
3657
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003658static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3659{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003660#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01003661 nf_conntrack_put(skb_nfct(dst));
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003662#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003663#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003664 nf_bridge_put(dst->nf_bridge);
3665#endif
Eric Dumazetb1937222014-09-28 22:18:47 -07003666 __nf_copy(dst, src, true);
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003667}
3668
James Morris984bc162006-06-09 00:29:17 -07003669#ifdef CONFIG_NETWORK_SECMARK
3670static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3671{
3672 to->secmark = from->secmark;
3673}
3674
3675static inline void skb_init_secmark(struct sk_buff *skb)
3676{
3677 skb->secmark = 0;
3678}
3679#else
3680static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3681{ }
3682
3683static inline void skb_init_secmark(struct sk_buff *skb)
3684{ }
3685#endif
3686
Eric W. Biederman574f7192014-04-01 12:20:24 -07003687static inline bool skb_irq_freeable(const struct sk_buff *skb)
3688{
3689 return !skb->destructor &&
3690#if IS_ENABLED(CONFIG_XFRM)
3691 !skb->sp &&
3692#endif
Florian Westphalcb9c6832017-01-23 18:21:56 +01003693 !skb_nfct(skb) &&
Eric W. Biederman574f7192014-04-01 12:20:24 -07003694 !skb->_skb_refdst &&
3695 !skb_has_frag_list(skb);
3696}
3697
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003698static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3699{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003700 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003701}
3702
Stephen Hemminger92477442009-03-21 13:39:26 -07003703static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003704{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003705 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003706}
3707
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003708static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3709{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003710 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003711}
3712
David S. Millerd5a9e242009-01-27 16:22:11 -08003713static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3714{
3715 skb->queue_mapping = rx_queue + 1;
3716}
3717
Stephen Hemminger92477442009-03-21 13:39:26 -07003718static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003719{
3720 return skb->queue_mapping - 1;
3721}
3722
Stephen Hemminger92477442009-03-21 13:39:26 -07003723static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003724{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003725 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08003726}
3727
Julian Anastasov4ff06202017-02-06 23:14:12 +02003728static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
3729{
3730 skb->dst_pending_confirm = val;
3731}
3732
3733static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
3734{
3735 return skb->dst_pending_confirm != 0;
3736}
3737
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003738static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3739{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003740#ifdef CONFIG_XFRM
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003741 return skb->sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003742#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003743 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003744#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003745}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003746
Pravin B Shelar68c33162013-02-14 14:02:41 +00003747/* Keeps track of mac header offset relative to skb->head.
3748 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3749 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07003750 * tunnel skb it points to outer mac header.
3751 * Keeps track of level of encapsulation of network headers.
3752 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00003753struct skb_gso_cb {
Alexander Duyck802ab552016-04-10 21:45:03 -04003754 union {
3755 int mac_offset;
3756 int data_offset;
3757 };
Eric Dumazet3347c962013-10-19 11:42:56 -07003758 int encap_level;
Alexander Duyck76443452016-02-05 15:27:37 -08003759 __wsum csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003760 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00003761};
Konstantin Khlebnikov9207f9d2016-01-08 15:21:46 +03003762#define SKB_SGO_CB_OFFSET 32
3763#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
Pravin B Shelar68c33162013-02-14 14:02:41 +00003764
3765static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3766{
3767 return (skb_mac_header(inner_skb) - inner_skb->head) -
3768 SKB_GSO_CB(inner_skb)->mac_offset;
3769}
3770
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00003771static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3772{
3773 int new_headroom, headroom;
3774 int ret;
3775
3776 headroom = skb_headroom(skb);
3777 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3778 if (ret)
3779 return ret;
3780
3781 new_headroom = skb_headroom(skb);
3782 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3783 return 0;
3784}
3785
Alexander Duyck08b64fc2016-02-05 15:27:49 -08003786static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
3787{
3788 /* Do not update partial checksums if remote checksum is enabled. */
3789 if (skb->remcsum_offload)
3790 return;
3791
3792 SKB_GSO_CB(skb)->csum = res;
3793 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
3794}
3795
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003796/* Compute the checksum for a gso segment. First compute the checksum value
3797 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3798 * then add in skb->csum (checksum from csum_start to end of packet).
3799 * skb->csum and csum_start are then updated to reflect the checksum of the
3800 * resultant packet starting from the transport header-- the resultant checksum
3801 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3802 * header.
3803 */
3804static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3805{
Alexander Duyck76443452016-02-05 15:27:37 -08003806 unsigned char *csum_start = skb_transport_header(skb);
3807 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
3808 __wsum partial = SKB_GSO_CB(skb)->csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003809
Alexander Duyck76443452016-02-05 15:27:37 -08003810 SKB_GSO_CB(skb)->csum = res;
3811 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003812
Alexander Duyck76443452016-02-05 15:27:37 -08003813 return csum_fold(csum_partial(csum_start, plen, partial));
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003814}
3815
David S. Millerbdcc0922012-03-07 20:53:36 -05003816static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07003817{
3818 return skb_shinfo(skb)->gso_size;
3819}
3820
Eric Dumazet36a8f392013-09-29 01:21:32 -07003821/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05003822static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02003823{
3824 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3825}
3826
Daniel Borkmann5293efe2016-08-18 01:00:39 +02003827static inline void skb_gso_reset(struct sk_buff *skb)
3828{
3829 skb_shinfo(skb)->gso_size = 0;
3830 skb_shinfo(skb)->gso_segs = 0;
3831 skb_shinfo(skb)->gso_type = 0;
3832}
3833
Joe Perches7965bd42013-09-26 14:48:15 -07003834void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07003835
3836static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3837{
3838 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3839 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003840 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3841
Alexander Duyckb78462e2010-06-02 12:24:37 +00003842 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3843 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07003844 __skb_warn_lro_forwarding(skb);
3845 return true;
3846 }
3847 return false;
3848}
3849
Herbert Xu35fc92a2007-03-26 23:22:20 -07003850static inline void skb_forward_csum(struct sk_buff *skb)
3851{
3852 /* Unfortunately we don't support this one. Any brave souls? */
3853 if (skb->ip_summed == CHECKSUM_COMPLETE)
3854 skb->ip_summed = CHECKSUM_NONE;
3855}
3856
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003857/**
3858 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3859 * @skb: skb to check
3860 *
3861 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3862 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3863 * use this helper, to document places where we make this assertion.
3864 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003865static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003866{
3867#ifdef DEBUG
3868 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3869#endif
3870}
3871
Rusty Russellf35d9d82008-02-04 23:49:54 -05003872bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00003873
Paul Durranted1f50c2014-01-09 10:02:46 +00003874int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
Linus Lüssing9afd85c2015-05-02 14:01:07 +02003875struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
3876 unsigned int transport_len,
3877 __sum16(*skb_chkf)(struct sk_buff *skb));
Paul Durranted1f50c2014-01-09 10:02:46 +00003878
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00003879/**
3880 * skb_head_is_locked - Determine if the skb->head is locked down
3881 * @skb: skb to check
3882 *
3883 * The head on skbs build around a head frag can be removed if they are
3884 * not cloned. This function returns true if the skb head is locked down
3885 * due to either being allocated via kmalloc, or by being a clone with
3886 * multiple references to the head.
3887 */
3888static inline bool skb_head_is_locked(const struct sk_buff *skb)
3889{
3890 return !skb->head_frag || skb_cloned(skb);
3891}
Florian Westphalfe6cc552014-02-13 23:09:12 +01003892
3893/**
3894 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3895 *
3896 * @skb: GSO skb
3897 *
3898 * skb_gso_network_seglen is used to determine the real size of the
3899 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3900 *
3901 * The MAC/L2 header is not accounted for.
3902 */
3903static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3904{
3905 unsigned int hdr_len = skb_transport_header(skb) -
3906 skb_network_header(skb);
3907 return hdr_len + skb_gso_transport_seglen(skb);
3908}
Thomas Grafee122c72015-07-21 10:43:58 +02003909
Edward Cree179bc672016-02-11 20:48:04 +00003910/* Local Checksum Offload.
3911 * Compute outer checksum based on the assumption that the
3912 * inner checksum will be offloaded later.
Edward Creee8ae7b02016-02-11 21:03:37 +00003913 * See Documentation/networking/checksum-offloads.txt for
3914 * explanation of how this works.
Edward Cree179bc672016-02-11 20:48:04 +00003915 * Fill in outer checksum adjustment (e.g. with sum of outer
3916 * pseudo-header) before calling.
3917 * Also ensure that inner checksum is in linear data area.
3918 */
3919static inline __wsum lco_csum(struct sk_buff *skb)
3920{
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003921 unsigned char *csum_start = skb_checksum_start(skb);
3922 unsigned char *l4_hdr = skb_transport_header(skb);
3923 __wsum partial;
Edward Cree179bc672016-02-11 20:48:04 +00003924
3925 /* Start with complement of inner checksum adjustment */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003926 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
3927 skb->csum_offset));
3928
Edward Cree179bc672016-02-11 20:48:04 +00003929 /* Add in checksum of our headers (incl. outer checksum
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003930 * adjustment filled in by caller) and return result.
Edward Cree179bc672016-02-11 20:48:04 +00003931 */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003932 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
Edward Cree179bc672016-02-11 20:48:04 +00003933}
3934
Linus Torvalds1da177e2005-04-16 15:20:36 -07003935#endif /* __KERNEL__ */
3936#endif /* _LINUX_SKBUFF_H */