blob: 3149a88de548368ce3c8d88947034ba7d5615683 [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
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700588 * @nfctinfo: Relationship of this skb to the connection
Linus Torvalds1da177e2005-04-16 15:20:36 -0700589 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700590 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700591 * @ipvs_property: skbuff is owned by ipvs
Willem de Bruijne7246e12017-01-07 17:06:35 -0500592 * @tc_skip_classify: do not classify packet. set by IFB device
Willem de Bruijn8dc07fd2017-01-07 17:06:37 -0500593 * @tc_at_ingress: used within tc_classify to distinguish in/egress
Willem de Bruijnbc31c902017-01-07 17:06:38 -0500594 * @tc_redirected: packet was redirected by a tc action
595 * @tc_from_ingress: if tc_redirected, tc_at_ingress at time of redirect
Randy Dunlap31729362008-02-18 20:52:13 -0800596 * @peeked: this packet has been seen already, so stats have been
597 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700598 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700599 * @protocol: Packet protocol from driver
600 * @destructor: Destruct function
601 * @nfct: Associated connection, if any
Linus Torvalds1da177e2005-04-16 15:20:36 -0700602 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800603 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700604 * @tc_index: Traffic control index
Tom Herbert61b905d2014-03-24 15:34:47 -0700605 * @hash: the packet hash
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700606 * @queue_mapping: Queue mapping for multiqueue devices
David S. Miller0b725a22014-08-25 15:51:53 -0700607 * @xmit_more: More SKBs are pending for this queue
Randy Dunlap553a5672008-04-20 10:51:01 -0700608 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700609 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Tom Herbert61b905d2014-03-24 15:34:47 -0700610 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
Changli Gao4ca24622011-08-19 07:26:44 -0700611 * ports.
Tom Herberta3b18dd2014-07-01 21:33:17 -0700612 * @sw_hash: indicates hash was computed in software stack
Johannes Berg6e3e9392011-11-09 10:15:42 +0100613 * @wifi_acked_valid: wifi_acked was set
614 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000615 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
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)
674 struct nf_conntrack *nfct;
675#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;
727 __u8 nfctinfo:3;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700728
Eric Dumazetb1937222014-09-28 22:18:47 -0700729 __u8 nf_trace:1;
730 __u8 ip_summed:2;
731 __u8 ooo_okay:1;
732 __u8 l4_hash:1;
733 __u8 sw_hash:1;
734 __u8 wifi_acked_valid:1;
735 __u8 wifi_acked:1;
736
737 __u8 no_fcs:1;
738 /* Indicates the inner headers are valid in the skbuff. */
739 __u8 encapsulation:1;
740 __u8 encap_hdr_csum:1;
741 __u8 csum_valid:1;
742 __u8 csum_complete_sw:1;
743 __u8 csum_level:2;
744 __u8 csum_bad:1;
745
746#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
844/**
845 * skb_dst - returns skb dst_entry
846 * @skb: buffer
847 *
848 * Returns skb dst_entry, regardless of reference taken or not.
849 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000850static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
851{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000852 /* If refdst was not refcounted, check we still are in a
853 * rcu_read_lock section
854 */
855 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
856 !rcu_read_lock_held() &&
857 !rcu_read_lock_bh_held());
858 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000859}
860
Eric Dumazet7fee2262010-05-11 23:19:48 +0000861/**
862 * skb_dst_set - sets skb dst
863 * @skb: buffer
864 * @dst: dst entry
865 *
866 * Sets skb dst, assuming a reference was taken on dst and should
867 * be released by skb_dst_drop()
868 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000869static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
870{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000871 skb->_skb_refdst = (unsigned long)dst;
872}
873
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200874/**
875 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
876 * @skb: buffer
877 * @dst: dst entry
878 *
879 * Sets skb dst, assuming a reference was not taken on dst.
880 * If dst entry is cached, we do not take reference and dst_release
881 * will be avoided by refdst_drop. If dst entry is not cached, we take
882 * reference, so that last dst_release can destroy the dst immediately.
883 */
884static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
885{
Hannes Frederic Sowadbfc4fb2014-12-06 19:19:42 +0100886 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
887 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200888}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000889
890/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300891 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000892 * @skb: buffer
893 */
894static inline bool skb_dst_is_noref(const struct sk_buff *skb)
895{
896 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000897}
898
Eric Dumazet511c3f92009-06-02 05:14:27 +0000899static inline struct rtable *skb_rtable(const struct sk_buff *skb)
900{
Eric Dumazetadf30902009-06-02 05:19:30 +0000901 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000902}
903
Jamal Hadi Salim8b10cab2016-07-02 06:43:14 -0400904/* For mangling skb->pkt_type from user space side from applications
905 * such as nft, tc, etc, we only allow a conservative subset of
906 * possible pkt_types to be set.
907*/
908static inline bool skb_pkt_type_ok(u32 ptype)
909{
910 return ptype <= PACKET_OTHERHOST;
911}
912
Joe Perches7965bd42013-09-26 14:48:15 -0700913void kfree_skb(struct sk_buff *skb);
914void kfree_skb_list(struct sk_buff *segs);
915void skb_tx_error(struct sk_buff *skb);
916void consume_skb(struct sk_buff *skb);
917void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000918extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000919
Joe Perches7965bd42013-09-26 14:48:15 -0700920void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
921bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
922 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000923
Joe Perches7965bd42013-09-26 14:48:15 -0700924struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
925 int node);
Eric Dumazet2ea2f622015-04-24 16:05:01 -0700926struct sk_buff *__build_skb(void *data, unsigned int frag_size);
Joe Perches7965bd42013-09-26 14:48:15 -0700927struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700928static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100929 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700930{
Eric Dumazet564824b2010-10-11 19:05:25 +0000931 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700932}
933
Eric Dumazet2e4e4412014-09-17 04:49:49 -0700934struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
935 unsigned long data_len,
936 int max_page_order,
937 int *errcode,
938 gfp_t gfp_mask);
939
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700940/* Layout of fast clones : [skb1][skb2][fclone_ref] */
941struct sk_buff_fclones {
942 struct sk_buff skb1;
943
944 struct sk_buff skb2;
945
946 atomic_t fclone_ref;
947};
948
949/**
950 * skb_fclone_busy - check if fclone is busy
Stephen Hemminger293de7d2016-10-23 09:28:29 -0700951 * @sk: socket
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700952 * @skb: buffer
953 *
Masanari Iidabda13fe2015-12-13 16:53:02 +0900954 * Returns true if skb is a fast clone, and its clone is not freed.
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700955 * Some drivers call skb_orphan() in their ndo_start_xmit(),
956 * so we also check that this didnt happen.
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700957 */
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700958static inline bool skb_fclone_busy(const struct sock *sk,
959 const struct sk_buff *skb)
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700960{
961 const struct sk_buff_fclones *fclones;
962
963 fclones = container_of(skb, struct sk_buff_fclones, skb1);
964
965 return skb->fclone == SKB_FCLONE_ORIG &&
Eric Dumazet6ffe75e2014-12-03 17:04:39 -0800966 atomic_read(&fclones->fclone_ref) > 1 &&
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700967 fclones->skb2.sk == sk;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700968}
969
David S. Millerd179cd12005-08-17 14:57:30 -0700970static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100971 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700972{
Mel Gormanc93bdd02012-07-31 16:44:19 -0700973 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700974}
975
Joe Perches7965bd42013-09-26 14:48:15 -0700976struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
Patrick McHardy0ebd0ac2013-04-17 06:46:58 +0000977static inline struct sk_buff *alloc_skb_head(gfp_t priority)
978{
979 return __alloc_skb_head(priority, -1);
980}
981
Joe Perches7965bd42013-09-26 14:48:15 -0700982struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
983int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
984struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
985struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
Octavian Purdilabad93e92014-06-12 01:36:26 +0300986struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
987 gfp_t gfp_mask, bool fclone);
988static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
989 gfp_t gfp_mask)
990{
991 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
992}
Eric Dumazet117632e2011-12-03 21:39:53 +0000993
Joe Perches7965bd42013-09-26 14:48:15 -0700994int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
995struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
996 unsigned int headroom);
997struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
998 int newtailroom, gfp_t priority);
Fan Du25a91d82014-01-18 09:54:23 +0800999int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1000 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001001int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
1002 int len);
1003int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
1004int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +00001005#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001006
Joe Perches7965bd42013-09-26 14:48:15 -07001007int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
1008 int getfrag(void *from, char *to, int offset,
1009 int len, int odd, struct sk_buff *skb),
1010 void *from, int length);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001011
Hannes Frederic Sowabe12a1f2015-05-21 16:59:58 +02001012int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1013 int offset, size_t size);
1014
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08001015struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -07001016 __u32 lower_offset;
1017 __u32 upper_offset;
1018 __u32 frag_idx;
1019 __u32 stepped_offset;
1020 struct sk_buff *root_skb;
1021 struct sk_buff *cur_skb;
1022 __u8 *frag_data;
1023};
1024
Joe Perches7965bd42013-09-26 14:48:15 -07001025void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1026 unsigned int to, struct skb_seq_state *st);
1027unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1028 struct skb_seq_state *st);
1029void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -07001030
Joe Perches7965bd42013-09-26 14:48:15 -07001031unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
Bojan Prtvar059a2442015-02-22 11:46:35 +01001032 unsigned int to, struct ts_config *config);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -07001033
Tom Herbert09323cc2013-12-15 22:16:19 -08001034/*
1035 * Packet hash types specify the type of hash in skb_set_hash.
1036 *
1037 * Hash types refer to the protocol layer addresses which are used to
1038 * construct a packet's hash. The hashes are used to differentiate or identify
1039 * flows of the protocol layer for the hash type. Hash types are either
1040 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1041 *
1042 * Properties of hashes:
1043 *
1044 * 1) Two packets in different flows have different hash values
1045 * 2) Two packets in the same flow should have the same hash value
1046 *
1047 * A hash at a higher layer is considered to be more specific. A driver should
1048 * set the most specific hash possible.
1049 *
1050 * A driver cannot indicate a more specific hash than the layer at which a hash
1051 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1052 *
1053 * A driver may indicate a hash level which is less specific than the
1054 * actual layer the hash was computed on. For instance, a hash computed
1055 * at L4 may be considered an L3 hash. This should only be done if the
1056 * driver can't unambiguously determine that the HW computed the hash at
1057 * the higher layer. Note that the "should" in the second property above
1058 * permits this.
1059 */
1060enum pkt_hash_types {
1061 PKT_HASH_TYPE_NONE, /* Undefined type */
1062 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1063 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1064 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1065};
1066
Tom Herbertbcc83832015-09-01 09:24:24 -07001067static inline void skb_clear_hash(struct sk_buff *skb)
1068{
1069 skb->hash = 0;
1070 skb->sw_hash = 0;
1071 skb->l4_hash = 0;
1072}
1073
1074static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1075{
1076 if (!skb->l4_hash)
1077 skb_clear_hash(skb);
1078}
1079
1080static inline void
1081__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1082{
1083 skb->l4_hash = is_l4;
1084 skb->sw_hash = is_sw;
1085 skb->hash = hash;
1086}
1087
Tom Herbert09323cc2013-12-15 22:16:19 -08001088static inline void
1089skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1090{
Tom Herbertbcc83832015-09-01 09:24:24 -07001091 /* Used by drivers to set hash from HW */
1092 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1093}
1094
1095static inline void
1096__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1097{
1098 __skb_set_hash(skb, hash, true, is_l4);
Tom Herbert09323cc2013-12-15 22:16:19 -08001099}
1100
Tom Herberte5276932015-09-01 09:24:23 -07001101void __skb_get_hash(struct sk_buff *skb);
Florian Westphalb9177832016-10-26 18:49:46 +02001102u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
Tom Herberte5276932015-09-01 09:24:23 -07001103u32 skb_get_poff(const struct sk_buff *skb);
1104u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1105 const struct flow_keys *keys, int hlen);
1106__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1107 void *data, int hlen_proto);
1108
1109static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1110 int thoff, u8 ip_proto)
1111{
1112 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1113}
1114
1115void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1116 const struct flow_dissector_key *key,
1117 unsigned int key_count);
1118
1119bool __skb_flow_dissect(const struct sk_buff *skb,
1120 struct flow_dissector *flow_dissector,
1121 void *target_container,
Tom Herbertcd79a232015-09-01 09:24:27 -07001122 void *data, __be16 proto, int nhoff, int hlen,
1123 unsigned int flags);
Tom Herberte5276932015-09-01 09:24:23 -07001124
1125static inline bool skb_flow_dissect(const struct sk_buff *skb,
1126 struct flow_dissector *flow_dissector,
Tom Herbertcd79a232015-09-01 09:24:27 -07001127 void *target_container, unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001128{
1129 return __skb_flow_dissect(skb, flow_dissector, target_container,
Tom Herbertcd79a232015-09-01 09:24:27 -07001130 NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001131}
1132
1133static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
Tom Herbertcd79a232015-09-01 09:24:27 -07001134 struct flow_keys *flow,
1135 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001136{
1137 memset(flow, 0, sizeof(*flow));
1138 return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
Tom Herbertcd79a232015-09-01 09:24:27 -07001139 NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001140}
1141
1142static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
1143 void *data, __be16 proto,
Tom Herbertcd79a232015-09-01 09:24:27 -07001144 int nhoff, int hlen,
1145 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001146{
1147 memset(flow, 0, sizeof(*flow));
1148 return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
Tom Herbertcd79a232015-09-01 09:24:27 -07001149 data, proto, nhoff, hlen, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001150}
1151
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001152static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001153{
Tom Herberta3b18dd2014-07-01 21:33:17 -07001154 if (!skb->l4_hash && !skb->sw_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001155 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001156
Tom Herbert61b905d2014-03-24 15:34:47 -07001157 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001158}
1159
David S. Miller20a17bf2015-09-01 21:19:17 -07001160__u32 __skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6);
Tom Herbertf70ea012015-07-31 16:52:10 -07001161
David S. Miller20a17bf2015-09-01 21:19:17 -07001162static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
Tom Herbertf70ea012015-07-31 16:52:10 -07001163{
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001164 if (!skb->l4_hash && !skb->sw_hash) {
1165 struct flow_keys keys;
Tom Herbertde4c1f82015-09-01 18:11:04 -07001166 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001167
Tom Herbertde4c1f82015-09-01 18:11:04 -07001168 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001169 }
Tom Herbertf70ea012015-07-31 16:52:10 -07001170
1171 return skb->hash;
1172}
1173
David S. Miller20a17bf2015-09-01 21:19:17 -07001174__u32 __skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl);
Tom Herbertf70ea012015-07-31 16:52:10 -07001175
David S. Miller20a17bf2015-09-01 21:19:17 -07001176static inline __u32 skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl4)
Tom Herbertf70ea012015-07-31 16:52:10 -07001177{
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001178 if (!skb->l4_hash && !skb->sw_hash) {
1179 struct flow_keys keys;
Tom Herbertde4c1f82015-09-01 18:11:04 -07001180 __u32 hash = __get_hash_from_flowi4(fl4, &keys);
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001181
Tom Herbertde4c1f82015-09-01 18:11:04 -07001182 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001183 }
Tom Herbertf70ea012015-07-31 16:52:10 -07001184
1185 return skb->hash;
1186}
1187
Tom Herbert50fb7992015-05-01 11:30:12 -07001188__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
1189
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001190static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1191{
Tom Herbert61b905d2014-03-24 15:34:47 -07001192 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001193}
1194
Tom Herbert3df7a742013-12-15 22:16:29 -08001195static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1196{
Tom Herbert61b905d2014-03-24 15:34:47 -07001197 to->hash = from->hash;
Tom Herberta3b18dd2014-07-01 21:33:17 -07001198 to->sw_hash = from->sw_hash;
Tom Herbert61b905d2014-03-24 15:34:47 -07001199 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -08001200};
1201
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001202#ifdef NET_SKBUFF_DATA_USES_OFFSET
1203static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1204{
1205 return skb->head + skb->end;
1206}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001207
1208static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1209{
1210 return skb->end;
1211}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001212#else
1213static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1214{
1215 return skb->end;
1216}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001217
1218static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1219{
1220 return skb->end - skb->head;
1221}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001222#endif
1223
Linus Torvalds1da177e2005-04-16 15:20:36 -07001224/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001225#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001226
Patrick Ohlyac45f602009-02-12 05:03:37 +00001227static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1228{
1229 return &skb_shinfo(skb)->hwtstamps;
1230}
1231
Linus Torvalds1da177e2005-04-16 15:20:36 -07001232/**
1233 * skb_queue_empty - check if a queue is empty
1234 * @list: queue head
1235 *
1236 * Returns true if the queue is empty, false otherwise.
1237 */
1238static inline int skb_queue_empty(const struct sk_buff_head *list)
1239{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001240 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001241}
1242
1243/**
David S. Millerfc7ebb22008-09-23 00:34:07 -07001244 * skb_queue_is_last - check if skb is the last entry in the queue
1245 * @list: queue head
1246 * @skb: buffer
1247 *
1248 * Returns true if @skb is the last buffer on the list.
1249 */
1250static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1251 const struct sk_buff *skb)
1252{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001253 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -07001254}
1255
1256/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001257 * skb_queue_is_first - check if skb is the first entry in the queue
1258 * @list: queue head
1259 * @skb: buffer
1260 *
1261 * Returns true if @skb is the first buffer on the list.
1262 */
1263static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1264 const struct sk_buff *skb)
1265{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001266 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001267}
1268
1269/**
David S. Miller249c8b42008-09-23 00:44:42 -07001270 * skb_queue_next - return the next packet in the queue
1271 * @list: queue head
1272 * @skb: current buffer
1273 *
1274 * Return the next packet in @list after @skb. It is only valid to
1275 * call this if skb_queue_is_last() evaluates to false.
1276 */
1277static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1278 const struct sk_buff *skb)
1279{
1280 /* This BUG_ON may seem severe, but if we just return then we
1281 * are going to dereference garbage.
1282 */
1283 BUG_ON(skb_queue_is_last(list, skb));
1284 return skb->next;
1285}
1286
1287/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001288 * skb_queue_prev - return the prev packet in the queue
1289 * @list: queue head
1290 * @skb: current buffer
1291 *
1292 * Return the prev packet in @list before @skb. It is only valid to
1293 * call this if skb_queue_is_first() evaluates to false.
1294 */
1295static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1296 const struct sk_buff *skb)
1297{
1298 /* This BUG_ON may seem severe, but if we just return then we
1299 * are going to dereference garbage.
1300 */
1301 BUG_ON(skb_queue_is_first(list, skb));
1302 return skb->prev;
1303}
1304
1305/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001306 * skb_get - reference buffer
1307 * @skb: buffer to reference
1308 *
1309 * Makes another reference to a socket buffer and returns a pointer
1310 * to the buffer.
1311 */
1312static inline struct sk_buff *skb_get(struct sk_buff *skb)
1313{
1314 atomic_inc(&skb->users);
1315 return skb;
1316}
1317
1318/*
1319 * If users == 1, we are the only owner and are can avoid redundant
1320 * atomic change.
1321 */
1322
1323/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001324 * skb_cloned - is the buffer a clone
1325 * @skb: buffer to check
1326 *
1327 * Returns true if the buffer was generated with skb_clone() and is
1328 * one of multiple shared copies of the buffer. Cloned buffers are
1329 * shared data so must not be written to under normal circumstances.
1330 */
1331static inline int skb_cloned(const struct sk_buff *skb)
1332{
1333 return skb->cloned &&
1334 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1335}
1336
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001337static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1338{
Mel Gormand0164ad2015-11-06 16:28:21 -08001339 might_sleep_if(gfpflags_allow_blocking(pri));
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001340
1341 if (skb_cloned(skb))
1342 return pskb_expand_head(skb, 0, 0, pri);
1343
1344 return 0;
1345}
1346
Linus Torvalds1da177e2005-04-16 15:20:36 -07001347/**
1348 * skb_header_cloned - is the header a clone
1349 * @skb: buffer to check
1350 *
1351 * Returns true if modifying the header part of the buffer requires
1352 * the data to be copied.
1353 */
1354static inline int skb_header_cloned(const struct sk_buff *skb)
1355{
1356 int dataref;
1357
1358 if (!skb->cloned)
1359 return 0;
1360
1361 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1362 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1363 return dataref != 1;
1364}
1365
Eric Dumazet9580bf22016-04-30 10:19:29 -07001366static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1367{
1368 might_sleep_if(gfpflags_allow_blocking(pri));
1369
1370 if (skb_header_cloned(skb))
1371 return pskb_expand_head(skb, 0, 0, pri);
1372
1373 return 0;
1374}
1375
Linus Torvalds1da177e2005-04-16 15:20:36 -07001376/**
1377 * skb_header_release - release reference to header
1378 * @skb: buffer to operate on
1379 *
1380 * Drop a reference to the header part of the buffer. This is done
1381 * by acquiring a payload reference. You must not read from the header
1382 * part of skb->data after this.
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001383 * Note : Check if you can use __skb_header_release() instead.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001384 */
1385static inline void skb_header_release(struct sk_buff *skb)
1386{
1387 BUG_ON(skb->nohdr);
1388 skb->nohdr = 1;
1389 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1390}
1391
1392/**
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001393 * __skb_header_release - release reference to header
1394 * @skb: buffer to operate on
1395 *
1396 * Variant of skb_header_release() assuming skb is private to caller.
1397 * We can avoid one atomic operation.
1398 */
1399static inline void __skb_header_release(struct sk_buff *skb)
1400{
1401 skb->nohdr = 1;
1402 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1403}
1404
1405
1406/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001407 * skb_shared - is the buffer shared
1408 * @skb: buffer to check
1409 *
1410 * Returns true if more than one person has a reference to this
1411 * buffer.
1412 */
1413static inline int skb_shared(const struct sk_buff *skb)
1414{
1415 return atomic_read(&skb->users) != 1;
1416}
1417
1418/**
1419 * skb_share_check - check if buffer is shared and if so clone it
1420 * @skb: buffer to check
1421 * @pri: priority for memory allocation
1422 *
1423 * If the buffer is shared the buffer is cloned and the old copy
1424 * drops a reference. A new clone with a single reference is returned.
1425 * If the buffer is not shared the original buffer is returned. When
1426 * being called from interrupt status or with spinlocks held pri must
1427 * be GFP_ATOMIC.
1428 *
1429 * NULL is returned on a memory allocation failure.
1430 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001431static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001432{
Mel Gormand0164ad2015-11-06 16:28:21 -08001433 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001434 if (skb_shared(skb)) {
1435 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001436
1437 if (likely(nskb))
1438 consume_skb(skb);
1439 else
1440 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001441 skb = nskb;
1442 }
1443 return skb;
1444}
1445
1446/*
1447 * Copy shared buffers into a new sk_buff. We effectively do COW on
1448 * packets to handle cases where we have a local reader and forward
1449 * and a couple of other messy ones. The normal one is tcpdumping
1450 * a packet thats being forwarded.
1451 */
1452
1453/**
1454 * skb_unshare - make a copy of a shared buffer
1455 * @skb: buffer to check
1456 * @pri: priority for memory allocation
1457 *
1458 * If the socket buffer is a clone then this function creates a new
1459 * copy of the data, drops a reference count on the old copy and returns
1460 * the new copy with the reference count at 1. If the buffer is not a clone
1461 * the original buffer is returned. When called with a spinlock held or
1462 * from interrupt state @pri must be %GFP_ATOMIC
1463 *
1464 * %NULL is returned on a memory allocation failure.
1465 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001466static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001467 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001468{
Mel Gormand0164ad2015-11-06 16:28:21 -08001469 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001470 if (skb_cloned(skb)) {
1471 struct sk_buff *nskb = skb_copy(skb, pri);
Alexander Aring31eff812014-10-10 23:10:47 +02001472
1473 /* Free our shared copy */
1474 if (likely(nskb))
1475 consume_skb(skb);
1476 else
1477 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001478 skb = nskb;
1479 }
1480 return skb;
1481}
1482
1483/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001484 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001485 * @list_: list to peek at
1486 *
1487 * Peek an &sk_buff. Unlike most other operations you _MUST_
1488 * be careful with this one. A peek leaves the buffer on the
1489 * list and someone else may run off with it. You must hold
1490 * the appropriate locks or have a private queue to do this.
1491 *
1492 * Returns %NULL for an empty list or a pointer to the head element.
1493 * The reference count is not incremented and the reference is therefore
1494 * volatile. Use with caution.
1495 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001496static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001497{
Eric Dumazet18d07002012-04-30 16:31:46 +00001498 struct sk_buff *skb = list_->next;
1499
1500 if (skb == (struct sk_buff *)list_)
1501 skb = NULL;
1502 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001503}
1504
1505/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001506 * skb_peek_next - peek skb following the given one from a queue
1507 * @skb: skb to start from
1508 * @list_: list to peek at
1509 *
1510 * Returns %NULL when the end of the list is met or a pointer to the
1511 * next element. The reference count is not incremented and the
1512 * reference is therefore volatile. Use with caution.
1513 */
1514static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1515 const struct sk_buff_head *list_)
1516{
1517 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001518
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001519 if (next == (struct sk_buff *)list_)
1520 next = NULL;
1521 return next;
1522}
1523
1524/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001525 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001526 * @list_: list to peek at
1527 *
1528 * Peek an &sk_buff. Unlike most other operations you _MUST_
1529 * be careful with this one. A peek leaves the buffer on the
1530 * list and someone else may run off with it. You must hold
1531 * the appropriate locks or have a private queue to do this.
1532 *
1533 * Returns %NULL for an empty list or a pointer to the tail element.
1534 * The reference count is not incremented and the reference is therefore
1535 * volatile. Use with caution.
1536 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001537static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001538{
Eric Dumazet18d07002012-04-30 16:31:46 +00001539 struct sk_buff *skb = list_->prev;
1540
1541 if (skb == (struct sk_buff *)list_)
1542 skb = NULL;
1543 return skb;
1544
Linus Torvalds1da177e2005-04-16 15:20:36 -07001545}
1546
1547/**
1548 * skb_queue_len - get queue length
1549 * @list_: list to measure
1550 *
1551 * Return the length of an &sk_buff queue.
1552 */
1553static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1554{
1555 return list_->qlen;
1556}
1557
David S. Miller67fed452008-09-21 22:36:24 -07001558/**
1559 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1560 * @list: queue to initialize
1561 *
1562 * This initializes only the list and queue length aspects of
1563 * an sk_buff_head object. This allows to initialize the list
1564 * aspects of an sk_buff_head without reinitializing things like
1565 * the spinlock. It can also be used for on-stack sk_buff_head
1566 * objects where the spinlock is known to not be used.
1567 */
1568static inline void __skb_queue_head_init(struct sk_buff_head *list)
1569{
1570 list->prev = list->next = (struct sk_buff *)list;
1571 list->qlen = 0;
1572}
1573
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001574/*
1575 * This function creates a split out lock class for each invocation;
1576 * this is needed for now since a whole lot of users of the skb-queue
1577 * infrastructure in drivers have different locking usage (in hardirq)
1578 * than the networking core (in softirq only). In the long run either the
1579 * network layer or drivers should need annotation to consolidate the
1580 * main types of usage into 3 classes.
1581 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582static inline void skb_queue_head_init(struct sk_buff_head *list)
1583{
1584 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001585 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001586}
1587
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001588static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1589 struct lock_class_key *class)
1590{
1591 skb_queue_head_init(list);
1592 lockdep_set_class(&list->lock, class);
1593}
1594
Linus Torvalds1da177e2005-04-16 15:20:36 -07001595/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001596 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597 *
1598 * The "__skb_xxxx()" functions are the non-atomic ones that
1599 * can only be called with interrupts disabled.
1600 */
Joe Perches7965bd42013-09-26 14:48:15 -07001601void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1602 struct sk_buff_head *list);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001603static inline void __skb_insert(struct sk_buff *newsk,
1604 struct sk_buff *prev, struct sk_buff *next,
1605 struct sk_buff_head *list)
1606{
1607 newsk->next = next;
1608 newsk->prev = prev;
1609 next->prev = prev->next = newsk;
1610 list->qlen++;
1611}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001612
David S. Miller67fed452008-09-21 22:36:24 -07001613static inline void __skb_queue_splice(const struct sk_buff_head *list,
1614 struct sk_buff *prev,
1615 struct sk_buff *next)
1616{
1617 struct sk_buff *first = list->next;
1618 struct sk_buff *last = list->prev;
1619
1620 first->prev = prev;
1621 prev->next = first;
1622
1623 last->next = next;
1624 next->prev = last;
1625}
1626
1627/**
1628 * skb_queue_splice - join two skb lists, this is designed for stacks
1629 * @list: the new list to add
1630 * @head: the place to add it in the first list
1631 */
1632static inline void skb_queue_splice(const struct sk_buff_head *list,
1633 struct sk_buff_head *head)
1634{
1635 if (!skb_queue_empty(list)) {
1636 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001637 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001638 }
1639}
1640
1641/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001642 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001643 * @list: the new list to add
1644 * @head: the place to add it in the first list
1645 *
1646 * The list at @list is reinitialised
1647 */
1648static inline void skb_queue_splice_init(struct sk_buff_head *list,
1649 struct sk_buff_head *head)
1650{
1651 if (!skb_queue_empty(list)) {
1652 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001653 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001654 __skb_queue_head_init(list);
1655 }
1656}
1657
1658/**
1659 * skb_queue_splice_tail - join two skb lists, each list being a queue
1660 * @list: the new list to add
1661 * @head: the place to add it in the first list
1662 */
1663static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1664 struct sk_buff_head *head)
1665{
1666 if (!skb_queue_empty(list)) {
1667 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001668 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001669 }
1670}
1671
1672/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001673 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001674 * @list: the new list to add
1675 * @head: the place to add it in the first list
1676 *
1677 * Each of the lists is a queue.
1678 * The list at @list is reinitialised
1679 */
1680static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1681 struct sk_buff_head *head)
1682{
1683 if (!skb_queue_empty(list)) {
1684 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001685 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001686 __skb_queue_head_init(list);
1687 }
1688}
1689
Linus Torvalds1da177e2005-04-16 15:20:36 -07001690/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001691 * __skb_queue_after - queue a buffer at the list head
1692 * @list: list to use
1693 * @prev: place after this buffer
1694 * @newsk: buffer to queue
1695 *
1696 * Queue a buffer int the middle of a list. This function takes no locks
1697 * and you must therefore hold required locks before calling it.
1698 *
1699 * A buffer cannot be placed on two lists at the same time.
1700 */
1701static inline void __skb_queue_after(struct sk_buff_head *list,
1702 struct sk_buff *prev,
1703 struct sk_buff *newsk)
1704{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001705 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001706}
1707
Joe Perches7965bd42013-09-26 14:48:15 -07001708void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1709 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001710
Gerrit Renkerf5572852008-04-14 00:05:28 -07001711static inline void __skb_queue_before(struct sk_buff_head *list,
1712 struct sk_buff *next,
1713 struct sk_buff *newsk)
1714{
1715 __skb_insert(newsk, next->prev, next, list);
1716}
1717
Stephen Hemminger300ce172005-10-30 13:47:34 -08001718/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001719 * __skb_queue_head - queue a buffer at the list head
1720 * @list: list to use
1721 * @newsk: buffer to queue
1722 *
1723 * Queue a buffer at the start of a list. This function takes no locks
1724 * and you must therefore hold required locks before calling it.
1725 *
1726 * A buffer cannot be placed on two lists at the same time.
1727 */
Joe Perches7965bd42013-09-26 14:48:15 -07001728void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001729static inline void __skb_queue_head(struct sk_buff_head *list,
1730 struct sk_buff *newsk)
1731{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001732 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733}
1734
1735/**
1736 * __skb_queue_tail - queue a buffer at the list tail
1737 * @list: list to use
1738 * @newsk: buffer to queue
1739 *
1740 * Queue a buffer at the end of a list. This function takes no locks
1741 * and you must therefore hold required locks before calling it.
1742 *
1743 * A buffer cannot be placed on two lists at the same time.
1744 */
Joe Perches7965bd42013-09-26 14:48:15 -07001745void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001746static inline void __skb_queue_tail(struct sk_buff_head *list,
1747 struct sk_buff *newsk)
1748{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001749 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001750}
1751
Linus Torvalds1da177e2005-04-16 15:20:36 -07001752/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001753 * remove sk_buff from list. _Must_ be called atomically, and with
1754 * the list known..
1755 */
Joe Perches7965bd42013-09-26 14:48:15 -07001756void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001757static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1758{
1759 struct sk_buff *next, *prev;
1760
1761 list->qlen--;
1762 next = skb->next;
1763 prev = skb->prev;
1764 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001765 next->prev = prev;
1766 prev->next = next;
1767}
1768
Gerrit Renkerf525c062008-04-14 00:04:12 -07001769/**
1770 * __skb_dequeue - remove from the head of the queue
1771 * @list: list to dequeue from
1772 *
1773 * Remove the head of the list. This function does not take any locks
1774 * so must be used with appropriate locks held only. The head item is
1775 * returned or %NULL if the list is empty.
1776 */
Joe Perches7965bd42013-09-26 14:48:15 -07001777struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Gerrit Renkerf525c062008-04-14 00:04:12 -07001778static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1779{
1780 struct sk_buff *skb = skb_peek(list);
1781 if (skb)
1782 __skb_unlink(skb, list);
1783 return skb;
1784}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001785
1786/**
1787 * __skb_dequeue_tail - remove from the tail of the queue
1788 * @list: list to dequeue from
1789 *
1790 * Remove the tail of the list. This function does not take any locks
1791 * so must be used with appropriate locks held only. The tail item is
1792 * returned or %NULL if the list is empty.
1793 */
Joe Perches7965bd42013-09-26 14:48:15 -07001794struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001795static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1796{
1797 struct sk_buff *skb = skb_peek_tail(list);
1798 if (skb)
1799 __skb_unlink(skb, list);
1800 return skb;
1801}
1802
1803
David S. Millerbdcc0922012-03-07 20:53:36 -05001804static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001805{
1806 return skb->data_len;
1807}
1808
1809static inline unsigned int skb_headlen(const struct sk_buff *skb)
1810{
1811 return skb->len - skb->data_len;
1812}
1813
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03001814static inline unsigned int skb_pagelen(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001815{
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03001816 unsigned int i, len = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001817
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03001818 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001819 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001820 return len + skb_headlen(skb);
1821}
1822
Ian Campbell131ea662011-08-19 06:25:00 +00001823/**
1824 * __skb_fill_page_desc - initialise a paged fragment in an skb
1825 * @skb: buffer containing fragment to be initialised
1826 * @i: paged fragment index to initialise
1827 * @page: the page to use for this fragment
1828 * @off: the offset to the data with @page
1829 * @size: the length of the data
1830 *
1831 * Initialises the @i'th fragment of @skb to point to &size bytes at
1832 * offset @off within @page.
1833 *
1834 * Does not take any additional reference on the fragment.
1835 */
1836static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1837 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001838{
1839 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1840
Mel Gormanc48a11c2012-07-31 16:44:23 -07001841 /*
Michal Hocko2f064f32015-08-21 14:11:51 -07001842 * Propagate page pfmemalloc to the skb if we can. The problem is
1843 * that not all callers have unique ownership of the page but rely
1844 * on page_is_pfmemalloc doing the right thing(tm).
Mel Gormanc48a11c2012-07-31 16:44:23 -07001845 */
Ian Campbella8605c62011-10-19 23:01:49 +00001846 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001847 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001848 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001849
1850 page = compound_head(page);
Michal Hocko2f064f32015-08-21 14:11:51 -07001851 if (page_is_pfmemalloc(page))
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001852 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00001853}
1854
1855/**
1856 * skb_fill_page_desc - initialise a paged fragment in an skb
1857 * @skb: buffer containing fragment to be initialised
1858 * @i: paged fragment index to initialise
1859 * @page: the page to use for this fragment
1860 * @off: the offset to the data with @page
1861 * @size: the length of the data
1862 *
1863 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01001864 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00001865 * addition updates @skb such that @i is the last fragment.
1866 *
1867 * Does not take any additional reference on the fragment.
1868 */
1869static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1870 struct page *page, int off, int size)
1871{
1872 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001873 skb_shinfo(skb)->nr_frags = i + 1;
1874}
1875
Joe Perches7965bd42013-09-26 14:48:15 -07001876void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1877 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001878
Jason Wangf8e617e2013-11-01 14:07:47 +08001879void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1880 unsigned int truesize);
1881
Linus Torvalds1da177e2005-04-16 15:20:36 -07001882#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001883#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001884#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1885
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001886#ifdef NET_SKBUFF_DATA_USES_OFFSET
1887static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1888{
1889 return skb->head + skb->tail;
1890}
1891
1892static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1893{
1894 skb->tail = skb->data - skb->head;
1895}
1896
1897static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1898{
1899 skb_reset_tail_pointer(skb);
1900 skb->tail += offset;
1901}
Simon Horman7cc46192013-05-28 20:34:29 +00001902
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001903#else /* NET_SKBUFF_DATA_USES_OFFSET */
1904static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1905{
1906 return skb->tail;
1907}
1908
1909static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1910{
1911 skb->tail = skb->data;
1912}
1913
1914static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1915{
1916 skb->tail = skb->data + offset;
1917}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001918
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001919#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1920
Linus Torvalds1da177e2005-04-16 15:20:36 -07001921/*
1922 * Add data to an sk_buff
1923 */
Mathias Krause0c7ddf32013-11-07 14:18:24 +01001924unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001925unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001926static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1927{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001928 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001929 SKB_LINEAR_ASSERT(skb);
1930 skb->tail += len;
1931 skb->len += len;
1932 return tmp;
1933}
1934
Joe Perches7965bd42013-09-26 14:48:15 -07001935unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001936static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1937{
1938 skb->data -= len;
1939 skb->len += len;
1940 return skb->data;
1941}
1942
Joe Perches7965bd42013-09-26 14:48:15 -07001943unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001944static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1945{
1946 skb->len -= len;
1947 BUG_ON(skb->len < skb->data_len);
1948 return skb->data += len;
1949}
1950
David S. Miller47d29642010-05-02 02:21:44 -07001951static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1952{
1953 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1954}
1955
Joe Perches7965bd42013-09-26 14:48:15 -07001956unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001957
1958static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1959{
1960 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001961 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001962 return NULL;
1963 skb->len -= len;
1964 return skb->data += len;
1965}
1966
1967static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1968{
1969 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1970}
1971
1972static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1973{
1974 if (likely(len <= skb_headlen(skb)))
1975 return 1;
1976 if (unlikely(len > skb->len))
1977 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001978 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001979}
1980
Eric Dumazetc8c8b122016-12-07 09:19:33 -08001981void skb_condense(struct sk_buff *skb);
1982
Linus Torvalds1da177e2005-04-16 15:20:36 -07001983/**
1984 * skb_headroom - bytes at buffer head
1985 * @skb: buffer to check
1986 *
1987 * Return the number of bytes of free space at the head of an &sk_buff.
1988 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001989static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001990{
1991 return skb->data - skb->head;
1992}
1993
1994/**
1995 * skb_tailroom - bytes at buffer end
1996 * @skb: buffer to check
1997 *
1998 * Return the number of bytes of free space at the tail of an sk_buff
1999 */
2000static inline int skb_tailroom(const struct sk_buff *skb)
2001{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07002002 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002003}
2004
2005/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00002006 * skb_availroom - bytes at buffer end
2007 * @skb: buffer to check
2008 *
2009 * Return the number of bytes of free space at the tail of an sk_buff
2010 * allocated by sk_stream_alloc()
2011 */
2012static inline int skb_availroom(const struct sk_buff *skb)
2013{
Eric Dumazet16fad692013-03-14 05:40:32 +00002014 if (skb_is_nonlinear(skb))
2015 return 0;
2016
2017 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00002018}
2019
2020/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002021 * skb_reserve - adjust headroom
2022 * @skb: buffer to alter
2023 * @len: bytes to move
2024 *
2025 * Increase the headroom of an empty &sk_buff by reducing the tail
2026 * room. This is only allowed for an empty buffer.
2027 */
David S. Miller8243126c2006-01-17 02:54:21 -08002028static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002029{
2030 skb->data += len;
2031 skb->tail += len;
2032}
2033
Benjamin Poirier1837b2e2016-02-29 15:03:33 -08002034/**
2035 * skb_tailroom_reserve - adjust reserved_tailroom
2036 * @skb: buffer to alter
2037 * @mtu: maximum amount of headlen permitted
2038 * @needed_tailroom: minimum amount of reserved_tailroom
2039 *
2040 * Set reserved_tailroom so that headlen can be as large as possible but
2041 * not larger than mtu and tailroom cannot be smaller than
2042 * needed_tailroom.
2043 * The required headroom should already have been reserved before using
2044 * this function.
2045 */
2046static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2047 unsigned int needed_tailroom)
2048{
2049 SKB_LINEAR_ASSERT(skb);
2050 if (mtu < skb_tailroom(skb) - needed_tailroom)
2051 /* use at most mtu */
2052 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2053 else
2054 /* use up to all available space */
2055 skb->reserved_tailroom = needed_tailroom;
2056}
2057
Tom Herbert8bce6d72014-09-29 20:22:29 -07002058#define ENCAP_TYPE_ETHER 0
2059#define ENCAP_TYPE_IPPROTO 1
2060
2061static inline void skb_set_inner_protocol(struct sk_buff *skb,
2062 __be16 protocol)
2063{
2064 skb->inner_protocol = protocol;
2065 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2066}
2067
2068static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2069 __u8 ipproto)
2070{
2071 skb->inner_ipproto = ipproto;
2072 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2073}
2074
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002075static inline void skb_reset_inner_headers(struct sk_buff *skb)
2076{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002077 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002078 skb->inner_network_header = skb->network_header;
2079 skb->inner_transport_header = skb->transport_header;
2080}
2081
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00002082static inline void skb_reset_mac_len(struct sk_buff *skb)
2083{
2084 skb->mac_len = skb->network_header - skb->mac_header;
2085}
2086
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002087static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2088 *skb)
2089{
2090 return skb->head + skb->inner_transport_header;
2091}
2092
Tom Herbert55dc5a92015-12-14 11:19:40 -08002093static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2094{
2095 return skb_inner_transport_header(skb) - skb->data;
2096}
2097
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002098static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2099{
2100 skb->inner_transport_header = skb->data - skb->head;
2101}
2102
2103static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2104 const int offset)
2105{
2106 skb_reset_inner_transport_header(skb);
2107 skb->inner_transport_header += offset;
2108}
2109
2110static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2111{
2112 return skb->head + skb->inner_network_header;
2113}
2114
2115static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2116{
2117 skb->inner_network_header = skb->data - skb->head;
2118}
2119
2120static inline void skb_set_inner_network_header(struct sk_buff *skb,
2121 const int offset)
2122{
2123 skb_reset_inner_network_header(skb);
2124 skb->inner_network_header += offset;
2125}
2126
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002127static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2128{
2129 return skb->head + skb->inner_mac_header;
2130}
2131
2132static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2133{
2134 skb->inner_mac_header = skb->data - skb->head;
2135}
2136
2137static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2138 const int offset)
2139{
2140 skb_reset_inner_mac_header(skb);
2141 skb->inner_mac_header += offset;
2142}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002143static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2144{
Cong Wang35d04612013-05-29 15:16:05 +08002145 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002146}
2147
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002148static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2149{
2150 return skb->head + skb->transport_header;
2151}
2152
2153static inline void skb_reset_transport_header(struct sk_buff *skb)
2154{
2155 skb->transport_header = skb->data - skb->head;
2156}
2157
2158static inline void skb_set_transport_header(struct sk_buff *skb,
2159 const int offset)
2160{
2161 skb_reset_transport_header(skb);
2162 skb->transport_header += offset;
2163}
2164
2165static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2166{
2167 return skb->head + skb->network_header;
2168}
2169
2170static inline void skb_reset_network_header(struct sk_buff *skb)
2171{
2172 skb->network_header = skb->data - skb->head;
2173}
2174
2175static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2176{
2177 skb_reset_network_header(skb);
2178 skb->network_header += offset;
2179}
2180
2181static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2182{
2183 return skb->head + skb->mac_header;
2184}
2185
2186static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2187{
Cong Wang35d04612013-05-29 15:16:05 +08002188 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002189}
2190
2191static inline void skb_reset_mac_header(struct sk_buff *skb)
2192{
2193 skb->mac_header = skb->data - skb->head;
2194}
2195
2196static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2197{
2198 skb_reset_mac_header(skb);
2199 skb->mac_header += offset;
2200}
2201
Timo Teräs0e3da5b2013-12-16 11:02:09 +02002202static inline void skb_pop_mac_header(struct sk_buff *skb)
2203{
2204 skb->mac_header = skb->network_header;
2205}
2206
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002207static inline void skb_probe_transport_header(struct sk_buff *skb,
2208 const int offset_hint)
2209{
2210 struct flow_keys keys;
2211
2212 if (skb_transport_header_was_set(skb))
2213 return;
Tom Herbertcd79a232015-09-01 09:24:27 -07002214 else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
Tom Herbert42aecaa2015-06-04 09:16:39 -07002215 skb_set_transport_header(skb, keys.control.thoff);
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002216 else
2217 skb_set_transport_header(skb, offset_hint);
2218}
2219
Eric Dumazet03606892012-02-23 10:55:02 +00002220static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2221{
2222 if (skb_mac_header_was_set(skb)) {
2223 const unsigned char *old_mac = skb_mac_header(skb);
2224
2225 skb_set_mac_header(skb, -skb->mac_len);
2226 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2227 }
2228}
2229
Michał Mirosław04fb4512010-12-14 15:24:08 +00002230static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2231{
2232 return skb->csum_start - skb_headroom(skb);
2233}
2234
Alexander Duyck08b64fc2016-02-05 15:27:49 -08002235static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2236{
2237 return skb->head + skb->csum_start;
2238}
2239
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002240static inline int skb_transport_offset(const struct sk_buff *skb)
2241{
2242 return skb_transport_header(skb) - skb->data;
2243}
2244
2245static inline u32 skb_network_header_len(const struct sk_buff *skb)
2246{
2247 return skb->transport_header - skb->network_header;
2248}
2249
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002250static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2251{
2252 return skb->inner_transport_header - skb->inner_network_header;
2253}
2254
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002255static inline int skb_network_offset(const struct sk_buff *skb)
2256{
2257 return skb_network_header(skb) - skb->data;
2258}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03002259
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002260static inline int skb_inner_network_offset(const struct sk_buff *skb)
2261{
2262 return skb_inner_network_header(skb) - skb->data;
2263}
2264
Changli Gaof9599ce2010-08-04 04:43:44 +00002265static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2266{
2267 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2268}
2269
Linus Torvalds1da177e2005-04-16 15:20:36 -07002270/*
2271 * CPUs often take a performance hit when accessing unaligned memory
2272 * locations. The actual performance hit varies, it can be small if the
2273 * hardware handles it or large if we have to take an exception and fix it
2274 * in software.
2275 *
2276 * Since an ethernet header is 14 bytes network drivers often end up with
2277 * the IP header at an unaligned offset. The IP header can be aligned by
2278 * shifting the start of the packet by 2 bytes. Drivers should do this
2279 * with:
2280 *
Tobias Klauser8660c122009-07-13 22:48:16 +00002281 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002282 *
2283 * The downside to this alignment of the IP header is that the DMA is now
2284 * unaligned. On some architectures the cost of an unaligned DMA is high
2285 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00002286 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002287 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2288 * to be overridden.
2289 */
2290#ifndef NET_IP_ALIGN
2291#define NET_IP_ALIGN 2
2292#endif
2293
Anton Blanchard025be812006-03-31 02:27:06 -08002294/*
2295 * The networking layer reserves some headroom in skb data (via
2296 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2297 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08002298 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08002299 *
2300 * Unfortunately this headroom changes the DMA alignment of the resulting
2301 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2302 * on some architectures. An architecture can override this value,
2303 * perhaps setting it to a cacheline in size (since that will maintain
2304 * cacheline alignment of the DMA). It must be a power of 2.
2305 *
David S. Millerd6301d32009-02-08 19:24:13 -08002306 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08002307 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07002308 *
2309 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2310 * to reduce average number of cache lines per packet.
2311 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07002312 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08002313 */
2314#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07002315#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08002316#endif
2317
Joe Perches7965bd42013-09-26 14:48:15 -07002318int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002319
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002320static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002321{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00002322 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07002323 WARN_ON(1);
2324 return;
2325 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002326 skb->len = len;
2327 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002328}
2329
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002330static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2331{
2332 __skb_set_length(skb, len);
2333}
2334
Joe Perches7965bd42013-09-26 14:48:15 -07002335void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002336
2337static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2338{
Herbert Xu3cc0e872006-06-09 16:13:38 -07002339 if (skb->data_len)
2340 return ___pskb_trim(skb, len);
2341 __skb_trim(skb, len);
2342 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002343}
2344
2345static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2346{
2347 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2348}
2349
2350/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07002351 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2352 * @skb: buffer to alter
2353 * @len: new length
2354 *
2355 * This is identical to pskb_trim except that the caller knows that
2356 * the skb is not cloned so we should never get an error due to out-
2357 * of-memory.
2358 */
2359static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2360{
2361 int err = pskb_trim(skb, len);
2362 BUG_ON(err);
2363}
2364
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002365static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2366{
2367 unsigned int diff = len - skb->len;
2368
2369 if (skb_tailroom(skb) < diff) {
2370 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2371 GFP_ATOMIC);
2372 if (ret)
2373 return ret;
2374 }
2375 __skb_set_length(skb, len);
2376 return 0;
2377}
2378
Herbert Xue9fa4f72006-08-13 20:12:58 -07002379/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002380 * skb_orphan - orphan a buffer
2381 * @skb: buffer to orphan
2382 *
2383 * If a buffer currently has an owner then we call the owner's
2384 * destructor function and make the @skb unowned. The buffer continues
2385 * to exist but is no longer charged to its former owner.
2386 */
2387static inline void skb_orphan(struct sk_buff *skb)
2388{
Eric Dumazetc34a7612013-07-30 16:11:15 -07002389 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002390 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002391 skb->destructor = NULL;
2392 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07002393 } else {
2394 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002395 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002396}
2397
2398/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002399 * skb_orphan_frags - orphan the frags contained in a buffer
2400 * @skb: buffer to orphan frags from
2401 * @gfp_mask: allocation mask for replacement pages
2402 *
2403 * For each frag in the SKB which needs a destructor (i.e. has an
2404 * owner) create a copy of that frag and release the original
2405 * page by calling the destructor.
2406 */
2407static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2408{
2409 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2410 return 0;
2411 return skb_copy_ubufs(skb, gfp_mask);
2412}
2413
2414/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002415 * __skb_queue_purge - empty a list
2416 * @list: list to empty
2417 *
2418 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2419 * the list and one reference dropped. This function does not take the
2420 * list lock and the caller must hold the relevant locks to use it.
2421 */
Joe Perches7965bd42013-09-26 14:48:15 -07002422void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002423static inline void __skb_queue_purge(struct sk_buff_head *list)
2424{
2425 struct sk_buff *skb;
2426 while ((skb = __skb_dequeue(list)) != NULL)
2427 kfree_skb(skb);
2428}
2429
Yaogong Wang9f5afea2016-09-07 14:49:28 -07002430void skb_rbtree_purge(struct rb_root *root);
2431
Joe Perches7965bd42013-09-26 14:48:15 -07002432void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002433
Joe Perches7965bd42013-09-26 14:48:15 -07002434struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2435 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07002436
2437/**
2438 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2439 * @dev: network device to receive on
2440 * @length: length to allocate
2441 *
2442 * Allocate a new &sk_buff and assign it a usage count of one. The
2443 * buffer has unspecified headroom built in. Users should allocate
2444 * the headroom they think they need without accounting for the
2445 * built in space. The built in space is used for optimisations.
2446 *
2447 * %NULL is returned if there is no free memory. Although this function
2448 * allocates memory it can be called from an interrupt.
2449 */
2450static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002451 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002452{
2453 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2454}
2455
Eric Dumazet6f532612012-05-18 05:12:12 +00002456/* legacy helper around __netdev_alloc_skb() */
2457static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2458 gfp_t gfp_mask)
2459{
2460 return __netdev_alloc_skb(NULL, length, gfp_mask);
2461}
2462
2463/* legacy helper around netdev_alloc_skb() */
2464static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2465{
2466 return netdev_alloc_skb(NULL, length);
2467}
2468
2469
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002470static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2471 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002472{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002473 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002474
2475 if (NET_IP_ALIGN && skb)
2476 skb_reserve(skb, NET_IP_ALIGN);
2477 return skb;
2478}
2479
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002480static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2481 unsigned int length)
2482{
2483 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2484}
2485
Alexander Duyck181edb22015-05-06 21:12:03 -07002486static inline void skb_free_frag(void *addr)
2487{
2488 __free_page_frag(addr);
2489}
2490
Alexander Duyckffde7322014-12-09 19:40:42 -08002491void *napi_alloc_frag(unsigned int fragsz);
Alexander Duyckfd11a832014-12-09 19:40:49 -08002492struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2493 unsigned int length, gfp_t gfp_mask);
2494static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2495 unsigned int length)
2496{
2497 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2498}
Jesper Dangaard Brouer795bb1c2016-02-08 13:14:59 +01002499void napi_consume_skb(struct sk_buff *skb, int budget);
2500
2501void __kfree_skb_flush(void);
Jesper Dangaard Brouer15fad712016-02-08 13:15:04 +01002502void __kfree_skb_defer(struct sk_buff *skb);
Alexander Duyckffde7322014-12-09 19:40:42 -08002503
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002504/**
Alexander Duyck71dfda52014-11-11 09:26:34 -08002505 * __dev_alloc_pages - allocate page for network Rx
2506 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2507 * @order: size of the allocation
2508 *
2509 * Allocate a new page.
2510 *
2511 * %NULL is returned if there is no free memory.
2512*/
2513static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2514 unsigned int order)
2515{
2516 /* This piece of code contains several assumptions.
2517 * 1. This is for device Rx, therefor a cold page is preferred.
2518 * 2. The expectation is the user wants a compound page.
2519 * 3. If requesting a order 0 page it will not be compound
2520 * due to the check to see if order has a value in prep_new_page
2521 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2522 * code in gfp_to_alloc_flags that should be enforcing this.
2523 */
2524 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2525
2526 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2527}
2528
2529static inline struct page *dev_alloc_pages(unsigned int order)
2530{
Neil Horman95829b32016-05-19 11:30:54 -04002531 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002532}
2533
2534/**
2535 * __dev_alloc_page - allocate a page for network Rx
2536 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2537 *
2538 * Allocate a new page.
2539 *
2540 * %NULL is returned if there is no free memory.
2541 */
2542static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2543{
2544 return __dev_alloc_pages(gfp_mask, 0);
2545}
2546
2547static inline struct page *dev_alloc_page(void)
2548{
Neil Horman95829b32016-05-19 11:30:54 -04002549 return dev_alloc_pages(0);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002550}
2551
2552/**
Mel Gorman06140022012-07-31 16:44:24 -07002553 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2554 * @page: The page that was allocated from skb_alloc_page
2555 * @skb: The skb that may need pfmemalloc set
2556 */
2557static inline void skb_propagate_pfmemalloc(struct page *page,
2558 struct sk_buff *skb)
2559{
Michal Hocko2f064f32015-08-21 14:11:51 -07002560 if (page_is_pfmemalloc(page))
Mel Gorman06140022012-07-31 16:44:24 -07002561 skb->pfmemalloc = true;
2562}
2563
Eric Dumazet564824b2010-10-11 19:05:25 +00002564/**
Masanari Iidae2278672014-02-18 22:54:36 +09002565 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002566 * @frag: the paged fragment
2567 *
2568 * Returns the &struct page associated with @frag.
2569 */
2570static inline struct page *skb_frag_page(const skb_frag_t *frag)
2571{
Ian Campbella8605c62011-10-19 23:01:49 +00002572 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00002573}
2574
2575/**
2576 * __skb_frag_ref - take an addition reference on a paged fragment.
2577 * @frag: the paged fragment
2578 *
2579 * Takes an additional reference on the paged fragment @frag.
2580 */
2581static inline void __skb_frag_ref(skb_frag_t *frag)
2582{
2583 get_page(skb_frag_page(frag));
2584}
2585
2586/**
2587 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2588 * @skb: the buffer
2589 * @f: the fragment offset.
2590 *
2591 * Takes an additional reference on the @f'th paged fragment of @skb.
2592 */
2593static inline void skb_frag_ref(struct sk_buff *skb, int f)
2594{
2595 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2596}
2597
2598/**
2599 * __skb_frag_unref - release a reference on a paged fragment.
2600 * @frag: the paged fragment
2601 *
2602 * Releases a reference on the paged fragment @frag.
2603 */
2604static inline void __skb_frag_unref(skb_frag_t *frag)
2605{
2606 put_page(skb_frag_page(frag));
2607}
2608
2609/**
2610 * skb_frag_unref - release a reference on a paged fragment of an skb.
2611 * @skb: the buffer
2612 * @f: the fragment offset
2613 *
2614 * Releases a reference on the @f'th paged fragment of @skb.
2615 */
2616static inline void skb_frag_unref(struct sk_buff *skb, int f)
2617{
2618 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2619}
2620
2621/**
2622 * skb_frag_address - gets the address of the data contained in a paged fragment
2623 * @frag: the paged fragment buffer
2624 *
2625 * Returns the address of the data within @frag. The page must already
2626 * be mapped.
2627 */
2628static inline void *skb_frag_address(const skb_frag_t *frag)
2629{
2630 return page_address(skb_frag_page(frag)) + frag->page_offset;
2631}
2632
2633/**
2634 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2635 * @frag: the paged fragment buffer
2636 *
2637 * Returns the address of the data within @frag. Checks that the page
2638 * is mapped and returns %NULL otherwise.
2639 */
2640static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2641{
2642 void *ptr = page_address(skb_frag_page(frag));
2643 if (unlikely(!ptr))
2644 return NULL;
2645
2646 return ptr + frag->page_offset;
2647}
2648
2649/**
2650 * __skb_frag_set_page - sets the page contained in a paged fragment
2651 * @frag: the paged fragment
2652 * @page: the page to set
2653 *
2654 * Sets the fragment @frag to contain @page.
2655 */
2656static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2657{
Ian Campbella8605c62011-10-19 23:01:49 +00002658 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002659}
2660
2661/**
2662 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2663 * @skb: the buffer
2664 * @f: the fragment offset
2665 * @page: the page to set
2666 *
2667 * Sets the @f'th fragment of @skb to contain @page.
2668 */
2669static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2670 struct page *page)
2671{
2672 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2673}
2674
Eric Dumazet400dfd32013-10-17 16:27:07 -07002675bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2676
Ian Campbell131ea662011-08-19 06:25:00 +00002677/**
2678 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002679 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002680 * @frag: the paged fragment to map
2681 * @offset: the offset within the fragment (starting at the
2682 * fragment's own offset)
2683 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002684 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00002685 *
2686 * Maps the page associated with @frag to @device.
2687 */
2688static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2689 const skb_frag_t *frag,
2690 size_t offset, size_t size,
2691 enum dma_data_direction dir)
2692{
2693 return dma_map_page(dev, skb_frag_page(frag),
2694 frag->page_offset + offset, size, dir);
2695}
2696
Eric Dumazet117632e2011-12-03 21:39:53 +00002697static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2698 gfp_t gfp_mask)
2699{
2700 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2701}
2702
Octavian Purdilabad93e92014-06-12 01:36:26 +03002703
2704static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2705 gfp_t gfp_mask)
2706{
2707 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2708}
2709
2710
Ian Campbell131ea662011-08-19 06:25:00 +00002711/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002712 * skb_clone_writable - is the header of a clone writable
2713 * @skb: buffer to check
2714 * @len: length up to which to write
2715 *
2716 * Returns true if modifying the header part of the cloned buffer
2717 * does not requires the data to be copied.
2718 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002719static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002720{
2721 return !skb_header_cloned(skb) &&
2722 skb_headroom(skb) + len <= skb->hdr_len;
2723}
2724
Daniel Borkmann36976492016-02-19 23:05:25 +01002725static inline int skb_try_make_writable(struct sk_buff *skb,
2726 unsigned int write_len)
2727{
2728 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
2729 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
2730}
2731
Herbert Xud9cc2042007-09-16 16:21:16 -07002732static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2733 int cloned)
2734{
2735 int delta = 0;
2736
Herbert Xud9cc2042007-09-16 16:21:16 -07002737 if (headroom > skb_headroom(skb))
2738 delta = headroom - skb_headroom(skb);
2739
2740 if (delta || cloned)
2741 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2742 GFP_ATOMIC);
2743 return 0;
2744}
2745
Patrick McHardy334a8132007-06-25 04:35:20 -07002746/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002747 * skb_cow - copy header of skb when it is required
2748 * @skb: buffer to cow
2749 * @headroom: needed headroom
2750 *
2751 * If the skb passed lacks sufficient headroom or its data part
2752 * is shared, data is reallocated. If reallocation fails, an error
2753 * is returned and original skb is not changed.
2754 *
2755 * The result is skb with writable area skb->head...skb->tail
2756 * and at least @headroom of space at head.
2757 */
2758static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2759{
Herbert Xud9cc2042007-09-16 16:21:16 -07002760 return __skb_cow(skb, headroom, skb_cloned(skb));
2761}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002762
Herbert Xud9cc2042007-09-16 16:21:16 -07002763/**
2764 * skb_cow_head - skb_cow but only making the head writable
2765 * @skb: buffer to cow
2766 * @headroom: needed headroom
2767 *
2768 * This function is identical to skb_cow except that we replace the
2769 * skb_cloned check by skb_header_cloned. It should be used when
2770 * you only need to push on some header and do not need to modify
2771 * the data.
2772 */
2773static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2774{
2775 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002776}
2777
2778/**
2779 * skb_padto - pad an skbuff up to a minimal size
2780 * @skb: buffer to pad
2781 * @len: minimal length
2782 *
2783 * Pads up a buffer to ensure the trailing bytes exist and are
2784 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002785 * is untouched. Otherwise it is extended. Returns zero on
2786 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002787 */
Herbert Xu5b057c62006-06-23 02:06:41 -07002788static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002789{
2790 unsigned int size = skb->len;
2791 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002792 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002793 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002794}
2795
Alexander Duyck9c0c1122014-12-03 08:17:33 -08002796/**
2797 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2798 * @skb: buffer to pad
2799 * @len: minimal length
2800 *
2801 * Pads up a buffer to ensure the trailing bytes exist and are
2802 * blanked. If the buffer already contains sufficient data it
2803 * is untouched. Otherwise it is extended. Returns zero on
2804 * success. The skb is freed on error.
2805 */
2806static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2807{
2808 unsigned int size = skb->len;
2809
2810 if (unlikely(size < len)) {
2811 len -= size;
2812 if (skb_pad(skb, len))
2813 return -ENOMEM;
2814 __skb_put(skb, len);
2815 }
2816 return 0;
2817}
2818
Linus Torvalds1da177e2005-04-16 15:20:36 -07002819static inline int skb_add_data(struct sk_buff *skb,
Al Viroaf2b0402014-11-27 21:44:24 -05002820 struct iov_iter *from, int copy)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002821{
2822 const int off = skb->len;
2823
2824 if (skb->ip_summed == CHECKSUM_NONE) {
Al Viroaf2b0402014-11-27 21:44:24 -05002825 __wsum csum = 0;
Al Viro15e6cb42016-11-01 22:42:45 -04002826 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
2827 &csum, from)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002828 skb->csum = csum_block_add(skb->csum, csum, off);
2829 return 0;
2830 }
Al Viro15e6cb42016-11-01 22:42:45 -04002831 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002832 return 0;
2833
2834 __skb_trim(skb, off);
2835 return -EFAULT;
2836}
2837
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002838static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2839 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002840{
2841 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002842 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002843
Ian Campbellea2ab692011-08-22 23:44:58 +00002844 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002845 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002846 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002847 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002848}
2849
Herbert Xu364c6ba2006-06-09 16:10:40 -07002850static inline int __skb_linearize(struct sk_buff *skb)
2851{
2852 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2853}
2854
Linus Torvalds1da177e2005-04-16 15:20:36 -07002855/**
2856 * skb_linearize - convert paged skb to linear one
2857 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002858 *
2859 * If there is no free memory -ENOMEM is returned, otherwise zero
2860 * is returned and the old skb data released.
2861 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002862static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002863{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002864 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2865}
2866
2867/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00002868 * skb_has_shared_frag - can any frag be overwritten
2869 * @skb: buffer to test
2870 *
2871 * Return true if the skb has at least one frag that might be modified
2872 * by an external entity (as in vmsplice()/sendfile())
2873 */
2874static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2875{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00002876 return skb_is_nonlinear(skb) &&
2877 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00002878}
2879
2880/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07002881 * skb_linearize_cow - make sure skb is linear and writable
2882 * @skb: buffer to process
2883 *
2884 * If there is no free memory -ENOMEM is returned, otherwise zero
2885 * is returned and the old skb data released.
2886 */
2887static inline int skb_linearize_cow(struct sk_buff *skb)
2888{
2889 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2890 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002891}
2892
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02002893static __always_inline void
2894__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
2895 unsigned int off)
2896{
2897 if (skb->ip_summed == CHECKSUM_COMPLETE)
2898 skb->csum = csum_block_sub(skb->csum,
2899 csum_partial(start, len, 0), off);
2900 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
2901 skb_checksum_start_offset(skb) < 0)
2902 skb->ip_summed = CHECKSUM_NONE;
2903}
2904
Linus Torvalds1da177e2005-04-16 15:20:36 -07002905/**
2906 * skb_postpull_rcsum - update checksum for received skb after pull
2907 * @skb: buffer to update
2908 * @start: start of data before pull
2909 * @len: length of data pulled
2910 *
2911 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002912 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2913 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002914 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002915static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002916 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002917{
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02002918 __skb_postpull_rcsum(skb, start, len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002919}
2920
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02002921static __always_inline void
2922__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
2923 unsigned int off)
2924{
2925 if (skb->ip_summed == CHECKSUM_COMPLETE)
2926 skb->csum = csum_block_add(skb->csum,
2927 csum_partial(start, len, 0), off);
2928}
Herbert Xucbb042f2006-03-20 22:43:56 -08002929
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02002930/**
2931 * skb_postpush_rcsum - update checksum for received skb after push
2932 * @skb: buffer to update
2933 * @start: start of data after push
2934 * @len: length of data pushed
2935 *
2936 * After doing a push on a received packet, you need to call this to
2937 * update the CHECKSUM_COMPLETE checksum.
2938 */
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01002939static inline void skb_postpush_rcsum(struct sk_buff *skb,
2940 const void *start, unsigned int len)
2941{
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02002942 __skb_postpush_rcsum(skb, start, len, 0);
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01002943}
2944
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02002945unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2946
David S. Miller7ce5a272013-12-02 17:26:05 -05002947/**
WANG Cong82a31b92016-06-30 10:15:22 -07002948 * skb_push_rcsum - push skb and update receive checksum
2949 * @skb: buffer to update
2950 * @len: length of data pulled
2951 *
2952 * This function performs an skb_push on the packet and updates
2953 * the CHECKSUM_COMPLETE checksum. It should be used on
2954 * receive path processing instead of skb_push unless you know
2955 * that the checksum difference is zero (e.g., a valid IP header)
2956 * or you are setting ip_summed to CHECKSUM_NONE.
2957 */
2958static inline unsigned char *skb_push_rcsum(struct sk_buff *skb,
2959 unsigned int len)
2960{
2961 skb_push(skb, len);
2962 skb_postpush_rcsum(skb, skb->data, len);
2963 return skb->data;
2964}
2965
2966/**
David S. Miller7ce5a272013-12-02 17:26:05 -05002967 * pskb_trim_rcsum - trim received skb and update checksum
2968 * @skb: buffer to trim
2969 * @len: new length
2970 *
2971 * This is exactly the same as pskb_trim except that it ensures the
2972 * checksum of received packets are still valid after the operation.
2973 */
2974
2975static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2976{
2977 if (likely(len >= skb->len))
2978 return 0;
2979 if (skb->ip_summed == CHECKSUM_COMPLETE)
2980 skb->ip_summed = CHECKSUM_NONE;
2981 return __pskb_trim(skb, len);
2982}
2983
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002984static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2985{
2986 if (skb->ip_summed == CHECKSUM_COMPLETE)
2987 skb->ip_summed = CHECKSUM_NONE;
2988 __skb_trim(skb, len);
2989 return 0;
2990}
2991
2992static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
2993{
2994 if (skb->ip_summed == CHECKSUM_COMPLETE)
2995 skb->ip_summed = CHECKSUM_NONE;
2996 return __skb_grow(skb, len);
2997}
2998
Linus Torvalds1da177e2005-04-16 15:20:36 -07002999#define skb_queue_walk(queue, skb) \
3000 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003001 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07003002 skb = skb->next)
3003
James Chapman46f89142007-04-30 00:07:31 -07003004#define skb_queue_walk_safe(queue, skb, tmp) \
3005 for (skb = (queue)->next, tmp = skb->next; \
3006 skb != (struct sk_buff *)(queue); \
3007 skb = tmp, tmp = skb->next)
3008
David S. Miller1164f522008-09-23 00:49:44 -07003009#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003010 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07003011 skb = skb->next)
3012
3013#define skb_queue_walk_from_safe(queue, skb, tmp) \
3014 for (tmp = skb->next; \
3015 skb != (struct sk_buff *)(queue); \
3016 skb = tmp, tmp = skb->next)
3017
Stephen Hemminger300ce172005-10-30 13:47:34 -08003018#define skb_queue_reverse_walk(queue, skb) \
3019 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003020 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08003021 skb = skb->prev)
3022
David S. Miller686a2952011-01-20 22:47:32 -08003023#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3024 for (skb = (queue)->prev, tmp = skb->prev; \
3025 skb != (struct sk_buff *)(queue); \
3026 skb = tmp, tmp = skb->prev)
3027
3028#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3029 for (tmp = skb->prev; \
3030 skb != (struct sk_buff *)(queue); \
3031 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003032
David S. Miller21dc3302010-08-23 00:13:46 -07003033static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07003034{
3035 return skb_shinfo(skb)->frag_list != NULL;
3036}
3037
3038static inline void skb_frag_list_init(struct sk_buff *skb)
3039{
3040 skb_shinfo(skb)->frag_list = NULL;
3041}
3042
David S. Milleree039872009-06-09 00:17:13 -07003043#define skb_walk_frags(skb, iter) \
3044 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3045
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003046
3047int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
3048 const struct sk_buff *skb);
3049struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
Paolo Abeni7c13f972016-11-04 11:28:59 +01003050 void (*destructor)(struct sock *sk,
3051 struct sk_buff *skb),
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003052 int *peeked, int *off, int *err,
3053 struct sk_buff **last);
Joe Perches7965bd42013-09-26 14:48:15 -07003054struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
Paolo Abeni7c13f972016-11-04 11:28:59 +01003055 void (*destructor)(struct sock *sk,
3056 struct sk_buff *skb),
Joe Perches7965bd42013-09-26 14:48:15 -07003057 int *peeked, int *off, int *err);
3058struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3059 int *err);
3060unsigned int datagram_poll(struct file *file, struct socket *sock,
3061 struct poll_table_struct *wait);
Al Viroc0371da2014-11-24 10:42:55 -05003062int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3063 struct iov_iter *to, int size);
David S. Miller51f3d022014-11-05 16:46:40 -05003064static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3065 struct msghdr *msg, int size)
3066{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003067 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
David S. Miller51f3d022014-11-05 16:46:40 -05003068}
Al Viroe5a4b0b2014-11-24 18:17:55 -05003069int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3070 struct msghdr *msg);
Al Viro3a654f92014-06-19 14:15:22 -04003071int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3072 struct iov_iter *from, int len);
Al Viro3a654f92014-06-19 14:15:22 -04003073int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
Joe Perches7965bd42013-09-26 14:48:15 -07003074void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
samanthakumar627d2d62016-04-05 12:41:16 -04003075void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3076static inline void skb_free_datagram_locked(struct sock *sk,
3077 struct sk_buff *skb)
3078{
3079 __skb_free_datagram_locked(sk, skb, 0);
3080}
Joe Perches7965bd42013-09-26 14:48:15 -07003081int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07003082int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3083int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3084__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3085 int len, __wsum csum);
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +02003086int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
Joe Perches7965bd42013-09-26 14:48:15 -07003087 struct pipe_inode_info *pipe, unsigned int len,
Al Viro25869262016-09-17 21:02:10 -04003088 unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07003089void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01003090unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00003091int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3092 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07003093void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3094int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3095void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Florian Westphalde960aa2014-01-26 10:58:16 +01003096unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
Marcelo Ricardo Leitnerae7ef812016-06-02 15:05:41 -03003097bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu);
Joe Perches7965bd42013-09-26 14:48:15 -07003098struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Vlad Yasevich0d5501c2014-08-08 14:42:13 -04003099struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
Jiri Pirkoe2195122014-11-19 14:05:01 +01003100int skb_ensure_writable(struct sk_buff *skb, int write_len);
Shmulik Ladkanibfca4c52016-09-19 19:11:09 +03003101int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
Jiri Pirko93515d52014-11-19 14:05:02 +01003102int skb_vlan_pop(struct sk_buff *skb);
3103int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
Sowmini Varadhan6fa01cc2016-04-22 18:36:35 -07003104struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3105 gfp_t gfp);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003106
Al Viro6ce8e9c2014-04-06 21:25:44 -04003107static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3108{
Al Viro21226ab2014-11-28 15:48:29 -05003109 return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro6ce8e9c2014-04-06 21:25:44 -04003110}
3111
Al Viro7eab8d92014-04-06 21:51:23 -04003112static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3113{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003114 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro7eab8d92014-04-06 21:51:23 -04003115}
3116
Daniel Borkmann2817a332013-10-30 11:50:51 +01003117struct skb_checksum_ops {
3118 __wsum (*update)(const void *mem, int len, __wsum wsum);
3119 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3120};
3121
3122__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3123 __wsum csum, const struct skb_checksum_ops *ops);
3124__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3125 __wsum csum);
3126
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003127static inline void * __must_check
3128__skb_header_pointer(const struct sk_buff *skb, int offset,
3129 int len, void *data, int hlen, void *buffer)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003130{
Patrick McHardy55820ee2005-07-05 14:08:10 -07003131 if (hlen - offset >= len)
David S. Miller690e36e2014-08-23 12:13:41 -07003132 return data + offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003133
David S. Miller690e36e2014-08-23 12:13:41 -07003134 if (!skb ||
3135 skb_copy_bits(skb, offset, buffer, len) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003136 return NULL;
3137
3138 return buffer;
3139}
3140
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003141static inline void * __must_check
3142skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
David S. Miller690e36e2014-08-23 12:13:41 -07003143{
3144 return __skb_header_pointer(skb, offset, len, skb->data,
3145 skb_headlen(skb), buffer);
3146}
3147
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01003148/**
3149 * skb_needs_linearize - check if we need to linearize a given skb
3150 * depending on the given device features.
3151 * @skb: socket buffer to check
3152 * @features: net device features
3153 *
3154 * Returns true if either:
3155 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3156 * 2. skb is fragmented and the device does not support SG.
3157 */
3158static inline bool skb_needs_linearize(struct sk_buff *skb,
3159 netdev_features_t features)
3160{
3161 return skb_is_nonlinear(skb) &&
3162 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3163 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3164}
3165
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03003166static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3167 void *to,
3168 const unsigned int len)
3169{
3170 memcpy(to, skb->data, len);
3171}
3172
3173static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3174 const int offset, void *to,
3175 const unsigned int len)
3176{
3177 memcpy(to, skb->data + offset, len);
3178}
3179
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03003180static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3181 const void *from,
3182 const unsigned int len)
3183{
3184 memcpy(skb->data, from, len);
3185}
3186
3187static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3188 const int offset,
3189 const void *from,
3190 const unsigned int len)
3191{
3192 memcpy(skb->data + offset, from, len);
3193}
3194
Joe Perches7965bd42013-09-26 14:48:15 -07003195void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003196
Patrick Ohlyac45f602009-02-12 05:03:37 +00003197static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3198{
3199 return skb->tstamp;
3200}
3201
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003202/**
3203 * skb_get_timestamp - get timestamp from a skb
3204 * @skb: skb to get stamp from
3205 * @stamp: pointer to struct timeval to store stamp in
3206 *
3207 * Timestamps are stored in the skb as offsets to a base timestamp.
3208 * This function converts the offset back to a struct timeval and stores
3209 * it in stamp.
3210 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00003211static inline void skb_get_timestamp(const struct sk_buff *skb,
3212 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003213{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003214 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003215}
3216
Patrick Ohlyac45f602009-02-12 05:03:37 +00003217static inline void skb_get_timestampns(const struct sk_buff *skb,
3218 struct timespec *stamp)
3219{
3220 *stamp = ktime_to_timespec(skb->tstamp);
3221}
3222
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003223static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003224{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003225 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003226}
3227
Stephen Hemminger164891a2007-04-23 22:26:16 -07003228static inline ktime_t net_timedelta(ktime_t t)
3229{
3230 return ktime_sub(ktime_get_real(), t);
3231}
3232
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003233static inline ktime_t net_invalid_timestamp(void)
3234{
Thomas Gleixner8b0e1952016-12-25 12:30:41 +01003235 return 0;
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003236}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003237
Alexander Duyck62bccb82014-09-04 13:31:35 -04003238struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3239
Richard Cochranc1f19b52010-07-17 08:49:36 +00003240#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3241
Joe Perches7965bd42013-09-26 14:48:15 -07003242void skb_clone_tx_timestamp(struct sk_buff *skb);
3243bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00003244
3245#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3246
3247static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3248{
3249}
3250
3251static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3252{
3253 return false;
3254}
3255
3256#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3257
3258/**
3259 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3260 *
Richard Cochranda92b192011-10-21 00:49:15 +00003261 * PHY drivers may accept clones of transmitted packets for
3262 * timestamping via their phy_driver.txtstamp method. These drivers
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003263 * must call this function to return the skb back to the stack with a
3264 * timestamp.
Richard Cochranda92b192011-10-21 00:49:15 +00003265 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00003266 * @skb: clone of the the original outgoing packet
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003267 * @hwtstamps: hardware time stamps
Richard Cochranc1f19b52010-07-17 08:49:36 +00003268 *
3269 */
3270void skb_complete_tx_timestamp(struct sk_buff *skb,
3271 struct skb_shared_hwtstamps *hwtstamps);
3272
Willem de Bruijne7fd2882014-08-04 22:11:48 -04003273void __skb_tstamp_tx(struct sk_buff *orig_skb,
3274 struct skb_shared_hwtstamps *hwtstamps,
3275 struct sock *sk, int tstype);
3276
Patrick Ohlyac45f602009-02-12 05:03:37 +00003277/**
3278 * skb_tstamp_tx - queue clone of skb with send time stamps
3279 * @orig_skb: the original outgoing packet
3280 * @hwtstamps: hardware time stamps, may be NULL if not available
3281 *
3282 * If the skb has a socket associated, then this function clones the
3283 * skb (thus sharing the actual data and optional structures), stores
3284 * the optional hardware time stamping information (if non NULL) or
3285 * generates a software time stamp (otherwise), then queues the clone
3286 * to the error queue of the socket. Errors are silently ignored.
3287 */
Joe Perches7965bd42013-09-26 14:48:15 -07003288void skb_tstamp_tx(struct sk_buff *orig_skb,
3289 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00003290
Richard Cochran4507a712010-07-17 08:48:28 +00003291static inline void sw_tx_timestamp(struct sk_buff *skb)
3292{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00003293 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
3294 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00003295 skb_tstamp_tx(skb, NULL);
3296}
3297
3298/**
3299 * skb_tx_timestamp() - Driver hook for transmit timestamping
3300 *
3301 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00003302 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00003303 *
David S. Miller73409f32013-12-27 13:04:33 -05003304 * Specifically, one should make absolutely sure that this function is
3305 * called before TX completion of this packet can trigger. Otherwise
3306 * the packet could potentially already be freed.
3307 *
Richard Cochran4507a712010-07-17 08:48:28 +00003308 * @skb: A socket buffer.
3309 */
3310static inline void skb_tx_timestamp(struct sk_buff *skb)
3311{
Richard Cochranc1f19b52010-07-17 08:49:36 +00003312 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00003313 sw_tx_timestamp(skb);
3314}
3315
Johannes Berg6e3e9392011-11-09 10:15:42 +01003316/**
3317 * skb_complete_wifi_ack - deliver skb with wifi status
3318 *
3319 * @skb: the original outgoing packet
3320 * @acked: ack status
3321 *
3322 */
3323void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3324
Joe Perches7965bd42013-09-26 14:48:15 -07003325__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3326__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003327
Herbert Xu60476372007-04-09 11:59:39 -07003328static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3329{
Tom Herbert6edec0e2015-02-10 16:30:28 -08003330 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3331 skb->csum_valid ||
3332 (skb->ip_summed == CHECKSUM_PARTIAL &&
3333 skb_checksum_start_offset(skb) >= 0));
Herbert Xu60476372007-04-09 11:59:39 -07003334}
3335
Herbert Xufb286bb2005-11-10 13:01:24 -08003336/**
3337 * skb_checksum_complete - Calculate checksum of an entire packet
3338 * @skb: packet to process
3339 *
3340 * This function calculates the checksum over the entire packet plus
3341 * the value of skb->csum. The latter can be used to supply the
3342 * checksum of a pseudo header as used by TCP/UDP. It returns the
3343 * checksum.
3344 *
3345 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3346 * this function can be used to verify that checksum on received
3347 * packets. In that case the function should return zero if the
3348 * checksum is correct. In particular, this function will return zero
3349 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3350 * hardware has already verified the correctness of the checksum.
3351 */
Al Viro4381ca32007-07-15 21:00:11 +01003352static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08003353{
Herbert Xu60476372007-04-09 11:59:39 -07003354 return skb_csum_unnecessary(skb) ?
3355 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003356}
3357
Tom Herbert77cffe22014-08-27 21:26:46 -07003358static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3359{
3360 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3361 if (skb->csum_level == 0)
3362 skb->ip_summed = CHECKSUM_NONE;
3363 else
3364 skb->csum_level--;
3365 }
3366}
3367
3368static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3369{
3370 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3371 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3372 skb->csum_level++;
3373 } else if (skb->ip_summed == CHECKSUM_NONE) {
3374 skb->ip_summed = CHECKSUM_UNNECESSARY;
3375 skb->csum_level = 0;
3376 }
3377}
3378
Tom Herbert5a212322014-08-31 15:12:41 -07003379static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
3380{
3381 /* Mark current checksum as bad (typically called from GRO
3382 * path). In the case that ip_summed is CHECKSUM_NONE
3383 * this must be the first checksum encountered in the packet.
3384 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
3385 * checksum after the last one validated. For UDP, a zero
3386 * checksum can not be marked as bad.
3387 */
3388
3389 if (skb->ip_summed == CHECKSUM_NONE ||
3390 skb->ip_summed == CHECKSUM_UNNECESSARY)
3391 skb->csum_bad = 1;
3392}
3393
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003394/* Check if we need to perform checksum complete validation.
3395 *
3396 * Returns true if checksum complete is needed, false otherwise
3397 * (either checksum is unnecessary or zero checksum is allowed).
3398 */
3399static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3400 bool zero_okay,
3401 __sum16 check)
3402{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003403 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3404 skb->csum_valid = 1;
Tom Herbert77cffe22014-08-27 21:26:46 -07003405 __skb_decr_checksum_unnecessary(skb);
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003406 return false;
3407 }
3408
3409 return true;
3410}
3411
3412/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3413 * in checksum_init.
3414 */
3415#define CHECKSUM_BREAK 76
3416
Tom Herbert4e18b9a2015-04-20 14:10:04 -07003417/* Unset checksum-complete
3418 *
3419 * Unset checksum complete can be done when packet is being modified
3420 * (uncompressed for instance) and checksum-complete value is
3421 * invalidated.
3422 */
3423static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3424{
3425 if (skb->ip_summed == CHECKSUM_COMPLETE)
3426 skb->ip_summed = CHECKSUM_NONE;
3427}
3428
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003429/* Validate (init) checksum based on checksum complete.
3430 *
3431 * Return values:
3432 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3433 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3434 * checksum is stored in skb->csum for use in __skb_checksum_complete
3435 * non-zero: value of invalid checksum
3436 *
3437 */
3438static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3439 bool complete,
3440 __wsum psum)
3441{
3442 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3443 if (!csum_fold(csum_add(psum, skb->csum))) {
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003444 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003445 return 0;
3446 }
Tom Herbert5a212322014-08-31 15:12:41 -07003447 } else if (skb->csum_bad) {
3448 /* ip_summed == CHECKSUM_NONE in this case */
Eric Dumazetc91d4602015-05-15 05:48:07 -07003449 return (__force __sum16)1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003450 }
3451
3452 skb->csum = psum;
3453
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003454 if (complete || skb->len <= CHECKSUM_BREAK) {
3455 __sum16 csum;
3456
3457 csum = __skb_checksum_complete(skb);
3458 skb->csum_valid = !csum;
3459 return csum;
3460 }
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003461
3462 return 0;
3463}
3464
3465static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3466{
3467 return 0;
3468}
3469
3470/* Perform checksum validate (init). Note that this is a macro since we only
3471 * want to calculate the pseudo header which is an input function if necessary.
3472 * First we try to validate without any computation (checksum unnecessary) and
3473 * then calculate based on checksum complete calling the function to compute
3474 * pseudo header.
3475 *
3476 * Return values:
3477 * 0: checksum is validated or try to in skb_checksum_complete
3478 * non-zero: value of invalid checksum
3479 */
3480#define __skb_checksum_validate(skb, proto, complete, \
3481 zero_okay, check, compute_pseudo) \
3482({ \
3483 __sum16 __ret = 0; \
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003484 skb->csum_valid = 0; \
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003485 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3486 __ret = __skb_checksum_validate_complete(skb, \
3487 complete, compute_pseudo(skb, proto)); \
3488 __ret; \
3489})
3490
3491#define skb_checksum_init(skb, proto, compute_pseudo) \
3492 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3493
3494#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3495 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3496
3497#define skb_checksum_validate(skb, proto, compute_pseudo) \
3498 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3499
3500#define skb_checksum_validate_zero_check(skb, proto, check, \
3501 compute_pseudo) \
Sabrina Dubroca096a4cf2015-02-06 18:54:19 +01003502 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003503
3504#define skb_checksum_simple_validate(skb) \
3505 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3506
Tom Herbertd96535a2014-08-31 15:12:42 -07003507static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3508{
3509 return (skb->ip_summed == CHECKSUM_NONE &&
3510 skb->csum_valid && !skb->csum_bad);
3511}
3512
3513static inline void __skb_checksum_convert(struct sk_buff *skb,
3514 __sum16 check, __wsum pseudo)
3515{
3516 skb->csum = ~pseudo;
3517 skb->ip_summed = CHECKSUM_COMPLETE;
3518}
3519
3520#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3521do { \
3522 if (__skb_checksum_convert_check(skb)) \
3523 __skb_checksum_convert(skb, check, \
3524 compute_pseudo(skb, proto)); \
3525} while (0)
3526
Tom Herbert15e23962015-02-10 16:30:31 -08003527static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3528 u16 start, u16 offset)
3529{
3530 skb->ip_summed = CHECKSUM_PARTIAL;
3531 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3532 skb->csum_offset = offset - start;
3533}
3534
Tom Herbertdcdc8992015-02-02 16:07:34 -08003535/* Update skbuf and packet to reflect the remote checksum offload operation.
3536 * When called, ptr indicates the starting point for skb->csum when
3537 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3538 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3539 */
3540static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
Tom Herbert15e23962015-02-10 16:30:31 -08003541 int start, int offset, bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08003542{
3543 __wsum delta;
3544
Tom Herbert15e23962015-02-10 16:30:31 -08003545 if (!nopartial) {
3546 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3547 return;
3548 }
3549
Tom Herbertdcdc8992015-02-02 16:07:34 -08003550 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3551 __skb_checksum_complete(skb);
3552 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3553 }
3554
3555 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3556
3557 /* Adjust skb->csum since we changed the packet */
3558 skb->csum = csum_add(skb->csum, delta);
3559}
3560
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003561#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07003562void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003563static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3564{
3565 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07003566 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003567}
3568static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3569{
3570 if (nfct)
3571 atomic_inc(&nfct->use);
3572}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003573#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003574#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003575static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3576{
3577 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3578 kfree(nf_bridge);
3579}
3580static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3581{
3582 if (nf_bridge)
3583 atomic_inc(&nf_bridge->use);
3584}
3585#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08003586static inline void nf_reset(struct sk_buff *skb)
3587{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003588#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08003589 nf_conntrack_put(skb->nfct);
3590 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003591#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003592#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Patrick McHardya193a4a2006-03-20 19:23:05 -08003593 nf_bridge_put(skb->nf_bridge);
3594 skb->nf_bridge = NULL;
3595#endif
3596}
3597
Patrick McHardy124dff02013-04-05 20:42:05 +02003598static inline void nf_reset_trace(struct sk_buff *skb)
3599{
Florian Westphal478b3602014-02-15 23:48:45 +01003600#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00003601 skb->nf_trace = 0;
3602#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003603}
3604
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003605/* Note: This doesn't put any conntrack and bridge info in dst. */
Eric Dumazetb1937222014-09-28 22:18:47 -07003606static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3607 bool copy)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003608{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003609#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003610 dst->nfct = src->nfct;
3611 nf_conntrack_get(src->nfct);
Eric Dumazetb1937222014-09-28 22:18:47 -07003612 if (copy)
3613 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003614#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003615#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003616 dst->nf_bridge = src->nf_bridge;
3617 nf_bridge_get(src->nf_bridge);
3618#endif
Florian Westphal478b3602014-02-15 23:48:45 +01003619#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Eric Dumazetb1937222014-09-28 22:18:47 -07003620 if (copy)
3621 dst->nf_trace = src->nf_trace;
Florian Westphal478b3602014-02-15 23:48:45 +01003622#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003623}
3624
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003625static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3626{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003627#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003628 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003629#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003630#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003631 nf_bridge_put(dst->nf_bridge);
3632#endif
Eric Dumazetb1937222014-09-28 22:18:47 -07003633 __nf_copy(dst, src, true);
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003634}
3635
James Morris984bc162006-06-09 00:29:17 -07003636#ifdef CONFIG_NETWORK_SECMARK
3637static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3638{
3639 to->secmark = from->secmark;
3640}
3641
3642static inline void skb_init_secmark(struct sk_buff *skb)
3643{
3644 skb->secmark = 0;
3645}
3646#else
3647static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3648{ }
3649
3650static inline void skb_init_secmark(struct sk_buff *skb)
3651{ }
3652#endif
3653
Eric W. Biederman574f7192014-04-01 12:20:24 -07003654static inline bool skb_irq_freeable(const struct sk_buff *skb)
3655{
3656 return !skb->destructor &&
3657#if IS_ENABLED(CONFIG_XFRM)
3658 !skb->sp &&
3659#endif
3660#if IS_ENABLED(CONFIG_NF_CONNTRACK)
3661 !skb->nfct &&
3662#endif
3663 !skb->_skb_refdst &&
3664 !skb_has_frag_list(skb);
3665}
3666
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003667static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3668{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003669 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003670}
3671
Stephen Hemminger92477442009-03-21 13:39:26 -07003672static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003673{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003674 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003675}
3676
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003677static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3678{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003679 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003680}
3681
David S. Millerd5a9e242009-01-27 16:22:11 -08003682static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3683{
3684 skb->queue_mapping = rx_queue + 1;
3685}
3686
Stephen Hemminger92477442009-03-21 13:39:26 -07003687static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003688{
3689 return skb->queue_mapping - 1;
3690}
3691
Stephen Hemminger92477442009-03-21 13:39:26 -07003692static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003693{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003694 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08003695}
3696
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003697static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3698{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003699#ifdef CONFIG_XFRM
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003700 return skb->sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003701#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003702 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003703#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003704}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003705
Pravin B Shelar68c33162013-02-14 14:02:41 +00003706/* Keeps track of mac header offset relative to skb->head.
3707 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3708 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07003709 * tunnel skb it points to outer mac header.
3710 * Keeps track of level of encapsulation of network headers.
3711 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00003712struct skb_gso_cb {
Alexander Duyck802ab552016-04-10 21:45:03 -04003713 union {
3714 int mac_offset;
3715 int data_offset;
3716 };
Eric Dumazet3347c962013-10-19 11:42:56 -07003717 int encap_level;
Alexander Duyck76443452016-02-05 15:27:37 -08003718 __wsum csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003719 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00003720};
Konstantin Khlebnikov9207f9d2016-01-08 15:21:46 +03003721#define SKB_SGO_CB_OFFSET 32
3722#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
Pravin B Shelar68c33162013-02-14 14:02:41 +00003723
3724static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3725{
3726 return (skb_mac_header(inner_skb) - inner_skb->head) -
3727 SKB_GSO_CB(inner_skb)->mac_offset;
3728}
3729
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00003730static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3731{
3732 int new_headroom, headroom;
3733 int ret;
3734
3735 headroom = skb_headroom(skb);
3736 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3737 if (ret)
3738 return ret;
3739
3740 new_headroom = skb_headroom(skb);
3741 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3742 return 0;
3743}
3744
Alexander Duyck08b64fc2016-02-05 15:27:49 -08003745static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
3746{
3747 /* Do not update partial checksums if remote checksum is enabled. */
3748 if (skb->remcsum_offload)
3749 return;
3750
3751 SKB_GSO_CB(skb)->csum = res;
3752 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
3753}
3754
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003755/* Compute the checksum for a gso segment. First compute the checksum value
3756 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3757 * then add in skb->csum (checksum from csum_start to end of packet).
3758 * skb->csum and csum_start are then updated to reflect the checksum of the
3759 * resultant packet starting from the transport header-- the resultant checksum
3760 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3761 * header.
3762 */
3763static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3764{
Alexander Duyck76443452016-02-05 15:27:37 -08003765 unsigned char *csum_start = skb_transport_header(skb);
3766 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
3767 __wsum partial = SKB_GSO_CB(skb)->csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003768
Alexander Duyck76443452016-02-05 15:27:37 -08003769 SKB_GSO_CB(skb)->csum = res;
3770 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003771
Alexander Duyck76443452016-02-05 15:27:37 -08003772 return csum_fold(csum_partial(csum_start, plen, partial));
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003773}
3774
David S. Millerbdcc0922012-03-07 20:53:36 -05003775static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07003776{
3777 return skb_shinfo(skb)->gso_size;
3778}
3779
Eric Dumazet36a8f392013-09-29 01:21:32 -07003780/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05003781static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02003782{
3783 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3784}
3785
Daniel Borkmann5293efe2016-08-18 01:00:39 +02003786static inline void skb_gso_reset(struct sk_buff *skb)
3787{
3788 skb_shinfo(skb)->gso_size = 0;
3789 skb_shinfo(skb)->gso_segs = 0;
3790 skb_shinfo(skb)->gso_type = 0;
3791}
3792
Joe Perches7965bd42013-09-26 14:48:15 -07003793void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07003794
3795static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3796{
3797 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3798 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003799 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3800
Alexander Duyckb78462e2010-06-02 12:24:37 +00003801 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3802 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07003803 __skb_warn_lro_forwarding(skb);
3804 return true;
3805 }
3806 return false;
3807}
3808
Herbert Xu35fc92a2007-03-26 23:22:20 -07003809static inline void skb_forward_csum(struct sk_buff *skb)
3810{
3811 /* Unfortunately we don't support this one. Any brave souls? */
3812 if (skb->ip_summed == CHECKSUM_COMPLETE)
3813 skb->ip_summed = CHECKSUM_NONE;
3814}
3815
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003816/**
3817 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3818 * @skb: skb to check
3819 *
3820 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3821 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3822 * use this helper, to document places where we make this assertion.
3823 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003824static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003825{
3826#ifdef DEBUG
3827 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3828#endif
3829}
3830
Rusty Russellf35d9d82008-02-04 23:49:54 -05003831bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00003832
Paul Durranted1f50c2014-01-09 10:02:46 +00003833int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
Linus Lüssing9afd85c2015-05-02 14:01:07 +02003834struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
3835 unsigned int transport_len,
3836 __sum16(*skb_chkf)(struct sk_buff *skb));
Paul Durranted1f50c2014-01-09 10:02:46 +00003837
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00003838/**
3839 * skb_head_is_locked - Determine if the skb->head is locked down
3840 * @skb: skb to check
3841 *
3842 * The head on skbs build around a head frag can be removed if they are
3843 * not cloned. This function returns true if the skb head is locked down
3844 * due to either being allocated via kmalloc, or by being a clone with
3845 * multiple references to the head.
3846 */
3847static inline bool skb_head_is_locked(const struct sk_buff *skb)
3848{
3849 return !skb->head_frag || skb_cloned(skb);
3850}
Florian Westphalfe6cc552014-02-13 23:09:12 +01003851
3852/**
3853 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3854 *
3855 * @skb: GSO skb
3856 *
3857 * skb_gso_network_seglen is used to determine the real size of the
3858 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3859 *
3860 * The MAC/L2 header is not accounted for.
3861 */
3862static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3863{
3864 unsigned int hdr_len = skb_transport_header(skb) -
3865 skb_network_header(skb);
3866 return hdr_len + skb_gso_transport_seglen(skb);
3867}
Thomas Grafee122c72015-07-21 10:43:58 +02003868
Edward Cree179bc672016-02-11 20:48:04 +00003869/* Local Checksum Offload.
3870 * Compute outer checksum based on the assumption that the
3871 * inner checksum will be offloaded later.
Edward Creee8ae7b02016-02-11 21:03:37 +00003872 * See Documentation/networking/checksum-offloads.txt for
3873 * explanation of how this works.
Edward Cree179bc672016-02-11 20:48:04 +00003874 * Fill in outer checksum adjustment (e.g. with sum of outer
3875 * pseudo-header) before calling.
3876 * Also ensure that inner checksum is in linear data area.
3877 */
3878static inline __wsum lco_csum(struct sk_buff *skb)
3879{
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003880 unsigned char *csum_start = skb_checksum_start(skb);
3881 unsigned char *l4_hdr = skb_transport_header(skb);
3882 __wsum partial;
Edward Cree179bc672016-02-11 20:48:04 +00003883
3884 /* Start with complement of inner checksum adjustment */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003885 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
3886 skb->csum_offset));
3887
Edward Cree179bc672016-02-11 20:48:04 +00003888 /* Add in checksum of our headers (incl. outer checksum
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003889 * adjustment filled in by caller) and return result.
Edward Cree179bc672016-02-11 20:48:04 +00003890 */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003891 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
Edward Cree179bc672016-02-11 20:48:04 +00003892}
3893
Linus Torvalds1da177e2005-04-16 15:20:36 -07003894#endif /* __KERNEL__ */
3895#endif /* _LINUX_SKBUFF_H */