blob: f1adddc1c5ac67c350e280d89abc3f3c10e2dbac [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>
Jiri Pirko1bd758e2015-05-12 14:56:07 +020037#include <net/flow_dissector.h>
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +020038#include <linux/splice.h>
Bernhard Thaler72b31f72015-05-30 15:27:40 +020039#include <linux/in6.h>
Jamal Hadi Salim8b10cab2016-07-02 06:43:14 -040040#include <linux/if_packet.h>
Tom Herbertf70ea012015-07-31 16:52:10 -070041#include <net/flow.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070042
Tom Herbert7a6ae712015-12-14 11:19:47 -080043/* The interface for checksum offload between the stack and networking drivers
44 * is as follows...
45 *
46 * A. IP checksum related features
47 *
48 * Drivers advertise checksum offload capabilities in the features of a device.
49 * From the stack's point of view these are capabilities offered by the driver,
50 * a driver typically only advertises features that it is capable of offloading
51 * to its device.
52 *
53 * The checksum related features are:
54 *
55 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
56 * IP (one's complement) checksum for any combination
57 * of protocols or protocol layering. The checksum is
58 * computed and set in a packet per the CHECKSUM_PARTIAL
59 * interface (see below).
60 *
61 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
62 * TCP or UDP packets over IPv4. These are specifically
63 * unencapsulated packets of the form IPv4|TCP or
64 * IPv4|UDP where the Protocol field in the IPv4 header
65 * is TCP or UDP. The IPv4 header may contain IP options
66 * This feature cannot be set in features for a device
67 * with NETIF_F_HW_CSUM also set. This feature is being
68 * DEPRECATED (see below).
69 *
70 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
71 * TCP or UDP packets over IPv6. These are specifically
72 * unencapsulated packets of the form IPv6|TCP or
73 * IPv4|UDP where the Next Header field in the IPv6
74 * header is either TCP or UDP. IPv6 extension headers
75 * are not supported with this feature. This feature
76 * cannot be set in features for a device with
77 * NETIF_F_HW_CSUM also set. This feature is being
78 * DEPRECATED (see below).
79 *
80 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
81 * This flag is used only used to disable the RX checksum
82 * feature for a device. The stack will accept receive
83 * checksum indication in packets received on a device
84 * regardless of whether NETIF_F_RXCSUM is set.
85 *
86 * B. Checksumming of received packets by device. Indication of checksum
87 * verification is in set skb->ip_summed. Possible values are:
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010088 *
89 * CHECKSUM_NONE:
90 *
Tom Herbert7a6ae712015-12-14 11:19:47 -080091 * Device did not checksum this packet e.g. due to lack of capabilities.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010092 * The packet contains full (though not verified) checksum in packet but
93 * not in skb->csum. Thus, skb->csum is undefined in this case.
94 *
95 * CHECKSUM_UNNECESSARY:
96 *
97 * The hardware you're dealing with doesn't calculate the full checksum
98 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
Tom Herbert77cffe22014-08-27 21:26:46 -070099 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
100 * if their checksums are okay. skb->csum is still undefined in this case
Tom Herbert7a6ae712015-12-14 11:19:47 -0800101 * though. A driver or device must never modify the checksum field in the
102 * packet even if checksum is verified.
Tom Herbert77cffe22014-08-27 21:26:46 -0700103 *
104 * CHECKSUM_UNNECESSARY is applicable to following protocols:
105 * TCP: IPv6 and IPv4.
106 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
107 * zero UDP checksum for either IPv4 or IPv6, the networking stack
108 * may perform further validation in this case.
109 * GRE: only if the checksum is present in the header.
110 * SCTP: indicates the CRC in SCTP header has been validated.
111 *
112 * skb->csum_level indicates the number of consecutive checksums found in
113 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
114 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
115 * and a device is able to verify the checksums for UDP (possibly zero),
116 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
117 * two. If the device were only able to verify the UDP checksum and not
118 * GRE, either because it doesn't support GRE checksum of because GRE
119 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
120 * not considered in this case).
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100121 *
122 * CHECKSUM_COMPLETE:
123 *
124 * This is the most generic way. The device supplied checksum of the _whole_
125 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
126 * hardware doesn't need to parse L3/L4 headers to implement this.
127 *
128 * Note: Even if device supports only some protocols, but is able to produce
129 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
130 *
131 * CHECKSUM_PARTIAL:
132 *
Tom Herbert6edec0e2015-02-10 16:30:28 -0800133 * A checksum is set up to be offloaded to a device as described in the
134 * output description for CHECKSUM_PARTIAL. This may occur on a packet
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100135 * received directly from another Linux OS, e.g., a virtualized Linux kernel
Tom Herbert6edec0e2015-02-10 16:30:28 -0800136 * on the same host, or it may be set in the input path in GRO or remote
137 * checksum offload. For the purposes of checksum verification, the checksum
138 * referred to by skb->csum_start + skb->csum_offset and any preceding
139 * checksums in the packet are considered verified. Any checksums in the
140 * packet that are after the checksum being offloaded are not considered to
141 * be verified.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100142 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800143 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
144 * in the skb->ip_summed for a packet. Values are:
145 *
146 * CHECKSUM_PARTIAL:
147 *
148 * The driver is required to checksum the packet as seen by hard_start_xmit()
149 * from skb->csum_start up to the end, and to record/write the checksum at
150 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
151 * csum_start and csum_offset values are valid values given the length and
152 * offset of the packet, however they should not attempt to validate that the
153 * checksum refers to a legitimate transport layer checksum-- it is the
154 * purview of the stack to validate that csum_start and csum_offset are set
155 * correctly.
156 *
157 * When the stack requests checksum offload for a packet, the driver MUST
158 * ensure that the checksum is set correctly. A driver can either offload the
159 * checksum calculation to the device, or call skb_checksum_help (in the case
160 * that the device does not support offload for a particular checksum).
161 *
162 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
163 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
164 * checksum offload capability. If a device has limited checksum capabilities
165 * (for instance can only perform NETIF_F_IP_CSUM or NETIF_F_IPV6_CSUM as
166 * described above) a helper function can be called to resolve
167 * CHECKSUM_PARTIAL. The helper functions are skb_csum_off_chk*. The helper
168 * function takes a spec argument that describes the protocol layer that is
169 * supported for checksum offload and can be called for each packet. If a
170 * packet does not match the specification for offload, skb_checksum_help
171 * is called to resolve the checksum.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100172 *
173 * CHECKSUM_NONE:
174 *
175 * The skb was already checksummed by the protocol, or a checksum is not
176 * required.
177 *
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100178 * CHECKSUM_UNNECESSARY:
179 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800180 * This has the same meaning on as CHECKSUM_NONE for checksum offload on
181 * output.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100182 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800183 * CHECKSUM_COMPLETE:
184 * Not used in checksum output. If a driver observes a packet with this value
185 * set in skbuff, if should treat as CHECKSUM_NONE being set.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100186 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800187 * D. Non-IP checksum (CRC) offloads
188 *
189 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
190 * offloading the SCTP CRC in a packet. To perform this offload the stack
191 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
192 * accordingly. Note the there is no indication in the skbuff that the
193 * CHECKSUM_PARTIAL refers to an SCTP checksum, a driver that supports
194 * both IP checksum offload and SCTP CRC offload must verify which offload
195 * is configured for a packet presumably by inspecting packet headers.
196 *
197 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
198 * offloading the FCOE CRC in a packet. To perform this offload the stack
199 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
200 * accordingly. Note the there is no indication in the skbuff that the
201 * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
202 * both IP checksum offload and FCOE CRC offload must verify which offload
203 * is configured for a packet presumably by inspecting packet headers.
204 *
205 * E. Checksumming on output with GSO.
206 *
207 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
208 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
209 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
210 * part of the GSO operation is implied. If a checksum is being offloaded
211 * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
212 * are set to refer to the outermost checksum being offload (two offloaded
213 * checksums are possible with UDP encapsulation).
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100214 */
215
Herbert Xu60476372007-04-09 11:59:39 -0700216/* Don't change this without changing skb_csum_unnecessary! */
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100217#define CHECKSUM_NONE 0
218#define CHECKSUM_UNNECESSARY 1
219#define CHECKSUM_COMPLETE 2
220#define CHECKSUM_PARTIAL 3
Linus Torvalds1da177e2005-04-16 15:20:36 -0700221
Tom Herbert77cffe22014-08-27 21:26:46 -0700222/* Maximum value in skb->csum_level */
223#define SKB_MAX_CSUM_LEVEL 3
224
Tobias Klauser0bec8c82014-07-22 12:06:23 +0200225#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
David S. Millerfc910a22007-03-25 20:27:59 -0700226#define SKB_WITH_OVERHEAD(X) \
Herbert Xudeea84b2007-10-21 16:27:46 -0700227 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
David S. Millerfc910a22007-03-25 20:27:59 -0700228#define SKB_MAX_ORDER(X, ORDER) \
229 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
231#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
232
Eric Dumazet87fb4b72011-10-13 07:28:54 +0000233/* return minimum truesize of one skb containing X bytes of data */
234#define SKB_TRUESIZE(X) ((X) + \
235 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
236 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
237
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238struct net_device;
David Howells716ea3a2007-04-02 20:19:53 -0700239struct scatterlist;
Jens Axboe9c55e012007-11-06 23:30:13 -0800240struct pipe_inode_info;
Herbert Xua8f820aa2014-11-07 21:22:22 +0800241struct iov_iter;
Alexander Duyckfd11a832014-12-09 19:40:49 -0800242struct napi_struct;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700243
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700244#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700245struct nf_conntrack {
246 atomic_t use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700247};
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700248#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +0200250#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251struct nf_bridge_info {
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000252 atomic_t use;
Florian Westphal3eaf4022015-04-02 14:31:44 +0200253 enum {
254 BRNF_PROTO_UNCHANGED,
255 BRNF_PROTO_8021Q,
256 BRNF_PROTO_PPPOE
Florian Westphal7fb48c52015-05-03 22:05:28 +0200257 } orig_proto:8;
Florian Westphal72b1e5e2015-07-23 16:21:30 +0200258 u8 pkt_otherhost:1;
259 u8 in_prerouting:1;
260 u8 bridged_dnat:1;
Bernhard Thaler411ffb42015-05-30 15:28:28 +0200261 __u16 frag_max_size;
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000262 struct net_device *physindev;
Florian Westphal63cdbc02015-09-14 17:06:27 +0200263
264 /* always valid & non-NULL from FORWARD on, for physdev match */
265 struct net_device *physoutdev;
Florian Westphal7fb48c52015-05-03 22:05:28 +0200266 union {
Florian Westphal72b1e5e2015-07-23 16:21:30 +0200267 /* prerouting: detect dnat in orig/reply direction */
Bernhard Thaler72b31f72015-05-30 15:27:40 +0200268 __be32 ipv4_daddr;
269 struct in6_addr ipv6_daddr;
Florian Westphal72b1e5e2015-07-23 16:21:30 +0200270
271 /* after prerouting + nat detected: store original source
272 * mac since neigh resolution overwrites it, only used while
273 * skb is out in neigh layer.
274 */
275 char neigh_header[8];
Bernhard Thaler72b31f72015-05-30 15:27:40 +0200276 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700277};
278#endif
279
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280struct sk_buff_head {
281 /* These two members must be first. */
282 struct sk_buff *next;
283 struct sk_buff *prev;
284
285 __u32 qlen;
286 spinlock_t lock;
287};
288
289struct sk_buff;
290
Ian Campbell9d4dde52011-12-22 23:39:14 +0000291/* To allow 64K frame to be packed as single skb without frag_list we
292 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
293 * buffers which do not start on a page boundary.
294 *
295 * Since GRO uses frags we allocate at least 16 regardless of page
296 * size.
Anton Blancharda715dea2011-03-27 14:57:26 +0000297 */
Ian Campbell9d4dde52011-12-22 23:39:14 +0000298#if (65536/PAGE_SIZE + 1) < 16
David S. Millereec00952011-03-29 23:34:08 -0700299#define MAX_SKB_FRAGS 16UL
Anton Blancharda715dea2011-03-27 14:57:26 +0000300#else
Ian Campbell9d4dde52011-12-22 23:39:14 +0000301#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
Anton Blancharda715dea2011-03-27 14:57:26 +0000302#endif
Hans Westgaard Ry5f74f822016-02-03 09:26:57 +0100303extern int sysctl_max_skb_frags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700304
Marcelo Ricardo Leitner3953c462016-06-02 15:05:40 -0300305/* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
306 * segment using its current segmentation instead.
307 */
308#define GSO_BY_FRAGS 0xFFFF
309
Linus Torvalds1da177e2005-04-16 15:20:36 -0700310typedef struct skb_frag_struct skb_frag_t;
311
312struct skb_frag_struct {
Ian Campbella8605c62011-10-19 23:01:49 +0000313 struct {
314 struct page *p;
315 } page;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000316#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
David S. Millera309bb02007-07-30 18:47:03 -0700317 __u32 page_offset;
318 __u32 size;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000319#else
320 __u16 page_offset;
321 __u16 size;
322#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700323};
324
Eric Dumazet9e903e02011-10-18 21:00:24 +0000325static inline unsigned int skb_frag_size(const skb_frag_t *frag)
326{
327 return frag->size;
328}
329
330static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
331{
332 frag->size = size;
333}
334
335static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
336{
337 frag->size += delta;
338}
339
340static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
341{
342 frag->size -= delta;
343}
344
Patrick Ohlyac45f602009-02-12 05:03:37 +0000345#define HAVE_HW_TIME_STAMP
346
347/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800348 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000349 * @hwtstamp: hardware time stamp transformed into duration
350 * since arbitrary point in time
Patrick Ohlyac45f602009-02-12 05:03:37 +0000351 *
352 * Software time stamps generated by ktime_get_real() are stored in
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400353 * skb->tstamp.
Patrick Ohlyac45f602009-02-12 05:03:37 +0000354 *
355 * hwtstamps can only be compared against other hwtstamps from
356 * the same device.
357 *
358 * This structure is attached to packets as part of the
359 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
360 */
361struct skb_shared_hwtstamps {
362 ktime_t hwtstamp;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000363};
364
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000365/* Definitions for tx_flags in struct skb_shared_info */
366enum {
367 /* generate hardware time stamp */
368 SKBTX_HW_TSTAMP = 1 << 0,
369
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400370 /* generate software time stamp when queueing packet to NIC */
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000371 SKBTX_SW_TSTAMP = 1 << 1,
372
373 /* device driver is going to provide hardware time stamp */
374 SKBTX_IN_PROGRESS = 1 << 2,
375
Shirley Maa6686f22011-07-06 12:22:12 +0000376 /* device driver supports TX zero-copy buffers */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000377 SKBTX_DEV_ZEROCOPY = 1 << 3,
Johannes Berg6e3e9392011-11-09 10:15:42 +0100378
379 /* generate wifi status information (where possible) */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000380 SKBTX_WIFI_STATUS = 1 << 4,
Pravin B Shelarc9af6db2013-02-11 09:27:41 +0000381
382 /* This indicates at least one fragment might be overwritten
383 * (as in vmsplice(), sendfile() ...)
384 * If we need to compute a TX checksum, we'll need to copy
385 * all frags to avoid possible bad checksum
386 */
387 SKBTX_SHARED_FRAG = 1 << 5,
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400388
389 /* generate software time stamp when entering packet scheduling */
390 SKBTX_SCHED_TSTAMP = 1 << 6,
Shirley Maa6686f22011-07-06 12:22:12 +0000391};
392
Willem de Bruijne1c8a602014-08-04 22:11:50 -0400393#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
Soheil Hassas Yeganeh0a2cf202016-04-27 23:39:01 -0400394 SKBTX_SCHED_TSTAMP)
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400395#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
396
Shirley Maa6686f22011-07-06 12:22:12 +0000397/*
398 * The callback notifies userspace to release buffers when skb DMA is done in
399 * lower device, the skb last reference should be 0 when calling this.
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000400 * The zerocopy_success argument is true if zero copy transmit occurred,
401 * false on data copy or out of memory error caused by data copy attempt.
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000402 * The ctx field is used to track device context.
403 * The desc field is used to track userspace buffer index.
Shirley Maa6686f22011-07-06 12:22:12 +0000404 */
405struct ubuf_info {
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000406 void (*callback)(struct ubuf_info *, bool zerocopy_success);
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000407 void *ctx;
Shirley Maa6686f22011-07-06 12:22:12 +0000408 unsigned long desc;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000409};
410
Linus Torvalds1da177e2005-04-16 15:20:36 -0700411/* This data is invariant across clones and lives at
412 * the end of the header data, ie. at skb->end.
413 */
414struct skb_shared_info {
Ian Campbell9f42f122012-01-05 07:13:39 +0000415 unsigned char nr_frags;
416 __u8 tx_flags;
Herbert Xu79671682006-06-22 02:40:14 -0700417 unsigned short gso_size;
418 /* Warning: this field is not always filled in (UFO)! */
419 unsigned short gso_segs;
420 unsigned short gso_type;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700421 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000422 struct skb_shared_hwtstamps hwtstamps;
Willem de Bruijn09c2d252014-08-04 22:11:47 -0400423 u32 tskey;
Ian Campbell9f42f122012-01-05 07:13:39 +0000424 __be32 ip6_frag_id;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700425
426 /*
427 * Warning : all fields before dataref are cleared in __alloc_skb()
428 */
429 atomic_t dataref;
430
Johann Baudy69e3c752009-05-18 22:11:22 -0700431 /* Intermediate layers must ensure that destructor_arg
432 * remains valid until skb destructor */
433 void * destructor_arg;
Shirley Maa6686f22011-07-06 12:22:12 +0000434
Eric Dumazetfed66382010-07-22 19:09:08 +0000435 /* must be last field, see pskb_expand_head() */
436 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700437};
438
439/* We divide dataref into two halves. The higher 16 bits hold references
440 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700441 * the entire skb->data. A clone of a headerless skb holds the length of
442 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700443 *
444 * All users must obey the rule that the skb->data reference count must be
445 * greater than or equal to the payload reference count.
446 *
447 * Holding a reference to the payload part means that the user does not
448 * care about modifications to the header part of skb->data.
449 */
450#define SKB_DATAREF_SHIFT 16
451#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
452
David S. Millerd179cd12005-08-17 14:57:30 -0700453
454enum {
Vijay Subramanianc8753d52014-10-02 10:00:43 -0700455 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
456 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
457 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
David S. Millerd179cd12005-08-17 14:57:30 -0700458};
459
Herbert Xu79671682006-06-22 02:40:14 -0700460enum {
461 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700462 SKB_GSO_UDP = 1 << 1,
Herbert Xu576a30e2006-06-27 13:22:38 -0700463
464 /* This indicates the skb is from an untrusted source. */
465 SKB_GSO_DODGY = 1 << 2,
Michael Chanb0da85372006-06-29 12:30:00 -0700466
467 /* This indicates the tcp segment has CWR set. */
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700468 SKB_GSO_TCP_ECN = 1 << 3,
469
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400470 SKB_GSO_TCP_FIXEDID = 1 << 4,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800471
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400472 SKB_GSO_TCPV6 = 1 << 5,
Pravin B Shelar68c33162013-02-14 14:02:41 +0000473
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400474 SKB_GSO_FCOE = 1 << 6,
Pravin B Shelar73136262013-03-07 13:21:51 +0000475
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400476 SKB_GSO_GRE = 1 << 7,
Simon Horman0d89d202013-05-23 21:02:52 +0000477
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400478 SKB_GSO_GRE_CSUM = 1 << 8,
Eric Dumazetcb32f512013-10-19 11:42:57 -0700479
Tom Herbert7e133182016-05-18 09:06:10 -0700480 SKB_GSO_IPXIP4 = 1 << 9,
Eric Dumazet61c1db72013-10-20 20:47:30 -0700481
Tom Herbert7e133182016-05-18 09:06:10 -0700482 SKB_GSO_IPXIP6 = 1 << 10,
Tom Herbert0f4f4ff2014-06-04 17:20:16 -0700483
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400484 SKB_GSO_UDP_TUNNEL = 1 << 11,
Tom Herbert4749c092014-06-04 17:20:23 -0700485
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400486 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 12,
487
Alexander Duyck802ab552016-04-10 21:45:03 -0400488 SKB_GSO_PARTIAL = 1 << 13,
489
490 SKB_GSO_TUNNEL_REMCSUM = 1 << 14,
Marcelo Ricardo Leitner90017ac2016-06-02 15:05:43 -0300491
492 SKB_GSO_SCTP = 1 << 15,
Herbert Xu79671682006-06-22 02:40:14 -0700493};
494
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700495#if BITS_PER_LONG > 32
496#define NET_SKBUFF_DATA_USES_OFFSET 1
497#endif
498
499#ifdef NET_SKBUFF_DATA_USES_OFFSET
500typedef unsigned int sk_buff_data_t;
501#else
502typedef unsigned char *sk_buff_data_t;
503#endif
504
Eric Dumazet363ec39232014-02-26 14:02:11 -0800505/**
506 * struct skb_mstamp - multi resolution time stamps
507 * @stamp_us: timestamp in us resolution
508 * @stamp_jiffies: timestamp in jiffies
509 */
510struct skb_mstamp {
511 union {
512 u64 v64;
513 struct {
514 u32 stamp_us;
515 u32 stamp_jiffies;
516 };
517 };
518};
519
520/**
521 * skb_mstamp_get - get current timestamp
522 * @cl: place to store timestamps
523 */
524static inline void skb_mstamp_get(struct skb_mstamp *cl)
525{
526 u64 val = local_clock();
527
528 do_div(val, NSEC_PER_USEC);
529 cl->stamp_us = (u32)val;
530 cl->stamp_jiffies = (u32)jiffies;
531}
532
533/**
534 * skb_mstamp_delta - compute the difference in usec between two skb_mstamp
535 * @t1: pointer to newest sample
536 * @t0: pointer to oldest sample
537 */
538static inline u32 skb_mstamp_us_delta(const struct skb_mstamp *t1,
539 const struct skb_mstamp *t0)
540{
541 s32 delta_us = t1->stamp_us - t0->stamp_us;
542 u32 delta_jiffies = t1->stamp_jiffies - t0->stamp_jiffies;
543
544 /* If delta_us is negative, this might be because interval is too big,
545 * or local_clock() drift is too big : fallback using jiffies.
546 */
547 if (delta_us <= 0 ||
548 delta_jiffies >= (INT_MAX / (USEC_PER_SEC / HZ)))
549
550 delta_us = jiffies_to_usecs(delta_jiffies);
551
552 return delta_us;
553}
554
Yuchung Cheng625a5e12015-10-16 21:57:45 -0700555static inline bool skb_mstamp_after(const struct skb_mstamp *t1,
556 const struct skb_mstamp *t0)
557{
558 s32 diff = t1->stamp_jiffies - t0->stamp_jiffies;
559
560 if (!diff)
561 diff = t1->stamp_us - t0->stamp_us;
562 return diff > 0;
563}
Eric Dumazet363ec39232014-02-26 14:02:11 -0800564
Linus Torvalds1da177e2005-04-16 15:20:36 -0700565/**
566 * struct sk_buff - socket buffer
567 * @next: Next buffer in list
568 * @prev: Previous buffer in list
Eric Dumazet363ec39232014-02-26 14:02:11 -0800569 * @tstamp: Time we arrived/left
Eric Dumazet56b17422014-11-03 08:19:53 -0800570 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700571 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700572 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700573 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000574 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700575 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700576 * @len: Length of actual data
577 * @data_len: Data length
578 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700579 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700580 * @csum: Checksum (must include start/offset pair)
581 * @csum_start: Offset from skb->head where checksumming should start
582 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700583 * @priority: Packet queueing priority
WANG Cong60ff7462014-05-04 16:39:18 -0700584 * @ignore_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700585 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700586 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700587 * @nohdr: Payload reference only, must not modify header
588 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700589 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700590 * @ipvs_property: skbuff is owned by ipvs
Willem de Bruijne7246e12017-01-07 17:06:35 -0500591 * @tc_skip_classify: do not classify packet. set by IFB device
Willem de Bruijn8dc07fd2017-01-07 17:06:37 -0500592 * @tc_at_ingress: used within tc_classify to distinguish in/egress
Willem de Bruijnbc31c902017-01-07 17:06:38 -0500593 * @tc_redirected: packet was redirected by a tc action
594 * @tc_from_ingress: if tc_redirected, tc_at_ingress at time of redirect
Randy Dunlap31729362008-02-18 20:52:13 -0800595 * @peeked: this packet has been seen already, so stats have been
596 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700597 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700598 * @protocol: Packet protocol from driver
599 * @destructor: Destruct function
Florian Westphala9e419d2017-01-23 18:21:59 +0100600 * @_nfct: Associated connection, if any (with nfctinfo bits)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700601 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800602 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700603 * @tc_index: Traffic control index
Tom Herbert61b905d2014-03-24 15:34:47 -0700604 * @hash: the packet hash
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700605 * @queue_mapping: Queue mapping for multiqueue devices
David S. Miller0b725a22014-08-25 15:51:53 -0700606 * @xmit_more: More SKBs are pending for this queue
Randy Dunlap553a5672008-04-20 10:51:01 -0700607 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700608 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Tom Herbert61b905d2014-03-24 15:34:47 -0700609 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
Changli Gao4ca24622011-08-19 07:26:44 -0700610 * ports.
Tom Herberta3b18dd2014-07-01 21:33:17 -0700611 * @sw_hash: indicates hash was computed in software stack
Johannes Berg6e3e9392011-11-09 10:15:42 +0100612 * @wifi_acked_valid: wifi_acked was set
613 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000614 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Julian Anastasov4ff06202017-02-06 23:14:12 +0200615 * @dst_pending_confirm: need to confirm neighbour
Eliezer Tamir06021292013-06-10 11:39:50 +0300616 * @napi_id: id of the NAPI struct this skb came from
James Morris984bc162006-06-09 00:29:17 -0700617 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700618 * @mark: Generic packet mark
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000619 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b2008-07-14 22:49:06 -0700620 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000621 * @inner_protocol: Protocol (encapsulation)
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000622 * @inner_transport_header: Inner transport layer header (encapsulation)
623 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000624 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700625 * @transport_header: Transport layer header
626 * @network_header: Network layer header
627 * @mac_header: Link layer header
628 * @tail: Tail pointer
629 * @end: End pointer
630 * @head: Head of buffer
631 * @data: Data head pointer
632 * @truesize: Buffer size
633 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700634 */
635
636struct sk_buff {
Eric Dumazet363ec39232014-02-26 14:02:11 -0800637 union {
Eric Dumazet56b17422014-11-03 08:19:53 -0800638 struct {
639 /* These two members must be first. */
640 struct sk_buff *next;
641 struct sk_buff *prev;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000642
Eric Dumazet56b17422014-11-03 08:19:53 -0800643 union {
644 ktime_t tstamp;
645 struct skb_mstamp skb_mstamp;
646 };
647 };
648 struct rb_node rbnode; /* used in netem & tcp stack */
649 };
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000650 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700651
Eric Dumazetc84d9492016-12-08 11:41:55 -0800652 union {
653 struct net_device *dev;
654 /* Some protocols might use this space to store information,
655 * while device pointer would be NULL.
656 * UDP receive path is one user.
657 */
658 unsigned long dev_scratch;
659 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700660 /*
661 * This is the control buffer. It is free to use for every
662 * layer. Please put your private variables there. If you
663 * want to keep them across layers you have to do a skb_clone()
664 * first. This is owned by whoever has the skb queued ATM.
665 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000666 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700667
Eric Dumazet7fee2262010-05-11 23:19:48 +0000668 unsigned long _skb_refdst;
Eric Dumazetb1937222014-09-28 22:18:47 -0700669 void (*destructor)(struct sk_buff *skb);
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000670#ifdef CONFIG_XFRM
671 struct sec_path *sp;
672#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700673#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +0100674 unsigned long _nfct;
Eric Dumazetb1937222014-09-28 22:18:47 -0700675#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +0200676#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Eric Dumazetb1937222014-09-28 22:18:47 -0700677 struct nf_bridge_info *nf_bridge;
678#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700679 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700680 data_len;
681 __u16 mac_len,
682 hdr_len;
Eric Dumazetb1937222014-09-28 22:18:47 -0700683
684 /* Following fields are _not_ copied in __copy_skb_header()
685 * Note that queue_mapping is here mostly to fill a hole.
686 */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200687 kmemcheck_bitfield_begin(flags1);
Eric Dumazetb1937222014-09-28 22:18:47 -0700688 __u16 queue_mapping;
Daniel Borkmann36bbef52016-09-20 00:26:13 +0200689
690/* if you move cloned around you also must adapt those constants */
691#ifdef __BIG_ENDIAN_BITFIELD
692#define CLONED_MASK (1 << 7)
693#else
694#define CLONED_MASK 1
695#endif
696#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
697
698 __u8 __cloned_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700699 __u8 cloned:1,
Harald Welte6869c4d2005-08-09 19:24:19 -0700700 nohdr:1,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800701 fclone:2,
Herbert Xua59322b2007-12-05 01:53:40 -0800702 peeked:1,
Eric Dumazetb1937222014-09-28 22:18:47 -0700703 head_frag:1,
Daniel Borkmann36bbef52016-09-20 00:26:13 +0200704 xmit_more:1,
705 __unused:1; /* one bit hole */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200706 kmemcheck_bitfield_end(flags1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700707
Eric Dumazetb1937222014-09-28 22:18:47 -0700708 /* fields enclosed in headers_start/headers_end are copied
709 * using a single memcpy() in __copy_skb_header()
710 */
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700711 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700712 __u32 headers_start[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700713 /* public: */
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200714
715/* if you move pkt_type around you also must adapt those constants */
716#ifdef __BIG_ENDIAN_BITFIELD
717#define PKT_TYPE_MAX (7 << 5)
718#else
719#define PKT_TYPE_MAX 7
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100720#endif
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200721#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
722
723 __u8 __pkt_type_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700724 __u8 pkt_type:3;
725 __u8 pfmemalloc:1;
726 __u8 ignore_df:1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700727
Eric Dumazetb1937222014-09-28 22:18:47 -0700728 __u8 nf_trace:1;
729 __u8 ip_summed:2;
730 __u8 ooo_okay:1;
731 __u8 l4_hash:1;
732 __u8 sw_hash:1;
733 __u8 wifi_acked_valid:1;
734 __u8 wifi_acked:1;
735
736 __u8 no_fcs:1;
737 /* Indicates the inner headers are valid in the skbuff. */
738 __u8 encapsulation:1;
739 __u8 encap_hdr_csum:1;
740 __u8 csum_valid:1;
741 __u8 csum_complete_sw:1;
742 __u8 csum_level:2;
743 __u8 csum_bad:1;
744
Julian Anastasov4ff06202017-02-06 23:14:12 +0200745 __u8 dst_pending_confirm:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700746#ifdef CONFIG_IPV6_NDISC_NODETYPE
747 __u8 ndisc_nodetype:2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700748#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700749 __u8 ipvs_property:1;
Tom Herbert8bce6d72014-09-29 20:22:29 -0700750 __u8 inner_protocol_type:1;
Tom Herberte585f232014-11-04 09:06:54 -0800751 __u8 remcsum_offload:1;
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200752#ifdef CONFIG_NET_SWITCHDEV
753 __u8 offload_fwd_mark:1;
754#endif
Willem de Bruijne7246e12017-01-07 17:06:35 -0500755#ifdef CONFIG_NET_CLS_ACT
756 __u8 tc_skip_classify:1;
Willem de Bruijn8dc07fd2017-01-07 17:06:37 -0500757 __u8 tc_at_ingress:1;
Willem de Bruijnbc31c902017-01-07 17:06:38 -0500758 __u8 tc_redirected:1;
759 __u8 tc_from_ingress:1;
Willem de Bruijne7246e12017-01-07 17:06:35 -0500760#endif
Alexander Duyck4031ae62012-01-27 06:22:53 +0000761
Linus Torvalds1da177e2005-04-16 15:20:36 -0700762#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700763 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700764#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200765
Eric Dumazetb1937222014-09-28 22:18:47 -0700766 union {
767 __wsum csum;
768 struct {
769 __u16 csum_start;
770 __u16 csum_offset;
771 };
772 };
773 __u32 priority;
774 int skb_iif;
775 __u32 hash;
776 __be16 vlan_proto;
777 __u16 vlan_tci;
Eric Dumazet2bd82482015-02-03 23:48:24 -0800778#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
779 union {
780 unsigned int napi_id;
781 unsigned int sender_cpu;
782 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700783#endif
James Morris984bc162006-06-09 00:29:17 -0700784#ifdef CONFIG_NETWORK_SECMARK
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200785 __u32 secmark;
James Morris984bc162006-06-09 00:29:17 -0700786#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -0700787
Neil Horman3b885782009-10-12 13:26:31 -0700788 union {
789 __u32 mark;
Eric Dumazet16fad692013-03-14 05:40:32 +0000790 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700791 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700792
Tom Herbert8bce6d72014-09-29 20:22:29 -0700793 union {
794 __be16 inner_protocol;
795 __u8 inner_ipproto;
796 };
797
Simon Horman1a37e412013-05-23 21:02:51 +0000798 __u16 inner_transport_header;
799 __u16 inner_network_header;
800 __u16 inner_mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700801
802 __be16 protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000803 __u16 transport_header;
804 __u16 network_header;
805 __u16 mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700806
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700807 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700808 __u32 headers_end[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700809 /* public: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700810
Linus Torvalds1da177e2005-04-16 15:20:36 -0700811 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700812 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700813 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700815 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700816 unsigned int truesize;
817 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700818};
819
820#ifdef __KERNEL__
821/*
822 * Handling routines are only of interest to the kernel
823 */
824#include <linux/slab.h>
825
Linus Torvalds1da177e2005-04-16 15:20:36 -0700826
Mel Gormanc93bdd02012-07-31 16:44:19 -0700827#define SKB_ALLOC_FCLONE 0x01
828#define SKB_ALLOC_RX 0x02
Alexander Duyckfd11a832014-12-09 19:40:49 -0800829#define SKB_ALLOC_NAPI 0x04
Mel Gormanc93bdd02012-07-31 16:44:19 -0700830
831/* Returns true if the skb was allocated from PFMEMALLOC reserves */
832static inline bool skb_pfmemalloc(const struct sk_buff *skb)
833{
834 return unlikely(skb->pfmemalloc);
835}
836
Eric Dumazet7fee2262010-05-11 23:19:48 +0000837/*
838 * skb might have a dst pointer attached, refcounted or not.
839 * _skb_refdst low order bit is set if refcount was _not_ taken
840 */
841#define SKB_DST_NOREF 1UL
842#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
843
Florian Westphala9e419d2017-01-23 18:21:59 +0100844#define SKB_NFCT_PTRMASK ~(7UL)
Eric Dumazet7fee2262010-05-11 23:19:48 +0000845/**
846 * skb_dst - returns skb dst_entry
847 * @skb: buffer
848 *
849 * Returns skb dst_entry, regardless of reference taken or not.
850 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000851static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
852{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000853 /* If refdst was not refcounted, check we still are in a
854 * rcu_read_lock section
855 */
856 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
857 !rcu_read_lock_held() &&
858 !rcu_read_lock_bh_held());
859 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000860}
861
Eric Dumazet7fee2262010-05-11 23:19:48 +0000862/**
863 * skb_dst_set - sets skb dst
864 * @skb: buffer
865 * @dst: dst entry
866 *
867 * Sets skb dst, assuming a reference was taken on dst and should
868 * be released by skb_dst_drop()
869 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000870static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
871{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000872 skb->_skb_refdst = (unsigned long)dst;
873}
874
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200875/**
876 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
877 * @skb: buffer
878 * @dst: dst entry
879 *
880 * Sets skb dst, assuming a reference was not taken on dst.
881 * If dst entry is cached, we do not take reference and dst_release
882 * will be avoided by refdst_drop. If dst entry is not cached, we take
883 * reference, so that last dst_release can destroy the dst immediately.
884 */
885static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
886{
Hannes Frederic Sowadbfc4fb2014-12-06 19:19:42 +0100887 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
888 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200889}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000890
891/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300892 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000893 * @skb: buffer
894 */
895static inline bool skb_dst_is_noref(const struct sk_buff *skb)
896{
897 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000898}
899
Eric Dumazet511c3f92009-06-02 05:14:27 +0000900static inline struct rtable *skb_rtable(const struct sk_buff *skb)
901{
Eric Dumazetadf30902009-06-02 05:19:30 +0000902 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000903}
904
Jamal Hadi Salim8b10cab2016-07-02 06:43:14 -0400905/* For mangling skb->pkt_type from user space side from applications
906 * such as nft, tc, etc, we only allow a conservative subset of
907 * possible pkt_types to be set.
908*/
909static inline bool skb_pkt_type_ok(u32 ptype)
910{
911 return ptype <= PACKET_OTHERHOST;
912}
913
Joe Perches7965bd42013-09-26 14:48:15 -0700914void kfree_skb(struct sk_buff *skb);
915void kfree_skb_list(struct sk_buff *segs);
916void skb_tx_error(struct sk_buff *skb);
917void consume_skb(struct sk_buff *skb);
918void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000919extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000920
Joe Perches7965bd42013-09-26 14:48:15 -0700921void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
922bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
923 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000924
Joe Perches7965bd42013-09-26 14:48:15 -0700925struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
926 int node);
Eric Dumazet2ea2f622015-04-24 16:05:01 -0700927struct sk_buff *__build_skb(void *data, unsigned int frag_size);
Joe Perches7965bd42013-09-26 14:48:15 -0700928struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700929static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100930 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700931{
Eric Dumazet564824b2010-10-11 19:05:25 +0000932 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700933}
934
Eric Dumazet2e4e4412014-09-17 04:49:49 -0700935struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
936 unsigned long data_len,
937 int max_page_order,
938 int *errcode,
939 gfp_t gfp_mask);
940
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700941/* Layout of fast clones : [skb1][skb2][fclone_ref] */
942struct sk_buff_fclones {
943 struct sk_buff skb1;
944
945 struct sk_buff skb2;
946
947 atomic_t fclone_ref;
948};
949
950/**
951 * skb_fclone_busy - check if fclone is busy
Stephen Hemminger293de7d2016-10-23 09:28:29 -0700952 * @sk: socket
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700953 * @skb: buffer
954 *
Masanari Iidabda13fe2015-12-13 16:53:02 +0900955 * Returns true if skb is a fast clone, and its clone is not freed.
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700956 * Some drivers call skb_orphan() in their ndo_start_xmit(),
957 * so we also check that this didnt happen.
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700958 */
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700959static inline bool skb_fclone_busy(const struct sock *sk,
960 const struct sk_buff *skb)
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700961{
962 const struct sk_buff_fclones *fclones;
963
964 fclones = container_of(skb, struct sk_buff_fclones, skb1);
965
966 return skb->fclone == SKB_FCLONE_ORIG &&
Eric Dumazet6ffe75e2014-12-03 17:04:39 -0800967 atomic_read(&fclones->fclone_ref) > 1 &&
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700968 fclones->skb2.sk == sk;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700969}
970
David S. Millerd179cd12005-08-17 14:57:30 -0700971static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100972 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700973{
Mel Gormanc93bdd02012-07-31 16:44:19 -0700974 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700975}
976
Joe Perches7965bd42013-09-26 14:48:15 -0700977struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
Patrick McHardy0ebd0ac2013-04-17 06:46:58 +0000978static inline struct sk_buff *alloc_skb_head(gfp_t priority)
979{
980 return __alloc_skb_head(priority, -1);
981}
982
Joe Perches7965bd42013-09-26 14:48:15 -0700983struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
984int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
985struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
986struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
Octavian Purdilabad93e92014-06-12 01:36:26 +0300987struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
988 gfp_t gfp_mask, bool fclone);
989static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
990 gfp_t gfp_mask)
991{
992 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
993}
Eric Dumazet117632e2011-12-03 21:39:53 +0000994
Joe Perches7965bd42013-09-26 14:48:15 -0700995int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
996struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
997 unsigned int headroom);
998struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
999 int newtailroom, gfp_t priority);
Fan Du25a91d82014-01-18 09:54:23 +08001000int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1001 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001002int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
1003 int len);
1004int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
1005int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +00001006#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007
Joe Perches7965bd42013-09-26 14:48:15 -07001008int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
1009 int getfrag(void *from, char *to, int offset,
1010 int len, int odd, struct sk_buff *skb),
1011 void *from, int length);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001012
Hannes Frederic Sowabe12a1f2015-05-21 16:59:58 +02001013int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1014 int offset, size_t size);
1015
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08001016struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -07001017 __u32 lower_offset;
1018 __u32 upper_offset;
1019 __u32 frag_idx;
1020 __u32 stepped_offset;
1021 struct sk_buff *root_skb;
1022 struct sk_buff *cur_skb;
1023 __u8 *frag_data;
1024};
1025
Joe Perches7965bd42013-09-26 14:48:15 -07001026void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1027 unsigned int to, struct skb_seq_state *st);
1028unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1029 struct skb_seq_state *st);
1030void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -07001031
Joe Perches7965bd42013-09-26 14:48:15 -07001032unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
Bojan Prtvar059a2442015-02-22 11:46:35 +01001033 unsigned int to, struct ts_config *config);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -07001034
Tom Herbert09323cc2013-12-15 22:16:19 -08001035/*
1036 * Packet hash types specify the type of hash in skb_set_hash.
1037 *
1038 * Hash types refer to the protocol layer addresses which are used to
1039 * construct a packet's hash. The hashes are used to differentiate or identify
1040 * flows of the protocol layer for the hash type. Hash types are either
1041 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1042 *
1043 * Properties of hashes:
1044 *
1045 * 1) Two packets in different flows have different hash values
1046 * 2) Two packets in the same flow should have the same hash value
1047 *
1048 * A hash at a higher layer is considered to be more specific. A driver should
1049 * set the most specific hash possible.
1050 *
1051 * A driver cannot indicate a more specific hash than the layer at which a hash
1052 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1053 *
1054 * A driver may indicate a hash level which is less specific than the
1055 * actual layer the hash was computed on. For instance, a hash computed
1056 * at L4 may be considered an L3 hash. This should only be done if the
1057 * driver can't unambiguously determine that the HW computed the hash at
1058 * the higher layer. Note that the "should" in the second property above
1059 * permits this.
1060 */
1061enum pkt_hash_types {
1062 PKT_HASH_TYPE_NONE, /* Undefined type */
1063 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1064 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1065 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1066};
1067
Tom Herbertbcc83832015-09-01 09:24:24 -07001068static inline void skb_clear_hash(struct sk_buff *skb)
1069{
1070 skb->hash = 0;
1071 skb->sw_hash = 0;
1072 skb->l4_hash = 0;
1073}
1074
1075static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1076{
1077 if (!skb->l4_hash)
1078 skb_clear_hash(skb);
1079}
1080
1081static inline void
1082__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1083{
1084 skb->l4_hash = is_l4;
1085 skb->sw_hash = is_sw;
1086 skb->hash = hash;
1087}
1088
Tom Herbert09323cc2013-12-15 22:16:19 -08001089static inline void
1090skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1091{
Tom Herbertbcc83832015-09-01 09:24:24 -07001092 /* Used by drivers to set hash from HW */
1093 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1094}
1095
1096static inline void
1097__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1098{
1099 __skb_set_hash(skb, hash, true, is_l4);
Tom Herbert09323cc2013-12-15 22:16:19 -08001100}
1101
Tom Herberte5276932015-09-01 09:24:23 -07001102void __skb_get_hash(struct sk_buff *skb);
Florian Westphalb9177832016-10-26 18:49:46 +02001103u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
Tom Herberte5276932015-09-01 09:24:23 -07001104u32 skb_get_poff(const struct sk_buff *skb);
1105u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1106 const struct flow_keys *keys, int hlen);
1107__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1108 void *data, int hlen_proto);
1109
1110static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1111 int thoff, u8 ip_proto)
1112{
1113 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1114}
1115
1116void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1117 const struct flow_dissector_key *key,
1118 unsigned int key_count);
1119
1120bool __skb_flow_dissect(const struct sk_buff *skb,
1121 struct flow_dissector *flow_dissector,
1122 void *target_container,
Tom Herbertcd79a232015-09-01 09:24:27 -07001123 void *data, __be16 proto, int nhoff, int hlen,
1124 unsigned int flags);
Tom Herberte5276932015-09-01 09:24:23 -07001125
1126static inline bool skb_flow_dissect(const struct sk_buff *skb,
1127 struct flow_dissector *flow_dissector,
Tom Herbertcd79a232015-09-01 09:24:27 -07001128 void *target_container, unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001129{
1130 return __skb_flow_dissect(skb, flow_dissector, target_container,
Tom Herbertcd79a232015-09-01 09:24:27 -07001131 NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001132}
1133
1134static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
Tom Herbertcd79a232015-09-01 09:24:27 -07001135 struct flow_keys *flow,
1136 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001137{
1138 memset(flow, 0, sizeof(*flow));
1139 return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
Tom Herbertcd79a232015-09-01 09:24:27 -07001140 NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001141}
1142
1143static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
1144 void *data, __be16 proto,
Tom Herbertcd79a232015-09-01 09:24:27 -07001145 int nhoff, int hlen,
1146 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001147{
1148 memset(flow, 0, sizeof(*flow));
1149 return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
Tom Herbertcd79a232015-09-01 09:24:27 -07001150 data, proto, nhoff, hlen, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001151}
1152
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001153static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001154{
Tom Herberta3b18dd2014-07-01 21:33:17 -07001155 if (!skb->l4_hash && !skb->sw_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001156 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001157
Tom Herbert61b905d2014-03-24 15:34:47 -07001158 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001159}
1160
David S. Miller20a17bf2015-09-01 21:19:17 -07001161__u32 __skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6);
Tom Herbertf70ea012015-07-31 16:52:10 -07001162
David S. Miller20a17bf2015-09-01 21:19:17 -07001163static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
Tom Herbertf70ea012015-07-31 16:52:10 -07001164{
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001165 if (!skb->l4_hash && !skb->sw_hash) {
1166 struct flow_keys keys;
Tom Herbertde4c1f82015-09-01 18:11:04 -07001167 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001168
Tom Herbertde4c1f82015-09-01 18:11:04 -07001169 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001170 }
Tom Herbertf70ea012015-07-31 16:52:10 -07001171
1172 return skb->hash;
1173}
1174
David S. Miller20a17bf2015-09-01 21:19:17 -07001175__u32 __skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl);
Tom Herbertf70ea012015-07-31 16:52:10 -07001176
David S. Miller20a17bf2015-09-01 21:19:17 -07001177static inline __u32 skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl4)
Tom Herbertf70ea012015-07-31 16:52:10 -07001178{
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001179 if (!skb->l4_hash && !skb->sw_hash) {
1180 struct flow_keys keys;
Tom Herbertde4c1f82015-09-01 18:11:04 -07001181 __u32 hash = __get_hash_from_flowi4(fl4, &keys);
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001182
Tom Herbertde4c1f82015-09-01 18:11:04 -07001183 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001184 }
Tom Herbertf70ea012015-07-31 16:52:10 -07001185
1186 return skb->hash;
1187}
1188
Tom Herbert50fb7992015-05-01 11:30:12 -07001189__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
1190
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001191static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1192{
Tom Herbert61b905d2014-03-24 15:34:47 -07001193 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001194}
1195
Tom Herbert3df7a742013-12-15 22:16:29 -08001196static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1197{
Tom Herbert61b905d2014-03-24 15:34:47 -07001198 to->hash = from->hash;
Tom Herberta3b18dd2014-07-01 21:33:17 -07001199 to->sw_hash = from->sw_hash;
Tom Herbert61b905d2014-03-24 15:34:47 -07001200 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -08001201};
1202
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001203#ifdef NET_SKBUFF_DATA_USES_OFFSET
1204static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1205{
1206 return skb->head + skb->end;
1207}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001208
1209static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1210{
1211 return skb->end;
1212}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001213#else
1214static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1215{
1216 return skb->end;
1217}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001218
1219static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1220{
1221 return skb->end - skb->head;
1222}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001223#endif
1224
Linus Torvalds1da177e2005-04-16 15:20:36 -07001225/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001226#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001227
Patrick Ohlyac45f602009-02-12 05:03:37 +00001228static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1229{
1230 return &skb_shinfo(skb)->hwtstamps;
1231}
1232
Linus Torvalds1da177e2005-04-16 15:20:36 -07001233/**
1234 * skb_queue_empty - check if a queue is empty
1235 * @list: queue head
1236 *
1237 * Returns true if the queue is empty, false otherwise.
1238 */
1239static inline int skb_queue_empty(const struct sk_buff_head *list)
1240{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001241 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001242}
1243
1244/**
David S. Millerfc7ebb22008-09-23 00:34:07 -07001245 * skb_queue_is_last - check if skb is the last entry in the queue
1246 * @list: queue head
1247 * @skb: buffer
1248 *
1249 * Returns true if @skb is the last buffer on the list.
1250 */
1251static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1252 const struct sk_buff *skb)
1253{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001254 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -07001255}
1256
1257/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001258 * skb_queue_is_first - check if skb is the first entry in the queue
1259 * @list: queue head
1260 * @skb: buffer
1261 *
1262 * Returns true if @skb is the first buffer on the list.
1263 */
1264static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1265 const struct sk_buff *skb)
1266{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001267 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001268}
1269
1270/**
David S. Miller249c8b42008-09-23 00:44:42 -07001271 * skb_queue_next - return the next packet in the queue
1272 * @list: queue head
1273 * @skb: current buffer
1274 *
1275 * Return the next packet in @list after @skb. It is only valid to
1276 * call this if skb_queue_is_last() evaluates to false.
1277 */
1278static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1279 const struct sk_buff *skb)
1280{
1281 /* This BUG_ON may seem severe, but if we just return then we
1282 * are going to dereference garbage.
1283 */
1284 BUG_ON(skb_queue_is_last(list, skb));
1285 return skb->next;
1286}
1287
1288/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001289 * skb_queue_prev - return the prev packet in the queue
1290 * @list: queue head
1291 * @skb: current buffer
1292 *
1293 * Return the prev packet in @list before @skb. It is only valid to
1294 * call this if skb_queue_is_first() evaluates to false.
1295 */
1296static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1297 const struct sk_buff *skb)
1298{
1299 /* This BUG_ON may seem severe, but if we just return then we
1300 * are going to dereference garbage.
1301 */
1302 BUG_ON(skb_queue_is_first(list, skb));
1303 return skb->prev;
1304}
1305
1306/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001307 * skb_get - reference buffer
1308 * @skb: buffer to reference
1309 *
1310 * Makes another reference to a socket buffer and returns a pointer
1311 * to the buffer.
1312 */
1313static inline struct sk_buff *skb_get(struct sk_buff *skb)
1314{
1315 atomic_inc(&skb->users);
1316 return skb;
1317}
1318
1319/*
1320 * If users == 1, we are the only owner and are can avoid redundant
1321 * atomic change.
1322 */
1323
1324/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001325 * skb_cloned - is the buffer a clone
1326 * @skb: buffer to check
1327 *
1328 * Returns true if the buffer was generated with skb_clone() and is
1329 * one of multiple shared copies of the buffer. Cloned buffers are
1330 * shared data so must not be written to under normal circumstances.
1331 */
1332static inline int skb_cloned(const struct sk_buff *skb)
1333{
1334 return skb->cloned &&
1335 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1336}
1337
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001338static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1339{
Mel Gormand0164ad2015-11-06 16:28:21 -08001340 might_sleep_if(gfpflags_allow_blocking(pri));
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001341
1342 if (skb_cloned(skb))
1343 return pskb_expand_head(skb, 0, 0, pri);
1344
1345 return 0;
1346}
1347
Linus Torvalds1da177e2005-04-16 15:20:36 -07001348/**
1349 * skb_header_cloned - is the header a clone
1350 * @skb: buffer to check
1351 *
1352 * Returns true if modifying the header part of the buffer requires
1353 * the data to be copied.
1354 */
1355static inline int skb_header_cloned(const struct sk_buff *skb)
1356{
1357 int dataref;
1358
1359 if (!skb->cloned)
1360 return 0;
1361
1362 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1363 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1364 return dataref != 1;
1365}
1366
Eric Dumazet9580bf22016-04-30 10:19:29 -07001367static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1368{
1369 might_sleep_if(gfpflags_allow_blocking(pri));
1370
1371 if (skb_header_cloned(skb))
1372 return pskb_expand_head(skb, 0, 0, pri);
1373
1374 return 0;
1375}
1376
Linus Torvalds1da177e2005-04-16 15:20:36 -07001377/**
1378 * skb_header_release - release reference to header
1379 * @skb: buffer to operate on
1380 *
1381 * Drop a reference to the header part of the buffer. This is done
1382 * by acquiring a payload reference. You must not read from the header
1383 * part of skb->data after this.
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001384 * Note : Check if you can use __skb_header_release() instead.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001385 */
1386static inline void skb_header_release(struct sk_buff *skb)
1387{
1388 BUG_ON(skb->nohdr);
1389 skb->nohdr = 1;
1390 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1391}
1392
1393/**
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001394 * __skb_header_release - release reference to header
1395 * @skb: buffer to operate on
1396 *
1397 * Variant of skb_header_release() assuming skb is private to caller.
1398 * We can avoid one atomic operation.
1399 */
1400static inline void __skb_header_release(struct sk_buff *skb)
1401{
1402 skb->nohdr = 1;
1403 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1404}
1405
1406
1407/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001408 * skb_shared - is the buffer shared
1409 * @skb: buffer to check
1410 *
1411 * Returns true if more than one person has a reference to this
1412 * buffer.
1413 */
1414static inline int skb_shared(const struct sk_buff *skb)
1415{
1416 return atomic_read(&skb->users) != 1;
1417}
1418
1419/**
1420 * skb_share_check - check if buffer is shared and if so clone it
1421 * @skb: buffer to check
1422 * @pri: priority for memory allocation
1423 *
1424 * If the buffer is shared the buffer is cloned and the old copy
1425 * drops a reference. A new clone with a single reference is returned.
1426 * If the buffer is not shared the original buffer is returned. When
1427 * being called from interrupt status or with spinlocks held pri must
1428 * be GFP_ATOMIC.
1429 *
1430 * NULL is returned on a memory allocation failure.
1431 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001432static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001433{
Mel Gormand0164ad2015-11-06 16:28:21 -08001434 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001435 if (skb_shared(skb)) {
1436 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001437
1438 if (likely(nskb))
1439 consume_skb(skb);
1440 else
1441 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001442 skb = nskb;
1443 }
1444 return skb;
1445}
1446
1447/*
1448 * Copy shared buffers into a new sk_buff. We effectively do COW on
1449 * packets to handle cases where we have a local reader and forward
1450 * and a couple of other messy ones. The normal one is tcpdumping
1451 * a packet thats being forwarded.
1452 */
1453
1454/**
1455 * skb_unshare - make a copy of a shared buffer
1456 * @skb: buffer to check
1457 * @pri: priority for memory allocation
1458 *
1459 * If the socket buffer is a clone then this function creates a new
1460 * copy of the data, drops a reference count on the old copy and returns
1461 * the new copy with the reference count at 1. If the buffer is not a clone
1462 * the original buffer is returned. When called with a spinlock held or
1463 * from interrupt state @pri must be %GFP_ATOMIC
1464 *
1465 * %NULL is returned on a memory allocation failure.
1466 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001467static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001468 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001469{
Mel Gormand0164ad2015-11-06 16:28:21 -08001470 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001471 if (skb_cloned(skb)) {
1472 struct sk_buff *nskb = skb_copy(skb, pri);
Alexander Aring31eff812014-10-10 23:10:47 +02001473
1474 /* Free our shared copy */
1475 if (likely(nskb))
1476 consume_skb(skb);
1477 else
1478 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001479 skb = nskb;
1480 }
1481 return skb;
1482}
1483
1484/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001485 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001486 * @list_: list to peek at
1487 *
1488 * Peek an &sk_buff. Unlike most other operations you _MUST_
1489 * be careful with this one. A peek leaves the buffer on the
1490 * list and someone else may run off with it. You must hold
1491 * the appropriate locks or have a private queue to do this.
1492 *
1493 * Returns %NULL for an empty list or a pointer to the head element.
1494 * The reference count is not incremented and the reference is therefore
1495 * volatile. Use with caution.
1496 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001497static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001498{
Eric Dumazet18d07002012-04-30 16:31:46 +00001499 struct sk_buff *skb = list_->next;
1500
1501 if (skb == (struct sk_buff *)list_)
1502 skb = NULL;
1503 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001504}
1505
1506/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001507 * skb_peek_next - peek skb following the given one from a queue
1508 * @skb: skb to start from
1509 * @list_: list to peek at
1510 *
1511 * Returns %NULL when the end of the list is met or a pointer to the
1512 * next element. The reference count is not incremented and the
1513 * reference is therefore volatile. Use with caution.
1514 */
1515static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1516 const struct sk_buff_head *list_)
1517{
1518 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001519
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001520 if (next == (struct sk_buff *)list_)
1521 next = NULL;
1522 return next;
1523}
1524
1525/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001526 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001527 * @list_: list to peek at
1528 *
1529 * Peek an &sk_buff. Unlike most other operations you _MUST_
1530 * be careful with this one. A peek leaves the buffer on the
1531 * list and someone else may run off with it. You must hold
1532 * the appropriate locks or have a private queue to do this.
1533 *
1534 * Returns %NULL for an empty list or a pointer to the tail element.
1535 * The reference count is not incremented and the reference is therefore
1536 * volatile. Use with caution.
1537 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001538static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001539{
Eric Dumazet18d07002012-04-30 16:31:46 +00001540 struct sk_buff *skb = list_->prev;
1541
1542 if (skb == (struct sk_buff *)list_)
1543 skb = NULL;
1544 return skb;
1545
Linus Torvalds1da177e2005-04-16 15:20:36 -07001546}
1547
1548/**
1549 * skb_queue_len - get queue length
1550 * @list_: list to measure
1551 *
1552 * Return the length of an &sk_buff queue.
1553 */
1554static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1555{
1556 return list_->qlen;
1557}
1558
David S. Miller67fed452008-09-21 22:36:24 -07001559/**
1560 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1561 * @list: queue to initialize
1562 *
1563 * This initializes only the list and queue length aspects of
1564 * an sk_buff_head object. This allows to initialize the list
1565 * aspects of an sk_buff_head without reinitializing things like
1566 * the spinlock. It can also be used for on-stack sk_buff_head
1567 * objects where the spinlock is known to not be used.
1568 */
1569static inline void __skb_queue_head_init(struct sk_buff_head *list)
1570{
1571 list->prev = list->next = (struct sk_buff *)list;
1572 list->qlen = 0;
1573}
1574
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001575/*
1576 * This function creates a split out lock class for each invocation;
1577 * this is needed for now since a whole lot of users of the skb-queue
1578 * infrastructure in drivers have different locking usage (in hardirq)
1579 * than the networking core (in softirq only). In the long run either the
1580 * network layer or drivers should need annotation to consolidate the
1581 * main types of usage into 3 classes.
1582 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001583static inline void skb_queue_head_init(struct sk_buff_head *list)
1584{
1585 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001586 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001587}
1588
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001589static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1590 struct lock_class_key *class)
1591{
1592 skb_queue_head_init(list);
1593 lockdep_set_class(&list->lock, class);
1594}
1595
Linus Torvalds1da177e2005-04-16 15:20:36 -07001596/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001597 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001598 *
1599 * The "__skb_xxxx()" functions are the non-atomic ones that
1600 * can only be called with interrupts disabled.
1601 */
Joe Perches7965bd42013-09-26 14:48:15 -07001602void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1603 struct sk_buff_head *list);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001604static inline void __skb_insert(struct sk_buff *newsk,
1605 struct sk_buff *prev, struct sk_buff *next,
1606 struct sk_buff_head *list)
1607{
1608 newsk->next = next;
1609 newsk->prev = prev;
1610 next->prev = prev->next = newsk;
1611 list->qlen++;
1612}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001613
David S. Miller67fed452008-09-21 22:36:24 -07001614static inline void __skb_queue_splice(const struct sk_buff_head *list,
1615 struct sk_buff *prev,
1616 struct sk_buff *next)
1617{
1618 struct sk_buff *first = list->next;
1619 struct sk_buff *last = list->prev;
1620
1621 first->prev = prev;
1622 prev->next = first;
1623
1624 last->next = next;
1625 next->prev = last;
1626}
1627
1628/**
1629 * skb_queue_splice - join two skb lists, this is designed for stacks
1630 * @list: the new list to add
1631 * @head: the place to add it in the first list
1632 */
1633static inline void skb_queue_splice(const struct sk_buff_head *list,
1634 struct sk_buff_head *head)
1635{
1636 if (!skb_queue_empty(list)) {
1637 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001638 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001639 }
1640}
1641
1642/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001643 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001644 * @list: the new list to add
1645 * @head: the place to add it in the first list
1646 *
1647 * The list at @list is reinitialised
1648 */
1649static inline void skb_queue_splice_init(struct sk_buff_head *list,
1650 struct sk_buff_head *head)
1651{
1652 if (!skb_queue_empty(list)) {
1653 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001654 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001655 __skb_queue_head_init(list);
1656 }
1657}
1658
1659/**
1660 * skb_queue_splice_tail - join two skb lists, each list being a queue
1661 * @list: the new list to add
1662 * @head: the place to add it in the first list
1663 */
1664static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1665 struct sk_buff_head *head)
1666{
1667 if (!skb_queue_empty(list)) {
1668 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001669 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001670 }
1671}
1672
1673/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001674 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001675 * @list: the new list to add
1676 * @head: the place to add it in the first list
1677 *
1678 * Each of the lists is a queue.
1679 * The list at @list is reinitialised
1680 */
1681static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1682 struct sk_buff_head *head)
1683{
1684 if (!skb_queue_empty(list)) {
1685 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001686 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001687 __skb_queue_head_init(list);
1688 }
1689}
1690
Linus Torvalds1da177e2005-04-16 15:20:36 -07001691/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001692 * __skb_queue_after - queue a buffer at the list head
1693 * @list: list to use
1694 * @prev: place after this buffer
1695 * @newsk: buffer to queue
1696 *
1697 * Queue a buffer int the middle of a list. This function takes no locks
1698 * and you must therefore hold required locks before calling it.
1699 *
1700 * A buffer cannot be placed on two lists at the same time.
1701 */
1702static inline void __skb_queue_after(struct sk_buff_head *list,
1703 struct sk_buff *prev,
1704 struct sk_buff *newsk)
1705{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001706 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001707}
1708
Joe Perches7965bd42013-09-26 14:48:15 -07001709void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1710 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001711
Gerrit Renkerf5572852008-04-14 00:05:28 -07001712static inline void __skb_queue_before(struct sk_buff_head *list,
1713 struct sk_buff *next,
1714 struct sk_buff *newsk)
1715{
1716 __skb_insert(newsk, next->prev, next, list);
1717}
1718
Stephen Hemminger300ce172005-10-30 13:47:34 -08001719/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001720 * __skb_queue_head - queue a buffer at the list head
1721 * @list: list to use
1722 * @newsk: buffer to queue
1723 *
1724 * Queue a buffer at the start of a list. This function takes no locks
1725 * and you must therefore hold required locks before calling it.
1726 *
1727 * A buffer cannot be placed on two lists at the same time.
1728 */
Joe Perches7965bd42013-09-26 14:48:15 -07001729void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001730static inline void __skb_queue_head(struct sk_buff_head *list,
1731 struct sk_buff *newsk)
1732{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001733 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001734}
1735
1736/**
1737 * __skb_queue_tail - queue a buffer at the list tail
1738 * @list: list to use
1739 * @newsk: buffer to queue
1740 *
1741 * Queue a buffer at the end of a list. This function takes no locks
1742 * and you must therefore hold required locks before calling it.
1743 *
1744 * A buffer cannot be placed on two lists at the same time.
1745 */
Joe Perches7965bd42013-09-26 14:48:15 -07001746void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001747static inline void __skb_queue_tail(struct sk_buff_head *list,
1748 struct sk_buff *newsk)
1749{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001750 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751}
1752
Linus Torvalds1da177e2005-04-16 15:20:36 -07001753/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001754 * remove sk_buff from list. _Must_ be called atomically, and with
1755 * the list known..
1756 */
Joe Perches7965bd42013-09-26 14:48:15 -07001757void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001758static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1759{
1760 struct sk_buff *next, *prev;
1761
1762 list->qlen--;
1763 next = skb->next;
1764 prev = skb->prev;
1765 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001766 next->prev = prev;
1767 prev->next = next;
1768}
1769
Gerrit Renkerf525c062008-04-14 00:04:12 -07001770/**
1771 * __skb_dequeue - remove from the head of the queue
1772 * @list: list to dequeue from
1773 *
1774 * Remove the head of the list. This function does not take any locks
1775 * so must be used with appropriate locks held only. The head item is
1776 * returned or %NULL if the list is empty.
1777 */
Joe Perches7965bd42013-09-26 14:48:15 -07001778struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Gerrit Renkerf525c062008-04-14 00:04:12 -07001779static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1780{
1781 struct sk_buff *skb = skb_peek(list);
1782 if (skb)
1783 __skb_unlink(skb, list);
1784 return skb;
1785}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001786
1787/**
1788 * __skb_dequeue_tail - remove from the tail of the queue
1789 * @list: list to dequeue from
1790 *
1791 * Remove the tail of the list. This function does not take any locks
1792 * so must be used with appropriate locks held only. The tail item is
1793 * returned or %NULL if the list is empty.
1794 */
Joe Perches7965bd42013-09-26 14:48:15 -07001795struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001796static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1797{
1798 struct sk_buff *skb = skb_peek_tail(list);
1799 if (skb)
1800 __skb_unlink(skb, list);
1801 return skb;
1802}
1803
1804
David S. Millerbdcc0922012-03-07 20:53:36 -05001805static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001806{
1807 return skb->data_len;
1808}
1809
1810static inline unsigned int skb_headlen(const struct sk_buff *skb)
1811{
1812 return skb->len - skb->data_len;
1813}
1814
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03001815static inline unsigned int skb_pagelen(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001816{
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03001817 unsigned int i, len = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001818
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03001819 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001820 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001821 return len + skb_headlen(skb);
1822}
1823
Ian Campbell131ea662011-08-19 06:25:00 +00001824/**
1825 * __skb_fill_page_desc - initialise a paged fragment in an skb
1826 * @skb: buffer containing fragment to be initialised
1827 * @i: paged fragment index to initialise
1828 * @page: the page to use for this fragment
1829 * @off: the offset to the data with @page
1830 * @size: the length of the data
1831 *
1832 * Initialises the @i'th fragment of @skb to point to &size bytes at
1833 * offset @off within @page.
1834 *
1835 * Does not take any additional reference on the fragment.
1836 */
1837static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1838 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001839{
1840 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1841
Mel Gormanc48a11c2012-07-31 16:44:23 -07001842 /*
Michal Hocko2f064f32015-08-21 14:11:51 -07001843 * Propagate page pfmemalloc to the skb if we can. The problem is
1844 * that not all callers have unique ownership of the page but rely
1845 * on page_is_pfmemalloc doing the right thing(tm).
Mel Gormanc48a11c2012-07-31 16:44:23 -07001846 */
Ian Campbella8605c62011-10-19 23:01:49 +00001847 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001848 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001849 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001850
1851 page = compound_head(page);
Michal Hocko2f064f32015-08-21 14:11:51 -07001852 if (page_is_pfmemalloc(page))
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001853 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00001854}
1855
1856/**
1857 * skb_fill_page_desc - initialise a paged fragment in an skb
1858 * @skb: buffer containing fragment to be initialised
1859 * @i: paged fragment index to initialise
1860 * @page: the page to use for this fragment
1861 * @off: the offset to the data with @page
1862 * @size: the length of the data
1863 *
1864 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01001865 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00001866 * addition updates @skb such that @i is the last fragment.
1867 *
1868 * Does not take any additional reference on the fragment.
1869 */
1870static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1871 struct page *page, int off, int size)
1872{
1873 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001874 skb_shinfo(skb)->nr_frags = i + 1;
1875}
1876
Joe Perches7965bd42013-09-26 14:48:15 -07001877void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1878 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001879
Jason Wangf8e617e2013-11-01 14:07:47 +08001880void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1881 unsigned int truesize);
1882
Linus Torvalds1da177e2005-04-16 15:20:36 -07001883#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001884#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001885#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1886
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001887#ifdef NET_SKBUFF_DATA_USES_OFFSET
1888static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1889{
1890 return skb->head + skb->tail;
1891}
1892
1893static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1894{
1895 skb->tail = skb->data - skb->head;
1896}
1897
1898static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1899{
1900 skb_reset_tail_pointer(skb);
1901 skb->tail += offset;
1902}
Simon Horman7cc46192013-05-28 20:34:29 +00001903
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001904#else /* NET_SKBUFF_DATA_USES_OFFSET */
1905static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1906{
1907 return skb->tail;
1908}
1909
1910static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1911{
1912 skb->tail = skb->data;
1913}
1914
1915static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1916{
1917 skb->tail = skb->data + offset;
1918}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001919
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001920#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1921
Linus Torvalds1da177e2005-04-16 15:20:36 -07001922/*
1923 * Add data to an sk_buff
1924 */
Mathias Krause0c7ddf32013-11-07 14:18:24 +01001925unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001926unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001927static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1928{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001929 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001930 SKB_LINEAR_ASSERT(skb);
1931 skb->tail += len;
1932 skb->len += len;
1933 return tmp;
1934}
1935
Joe Perches7965bd42013-09-26 14:48:15 -07001936unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001937static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1938{
1939 skb->data -= len;
1940 skb->len += len;
1941 return skb->data;
1942}
1943
Joe Perches7965bd42013-09-26 14:48:15 -07001944unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001945static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1946{
1947 skb->len -= len;
1948 BUG_ON(skb->len < skb->data_len);
1949 return skb->data += len;
1950}
1951
David S. Miller47d29642010-05-02 02:21:44 -07001952static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1953{
1954 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1955}
1956
Joe Perches7965bd42013-09-26 14:48:15 -07001957unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001958
1959static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1960{
1961 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001962 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001963 return NULL;
1964 skb->len -= len;
1965 return skb->data += len;
1966}
1967
1968static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1969{
1970 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1971}
1972
1973static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1974{
1975 if (likely(len <= skb_headlen(skb)))
1976 return 1;
1977 if (unlikely(len > skb->len))
1978 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001979 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001980}
1981
Eric Dumazetc8c8b122016-12-07 09:19:33 -08001982void skb_condense(struct sk_buff *skb);
1983
Linus Torvalds1da177e2005-04-16 15:20:36 -07001984/**
1985 * skb_headroom - bytes at buffer head
1986 * @skb: buffer to check
1987 *
1988 * Return the number of bytes of free space at the head of an &sk_buff.
1989 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001990static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001991{
1992 return skb->data - skb->head;
1993}
1994
1995/**
1996 * skb_tailroom - bytes at buffer end
1997 * @skb: buffer to check
1998 *
1999 * Return the number of bytes of free space at the tail of an sk_buff
2000 */
2001static inline int skb_tailroom(const struct sk_buff *skb)
2002{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07002003 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002004}
2005
2006/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00002007 * skb_availroom - bytes at buffer end
2008 * @skb: buffer to check
2009 *
2010 * Return the number of bytes of free space at the tail of an sk_buff
2011 * allocated by sk_stream_alloc()
2012 */
2013static inline int skb_availroom(const struct sk_buff *skb)
2014{
Eric Dumazet16fad692013-03-14 05:40:32 +00002015 if (skb_is_nonlinear(skb))
2016 return 0;
2017
2018 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00002019}
2020
2021/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002022 * skb_reserve - adjust headroom
2023 * @skb: buffer to alter
2024 * @len: bytes to move
2025 *
2026 * Increase the headroom of an empty &sk_buff by reducing the tail
2027 * room. This is only allowed for an empty buffer.
2028 */
David S. Miller8243126c2006-01-17 02:54:21 -08002029static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002030{
2031 skb->data += len;
2032 skb->tail += len;
2033}
2034
Benjamin Poirier1837b2e2016-02-29 15:03:33 -08002035/**
2036 * skb_tailroom_reserve - adjust reserved_tailroom
2037 * @skb: buffer to alter
2038 * @mtu: maximum amount of headlen permitted
2039 * @needed_tailroom: minimum amount of reserved_tailroom
2040 *
2041 * Set reserved_tailroom so that headlen can be as large as possible but
2042 * not larger than mtu and tailroom cannot be smaller than
2043 * needed_tailroom.
2044 * The required headroom should already have been reserved before using
2045 * this function.
2046 */
2047static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2048 unsigned int needed_tailroom)
2049{
2050 SKB_LINEAR_ASSERT(skb);
2051 if (mtu < skb_tailroom(skb) - needed_tailroom)
2052 /* use at most mtu */
2053 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2054 else
2055 /* use up to all available space */
2056 skb->reserved_tailroom = needed_tailroom;
2057}
2058
Tom Herbert8bce6d72014-09-29 20:22:29 -07002059#define ENCAP_TYPE_ETHER 0
2060#define ENCAP_TYPE_IPPROTO 1
2061
2062static inline void skb_set_inner_protocol(struct sk_buff *skb,
2063 __be16 protocol)
2064{
2065 skb->inner_protocol = protocol;
2066 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2067}
2068
2069static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2070 __u8 ipproto)
2071{
2072 skb->inner_ipproto = ipproto;
2073 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2074}
2075
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002076static inline void skb_reset_inner_headers(struct sk_buff *skb)
2077{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002078 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002079 skb->inner_network_header = skb->network_header;
2080 skb->inner_transport_header = skb->transport_header;
2081}
2082
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00002083static inline void skb_reset_mac_len(struct sk_buff *skb)
2084{
2085 skb->mac_len = skb->network_header - skb->mac_header;
2086}
2087
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002088static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2089 *skb)
2090{
2091 return skb->head + skb->inner_transport_header;
2092}
2093
Tom Herbert55dc5a92015-12-14 11:19:40 -08002094static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2095{
2096 return skb_inner_transport_header(skb) - skb->data;
2097}
2098
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002099static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2100{
2101 skb->inner_transport_header = skb->data - skb->head;
2102}
2103
2104static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2105 const int offset)
2106{
2107 skb_reset_inner_transport_header(skb);
2108 skb->inner_transport_header += offset;
2109}
2110
2111static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2112{
2113 return skb->head + skb->inner_network_header;
2114}
2115
2116static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2117{
2118 skb->inner_network_header = skb->data - skb->head;
2119}
2120
2121static inline void skb_set_inner_network_header(struct sk_buff *skb,
2122 const int offset)
2123{
2124 skb_reset_inner_network_header(skb);
2125 skb->inner_network_header += offset;
2126}
2127
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002128static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2129{
2130 return skb->head + skb->inner_mac_header;
2131}
2132
2133static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2134{
2135 skb->inner_mac_header = skb->data - skb->head;
2136}
2137
2138static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2139 const int offset)
2140{
2141 skb_reset_inner_mac_header(skb);
2142 skb->inner_mac_header += offset;
2143}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002144static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2145{
Cong Wang35d04612013-05-29 15:16:05 +08002146 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002147}
2148
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002149static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2150{
2151 return skb->head + skb->transport_header;
2152}
2153
2154static inline void skb_reset_transport_header(struct sk_buff *skb)
2155{
2156 skb->transport_header = skb->data - skb->head;
2157}
2158
2159static inline void skb_set_transport_header(struct sk_buff *skb,
2160 const int offset)
2161{
2162 skb_reset_transport_header(skb);
2163 skb->transport_header += offset;
2164}
2165
2166static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2167{
2168 return skb->head + skb->network_header;
2169}
2170
2171static inline void skb_reset_network_header(struct sk_buff *skb)
2172{
2173 skb->network_header = skb->data - skb->head;
2174}
2175
2176static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2177{
2178 skb_reset_network_header(skb);
2179 skb->network_header += offset;
2180}
2181
2182static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2183{
2184 return skb->head + skb->mac_header;
2185}
2186
2187static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2188{
Cong Wang35d04612013-05-29 15:16:05 +08002189 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002190}
2191
2192static inline void skb_reset_mac_header(struct sk_buff *skb)
2193{
2194 skb->mac_header = skb->data - skb->head;
2195}
2196
2197static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2198{
2199 skb_reset_mac_header(skb);
2200 skb->mac_header += offset;
2201}
2202
Timo Teräs0e3da5b2013-12-16 11:02:09 +02002203static inline void skb_pop_mac_header(struct sk_buff *skb)
2204{
2205 skb->mac_header = skb->network_header;
2206}
2207
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002208static inline void skb_probe_transport_header(struct sk_buff *skb,
2209 const int offset_hint)
2210{
2211 struct flow_keys keys;
2212
2213 if (skb_transport_header_was_set(skb))
2214 return;
Tom Herbertcd79a232015-09-01 09:24:27 -07002215 else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
Tom Herbert42aecaa2015-06-04 09:16:39 -07002216 skb_set_transport_header(skb, keys.control.thoff);
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002217 else
2218 skb_set_transport_header(skb, offset_hint);
2219}
2220
Eric Dumazet03606892012-02-23 10:55:02 +00002221static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2222{
2223 if (skb_mac_header_was_set(skb)) {
2224 const unsigned char *old_mac = skb_mac_header(skb);
2225
2226 skb_set_mac_header(skb, -skb->mac_len);
2227 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2228 }
2229}
2230
Michał Mirosław04fb4512010-12-14 15:24:08 +00002231static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2232{
2233 return skb->csum_start - skb_headroom(skb);
2234}
2235
Alexander Duyck08b64fc2016-02-05 15:27:49 -08002236static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2237{
2238 return skb->head + skb->csum_start;
2239}
2240
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002241static inline int skb_transport_offset(const struct sk_buff *skb)
2242{
2243 return skb_transport_header(skb) - skb->data;
2244}
2245
2246static inline u32 skb_network_header_len(const struct sk_buff *skb)
2247{
2248 return skb->transport_header - skb->network_header;
2249}
2250
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002251static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2252{
2253 return skb->inner_transport_header - skb->inner_network_header;
2254}
2255
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002256static inline int skb_network_offset(const struct sk_buff *skb)
2257{
2258 return skb_network_header(skb) - skb->data;
2259}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03002260
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002261static inline int skb_inner_network_offset(const struct sk_buff *skb)
2262{
2263 return skb_inner_network_header(skb) - skb->data;
2264}
2265
Changli Gaof9599ce2010-08-04 04:43:44 +00002266static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2267{
2268 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2269}
2270
Linus Torvalds1da177e2005-04-16 15:20:36 -07002271/*
2272 * CPUs often take a performance hit when accessing unaligned memory
2273 * locations. The actual performance hit varies, it can be small if the
2274 * hardware handles it or large if we have to take an exception and fix it
2275 * in software.
2276 *
2277 * Since an ethernet header is 14 bytes network drivers often end up with
2278 * the IP header at an unaligned offset. The IP header can be aligned by
2279 * shifting the start of the packet by 2 bytes. Drivers should do this
2280 * with:
2281 *
Tobias Klauser8660c122009-07-13 22:48:16 +00002282 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002283 *
2284 * The downside to this alignment of the IP header is that the DMA is now
2285 * unaligned. On some architectures the cost of an unaligned DMA is high
2286 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00002287 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002288 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2289 * to be overridden.
2290 */
2291#ifndef NET_IP_ALIGN
2292#define NET_IP_ALIGN 2
2293#endif
2294
Anton Blanchard025be812006-03-31 02:27:06 -08002295/*
2296 * The networking layer reserves some headroom in skb data (via
2297 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2298 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08002299 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08002300 *
2301 * Unfortunately this headroom changes the DMA alignment of the resulting
2302 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2303 * on some architectures. An architecture can override this value,
2304 * perhaps setting it to a cacheline in size (since that will maintain
2305 * cacheline alignment of the DMA). It must be a power of 2.
2306 *
David S. Millerd6301d32009-02-08 19:24:13 -08002307 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08002308 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07002309 *
2310 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2311 * to reduce average number of cache lines per packet.
2312 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07002313 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08002314 */
2315#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07002316#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08002317#endif
2318
Joe Perches7965bd42013-09-26 14:48:15 -07002319int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002320
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002321static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002322{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00002323 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07002324 WARN_ON(1);
2325 return;
2326 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002327 skb->len = len;
2328 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002329}
2330
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002331static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2332{
2333 __skb_set_length(skb, len);
2334}
2335
Joe Perches7965bd42013-09-26 14:48:15 -07002336void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002337
2338static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2339{
Herbert Xu3cc0e872006-06-09 16:13:38 -07002340 if (skb->data_len)
2341 return ___pskb_trim(skb, len);
2342 __skb_trim(skb, len);
2343 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002344}
2345
2346static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2347{
2348 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2349}
2350
2351/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07002352 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2353 * @skb: buffer to alter
2354 * @len: new length
2355 *
2356 * This is identical to pskb_trim except that the caller knows that
2357 * the skb is not cloned so we should never get an error due to out-
2358 * of-memory.
2359 */
2360static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2361{
2362 int err = pskb_trim(skb, len);
2363 BUG_ON(err);
2364}
2365
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002366static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2367{
2368 unsigned int diff = len - skb->len;
2369
2370 if (skb_tailroom(skb) < diff) {
2371 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2372 GFP_ATOMIC);
2373 if (ret)
2374 return ret;
2375 }
2376 __skb_set_length(skb, len);
2377 return 0;
2378}
2379
Herbert Xue9fa4f72006-08-13 20:12:58 -07002380/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002381 * skb_orphan - orphan a buffer
2382 * @skb: buffer to orphan
2383 *
2384 * If a buffer currently has an owner then we call the owner's
2385 * destructor function and make the @skb unowned. The buffer continues
2386 * to exist but is no longer charged to its former owner.
2387 */
2388static inline void skb_orphan(struct sk_buff *skb)
2389{
Eric Dumazetc34a7612013-07-30 16:11:15 -07002390 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002391 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002392 skb->destructor = NULL;
2393 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07002394 } else {
2395 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002396 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002397}
2398
2399/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002400 * skb_orphan_frags - orphan the frags contained in a buffer
2401 * @skb: buffer to orphan frags from
2402 * @gfp_mask: allocation mask for replacement pages
2403 *
2404 * For each frag in the SKB which needs a destructor (i.e. has an
2405 * owner) create a copy of that frag and release the original
2406 * page by calling the destructor.
2407 */
2408static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2409{
2410 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2411 return 0;
2412 return skb_copy_ubufs(skb, gfp_mask);
2413}
2414
2415/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002416 * __skb_queue_purge - empty a list
2417 * @list: list to empty
2418 *
2419 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2420 * the list and one reference dropped. This function does not take the
2421 * list lock and the caller must hold the relevant locks to use it.
2422 */
Joe Perches7965bd42013-09-26 14:48:15 -07002423void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002424static inline void __skb_queue_purge(struct sk_buff_head *list)
2425{
2426 struct sk_buff *skb;
2427 while ((skb = __skb_dequeue(list)) != NULL)
2428 kfree_skb(skb);
2429}
2430
Yaogong Wang9f5afea2016-09-07 14:49:28 -07002431void skb_rbtree_purge(struct rb_root *root);
2432
Joe Perches7965bd42013-09-26 14:48:15 -07002433void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002434
Joe Perches7965bd42013-09-26 14:48:15 -07002435struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2436 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07002437
2438/**
2439 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2440 * @dev: network device to receive on
2441 * @length: length to allocate
2442 *
2443 * Allocate a new &sk_buff and assign it a usage count of one. The
2444 * buffer has unspecified headroom built in. Users should allocate
2445 * the headroom they think they need without accounting for the
2446 * built in space. The built in space is used for optimisations.
2447 *
2448 * %NULL is returned if there is no free memory. Although this function
2449 * allocates memory it can be called from an interrupt.
2450 */
2451static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002452 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002453{
2454 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2455}
2456
Eric Dumazet6f532612012-05-18 05:12:12 +00002457/* legacy helper around __netdev_alloc_skb() */
2458static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2459 gfp_t gfp_mask)
2460{
2461 return __netdev_alloc_skb(NULL, length, gfp_mask);
2462}
2463
2464/* legacy helper around netdev_alloc_skb() */
2465static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2466{
2467 return netdev_alloc_skb(NULL, length);
2468}
2469
2470
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002471static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2472 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002473{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002474 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002475
2476 if (NET_IP_ALIGN && skb)
2477 skb_reserve(skb, NET_IP_ALIGN);
2478 return skb;
2479}
2480
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002481static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2482 unsigned int length)
2483{
2484 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2485}
2486
Alexander Duyck181edb22015-05-06 21:12:03 -07002487static inline void skb_free_frag(void *addr)
2488{
Alexander Duyck8c2dd3e2017-01-10 16:58:06 -08002489 page_frag_free(addr);
Alexander Duyck181edb22015-05-06 21:12:03 -07002490}
2491
Alexander Duyckffde7322014-12-09 19:40:42 -08002492void *napi_alloc_frag(unsigned int fragsz);
Alexander Duyckfd11a832014-12-09 19:40:49 -08002493struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2494 unsigned int length, gfp_t gfp_mask);
2495static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2496 unsigned int length)
2497{
2498 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2499}
Jesper Dangaard Brouer795bb1c2016-02-08 13:14:59 +01002500void napi_consume_skb(struct sk_buff *skb, int budget);
2501
2502void __kfree_skb_flush(void);
Jesper Dangaard Brouer15fad712016-02-08 13:15:04 +01002503void __kfree_skb_defer(struct sk_buff *skb);
Alexander Duyckffde7322014-12-09 19:40:42 -08002504
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002505/**
Alexander Duyck71dfda52014-11-11 09:26:34 -08002506 * __dev_alloc_pages - allocate page for network Rx
2507 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2508 * @order: size of the allocation
2509 *
2510 * Allocate a new page.
2511 *
2512 * %NULL is returned if there is no free memory.
2513*/
2514static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2515 unsigned int order)
2516{
2517 /* This piece of code contains several assumptions.
2518 * 1. This is for device Rx, therefor a cold page is preferred.
2519 * 2. The expectation is the user wants a compound page.
2520 * 3. If requesting a order 0 page it will not be compound
2521 * due to the check to see if order has a value in prep_new_page
2522 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2523 * code in gfp_to_alloc_flags that should be enforcing this.
2524 */
2525 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2526
2527 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2528}
2529
2530static inline struct page *dev_alloc_pages(unsigned int order)
2531{
Neil Horman95829b32016-05-19 11:30:54 -04002532 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002533}
2534
2535/**
2536 * __dev_alloc_page - allocate a page for network Rx
2537 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2538 *
2539 * Allocate a new page.
2540 *
2541 * %NULL is returned if there is no free memory.
2542 */
2543static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2544{
2545 return __dev_alloc_pages(gfp_mask, 0);
2546}
2547
2548static inline struct page *dev_alloc_page(void)
2549{
Neil Horman95829b32016-05-19 11:30:54 -04002550 return dev_alloc_pages(0);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002551}
2552
2553/**
Mel Gorman06140022012-07-31 16:44:24 -07002554 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2555 * @page: The page that was allocated from skb_alloc_page
2556 * @skb: The skb that may need pfmemalloc set
2557 */
2558static inline void skb_propagate_pfmemalloc(struct page *page,
2559 struct sk_buff *skb)
2560{
Michal Hocko2f064f32015-08-21 14:11:51 -07002561 if (page_is_pfmemalloc(page))
Mel Gorman06140022012-07-31 16:44:24 -07002562 skb->pfmemalloc = true;
2563}
2564
Eric Dumazet564824b2010-10-11 19:05:25 +00002565/**
Masanari Iidae2278672014-02-18 22:54:36 +09002566 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002567 * @frag: the paged fragment
2568 *
2569 * Returns the &struct page associated with @frag.
2570 */
2571static inline struct page *skb_frag_page(const skb_frag_t *frag)
2572{
Ian Campbella8605c62011-10-19 23:01:49 +00002573 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00002574}
2575
2576/**
2577 * __skb_frag_ref - take an addition reference on a paged fragment.
2578 * @frag: the paged fragment
2579 *
2580 * Takes an additional reference on the paged fragment @frag.
2581 */
2582static inline void __skb_frag_ref(skb_frag_t *frag)
2583{
2584 get_page(skb_frag_page(frag));
2585}
2586
2587/**
2588 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2589 * @skb: the buffer
2590 * @f: the fragment offset.
2591 *
2592 * Takes an additional reference on the @f'th paged fragment of @skb.
2593 */
2594static inline void skb_frag_ref(struct sk_buff *skb, int f)
2595{
2596 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2597}
2598
2599/**
2600 * __skb_frag_unref - release a reference on a paged fragment.
2601 * @frag: the paged fragment
2602 *
2603 * Releases a reference on the paged fragment @frag.
2604 */
2605static inline void __skb_frag_unref(skb_frag_t *frag)
2606{
2607 put_page(skb_frag_page(frag));
2608}
2609
2610/**
2611 * skb_frag_unref - release a reference on a paged fragment of an skb.
2612 * @skb: the buffer
2613 * @f: the fragment offset
2614 *
2615 * Releases a reference on the @f'th paged fragment of @skb.
2616 */
2617static inline void skb_frag_unref(struct sk_buff *skb, int f)
2618{
2619 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2620}
2621
2622/**
2623 * skb_frag_address - gets the address of the data contained in a paged fragment
2624 * @frag: the paged fragment buffer
2625 *
2626 * Returns the address of the data within @frag. The page must already
2627 * be mapped.
2628 */
2629static inline void *skb_frag_address(const skb_frag_t *frag)
2630{
2631 return page_address(skb_frag_page(frag)) + frag->page_offset;
2632}
2633
2634/**
2635 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2636 * @frag: the paged fragment buffer
2637 *
2638 * Returns the address of the data within @frag. Checks that the page
2639 * is mapped and returns %NULL otherwise.
2640 */
2641static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2642{
2643 void *ptr = page_address(skb_frag_page(frag));
2644 if (unlikely(!ptr))
2645 return NULL;
2646
2647 return ptr + frag->page_offset;
2648}
2649
2650/**
2651 * __skb_frag_set_page - sets the page contained in a paged fragment
2652 * @frag: the paged fragment
2653 * @page: the page to set
2654 *
2655 * Sets the fragment @frag to contain @page.
2656 */
2657static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2658{
Ian Campbella8605c62011-10-19 23:01:49 +00002659 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002660}
2661
2662/**
2663 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2664 * @skb: the buffer
2665 * @f: the fragment offset
2666 * @page: the page to set
2667 *
2668 * Sets the @f'th fragment of @skb to contain @page.
2669 */
2670static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2671 struct page *page)
2672{
2673 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2674}
2675
Eric Dumazet400dfd32013-10-17 16:27:07 -07002676bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2677
Ian Campbell131ea662011-08-19 06:25:00 +00002678/**
2679 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002680 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002681 * @frag: the paged fragment to map
2682 * @offset: the offset within the fragment (starting at the
2683 * fragment's own offset)
2684 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002685 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00002686 *
2687 * Maps the page associated with @frag to @device.
2688 */
2689static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2690 const skb_frag_t *frag,
2691 size_t offset, size_t size,
2692 enum dma_data_direction dir)
2693{
2694 return dma_map_page(dev, skb_frag_page(frag),
2695 frag->page_offset + offset, size, dir);
2696}
2697
Eric Dumazet117632e2011-12-03 21:39:53 +00002698static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2699 gfp_t gfp_mask)
2700{
2701 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2702}
2703
Octavian Purdilabad93e92014-06-12 01:36:26 +03002704
2705static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2706 gfp_t gfp_mask)
2707{
2708 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2709}
2710
2711
Ian Campbell131ea662011-08-19 06:25:00 +00002712/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002713 * skb_clone_writable - is the header of a clone writable
2714 * @skb: buffer to check
2715 * @len: length up to which to write
2716 *
2717 * Returns true if modifying the header part of the cloned buffer
2718 * does not requires the data to be copied.
2719 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002720static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002721{
2722 return !skb_header_cloned(skb) &&
2723 skb_headroom(skb) + len <= skb->hdr_len;
2724}
2725
Daniel Borkmann36976492016-02-19 23:05:25 +01002726static inline int skb_try_make_writable(struct sk_buff *skb,
2727 unsigned int write_len)
2728{
2729 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
2730 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
2731}
2732
Herbert Xud9cc2042007-09-16 16:21:16 -07002733static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2734 int cloned)
2735{
2736 int delta = 0;
2737
Herbert Xud9cc2042007-09-16 16:21:16 -07002738 if (headroom > skb_headroom(skb))
2739 delta = headroom - skb_headroom(skb);
2740
2741 if (delta || cloned)
2742 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2743 GFP_ATOMIC);
2744 return 0;
2745}
2746
Patrick McHardy334a8132007-06-25 04:35:20 -07002747/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002748 * skb_cow - copy header of skb when it is required
2749 * @skb: buffer to cow
2750 * @headroom: needed headroom
2751 *
2752 * If the skb passed lacks sufficient headroom or its data part
2753 * is shared, data is reallocated. If reallocation fails, an error
2754 * is returned and original skb is not changed.
2755 *
2756 * The result is skb with writable area skb->head...skb->tail
2757 * and at least @headroom of space at head.
2758 */
2759static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2760{
Herbert Xud9cc2042007-09-16 16:21:16 -07002761 return __skb_cow(skb, headroom, skb_cloned(skb));
2762}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002763
Herbert Xud9cc2042007-09-16 16:21:16 -07002764/**
2765 * skb_cow_head - skb_cow but only making the head writable
2766 * @skb: buffer to cow
2767 * @headroom: needed headroom
2768 *
2769 * This function is identical to skb_cow except that we replace the
2770 * skb_cloned check by skb_header_cloned. It should be used when
2771 * you only need to push on some header and do not need to modify
2772 * the data.
2773 */
2774static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2775{
2776 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002777}
2778
2779/**
2780 * skb_padto - pad an skbuff up to a minimal size
2781 * @skb: buffer to pad
2782 * @len: minimal length
2783 *
2784 * Pads up a buffer to ensure the trailing bytes exist and are
2785 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002786 * is untouched. Otherwise it is extended. Returns zero on
2787 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002788 */
Herbert Xu5b057c62006-06-23 02:06:41 -07002789static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002790{
2791 unsigned int size = skb->len;
2792 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002793 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002794 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002795}
2796
Alexander Duyck9c0c1122014-12-03 08:17:33 -08002797/**
2798 * skb_put_padto - increase size and 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
2804 * is untouched. Otherwise it is extended. Returns zero on
2805 * success. The skb is freed on error.
2806 */
2807static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2808{
2809 unsigned int size = skb->len;
2810
2811 if (unlikely(size < len)) {
2812 len -= size;
2813 if (skb_pad(skb, len))
2814 return -ENOMEM;
2815 __skb_put(skb, len);
2816 }
2817 return 0;
2818}
2819
Linus Torvalds1da177e2005-04-16 15:20:36 -07002820static inline int skb_add_data(struct sk_buff *skb,
Al Viroaf2b0402014-11-27 21:44:24 -05002821 struct iov_iter *from, int copy)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002822{
2823 const int off = skb->len;
2824
2825 if (skb->ip_summed == CHECKSUM_NONE) {
Al Viroaf2b0402014-11-27 21:44:24 -05002826 __wsum csum = 0;
Al Viro15e6cb42016-11-01 22:42:45 -04002827 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
2828 &csum, from)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002829 skb->csum = csum_block_add(skb->csum, csum, off);
2830 return 0;
2831 }
Al Viro15e6cb42016-11-01 22:42:45 -04002832 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002833 return 0;
2834
2835 __skb_trim(skb, off);
2836 return -EFAULT;
2837}
2838
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002839static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2840 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002841{
2842 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002843 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002844
Ian Campbellea2ab692011-08-22 23:44:58 +00002845 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002846 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002847 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002848 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002849}
2850
Herbert Xu364c6ba2006-06-09 16:10:40 -07002851static inline int __skb_linearize(struct sk_buff *skb)
2852{
2853 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2854}
2855
Linus Torvalds1da177e2005-04-16 15:20:36 -07002856/**
2857 * skb_linearize - convert paged skb to linear one
2858 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002859 *
2860 * If there is no free memory -ENOMEM is returned, otherwise zero
2861 * is returned and the old skb data released.
2862 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002863static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002864{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002865 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2866}
2867
2868/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00002869 * skb_has_shared_frag - can any frag be overwritten
2870 * @skb: buffer to test
2871 *
2872 * Return true if the skb has at least one frag that might be modified
2873 * by an external entity (as in vmsplice()/sendfile())
2874 */
2875static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2876{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00002877 return skb_is_nonlinear(skb) &&
2878 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00002879}
2880
2881/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07002882 * skb_linearize_cow - make sure skb is linear and writable
2883 * @skb: buffer to process
2884 *
2885 * If there is no free memory -ENOMEM is returned, otherwise zero
2886 * is returned and the old skb data released.
2887 */
2888static inline int skb_linearize_cow(struct sk_buff *skb)
2889{
2890 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2891 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002892}
2893
Daniel Borkmann479ffcc2016-08-05 00:11:12 +02002894static __always_inline void
2895__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
2896 unsigned int off)
2897{
2898 if (skb->ip_summed == CHECKSUM_COMPLETE)
2899 skb->csum = csum_block_sub(skb->csum,
2900 csum_partial(start, len, 0), off);
2901 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
2902 skb_checksum_start_offset(skb) < 0)
2903 skb->ip_summed = CHECKSUM_NONE;
2904}
2905
Linus Torvalds1da177e2005-04-16 15:20:36 -07002906/**
2907 * skb_postpull_rcsum - update checksum for received skb after pull
2908 * @skb: buffer to update
2909 * @start: start of data before pull
2910 * @len: length of data pulled
2911 *
2912 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002913 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2914 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002915 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002916static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002917 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002918{
Daniel Borkmann479ffcc2016-08-05 00:11:12 +02002919 __skb_postpull_rcsum(skb, start, len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002920}
2921
Daniel Borkmann479ffcc2016-08-05 00:11:12 +02002922static __always_inline void
2923__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
2924 unsigned int off)
2925{
2926 if (skb->ip_summed == CHECKSUM_COMPLETE)
2927 skb->csum = csum_block_add(skb->csum,
2928 csum_partial(start, len, 0), off);
2929}
Herbert Xucbb042f2006-03-20 22:43:56 -08002930
Daniel Borkmann479ffcc2016-08-05 00:11:12 +02002931/**
2932 * skb_postpush_rcsum - update checksum for received skb after push
2933 * @skb: buffer to update
2934 * @start: start of data after push
2935 * @len: length of data pushed
2936 *
2937 * After doing a push on a received packet, you need to call this to
2938 * update the CHECKSUM_COMPLETE checksum.
2939 */
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01002940static inline void skb_postpush_rcsum(struct sk_buff *skb,
2941 const void *start, unsigned int len)
2942{
Daniel Borkmann479ffcc2016-08-05 00:11:12 +02002943 __skb_postpush_rcsum(skb, start, len, 0);
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01002944}
2945
Daniel Borkmann479ffcc2016-08-05 00:11:12 +02002946unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2947
David S. Miller7ce5a272013-12-02 17:26:05 -05002948/**
WANG Cong82a31b92016-06-30 10:15:22 -07002949 * skb_push_rcsum - push skb and update receive checksum
2950 * @skb: buffer to update
2951 * @len: length of data pulled
2952 *
2953 * This function performs an skb_push on the packet and updates
2954 * the CHECKSUM_COMPLETE checksum. It should be used on
2955 * receive path processing instead of skb_push unless you know
2956 * that the checksum difference is zero (e.g., a valid IP header)
2957 * or you are setting ip_summed to CHECKSUM_NONE.
2958 */
2959static inline unsigned char *skb_push_rcsum(struct sk_buff *skb,
2960 unsigned int len)
2961{
2962 skb_push(skb, len);
2963 skb_postpush_rcsum(skb, skb->data, len);
2964 return skb->data;
2965}
2966
2967/**
David S. Miller7ce5a272013-12-02 17:26:05 -05002968 * pskb_trim_rcsum - trim received skb and update checksum
2969 * @skb: buffer to trim
2970 * @len: new length
2971 *
2972 * This is exactly the same as pskb_trim except that it ensures the
2973 * checksum of received packets are still valid after the operation.
2974 */
2975
2976static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2977{
2978 if (likely(len >= skb->len))
2979 return 0;
2980 if (skb->ip_summed == CHECKSUM_COMPLETE)
2981 skb->ip_summed = CHECKSUM_NONE;
2982 return __pskb_trim(skb, len);
2983}
2984
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002985static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2986{
2987 if (skb->ip_summed == CHECKSUM_COMPLETE)
2988 skb->ip_summed = CHECKSUM_NONE;
2989 __skb_trim(skb, len);
2990 return 0;
2991}
2992
2993static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
2994{
2995 if (skb->ip_summed == CHECKSUM_COMPLETE)
2996 skb->ip_summed = CHECKSUM_NONE;
2997 return __skb_grow(skb, len);
2998}
2999
Linus Torvalds1da177e2005-04-16 15:20:36 -07003000#define skb_queue_walk(queue, skb) \
3001 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003002 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07003003 skb = skb->next)
3004
James Chapman46f89142007-04-30 00:07:31 -07003005#define skb_queue_walk_safe(queue, skb, tmp) \
3006 for (skb = (queue)->next, tmp = skb->next; \
3007 skb != (struct sk_buff *)(queue); \
3008 skb = tmp, tmp = skb->next)
3009
David S. Miller1164f522008-09-23 00:49:44 -07003010#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003011 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07003012 skb = skb->next)
3013
3014#define skb_queue_walk_from_safe(queue, skb, tmp) \
3015 for (tmp = skb->next; \
3016 skb != (struct sk_buff *)(queue); \
3017 skb = tmp, tmp = skb->next)
3018
Stephen Hemminger300ce172005-10-30 13:47:34 -08003019#define skb_queue_reverse_walk(queue, skb) \
3020 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003021 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08003022 skb = skb->prev)
3023
David S. Miller686a2952011-01-20 22:47:32 -08003024#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3025 for (skb = (queue)->prev, tmp = skb->prev; \
3026 skb != (struct sk_buff *)(queue); \
3027 skb = tmp, tmp = skb->prev)
3028
3029#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3030 for (tmp = skb->prev; \
3031 skb != (struct sk_buff *)(queue); \
3032 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003033
David S. Miller21dc3302010-08-23 00:13:46 -07003034static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07003035{
3036 return skb_shinfo(skb)->frag_list != NULL;
3037}
3038
3039static inline void skb_frag_list_init(struct sk_buff *skb)
3040{
3041 skb_shinfo(skb)->frag_list = NULL;
3042}
3043
David S. Milleree039872009-06-09 00:17:13 -07003044#define skb_walk_frags(skb, iter) \
3045 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3046
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003047
3048int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
3049 const struct sk_buff *skb);
3050struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
Paolo Abeni7c13f972016-11-04 11:28:59 +01003051 void (*destructor)(struct sock *sk,
3052 struct sk_buff *skb),
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003053 int *peeked, int *off, int *err,
3054 struct sk_buff **last);
Joe Perches7965bd42013-09-26 14:48:15 -07003055struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
Paolo Abeni7c13f972016-11-04 11:28:59 +01003056 void (*destructor)(struct sock *sk,
3057 struct sk_buff *skb),
Joe Perches7965bd42013-09-26 14:48:15 -07003058 int *peeked, int *off, int *err);
3059struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3060 int *err);
3061unsigned int datagram_poll(struct file *file, struct socket *sock,
3062 struct poll_table_struct *wait);
Al Viroc0371da2014-11-24 10:42:55 -05003063int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3064 struct iov_iter *to, int size);
David S. Miller51f3d022014-11-05 16:46:40 -05003065static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3066 struct msghdr *msg, int size)
3067{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003068 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
David S. Miller51f3d022014-11-05 16:46:40 -05003069}
Al Viroe5a4b0b2014-11-24 18:17:55 -05003070int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3071 struct msghdr *msg);
Al Viro3a654f92014-06-19 14:15:22 -04003072int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3073 struct iov_iter *from, int len);
Al Viro3a654f92014-06-19 14:15:22 -04003074int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
Joe Perches7965bd42013-09-26 14:48:15 -07003075void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
samanthakumar627d2d62016-04-05 12:41:16 -04003076void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3077static inline void skb_free_datagram_locked(struct sock *sk,
3078 struct sk_buff *skb)
3079{
3080 __skb_free_datagram_locked(sk, skb, 0);
3081}
Joe Perches7965bd42013-09-26 14:48:15 -07003082int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07003083int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3084int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3085__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3086 int len, __wsum csum);
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +02003087int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
Joe Perches7965bd42013-09-26 14:48:15 -07003088 struct pipe_inode_info *pipe, unsigned int len,
Al Viro25869262016-09-17 21:02:10 -04003089 unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07003090void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01003091unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00003092int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3093 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07003094void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3095int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3096void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Florian Westphalde960aa2014-01-26 10:58:16 +01003097unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
Marcelo Ricardo Leitnerae7ef812016-06-02 15:05:41 -03003098bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu);
Joe Perches7965bd42013-09-26 14:48:15 -07003099struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Vlad Yasevich0d5501c2014-08-08 14:42:13 -04003100struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
Jiri Pirkoe2195122014-11-19 14:05:01 +01003101int skb_ensure_writable(struct sk_buff *skb, int write_len);
Shmulik Ladkanibfca4c52016-09-19 19:11:09 +03003102int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
Jiri Pirko93515d52014-11-19 14:05:02 +01003103int skb_vlan_pop(struct sk_buff *skb);
3104int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
Sowmini Varadhan6fa01cc2016-04-22 18:36:35 -07003105struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3106 gfp_t gfp);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003107
Al Viro6ce8e9c2014-04-06 21:25:44 -04003108static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3109{
Al Viro21226ab2014-11-28 15:48:29 -05003110 return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro6ce8e9c2014-04-06 21:25:44 -04003111}
3112
Al Viro7eab8d92014-04-06 21:51:23 -04003113static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3114{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003115 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro7eab8d92014-04-06 21:51:23 -04003116}
3117
Daniel Borkmann2817a332013-10-30 11:50:51 +01003118struct skb_checksum_ops {
3119 __wsum (*update)(const void *mem, int len, __wsum wsum);
3120 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3121};
3122
3123__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3124 __wsum csum, const struct skb_checksum_ops *ops);
3125__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3126 __wsum csum);
3127
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003128static inline void * __must_check
3129__skb_header_pointer(const struct sk_buff *skb, int offset,
3130 int len, void *data, int hlen, void *buffer)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003131{
Patrick McHardy55820ee2005-07-05 14:08:10 -07003132 if (hlen - offset >= len)
David S. Miller690e36e2014-08-23 12:13:41 -07003133 return data + offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003134
David S. Miller690e36e2014-08-23 12:13:41 -07003135 if (!skb ||
3136 skb_copy_bits(skb, offset, buffer, len) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003137 return NULL;
3138
3139 return buffer;
3140}
3141
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003142static inline void * __must_check
3143skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
David S. Miller690e36e2014-08-23 12:13:41 -07003144{
3145 return __skb_header_pointer(skb, offset, len, skb->data,
3146 skb_headlen(skb), buffer);
3147}
3148
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01003149/**
3150 * skb_needs_linearize - check if we need to linearize a given skb
3151 * depending on the given device features.
3152 * @skb: socket buffer to check
3153 * @features: net device features
3154 *
3155 * Returns true if either:
3156 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3157 * 2. skb is fragmented and the device does not support SG.
3158 */
3159static inline bool skb_needs_linearize(struct sk_buff *skb,
3160 netdev_features_t features)
3161{
3162 return skb_is_nonlinear(skb) &&
3163 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3164 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3165}
3166
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03003167static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3168 void *to,
3169 const unsigned int len)
3170{
3171 memcpy(to, skb->data, len);
3172}
3173
3174static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3175 const int offset, void *to,
3176 const unsigned int len)
3177{
3178 memcpy(to, skb->data + offset, len);
3179}
3180
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03003181static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3182 const void *from,
3183 const unsigned int len)
3184{
3185 memcpy(skb->data, from, len);
3186}
3187
3188static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3189 const int offset,
3190 const void *from,
3191 const unsigned int len)
3192{
3193 memcpy(skb->data + offset, from, len);
3194}
3195
Joe Perches7965bd42013-09-26 14:48:15 -07003196void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003197
Patrick Ohlyac45f602009-02-12 05:03:37 +00003198static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3199{
3200 return skb->tstamp;
3201}
3202
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003203/**
3204 * skb_get_timestamp - get timestamp from a skb
3205 * @skb: skb to get stamp from
3206 * @stamp: pointer to struct timeval to store stamp in
3207 *
3208 * Timestamps are stored in the skb as offsets to a base timestamp.
3209 * This function converts the offset back to a struct timeval and stores
3210 * it in stamp.
3211 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00003212static inline void skb_get_timestamp(const struct sk_buff *skb,
3213 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003214{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003215 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003216}
3217
Patrick Ohlyac45f602009-02-12 05:03:37 +00003218static inline void skb_get_timestampns(const struct sk_buff *skb,
3219 struct timespec *stamp)
3220{
3221 *stamp = ktime_to_timespec(skb->tstamp);
3222}
3223
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003224static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003225{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003226 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003227}
3228
Stephen Hemminger164891a2007-04-23 22:26:16 -07003229static inline ktime_t net_timedelta(ktime_t t)
3230{
3231 return ktime_sub(ktime_get_real(), t);
3232}
3233
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003234static inline ktime_t net_invalid_timestamp(void)
3235{
Thomas Gleixner8b0e1952016-12-25 12:30:41 +01003236 return 0;
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003237}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003238
Alexander Duyck62bccb82014-09-04 13:31:35 -04003239struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3240
Richard Cochranc1f19b52010-07-17 08:49:36 +00003241#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3242
Joe Perches7965bd42013-09-26 14:48:15 -07003243void skb_clone_tx_timestamp(struct sk_buff *skb);
3244bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00003245
3246#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3247
3248static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3249{
3250}
3251
3252static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3253{
3254 return false;
3255}
3256
3257#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3258
3259/**
3260 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3261 *
Richard Cochranda92b192011-10-21 00:49:15 +00003262 * PHY drivers may accept clones of transmitted packets for
3263 * timestamping via their phy_driver.txtstamp method. These drivers
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003264 * must call this function to return the skb back to the stack with a
3265 * timestamp.
Richard Cochranda92b192011-10-21 00:49:15 +00003266 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00003267 * @skb: clone of the the original outgoing packet
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003268 * @hwtstamps: hardware time stamps
Richard Cochranc1f19b52010-07-17 08:49:36 +00003269 *
3270 */
3271void skb_complete_tx_timestamp(struct sk_buff *skb,
3272 struct skb_shared_hwtstamps *hwtstamps);
3273
Willem de Bruijne7fd2882014-08-04 22:11:48 -04003274void __skb_tstamp_tx(struct sk_buff *orig_skb,
3275 struct skb_shared_hwtstamps *hwtstamps,
3276 struct sock *sk, int tstype);
3277
Patrick Ohlyac45f602009-02-12 05:03:37 +00003278/**
3279 * skb_tstamp_tx - queue clone of skb with send time stamps
3280 * @orig_skb: the original outgoing packet
3281 * @hwtstamps: hardware time stamps, may be NULL if not available
3282 *
3283 * If the skb has a socket associated, then this function clones the
3284 * skb (thus sharing the actual data and optional structures), stores
3285 * the optional hardware time stamping information (if non NULL) or
3286 * generates a software time stamp (otherwise), then queues the clone
3287 * to the error queue of the socket. Errors are silently ignored.
3288 */
Joe Perches7965bd42013-09-26 14:48:15 -07003289void skb_tstamp_tx(struct sk_buff *orig_skb,
3290 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00003291
Richard Cochran4507a712010-07-17 08:48:28 +00003292static inline void sw_tx_timestamp(struct sk_buff *skb)
3293{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00003294 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
3295 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00003296 skb_tstamp_tx(skb, NULL);
3297}
3298
3299/**
3300 * skb_tx_timestamp() - Driver hook for transmit timestamping
3301 *
3302 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00003303 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00003304 *
David S. Miller73409f32013-12-27 13:04:33 -05003305 * Specifically, one should make absolutely sure that this function is
3306 * called before TX completion of this packet can trigger. Otherwise
3307 * the packet could potentially already be freed.
3308 *
Richard Cochran4507a712010-07-17 08:48:28 +00003309 * @skb: A socket buffer.
3310 */
3311static inline void skb_tx_timestamp(struct sk_buff *skb)
3312{
Richard Cochranc1f19b52010-07-17 08:49:36 +00003313 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00003314 sw_tx_timestamp(skb);
3315}
3316
Johannes Berg6e3e9392011-11-09 10:15:42 +01003317/**
3318 * skb_complete_wifi_ack - deliver skb with wifi status
3319 *
3320 * @skb: the original outgoing packet
3321 * @acked: ack status
3322 *
3323 */
3324void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3325
Joe Perches7965bd42013-09-26 14:48:15 -07003326__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3327__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003328
Herbert Xu60476372007-04-09 11:59:39 -07003329static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3330{
Tom Herbert6edec0e2015-02-10 16:30:28 -08003331 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3332 skb->csum_valid ||
3333 (skb->ip_summed == CHECKSUM_PARTIAL &&
3334 skb_checksum_start_offset(skb) >= 0));
Herbert Xu60476372007-04-09 11:59:39 -07003335}
3336
Herbert Xufb286bb2005-11-10 13:01:24 -08003337/**
3338 * skb_checksum_complete - Calculate checksum of an entire packet
3339 * @skb: packet to process
3340 *
3341 * This function calculates the checksum over the entire packet plus
3342 * the value of skb->csum. The latter can be used to supply the
3343 * checksum of a pseudo header as used by TCP/UDP. It returns the
3344 * checksum.
3345 *
3346 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3347 * this function can be used to verify that checksum on received
3348 * packets. In that case the function should return zero if the
3349 * checksum is correct. In particular, this function will return zero
3350 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3351 * hardware has already verified the correctness of the checksum.
3352 */
Al Viro4381ca32007-07-15 21:00:11 +01003353static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08003354{
Herbert Xu60476372007-04-09 11:59:39 -07003355 return skb_csum_unnecessary(skb) ?
3356 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003357}
3358
Tom Herbert77cffe22014-08-27 21:26:46 -07003359static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3360{
3361 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3362 if (skb->csum_level == 0)
3363 skb->ip_summed = CHECKSUM_NONE;
3364 else
3365 skb->csum_level--;
3366 }
3367}
3368
3369static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3370{
3371 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3372 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3373 skb->csum_level++;
3374 } else if (skb->ip_summed == CHECKSUM_NONE) {
3375 skb->ip_summed = CHECKSUM_UNNECESSARY;
3376 skb->csum_level = 0;
3377 }
3378}
3379
Tom Herbert5a212322014-08-31 15:12:41 -07003380static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
3381{
3382 /* Mark current checksum as bad (typically called from GRO
3383 * path). In the case that ip_summed is CHECKSUM_NONE
3384 * this must be the first checksum encountered in the packet.
3385 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
3386 * checksum after the last one validated. For UDP, a zero
3387 * checksum can not be marked as bad.
3388 */
3389
3390 if (skb->ip_summed == CHECKSUM_NONE ||
3391 skb->ip_summed == CHECKSUM_UNNECESSARY)
3392 skb->csum_bad = 1;
3393}
3394
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003395/* Check if we need to perform checksum complete validation.
3396 *
3397 * Returns true if checksum complete is needed, false otherwise
3398 * (either checksum is unnecessary or zero checksum is allowed).
3399 */
3400static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3401 bool zero_okay,
3402 __sum16 check)
3403{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003404 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3405 skb->csum_valid = 1;
Tom Herbert77cffe22014-08-27 21:26:46 -07003406 __skb_decr_checksum_unnecessary(skb);
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003407 return false;
3408 }
3409
3410 return true;
3411}
3412
3413/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3414 * in checksum_init.
3415 */
3416#define CHECKSUM_BREAK 76
3417
Tom Herbert4e18b9a2015-04-20 14:10:04 -07003418/* Unset checksum-complete
3419 *
3420 * Unset checksum complete can be done when packet is being modified
3421 * (uncompressed for instance) and checksum-complete value is
3422 * invalidated.
3423 */
3424static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3425{
3426 if (skb->ip_summed == CHECKSUM_COMPLETE)
3427 skb->ip_summed = CHECKSUM_NONE;
3428}
3429
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003430/* Validate (init) checksum based on checksum complete.
3431 *
3432 * Return values:
3433 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3434 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3435 * checksum is stored in skb->csum for use in __skb_checksum_complete
3436 * non-zero: value of invalid checksum
3437 *
3438 */
3439static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3440 bool complete,
3441 __wsum psum)
3442{
3443 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3444 if (!csum_fold(csum_add(psum, skb->csum))) {
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003445 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003446 return 0;
3447 }
Tom Herbert5a212322014-08-31 15:12:41 -07003448 } else if (skb->csum_bad) {
3449 /* ip_summed == CHECKSUM_NONE in this case */
Eric Dumazetc91d4602015-05-15 05:48:07 -07003450 return (__force __sum16)1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003451 }
3452
3453 skb->csum = psum;
3454
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003455 if (complete || skb->len <= CHECKSUM_BREAK) {
3456 __sum16 csum;
3457
3458 csum = __skb_checksum_complete(skb);
3459 skb->csum_valid = !csum;
3460 return csum;
3461 }
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003462
3463 return 0;
3464}
3465
3466static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3467{
3468 return 0;
3469}
3470
3471/* Perform checksum validate (init). Note that this is a macro since we only
3472 * want to calculate the pseudo header which is an input function if necessary.
3473 * First we try to validate without any computation (checksum unnecessary) and
3474 * then calculate based on checksum complete calling the function to compute
3475 * pseudo header.
3476 *
3477 * Return values:
3478 * 0: checksum is validated or try to in skb_checksum_complete
3479 * non-zero: value of invalid checksum
3480 */
3481#define __skb_checksum_validate(skb, proto, complete, \
3482 zero_okay, check, compute_pseudo) \
3483({ \
3484 __sum16 __ret = 0; \
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003485 skb->csum_valid = 0; \
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003486 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3487 __ret = __skb_checksum_validate_complete(skb, \
3488 complete, compute_pseudo(skb, proto)); \
3489 __ret; \
3490})
3491
3492#define skb_checksum_init(skb, proto, compute_pseudo) \
3493 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3494
3495#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3496 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3497
3498#define skb_checksum_validate(skb, proto, compute_pseudo) \
3499 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3500
3501#define skb_checksum_validate_zero_check(skb, proto, check, \
3502 compute_pseudo) \
Sabrina Dubroca096a4cf2015-02-06 18:54:19 +01003503 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003504
3505#define skb_checksum_simple_validate(skb) \
3506 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3507
Tom Herbertd96535a2014-08-31 15:12:42 -07003508static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3509{
3510 return (skb->ip_summed == CHECKSUM_NONE &&
3511 skb->csum_valid && !skb->csum_bad);
3512}
3513
3514static inline void __skb_checksum_convert(struct sk_buff *skb,
3515 __sum16 check, __wsum pseudo)
3516{
3517 skb->csum = ~pseudo;
3518 skb->ip_summed = CHECKSUM_COMPLETE;
3519}
3520
3521#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3522do { \
3523 if (__skb_checksum_convert_check(skb)) \
3524 __skb_checksum_convert(skb, check, \
3525 compute_pseudo(skb, proto)); \
3526} while (0)
3527
Tom Herbert15e23962015-02-10 16:30:31 -08003528static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3529 u16 start, u16 offset)
3530{
3531 skb->ip_summed = CHECKSUM_PARTIAL;
3532 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3533 skb->csum_offset = offset - start;
3534}
3535
Tom Herbertdcdc8992015-02-02 16:07:34 -08003536/* Update skbuf and packet to reflect the remote checksum offload operation.
3537 * When called, ptr indicates the starting point for skb->csum when
3538 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3539 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3540 */
3541static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
Tom Herbert15e23962015-02-10 16:30:31 -08003542 int start, int offset, bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08003543{
3544 __wsum delta;
3545
Tom Herbert15e23962015-02-10 16:30:31 -08003546 if (!nopartial) {
3547 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3548 return;
3549 }
3550
Tom Herbertdcdc8992015-02-02 16:07:34 -08003551 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3552 __skb_checksum_complete(skb);
3553 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3554 }
3555
3556 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3557
3558 /* Adjust skb->csum since we changed the packet */
3559 skb->csum = csum_add(skb->csum, delta);
3560}
3561
Florian Westphalcb9c6832017-01-23 18:21:56 +01003562static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
3563{
3564#if IS_ENABLED(CONFIG_NF_CONNTRACK)
Florian Westphala9e419d2017-01-23 18:21:59 +01003565 return (void *)(skb->_nfct & SKB_NFCT_PTRMASK);
Florian Westphalcb9c6832017-01-23 18:21:56 +01003566#else
3567 return NULL;
3568#endif
3569}
3570
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003571#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07003572void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003573static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3574{
3575 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07003576 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003577}
3578static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3579{
3580 if (nfct)
3581 atomic_inc(&nfct->use);
3582}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003583#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003584#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003585static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3586{
3587 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3588 kfree(nf_bridge);
3589}
3590static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3591{
3592 if (nf_bridge)
3593 atomic_inc(&nf_bridge->use);
3594}
3595#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08003596static inline void nf_reset(struct sk_buff *skb)
3597{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003598#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01003599 nf_conntrack_put(skb_nfct(skb));
3600 skb->_nfct = 0;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003601#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003602#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Patrick McHardya193a4a2006-03-20 19:23:05 -08003603 nf_bridge_put(skb->nf_bridge);
3604 skb->nf_bridge = NULL;
3605#endif
3606}
3607
Patrick McHardy124dff02013-04-05 20:42:05 +02003608static inline void nf_reset_trace(struct sk_buff *skb)
3609{
Florian Westphal478b3602014-02-15 23:48:45 +01003610#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00003611 skb->nf_trace = 0;
3612#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003613}
3614
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003615/* Note: This doesn't put any conntrack and bridge info in dst. */
Eric Dumazetb1937222014-09-28 22:18:47 -07003616static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3617 bool copy)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003618{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003619#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01003620 dst->_nfct = src->_nfct;
3621 nf_conntrack_get(skb_nfct(src));
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003622#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003623#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003624 dst->nf_bridge = src->nf_bridge;
3625 nf_bridge_get(src->nf_bridge);
3626#endif
Florian Westphal478b3602014-02-15 23:48:45 +01003627#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Eric Dumazetb1937222014-09-28 22:18:47 -07003628 if (copy)
3629 dst->nf_trace = src->nf_trace;
Florian Westphal478b3602014-02-15 23:48:45 +01003630#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003631}
3632
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003633static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3634{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003635#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01003636 nf_conntrack_put(skb_nfct(dst));
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003637#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003638#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003639 nf_bridge_put(dst->nf_bridge);
3640#endif
Eric Dumazetb1937222014-09-28 22:18:47 -07003641 __nf_copy(dst, src, true);
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003642}
3643
James Morris984bc162006-06-09 00:29:17 -07003644#ifdef CONFIG_NETWORK_SECMARK
3645static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3646{
3647 to->secmark = from->secmark;
3648}
3649
3650static inline void skb_init_secmark(struct sk_buff *skb)
3651{
3652 skb->secmark = 0;
3653}
3654#else
3655static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3656{ }
3657
3658static inline void skb_init_secmark(struct sk_buff *skb)
3659{ }
3660#endif
3661
Eric W. Biederman574f7192014-04-01 12:20:24 -07003662static inline bool skb_irq_freeable(const struct sk_buff *skb)
3663{
3664 return !skb->destructor &&
3665#if IS_ENABLED(CONFIG_XFRM)
3666 !skb->sp &&
3667#endif
Florian Westphalcb9c6832017-01-23 18:21:56 +01003668 !skb_nfct(skb) &&
Eric W. Biederman574f7192014-04-01 12:20:24 -07003669 !skb->_skb_refdst &&
3670 !skb_has_frag_list(skb);
3671}
3672
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003673static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3674{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003675 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003676}
3677
Stephen Hemminger92477442009-03-21 13:39:26 -07003678static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003679{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003680 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003681}
3682
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003683static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3684{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003685 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003686}
3687
David S. Millerd5a9e242009-01-27 16:22:11 -08003688static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3689{
3690 skb->queue_mapping = rx_queue + 1;
3691}
3692
Stephen Hemminger92477442009-03-21 13:39:26 -07003693static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003694{
3695 return skb->queue_mapping - 1;
3696}
3697
Stephen Hemminger92477442009-03-21 13:39:26 -07003698static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003699{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003700 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08003701}
3702
Julian Anastasov4ff06202017-02-06 23:14:12 +02003703static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
3704{
3705 skb->dst_pending_confirm = val;
3706}
3707
3708static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
3709{
3710 return skb->dst_pending_confirm != 0;
3711}
3712
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003713static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3714{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003715#ifdef CONFIG_XFRM
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003716 return skb->sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003717#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003718 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003719#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003720}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003721
Pravin B Shelar68c33162013-02-14 14:02:41 +00003722/* Keeps track of mac header offset relative to skb->head.
3723 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3724 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07003725 * tunnel skb it points to outer mac header.
3726 * Keeps track of level of encapsulation of network headers.
3727 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00003728struct skb_gso_cb {
Alexander Duyck802ab552016-04-10 21:45:03 -04003729 union {
3730 int mac_offset;
3731 int data_offset;
3732 };
Eric Dumazet3347c962013-10-19 11:42:56 -07003733 int encap_level;
Alexander Duyck76443452016-02-05 15:27:37 -08003734 __wsum csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003735 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00003736};
Konstantin Khlebnikov9207f9d2016-01-08 15:21:46 +03003737#define SKB_SGO_CB_OFFSET 32
3738#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
Pravin B Shelar68c33162013-02-14 14:02:41 +00003739
3740static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3741{
3742 return (skb_mac_header(inner_skb) - inner_skb->head) -
3743 SKB_GSO_CB(inner_skb)->mac_offset;
3744}
3745
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00003746static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3747{
3748 int new_headroom, headroom;
3749 int ret;
3750
3751 headroom = skb_headroom(skb);
3752 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3753 if (ret)
3754 return ret;
3755
3756 new_headroom = skb_headroom(skb);
3757 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3758 return 0;
3759}
3760
Alexander Duyck08b64fc2016-02-05 15:27:49 -08003761static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
3762{
3763 /* Do not update partial checksums if remote checksum is enabled. */
3764 if (skb->remcsum_offload)
3765 return;
3766
3767 SKB_GSO_CB(skb)->csum = res;
3768 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
3769}
3770
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003771/* Compute the checksum for a gso segment. First compute the checksum value
3772 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3773 * then add in skb->csum (checksum from csum_start to end of packet).
3774 * skb->csum and csum_start are then updated to reflect the checksum of the
3775 * resultant packet starting from the transport header-- the resultant checksum
3776 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3777 * header.
3778 */
3779static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3780{
Alexander Duyck76443452016-02-05 15:27:37 -08003781 unsigned char *csum_start = skb_transport_header(skb);
3782 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
3783 __wsum partial = SKB_GSO_CB(skb)->csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003784
Alexander Duyck76443452016-02-05 15:27:37 -08003785 SKB_GSO_CB(skb)->csum = res;
3786 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003787
Alexander Duyck76443452016-02-05 15:27:37 -08003788 return csum_fold(csum_partial(csum_start, plen, partial));
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003789}
3790
David S. Millerbdcc0922012-03-07 20:53:36 -05003791static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07003792{
3793 return skb_shinfo(skb)->gso_size;
3794}
3795
Eric Dumazet36a8f392013-09-29 01:21:32 -07003796/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05003797static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02003798{
3799 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3800}
3801
Daniel Borkmann5293efe2016-08-18 01:00:39 +02003802static inline void skb_gso_reset(struct sk_buff *skb)
3803{
3804 skb_shinfo(skb)->gso_size = 0;
3805 skb_shinfo(skb)->gso_segs = 0;
3806 skb_shinfo(skb)->gso_type = 0;
3807}
3808
Joe Perches7965bd42013-09-26 14:48:15 -07003809void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07003810
3811static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3812{
3813 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3814 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003815 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3816
Alexander Duyckb78462e2010-06-02 12:24:37 +00003817 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3818 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07003819 __skb_warn_lro_forwarding(skb);
3820 return true;
3821 }
3822 return false;
3823}
3824
Herbert Xu35fc92a2007-03-26 23:22:20 -07003825static inline void skb_forward_csum(struct sk_buff *skb)
3826{
3827 /* Unfortunately we don't support this one. Any brave souls? */
3828 if (skb->ip_summed == CHECKSUM_COMPLETE)
3829 skb->ip_summed = CHECKSUM_NONE;
3830}
3831
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003832/**
3833 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3834 * @skb: skb to check
3835 *
3836 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3837 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3838 * use this helper, to document places where we make this assertion.
3839 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003840static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003841{
3842#ifdef DEBUG
3843 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3844#endif
3845}
3846
Rusty Russellf35d9d82008-02-04 23:49:54 -05003847bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00003848
Paul Durranted1f50c2014-01-09 10:02:46 +00003849int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
Linus Lüssing9afd85c2015-05-02 14:01:07 +02003850struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
3851 unsigned int transport_len,
3852 __sum16(*skb_chkf)(struct sk_buff *skb));
Paul Durranted1f50c2014-01-09 10:02:46 +00003853
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00003854/**
3855 * skb_head_is_locked - Determine if the skb->head is locked down
3856 * @skb: skb to check
3857 *
3858 * The head on skbs build around a head frag can be removed if they are
3859 * not cloned. This function returns true if the skb head is locked down
3860 * due to either being allocated via kmalloc, or by being a clone with
3861 * multiple references to the head.
3862 */
3863static inline bool skb_head_is_locked(const struct sk_buff *skb)
3864{
3865 return !skb->head_frag || skb_cloned(skb);
3866}
Florian Westphalfe6cc552014-02-13 23:09:12 +01003867
3868/**
3869 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3870 *
3871 * @skb: GSO skb
3872 *
3873 * skb_gso_network_seglen is used to determine the real size of the
3874 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3875 *
3876 * The MAC/L2 header is not accounted for.
3877 */
3878static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3879{
3880 unsigned int hdr_len = skb_transport_header(skb) -
3881 skb_network_header(skb);
3882 return hdr_len + skb_gso_transport_seglen(skb);
3883}
Thomas Grafee122c72015-07-21 10:43:58 +02003884
Edward Cree179bc672016-02-11 20:48:04 +00003885/* Local Checksum Offload.
3886 * Compute outer checksum based on the assumption that the
3887 * inner checksum will be offloaded later.
Edward Creee8ae7b02016-02-11 21:03:37 +00003888 * See Documentation/networking/checksum-offloads.txt for
3889 * explanation of how this works.
Edward Cree179bc672016-02-11 20:48:04 +00003890 * Fill in outer checksum adjustment (e.g. with sum of outer
3891 * pseudo-header) before calling.
3892 * Also ensure that inner checksum is in linear data area.
3893 */
3894static inline __wsum lco_csum(struct sk_buff *skb)
3895{
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003896 unsigned char *csum_start = skb_checksum_start(skb);
3897 unsigned char *l4_hdr = skb_transport_header(skb);
3898 __wsum partial;
Edward Cree179bc672016-02-11 20:48:04 +00003899
3900 /* Start with complement of inner checksum adjustment */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003901 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
3902 skb->csum_offset));
3903
Edward Cree179bc672016-02-11 20:48:04 +00003904 /* Add in checksum of our headers (incl. outer checksum
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003905 * adjustment filled in by caller) and return result.
Edward Cree179bc672016-02-11 20:48:04 +00003906 */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003907 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
Edward Cree179bc672016-02-11 20:48:04 +00003908}
3909
Linus Torvalds1da177e2005-04-16 15:20:36 -07003910#endif /* __KERNEL__ */
3911#endif /* _LINUX_SKBUFF_H */