blob: 601258f6e62153f3814e2a38e4c21ad97fc7ee87 [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
Randy Dunlap31729362008-02-18 20:52:13 -0800592 * @peeked: this packet has been seen already, so stats have been
593 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700594 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700595 * @protocol: Packet protocol from driver
596 * @destructor: Destruct function
597 * @nfct: Associated connection, if any
Linus Torvalds1da177e2005-04-16 15:20:36 -0700598 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800599 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700600 * @tc_index: Traffic control index
601 * @tc_verd: traffic control verdict
Tom Herbert61b905d2014-03-24 15:34:47 -0700602 * @hash: the packet hash
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700603 * @queue_mapping: Queue mapping for multiqueue devices
David S. Miller0b725a22014-08-25 15:51:53 -0700604 * @xmit_more: More SKBs are pending for this queue
Randy Dunlap553a5672008-04-20 10:51:01 -0700605 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700606 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Tom Herbert61b905d2014-03-24 15:34:47 -0700607 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
Changli Gao4ca24622011-08-19 07:26:44 -0700608 * ports.
Tom Herberta3b18dd2014-07-01 21:33:17 -0700609 * @sw_hash: indicates hash was computed in software stack
Johannes Berg6e3e9392011-11-09 10:15:42 +0100610 * @wifi_acked_valid: wifi_acked was set
611 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000612 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Eliezer Tamir06021292013-06-10 11:39:50 +0300613 * @napi_id: id of the NAPI struct this skb came from
James Morris984bc162006-06-09 00:29:17 -0700614 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700615 * @mark: Generic packet mark
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000616 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b02008-07-14 22:49:06 -0700617 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000618 * @inner_protocol: Protocol (encapsulation)
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000619 * @inner_transport_header: Inner transport layer header (encapsulation)
620 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000621 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700622 * @transport_header: Transport layer header
623 * @network_header: Network layer header
624 * @mac_header: Link layer header
625 * @tail: Tail pointer
626 * @end: End pointer
627 * @head: Head of buffer
628 * @data: Data head pointer
629 * @truesize: Buffer size
630 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631 */
632
633struct sk_buff {
Eric Dumazet363ec39232014-02-26 14:02:11 -0800634 union {
Eric Dumazet56b17422014-11-03 08:19:53 -0800635 struct {
636 /* These two members must be first. */
637 struct sk_buff *next;
638 struct sk_buff *prev;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000639
Eric Dumazet56b17422014-11-03 08:19:53 -0800640 union {
641 ktime_t tstamp;
642 struct skb_mstamp skb_mstamp;
643 };
644 };
645 struct rb_node rbnode; /* used in netem & tcp stack */
646 };
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000647 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648 struct net_device *dev;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650 /*
651 * This is the control buffer. It is free to use for every
652 * layer. Please put your private variables there. If you
653 * want to keep them across layers you have to do a skb_clone()
654 * first. This is owned by whoever has the skb queued ATM.
655 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000656 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657
Eric Dumazet7fee2262010-05-11 23:19:48 +0000658 unsigned long _skb_refdst;
Eric Dumazetb1937222014-09-28 22:18:47 -0700659 void (*destructor)(struct sk_buff *skb);
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000660#ifdef CONFIG_XFRM
661 struct sec_path *sp;
662#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700663#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
664 struct nf_conntrack *nfct;
665#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +0200666#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Eric Dumazetb1937222014-09-28 22:18:47 -0700667 struct nf_bridge_info *nf_bridge;
668#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700669 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700670 data_len;
671 __u16 mac_len,
672 hdr_len;
Eric Dumazetb1937222014-09-28 22:18:47 -0700673
674 /* Following fields are _not_ copied in __copy_skb_header()
675 * Note that queue_mapping is here mostly to fill a hole.
676 */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200677 kmemcheck_bitfield_begin(flags1);
Eric Dumazetb1937222014-09-28 22:18:47 -0700678 __u16 queue_mapping;
Daniel Borkmann36bbef52016-09-20 00:26:13 +0200679
680/* if you move cloned around you also must adapt those constants */
681#ifdef __BIG_ENDIAN_BITFIELD
682#define CLONED_MASK (1 << 7)
683#else
684#define CLONED_MASK 1
685#endif
686#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
687
688 __u8 __cloned_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700689 __u8 cloned:1,
Harald Welte6869c4d2005-08-09 19:24:19 -0700690 nohdr:1,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800691 fclone:2,
Herbert Xua59322b2007-12-05 01:53:40 -0800692 peeked:1,
Eric Dumazetb1937222014-09-28 22:18:47 -0700693 head_frag:1,
Daniel Borkmann36bbef52016-09-20 00:26:13 +0200694 xmit_more:1,
695 __unused:1; /* one bit hole */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200696 kmemcheck_bitfield_end(flags1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700697
Eric Dumazetb1937222014-09-28 22:18:47 -0700698 /* fields enclosed in headers_start/headers_end are copied
699 * using a single memcpy() in __copy_skb_header()
700 */
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700701 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700702 __u32 headers_start[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700703 /* public: */
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200704
705/* if you move pkt_type around you also must adapt those constants */
706#ifdef __BIG_ENDIAN_BITFIELD
707#define PKT_TYPE_MAX (7 << 5)
708#else
709#define PKT_TYPE_MAX 7
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100710#endif
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200711#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
712
713 __u8 __pkt_type_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700714 __u8 pkt_type:3;
715 __u8 pfmemalloc:1;
716 __u8 ignore_df:1;
717 __u8 nfctinfo:3;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718
Eric Dumazetb1937222014-09-28 22:18:47 -0700719 __u8 nf_trace:1;
720 __u8 ip_summed:2;
721 __u8 ooo_okay:1;
722 __u8 l4_hash:1;
723 __u8 sw_hash:1;
724 __u8 wifi_acked_valid:1;
725 __u8 wifi_acked:1;
726
727 __u8 no_fcs:1;
728 /* Indicates the inner headers are valid in the skbuff. */
729 __u8 encapsulation:1;
730 __u8 encap_hdr_csum:1;
731 __u8 csum_valid:1;
732 __u8 csum_complete_sw:1;
733 __u8 csum_level:2;
734 __u8 csum_bad:1;
735
736#ifdef CONFIG_IPV6_NDISC_NODETYPE
737 __u8 ndisc_nodetype:2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700738#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700739 __u8 ipvs_property:1;
Tom Herbert8bce6d72014-09-29 20:22:29 -0700740 __u8 inner_protocol_type:1;
Tom Herberte585f232014-11-04 09:06:54 -0800741 __u8 remcsum_offload:1;
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200742#ifdef CONFIG_NET_SWITCHDEV
743 __u8 offload_fwd_mark:1;
744#endif
745 /* 2, 4 or 5 bit hole */
Alexander Duyck4031ae62012-01-27 06:22:53 +0000746
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700748 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700749#ifdef CONFIG_NET_CLS_ACT
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700750 __u16 tc_verd; /* traffic control verdict */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700751#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700752#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200753
Eric Dumazetb1937222014-09-28 22:18:47 -0700754 union {
755 __wsum csum;
756 struct {
757 __u16 csum_start;
758 __u16 csum_offset;
759 };
760 };
761 __u32 priority;
762 int skb_iif;
763 __u32 hash;
764 __be16 vlan_proto;
765 __u16 vlan_tci;
Eric Dumazet2bd82482015-02-03 23:48:24 -0800766#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
767 union {
768 unsigned int napi_id;
769 unsigned int sender_cpu;
770 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700771#endif
James Morris984bc162006-06-09 00:29:17 -0700772#ifdef CONFIG_NETWORK_SECMARK
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200773 __u32 secmark;
James Morris984bc162006-06-09 00:29:17 -0700774#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -0700775
Neil Horman3b885782009-10-12 13:26:31 -0700776 union {
777 __u32 mark;
Eric Dumazet16fad692013-03-14 05:40:32 +0000778 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700779 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700780
Tom Herbert8bce6d72014-09-29 20:22:29 -0700781 union {
782 __be16 inner_protocol;
783 __u8 inner_ipproto;
784 };
785
Simon Horman1a37e412013-05-23 21:02:51 +0000786 __u16 inner_transport_header;
787 __u16 inner_network_header;
788 __u16 inner_mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700789
790 __be16 protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000791 __u16 transport_header;
792 __u16 network_header;
793 __u16 mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700794
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700795 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700796 __u32 headers_end[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700797 /* public: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700798
Linus Torvalds1da177e2005-04-16 15:20:36 -0700799 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700800 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700801 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700802 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700803 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700804 unsigned int truesize;
805 atomic_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700806};
807
808#ifdef __KERNEL__
809/*
810 * Handling routines are only of interest to the kernel
811 */
812#include <linux/slab.h>
813
Linus Torvalds1da177e2005-04-16 15:20:36 -0700814
Mel Gormanc93bdd02012-07-31 16:44:19 -0700815#define SKB_ALLOC_FCLONE 0x01
816#define SKB_ALLOC_RX 0x02
Alexander Duyckfd11a832014-12-09 19:40:49 -0800817#define SKB_ALLOC_NAPI 0x04
Mel Gormanc93bdd02012-07-31 16:44:19 -0700818
819/* Returns true if the skb was allocated from PFMEMALLOC reserves */
820static inline bool skb_pfmemalloc(const struct sk_buff *skb)
821{
822 return unlikely(skb->pfmemalloc);
823}
824
Eric Dumazet7fee2262010-05-11 23:19:48 +0000825/*
826 * skb might have a dst pointer attached, refcounted or not.
827 * _skb_refdst low order bit is set if refcount was _not_ taken
828 */
829#define SKB_DST_NOREF 1UL
830#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
831
832/**
833 * skb_dst - returns skb dst_entry
834 * @skb: buffer
835 *
836 * Returns skb dst_entry, regardless of reference taken or not.
837 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000838static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
839{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000840 /* If refdst was not refcounted, check we still are in a
841 * rcu_read_lock section
842 */
843 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
844 !rcu_read_lock_held() &&
845 !rcu_read_lock_bh_held());
846 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000847}
848
Eric Dumazet7fee2262010-05-11 23:19:48 +0000849/**
850 * skb_dst_set - sets skb dst
851 * @skb: buffer
852 * @dst: dst entry
853 *
854 * Sets skb dst, assuming a reference was taken on dst and should
855 * be released by skb_dst_drop()
856 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000857static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
858{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000859 skb->_skb_refdst = (unsigned long)dst;
860}
861
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200862/**
863 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
864 * @skb: buffer
865 * @dst: dst entry
866 *
867 * Sets skb dst, assuming a reference was not taken on dst.
868 * If dst entry is cached, we do not take reference and dst_release
869 * will be avoided by refdst_drop. If dst entry is not cached, we take
870 * reference, so that last dst_release can destroy the dst immediately.
871 */
872static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
873{
Hannes Frederic Sowadbfc4fb2014-12-06 19:19:42 +0100874 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
875 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200876}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000877
878/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300879 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000880 * @skb: buffer
881 */
882static inline bool skb_dst_is_noref(const struct sk_buff *skb)
883{
884 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000885}
886
Eric Dumazet511c3f92009-06-02 05:14:27 +0000887static inline struct rtable *skb_rtable(const struct sk_buff *skb)
888{
Eric Dumazetadf30902009-06-02 05:19:30 +0000889 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000890}
891
Jamal Hadi Salim8b10cab2016-07-02 06:43:14 -0400892/* For mangling skb->pkt_type from user space side from applications
893 * such as nft, tc, etc, we only allow a conservative subset of
894 * possible pkt_types to be set.
895*/
896static inline bool skb_pkt_type_ok(u32 ptype)
897{
898 return ptype <= PACKET_OTHERHOST;
899}
900
Joe Perches7965bd42013-09-26 14:48:15 -0700901void kfree_skb(struct sk_buff *skb);
902void kfree_skb_list(struct sk_buff *segs);
903void skb_tx_error(struct sk_buff *skb);
904void consume_skb(struct sk_buff *skb);
905void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000906extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000907
Joe Perches7965bd42013-09-26 14:48:15 -0700908void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
909bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
910 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000911
Joe Perches7965bd42013-09-26 14:48:15 -0700912struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
913 int node);
Eric Dumazet2ea2f622015-04-24 16:05:01 -0700914struct sk_buff *__build_skb(void *data, unsigned int frag_size);
Joe Perches7965bd42013-09-26 14:48:15 -0700915struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700916static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100917 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700918{
Eric Dumazet564824b2010-10-11 19:05:25 +0000919 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700920}
921
Eric Dumazet2e4e4412014-09-17 04:49:49 -0700922struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
923 unsigned long data_len,
924 int max_page_order,
925 int *errcode,
926 gfp_t gfp_mask);
927
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700928/* Layout of fast clones : [skb1][skb2][fclone_ref] */
929struct sk_buff_fclones {
930 struct sk_buff skb1;
931
932 struct sk_buff skb2;
933
934 atomic_t fclone_ref;
935};
936
937/**
938 * skb_fclone_busy - check if fclone is busy
939 * @skb: buffer
940 *
Masanari Iidabda13fe2015-12-13 16:53:02 +0900941 * Returns true if skb is a fast clone, and its clone is not freed.
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700942 * Some drivers call skb_orphan() in their ndo_start_xmit(),
943 * so we also check that this didnt happen.
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700944 */
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700945static inline bool skb_fclone_busy(const struct sock *sk,
946 const struct sk_buff *skb)
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700947{
948 const struct sk_buff_fclones *fclones;
949
950 fclones = container_of(skb, struct sk_buff_fclones, skb1);
951
952 return skb->fclone == SKB_FCLONE_ORIG &&
Eric Dumazet6ffe75e2014-12-03 17:04:39 -0800953 atomic_read(&fclones->fclone_ref) > 1 &&
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700954 fclones->skb2.sk == sk;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700955}
956
David S. Millerd179cd12005-08-17 14:57:30 -0700957static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100958 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700959{
Mel Gormanc93bdd02012-07-31 16:44:19 -0700960 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700961}
962
Joe Perches7965bd42013-09-26 14:48:15 -0700963struct sk_buff *__alloc_skb_head(gfp_t priority, int node);
Patrick McHardy0ebd0ac2013-04-17 06:46:58 +0000964static inline struct sk_buff *alloc_skb_head(gfp_t priority)
965{
966 return __alloc_skb_head(priority, -1);
967}
968
Joe Perches7965bd42013-09-26 14:48:15 -0700969struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
970int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
971struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
972struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
Octavian Purdilabad93e92014-06-12 01:36:26 +0300973struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
974 gfp_t gfp_mask, bool fclone);
975static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
976 gfp_t gfp_mask)
977{
978 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
979}
Eric Dumazet117632e2011-12-03 21:39:53 +0000980
Joe Perches7965bd42013-09-26 14:48:15 -0700981int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
982struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
983 unsigned int headroom);
984struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
985 int newtailroom, gfp_t priority);
Fan Du25a91d82014-01-18 09:54:23 +0800986int skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
987 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -0700988int skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg, int offset,
989 int len);
990int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
991int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +0000992#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700993
Joe Perches7965bd42013-09-26 14:48:15 -0700994int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
995 int getfrag(void *from, char *to, int offset,
996 int len, int odd, struct sk_buff *skb),
997 void *from, int length);
Ananda Rajue89e9cf2005-10-18 15:46:41 -0700998
Hannes Frederic Sowabe12a1f2015-05-21 16:59:58 +0200999int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1000 int offset, size_t size);
1001
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08001002struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -07001003 __u32 lower_offset;
1004 __u32 upper_offset;
1005 __u32 frag_idx;
1006 __u32 stepped_offset;
1007 struct sk_buff *root_skb;
1008 struct sk_buff *cur_skb;
1009 __u8 *frag_data;
1010};
1011
Joe Perches7965bd42013-09-26 14:48:15 -07001012void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1013 unsigned int to, struct skb_seq_state *st);
1014unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1015 struct skb_seq_state *st);
1016void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -07001017
Joe Perches7965bd42013-09-26 14:48:15 -07001018unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
Bojan Prtvar059a2442015-02-22 11:46:35 +01001019 unsigned int to, struct ts_config *config);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -07001020
Tom Herbert09323cc2013-12-15 22:16:19 -08001021/*
1022 * Packet hash types specify the type of hash in skb_set_hash.
1023 *
1024 * Hash types refer to the protocol layer addresses which are used to
1025 * construct a packet's hash. The hashes are used to differentiate or identify
1026 * flows of the protocol layer for the hash type. Hash types are either
1027 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1028 *
1029 * Properties of hashes:
1030 *
1031 * 1) Two packets in different flows have different hash values
1032 * 2) Two packets in the same flow should have the same hash value
1033 *
1034 * A hash at a higher layer is considered to be more specific. A driver should
1035 * set the most specific hash possible.
1036 *
1037 * A driver cannot indicate a more specific hash than the layer at which a hash
1038 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1039 *
1040 * A driver may indicate a hash level which is less specific than the
1041 * actual layer the hash was computed on. For instance, a hash computed
1042 * at L4 may be considered an L3 hash. This should only be done if the
1043 * driver can't unambiguously determine that the HW computed the hash at
1044 * the higher layer. Note that the "should" in the second property above
1045 * permits this.
1046 */
1047enum pkt_hash_types {
1048 PKT_HASH_TYPE_NONE, /* Undefined type */
1049 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1050 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1051 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1052};
1053
Tom Herbertbcc83832015-09-01 09:24:24 -07001054static inline void skb_clear_hash(struct sk_buff *skb)
1055{
1056 skb->hash = 0;
1057 skb->sw_hash = 0;
1058 skb->l4_hash = 0;
1059}
1060
1061static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1062{
1063 if (!skb->l4_hash)
1064 skb_clear_hash(skb);
1065}
1066
1067static inline void
1068__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1069{
1070 skb->l4_hash = is_l4;
1071 skb->sw_hash = is_sw;
1072 skb->hash = hash;
1073}
1074
Tom Herbert09323cc2013-12-15 22:16:19 -08001075static inline void
1076skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1077{
Tom Herbertbcc83832015-09-01 09:24:24 -07001078 /* Used by drivers to set hash from HW */
1079 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1080}
1081
1082static inline void
1083__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1084{
1085 __skb_set_hash(skb, hash, true, is_l4);
Tom Herbert09323cc2013-12-15 22:16:19 -08001086}
1087
Tom Herberte5276932015-09-01 09:24:23 -07001088void __skb_get_hash(struct sk_buff *skb);
David S. Millereb70db82016-07-01 16:07:50 -04001089u32 __skb_get_hash_symmetric(struct sk_buff *skb);
Tom Herberte5276932015-09-01 09:24:23 -07001090u32 skb_get_poff(const struct sk_buff *skb);
1091u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1092 const struct flow_keys *keys, int hlen);
1093__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1094 void *data, int hlen_proto);
1095
1096static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1097 int thoff, u8 ip_proto)
1098{
1099 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1100}
1101
1102void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1103 const struct flow_dissector_key *key,
1104 unsigned int key_count);
1105
1106bool __skb_flow_dissect(const struct sk_buff *skb,
1107 struct flow_dissector *flow_dissector,
1108 void *target_container,
Tom Herbertcd79a232015-09-01 09:24:27 -07001109 void *data, __be16 proto, int nhoff, int hlen,
1110 unsigned int flags);
Tom Herberte5276932015-09-01 09:24:23 -07001111
1112static inline bool skb_flow_dissect(const struct sk_buff *skb,
1113 struct flow_dissector *flow_dissector,
Tom Herbertcd79a232015-09-01 09:24:27 -07001114 void *target_container, unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001115{
1116 return __skb_flow_dissect(skb, flow_dissector, target_container,
Tom Herbertcd79a232015-09-01 09:24:27 -07001117 NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001118}
1119
1120static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
Tom Herbertcd79a232015-09-01 09:24:27 -07001121 struct flow_keys *flow,
1122 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001123{
1124 memset(flow, 0, sizeof(*flow));
1125 return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
Tom Herbertcd79a232015-09-01 09:24:27 -07001126 NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001127}
1128
1129static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
1130 void *data, __be16 proto,
Tom Herbertcd79a232015-09-01 09:24:27 -07001131 int nhoff, int hlen,
1132 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001133{
1134 memset(flow, 0, sizeof(*flow));
1135 return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
Tom Herbertcd79a232015-09-01 09:24:27 -07001136 data, proto, nhoff, hlen, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001137}
1138
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001139static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001140{
Tom Herberta3b18dd2014-07-01 21:33:17 -07001141 if (!skb->l4_hash && !skb->sw_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001142 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001143
Tom Herbert61b905d2014-03-24 15:34:47 -07001144 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001145}
1146
David S. Miller20a17bf2015-09-01 21:19:17 -07001147__u32 __skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6);
Tom Herbertf70ea012015-07-31 16:52:10 -07001148
David S. Miller20a17bf2015-09-01 21:19:17 -07001149static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
Tom Herbertf70ea012015-07-31 16:52:10 -07001150{
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001151 if (!skb->l4_hash && !skb->sw_hash) {
1152 struct flow_keys keys;
Tom Herbertde4c1f82015-09-01 18:11:04 -07001153 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001154
Tom Herbertde4c1f82015-09-01 18:11:04 -07001155 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001156 }
Tom Herbertf70ea012015-07-31 16:52:10 -07001157
1158 return skb->hash;
1159}
1160
David S. Miller20a17bf2015-09-01 21:19:17 -07001161__u32 __skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl);
Tom Herbertf70ea012015-07-31 16:52:10 -07001162
David S. Miller20a17bf2015-09-01 21:19:17 -07001163static inline __u32 skb_get_hash_flowi4(struct sk_buff *skb, const struct flowi4 *fl4)
Tom Herbertf70ea012015-07-31 16:52:10 -07001164{
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001165 if (!skb->l4_hash && !skb->sw_hash) {
1166 struct flow_keys keys;
Tom Herbertde4c1f82015-09-01 18:11:04 -07001167 __u32 hash = __get_hash_from_flowi4(fl4, &keys);
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001168
Tom Herbertde4c1f82015-09-01 18:11:04 -07001169 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001170 }
Tom Herbertf70ea012015-07-31 16:52:10 -07001171
1172 return skb->hash;
1173}
1174
Tom Herbert50fb7992015-05-01 11:30:12 -07001175__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
1176
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001177static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1178{
Tom Herbert61b905d2014-03-24 15:34:47 -07001179 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001180}
1181
Tom Herbert3df7a742013-12-15 22:16:29 -08001182static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1183{
Tom Herbert61b905d2014-03-24 15:34:47 -07001184 to->hash = from->hash;
Tom Herberta3b18dd2014-07-01 21:33:17 -07001185 to->sw_hash = from->sw_hash;
Tom Herbert61b905d2014-03-24 15:34:47 -07001186 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -08001187};
1188
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001189#ifdef NET_SKBUFF_DATA_USES_OFFSET
1190static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1191{
1192 return skb->head + skb->end;
1193}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001194
1195static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1196{
1197 return skb->end;
1198}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001199#else
1200static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1201{
1202 return skb->end;
1203}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001204
1205static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1206{
1207 return skb->end - skb->head;
1208}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001209#endif
1210
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001212#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001213
Patrick Ohlyac45f602009-02-12 05:03:37 +00001214static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1215{
1216 return &skb_shinfo(skb)->hwtstamps;
1217}
1218
Linus Torvalds1da177e2005-04-16 15:20:36 -07001219/**
1220 * skb_queue_empty - check if a queue is empty
1221 * @list: queue head
1222 *
1223 * Returns true if the queue is empty, false otherwise.
1224 */
1225static inline int skb_queue_empty(const struct sk_buff_head *list)
1226{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001227 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001228}
1229
1230/**
David S. Millerfc7ebb22008-09-23 00:34:07 -07001231 * skb_queue_is_last - check if skb is the last entry in the queue
1232 * @list: queue head
1233 * @skb: buffer
1234 *
1235 * Returns true if @skb is the last buffer on the list.
1236 */
1237static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1238 const struct sk_buff *skb)
1239{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001240 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -07001241}
1242
1243/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001244 * skb_queue_is_first - check if skb is the first entry in the queue
1245 * @list: queue head
1246 * @skb: buffer
1247 *
1248 * Returns true if @skb is the first buffer on the list.
1249 */
1250static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1251 const struct sk_buff *skb)
1252{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001253 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001254}
1255
1256/**
David S. Miller249c8b42008-09-23 00:44:42 -07001257 * skb_queue_next - return the next packet in the queue
1258 * @list: queue head
1259 * @skb: current buffer
1260 *
1261 * Return the next packet in @list after @skb. It is only valid to
1262 * call this if skb_queue_is_last() evaluates to false.
1263 */
1264static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1265 const struct sk_buff *skb)
1266{
1267 /* This BUG_ON may seem severe, but if we just return then we
1268 * are going to dereference garbage.
1269 */
1270 BUG_ON(skb_queue_is_last(list, skb));
1271 return skb->next;
1272}
1273
1274/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001275 * skb_queue_prev - return the prev packet in the queue
1276 * @list: queue head
1277 * @skb: current buffer
1278 *
1279 * Return the prev packet in @list before @skb. It is only valid to
1280 * call this if skb_queue_is_first() evaluates to false.
1281 */
1282static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1283 const struct sk_buff *skb)
1284{
1285 /* This BUG_ON may seem severe, but if we just return then we
1286 * are going to dereference garbage.
1287 */
1288 BUG_ON(skb_queue_is_first(list, skb));
1289 return skb->prev;
1290}
1291
1292/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001293 * skb_get - reference buffer
1294 * @skb: buffer to reference
1295 *
1296 * Makes another reference to a socket buffer and returns a pointer
1297 * to the buffer.
1298 */
1299static inline struct sk_buff *skb_get(struct sk_buff *skb)
1300{
1301 atomic_inc(&skb->users);
1302 return skb;
1303}
1304
1305/*
1306 * If users == 1, we are the only owner and are can avoid redundant
1307 * atomic change.
1308 */
1309
1310/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001311 * skb_cloned - is the buffer a clone
1312 * @skb: buffer to check
1313 *
1314 * Returns true if the buffer was generated with skb_clone() and is
1315 * one of multiple shared copies of the buffer. Cloned buffers are
1316 * shared data so must not be written to under normal circumstances.
1317 */
1318static inline int skb_cloned(const struct sk_buff *skb)
1319{
1320 return skb->cloned &&
1321 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1322}
1323
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001324static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1325{
Mel Gormand0164ad2015-11-06 16:28:21 -08001326 might_sleep_if(gfpflags_allow_blocking(pri));
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001327
1328 if (skb_cloned(skb))
1329 return pskb_expand_head(skb, 0, 0, pri);
1330
1331 return 0;
1332}
1333
Linus Torvalds1da177e2005-04-16 15:20:36 -07001334/**
1335 * skb_header_cloned - is the header a clone
1336 * @skb: buffer to check
1337 *
1338 * Returns true if modifying the header part of the buffer requires
1339 * the data to be copied.
1340 */
1341static inline int skb_header_cloned(const struct sk_buff *skb)
1342{
1343 int dataref;
1344
1345 if (!skb->cloned)
1346 return 0;
1347
1348 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1349 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1350 return dataref != 1;
1351}
1352
Eric Dumazet9580bf22016-04-30 10:19:29 -07001353static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1354{
1355 might_sleep_if(gfpflags_allow_blocking(pri));
1356
1357 if (skb_header_cloned(skb))
1358 return pskb_expand_head(skb, 0, 0, pri);
1359
1360 return 0;
1361}
1362
Linus Torvalds1da177e2005-04-16 15:20:36 -07001363/**
1364 * skb_header_release - release reference to header
1365 * @skb: buffer to operate on
1366 *
1367 * Drop a reference to the header part of the buffer. This is done
1368 * by acquiring a payload reference. You must not read from the header
1369 * part of skb->data after this.
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001370 * Note : Check if you can use __skb_header_release() instead.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001371 */
1372static inline void skb_header_release(struct sk_buff *skb)
1373{
1374 BUG_ON(skb->nohdr);
1375 skb->nohdr = 1;
1376 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1377}
1378
1379/**
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001380 * __skb_header_release - release reference to header
1381 * @skb: buffer to operate on
1382 *
1383 * Variant of skb_header_release() assuming skb is private to caller.
1384 * We can avoid one atomic operation.
1385 */
1386static inline void __skb_header_release(struct sk_buff *skb)
1387{
1388 skb->nohdr = 1;
1389 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1390}
1391
1392
1393/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001394 * skb_shared - is the buffer shared
1395 * @skb: buffer to check
1396 *
1397 * Returns true if more than one person has a reference to this
1398 * buffer.
1399 */
1400static inline int skb_shared(const struct sk_buff *skb)
1401{
1402 return atomic_read(&skb->users) != 1;
1403}
1404
1405/**
1406 * skb_share_check - check if buffer is shared and if so clone it
1407 * @skb: buffer to check
1408 * @pri: priority for memory allocation
1409 *
1410 * If the buffer is shared the buffer is cloned and the old copy
1411 * drops a reference. A new clone with a single reference is returned.
1412 * If the buffer is not shared the original buffer is returned. When
1413 * being called from interrupt status or with spinlocks held pri must
1414 * be GFP_ATOMIC.
1415 *
1416 * NULL is returned on a memory allocation failure.
1417 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001418static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001419{
Mel Gormand0164ad2015-11-06 16:28:21 -08001420 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001421 if (skb_shared(skb)) {
1422 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001423
1424 if (likely(nskb))
1425 consume_skb(skb);
1426 else
1427 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001428 skb = nskb;
1429 }
1430 return skb;
1431}
1432
1433/*
1434 * Copy shared buffers into a new sk_buff. We effectively do COW on
1435 * packets to handle cases where we have a local reader and forward
1436 * and a couple of other messy ones. The normal one is tcpdumping
1437 * a packet thats being forwarded.
1438 */
1439
1440/**
1441 * skb_unshare - make a copy of a shared buffer
1442 * @skb: buffer to check
1443 * @pri: priority for memory allocation
1444 *
1445 * If the socket buffer is a clone then this function creates a new
1446 * copy of the data, drops a reference count on the old copy and returns
1447 * the new copy with the reference count at 1. If the buffer is not a clone
1448 * the original buffer is returned. When called with a spinlock held or
1449 * from interrupt state @pri must be %GFP_ATOMIC
1450 *
1451 * %NULL is returned on a memory allocation failure.
1452 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001453static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001454 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001455{
Mel Gormand0164ad2015-11-06 16:28:21 -08001456 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001457 if (skb_cloned(skb)) {
1458 struct sk_buff *nskb = skb_copy(skb, pri);
Alexander Aring31eff812014-10-10 23:10:47 +02001459
1460 /* Free our shared copy */
1461 if (likely(nskb))
1462 consume_skb(skb);
1463 else
1464 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001465 skb = nskb;
1466 }
1467 return skb;
1468}
1469
1470/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001471 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001472 * @list_: list to peek at
1473 *
1474 * Peek an &sk_buff. Unlike most other operations you _MUST_
1475 * be careful with this one. A peek leaves the buffer on the
1476 * list and someone else may run off with it. You must hold
1477 * the appropriate locks or have a private queue to do this.
1478 *
1479 * Returns %NULL for an empty list or a pointer to the head element.
1480 * The reference count is not incremented and the reference is therefore
1481 * volatile. Use with caution.
1482 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001483static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001484{
Eric Dumazet18d07002012-04-30 16:31:46 +00001485 struct sk_buff *skb = list_->next;
1486
1487 if (skb == (struct sk_buff *)list_)
1488 skb = NULL;
1489 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001490}
1491
1492/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001493 * skb_peek_next - peek skb following the given one from a queue
1494 * @skb: skb to start from
1495 * @list_: list to peek at
1496 *
1497 * Returns %NULL when the end of the list is met or a pointer to the
1498 * next element. The reference count is not incremented and the
1499 * reference is therefore volatile. Use with caution.
1500 */
1501static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1502 const struct sk_buff_head *list_)
1503{
1504 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001505
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001506 if (next == (struct sk_buff *)list_)
1507 next = NULL;
1508 return next;
1509}
1510
1511/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001512 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001513 * @list_: list to peek at
1514 *
1515 * Peek an &sk_buff. Unlike most other operations you _MUST_
1516 * be careful with this one. A peek leaves the buffer on the
1517 * list and someone else may run off with it. You must hold
1518 * the appropriate locks or have a private queue to do this.
1519 *
1520 * Returns %NULL for an empty list or a pointer to the tail element.
1521 * The reference count is not incremented and the reference is therefore
1522 * volatile. Use with caution.
1523 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001524static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001525{
Eric Dumazet18d07002012-04-30 16:31:46 +00001526 struct sk_buff *skb = list_->prev;
1527
1528 if (skb == (struct sk_buff *)list_)
1529 skb = NULL;
1530 return skb;
1531
Linus Torvalds1da177e2005-04-16 15:20:36 -07001532}
1533
1534/**
1535 * skb_queue_len - get queue length
1536 * @list_: list to measure
1537 *
1538 * Return the length of an &sk_buff queue.
1539 */
1540static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1541{
1542 return list_->qlen;
1543}
1544
David S. Miller67fed452008-09-21 22:36:24 -07001545/**
1546 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1547 * @list: queue to initialize
1548 *
1549 * This initializes only the list and queue length aspects of
1550 * an sk_buff_head object. This allows to initialize the list
1551 * aspects of an sk_buff_head without reinitializing things like
1552 * the spinlock. It can also be used for on-stack sk_buff_head
1553 * objects where the spinlock is known to not be used.
1554 */
1555static inline void __skb_queue_head_init(struct sk_buff_head *list)
1556{
1557 list->prev = list->next = (struct sk_buff *)list;
1558 list->qlen = 0;
1559}
1560
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001561/*
1562 * This function creates a split out lock class for each invocation;
1563 * this is needed for now since a whole lot of users of the skb-queue
1564 * infrastructure in drivers have different locking usage (in hardirq)
1565 * than the networking core (in softirq only). In the long run either the
1566 * network layer or drivers should need annotation to consolidate the
1567 * main types of usage into 3 classes.
1568 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001569static inline void skb_queue_head_init(struct sk_buff_head *list)
1570{
1571 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001572 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001573}
1574
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001575static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1576 struct lock_class_key *class)
1577{
1578 skb_queue_head_init(list);
1579 lockdep_set_class(&list->lock, class);
1580}
1581
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001583 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001584 *
1585 * The "__skb_xxxx()" functions are the non-atomic ones that
1586 * can only be called with interrupts disabled.
1587 */
Joe Perches7965bd42013-09-26 14:48:15 -07001588void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1589 struct sk_buff_head *list);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001590static inline void __skb_insert(struct sk_buff *newsk,
1591 struct sk_buff *prev, struct sk_buff *next,
1592 struct sk_buff_head *list)
1593{
1594 newsk->next = next;
1595 newsk->prev = prev;
1596 next->prev = prev->next = newsk;
1597 list->qlen++;
1598}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001599
David S. Miller67fed452008-09-21 22:36:24 -07001600static inline void __skb_queue_splice(const struct sk_buff_head *list,
1601 struct sk_buff *prev,
1602 struct sk_buff *next)
1603{
1604 struct sk_buff *first = list->next;
1605 struct sk_buff *last = list->prev;
1606
1607 first->prev = prev;
1608 prev->next = first;
1609
1610 last->next = next;
1611 next->prev = last;
1612}
1613
1614/**
1615 * skb_queue_splice - join two skb lists, this is designed for stacks
1616 * @list: the new list to add
1617 * @head: the place to add it in the first list
1618 */
1619static inline void skb_queue_splice(const struct sk_buff_head *list,
1620 struct sk_buff_head *head)
1621{
1622 if (!skb_queue_empty(list)) {
1623 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001624 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001625 }
1626}
1627
1628/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001629 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001630 * @list: the new list to add
1631 * @head: the place to add it in the first list
1632 *
1633 * The list at @list is reinitialised
1634 */
1635static inline void skb_queue_splice_init(struct sk_buff_head *list,
1636 struct sk_buff_head *head)
1637{
1638 if (!skb_queue_empty(list)) {
1639 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001640 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001641 __skb_queue_head_init(list);
1642 }
1643}
1644
1645/**
1646 * skb_queue_splice_tail - join two skb lists, each list being a queue
1647 * @list: the new list to add
1648 * @head: the place to add it in the first list
1649 */
1650static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1651 struct sk_buff_head *head)
1652{
1653 if (!skb_queue_empty(list)) {
1654 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001655 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001656 }
1657}
1658
1659/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001660 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001661 * @list: the new list to add
1662 * @head: the place to add it in the first list
1663 *
1664 * Each of the lists is a queue.
1665 * The list at @list is reinitialised
1666 */
1667static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1668 struct sk_buff_head *head)
1669{
1670 if (!skb_queue_empty(list)) {
1671 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001672 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001673 __skb_queue_head_init(list);
1674 }
1675}
1676
Linus Torvalds1da177e2005-04-16 15:20:36 -07001677/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001678 * __skb_queue_after - queue a buffer at the list head
1679 * @list: list to use
1680 * @prev: place after this buffer
1681 * @newsk: buffer to queue
1682 *
1683 * Queue a buffer int the middle of a list. This function takes no locks
1684 * and you must therefore hold required locks before calling it.
1685 *
1686 * A buffer cannot be placed on two lists at the same time.
1687 */
1688static inline void __skb_queue_after(struct sk_buff_head *list,
1689 struct sk_buff *prev,
1690 struct sk_buff *newsk)
1691{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001692 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001693}
1694
Joe Perches7965bd42013-09-26 14:48:15 -07001695void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1696 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001697
Gerrit Renkerf5572852008-04-14 00:05:28 -07001698static inline void __skb_queue_before(struct sk_buff_head *list,
1699 struct sk_buff *next,
1700 struct sk_buff *newsk)
1701{
1702 __skb_insert(newsk, next->prev, next, list);
1703}
1704
Stephen Hemminger300ce172005-10-30 13:47:34 -08001705/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001706 * __skb_queue_head - queue a buffer at the list head
1707 * @list: list to use
1708 * @newsk: buffer to queue
1709 *
1710 * Queue a buffer at the start of a list. This function takes no locks
1711 * and you must therefore hold required locks before calling it.
1712 *
1713 * A buffer cannot be placed on two lists at the same time.
1714 */
Joe Perches7965bd42013-09-26 14:48:15 -07001715void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001716static inline void __skb_queue_head(struct sk_buff_head *list,
1717 struct sk_buff *newsk)
1718{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001719 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001720}
1721
1722/**
1723 * __skb_queue_tail - queue a buffer at the list tail
1724 * @list: list to use
1725 * @newsk: buffer to queue
1726 *
1727 * Queue a buffer at the end of a list. This function takes no locks
1728 * and you must therefore hold required locks before calling it.
1729 *
1730 * A buffer cannot be placed on two lists at the same time.
1731 */
Joe Perches7965bd42013-09-26 14:48:15 -07001732void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733static inline void __skb_queue_tail(struct sk_buff_head *list,
1734 struct sk_buff *newsk)
1735{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001736 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001737}
1738
Linus Torvalds1da177e2005-04-16 15:20:36 -07001739/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001740 * remove sk_buff from list. _Must_ be called atomically, and with
1741 * the list known..
1742 */
Joe Perches7965bd42013-09-26 14:48:15 -07001743void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001744static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1745{
1746 struct sk_buff *next, *prev;
1747
1748 list->qlen--;
1749 next = skb->next;
1750 prev = skb->prev;
1751 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001752 next->prev = prev;
1753 prev->next = next;
1754}
1755
Gerrit Renkerf525c062008-04-14 00:04:12 -07001756/**
1757 * __skb_dequeue - remove from the head of the queue
1758 * @list: list to dequeue from
1759 *
1760 * Remove the head of the list. This function does not take any locks
1761 * so must be used with appropriate locks held only. The head item is
1762 * returned or %NULL if the list is empty.
1763 */
Joe Perches7965bd42013-09-26 14:48:15 -07001764struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Gerrit Renkerf525c062008-04-14 00:04:12 -07001765static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1766{
1767 struct sk_buff *skb = skb_peek(list);
1768 if (skb)
1769 __skb_unlink(skb, list);
1770 return skb;
1771}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001772
1773/**
1774 * __skb_dequeue_tail - remove from the tail of the queue
1775 * @list: list to dequeue from
1776 *
1777 * Remove the tail of the list. This function does not take any locks
1778 * so must be used with appropriate locks held only. The tail item is
1779 * returned or %NULL if the list is empty.
1780 */
Joe Perches7965bd42013-09-26 14:48:15 -07001781struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001782static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1783{
1784 struct sk_buff *skb = skb_peek_tail(list);
1785 if (skb)
1786 __skb_unlink(skb, list);
1787 return skb;
1788}
1789
1790
David S. Millerbdcc0922012-03-07 20:53:36 -05001791static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001792{
1793 return skb->data_len;
1794}
1795
1796static inline unsigned int skb_headlen(const struct sk_buff *skb)
1797{
1798 return skb->len - skb->data_len;
1799}
1800
1801static inline int skb_pagelen(const struct sk_buff *skb)
1802{
1803 int i, len = 0;
1804
1805 for (i = (int)skb_shinfo(skb)->nr_frags - 1; i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001806 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001807 return len + skb_headlen(skb);
1808}
1809
Ian Campbell131ea662011-08-19 06:25:00 +00001810/**
1811 * __skb_fill_page_desc - initialise a paged fragment in an skb
1812 * @skb: buffer containing fragment to be initialised
1813 * @i: paged fragment index to initialise
1814 * @page: the page to use for this fragment
1815 * @off: the offset to the data with @page
1816 * @size: the length of the data
1817 *
1818 * Initialises the @i'th fragment of @skb to point to &size bytes at
1819 * offset @off within @page.
1820 *
1821 * Does not take any additional reference on the fragment.
1822 */
1823static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1824 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001825{
1826 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1827
Mel Gormanc48a11c2012-07-31 16:44:23 -07001828 /*
Michal Hocko2f064f32015-08-21 14:11:51 -07001829 * Propagate page pfmemalloc to the skb if we can. The problem is
1830 * that not all callers have unique ownership of the page but rely
1831 * on page_is_pfmemalloc doing the right thing(tm).
Mel Gormanc48a11c2012-07-31 16:44:23 -07001832 */
Ian Campbella8605c62011-10-19 23:01:49 +00001833 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001834 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001835 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001836
1837 page = compound_head(page);
Michal Hocko2f064f32015-08-21 14:11:51 -07001838 if (page_is_pfmemalloc(page))
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001839 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00001840}
1841
1842/**
1843 * skb_fill_page_desc - initialise a paged fragment in an skb
1844 * @skb: buffer containing fragment to be initialised
1845 * @i: paged fragment index to initialise
1846 * @page: the page to use for this fragment
1847 * @off: the offset to the data with @page
1848 * @size: the length of the data
1849 *
1850 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01001851 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00001852 * addition updates @skb such that @i is the last fragment.
1853 *
1854 * Does not take any additional reference on the fragment.
1855 */
1856static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1857 struct page *page, int off, int size)
1858{
1859 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001860 skb_shinfo(skb)->nr_frags = i + 1;
1861}
1862
Joe Perches7965bd42013-09-26 14:48:15 -07001863void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1864 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001865
Jason Wangf8e617e2013-11-01 14:07:47 +08001866void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1867 unsigned int truesize);
1868
Linus Torvalds1da177e2005-04-16 15:20:36 -07001869#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001870#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001871#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1872
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001873#ifdef NET_SKBUFF_DATA_USES_OFFSET
1874static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1875{
1876 return skb->head + skb->tail;
1877}
1878
1879static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1880{
1881 skb->tail = skb->data - skb->head;
1882}
1883
1884static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1885{
1886 skb_reset_tail_pointer(skb);
1887 skb->tail += offset;
1888}
Simon Horman7cc46192013-05-28 20:34:29 +00001889
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001890#else /* NET_SKBUFF_DATA_USES_OFFSET */
1891static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1892{
1893 return skb->tail;
1894}
1895
1896static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1897{
1898 skb->tail = skb->data;
1899}
1900
1901static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1902{
1903 skb->tail = skb->data + offset;
1904}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001905
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001906#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1907
Linus Torvalds1da177e2005-04-16 15:20:36 -07001908/*
1909 * Add data to an sk_buff
1910 */
Mathias Krause0c7ddf32013-11-07 14:18:24 +01001911unsigned char *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001912unsigned char *skb_put(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001913static inline unsigned char *__skb_put(struct sk_buff *skb, unsigned int len)
1914{
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001915 unsigned char *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001916 SKB_LINEAR_ASSERT(skb);
1917 skb->tail += len;
1918 skb->len += len;
1919 return tmp;
1920}
1921
Joe Perches7965bd42013-09-26 14:48:15 -07001922unsigned char *skb_push(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001923static inline unsigned char *__skb_push(struct sk_buff *skb, unsigned int len)
1924{
1925 skb->data -= len;
1926 skb->len += len;
1927 return skb->data;
1928}
1929
Joe Perches7965bd42013-09-26 14:48:15 -07001930unsigned char *skb_pull(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001931static inline unsigned char *__skb_pull(struct sk_buff *skb, unsigned int len)
1932{
1933 skb->len -= len;
1934 BUG_ON(skb->len < skb->data_len);
1935 return skb->data += len;
1936}
1937
David S. Miller47d29642010-05-02 02:21:44 -07001938static inline unsigned char *skb_pull_inline(struct sk_buff *skb, unsigned int len)
1939{
1940 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1941}
1942
Joe Perches7965bd42013-09-26 14:48:15 -07001943unsigned char *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001944
1945static inline unsigned char *__pskb_pull(struct sk_buff *skb, unsigned int len)
1946{
1947 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001948 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001949 return NULL;
1950 skb->len -= len;
1951 return skb->data += len;
1952}
1953
1954static inline unsigned char *pskb_pull(struct sk_buff *skb, unsigned int len)
1955{
1956 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
1957}
1958
1959static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
1960{
1961 if (likely(len <= skb_headlen(skb)))
1962 return 1;
1963 if (unlikely(len > skb->len))
1964 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07001965 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001966}
1967
1968/**
1969 * skb_headroom - bytes at buffer head
1970 * @skb: buffer to check
1971 *
1972 * Return the number of bytes of free space at the head of an &sk_buff.
1973 */
Chuck Leverc2636b42007-10-23 21:07:32 -07001974static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001975{
1976 return skb->data - skb->head;
1977}
1978
1979/**
1980 * skb_tailroom - bytes at buffer end
1981 * @skb: buffer to check
1982 *
1983 * Return the number of bytes of free space at the tail of an sk_buff
1984 */
1985static inline int skb_tailroom(const struct sk_buff *skb)
1986{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001987 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001988}
1989
1990/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00001991 * skb_availroom - bytes at buffer end
1992 * @skb: buffer to check
1993 *
1994 * Return the number of bytes of free space at the tail of an sk_buff
1995 * allocated by sk_stream_alloc()
1996 */
1997static inline int skb_availroom(const struct sk_buff *skb)
1998{
Eric Dumazet16fad692013-03-14 05:40:32 +00001999 if (skb_is_nonlinear(skb))
2000 return 0;
2001
2002 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00002003}
2004
2005/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002006 * skb_reserve - adjust headroom
2007 * @skb: buffer to alter
2008 * @len: bytes to move
2009 *
2010 * Increase the headroom of an empty &sk_buff by reducing the tail
2011 * room. This is only allowed for an empty buffer.
2012 */
David S. Miller8243126c2006-01-17 02:54:21 -08002013static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002014{
2015 skb->data += len;
2016 skb->tail += len;
2017}
2018
Benjamin Poirier1837b2e2016-02-29 15:03:33 -08002019/**
2020 * skb_tailroom_reserve - adjust reserved_tailroom
2021 * @skb: buffer to alter
2022 * @mtu: maximum amount of headlen permitted
2023 * @needed_tailroom: minimum amount of reserved_tailroom
2024 *
2025 * Set reserved_tailroom so that headlen can be as large as possible but
2026 * not larger than mtu and tailroom cannot be smaller than
2027 * needed_tailroom.
2028 * The required headroom should already have been reserved before using
2029 * this function.
2030 */
2031static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2032 unsigned int needed_tailroom)
2033{
2034 SKB_LINEAR_ASSERT(skb);
2035 if (mtu < skb_tailroom(skb) - needed_tailroom)
2036 /* use at most mtu */
2037 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2038 else
2039 /* use up to all available space */
2040 skb->reserved_tailroom = needed_tailroom;
2041}
2042
Tom Herbert8bce6d72014-09-29 20:22:29 -07002043#define ENCAP_TYPE_ETHER 0
2044#define ENCAP_TYPE_IPPROTO 1
2045
2046static inline void skb_set_inner_protocol(struct sk_buff *skb,
2047 __be16 protocol)
2048{
2049 skb->inner_protocol = protocol;
2050 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2051}
2052
2053static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2054 __u8 ipproto)
2055{
2056 skb->inner_ipproto = ipproto;
2057 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2058}
2059
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002060static inline void skb_reset_inner_headers(struct sk_buff *skb)
2061{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002062 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002063 skb->inner_network_header = skb->network_header;
2064 skb->inner_transport_header = skb->transport_header;
2065}
2066
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00002067static inline void skb_reset_mac_len(struct sk_buff *skb)
2068{
2069 skb->mac_len = skb->network_header - skb->mac_header;
2070}
2071
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002072static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2073 *skb)
2074{
2075 return skb->head + skb->inner_transport_header;
2076}
2077
Tom Herbert55dc5a92015-12-14 11:19:40 -08002078static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2079{
2080 return skb_inner_transport_header(skb) - skb->data;
2081}
2082
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002083static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2084{
2085 skb->inner_transport_header = skb->data - skb->head;
2086}
2087
2088static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2089 const int offset)
2090{
2091 skb_reset_inner_transport_header(skb);
2092 skb->inner_transport_header += offset;
2093}
2094
2095static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2096{
2097 return skb->head + skb->inner_network_header;
2098}
2099
2100static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2101{
2102 skb->inner_network_header = skb->data - skb->head;
2103}
2104
2105static inline void skb_set_inner_network_header(struct sk_buff *skb,
2106 const int offset)
2107{
2108 skb_reset_inner_network_header(skb);
2109 skb->inner_network_header += offset;
2110}
2111
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002112static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2113{
2114 return skb->head + skb->inner_mac_header;
2115}
2116
2117static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2118{
2119 skb->inner_mac_header = skb->data - skb->head;
2120}
2121
2122static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2123 const int offset)
2124{
2125 skb_reset_inner_mac_header(skb);
2126 skb->inner_mac_header += offset;
2127}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002128static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2129{
Cong Wang35d04612013-05-29 15:16:05 +08002130 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002131}
2132
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002133static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2134{
2135 return skb->head + skb->transport_header;
2136}
2137
2138static inline void skb_reset_transport_header(struct sk_buff *skb)
2139{
2140 skb->transport_header = skb->data - skb->head;
2141}
2142
2143static inline void skb_set_transport_header(struct sk_buff *skb,
2144 const int offset)
2145{
2146 skb_reset_transport_header(skb);
2147 skb->transport_header += offset;
2148}
2149
2150static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2151{
2152 return skb->head + skb->network_header;
2153}
2154
2155static inline void skb_reset_network_header(struct sk_buff *skb)
2156{
2157 skb->network_header = skb->data - skb->head;
2158}
2159
2160static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2161{
2162 skb_reset_network_header(skb);
2163 skb->network_header += offset;
2164}
2165
2166static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2167{
2168 return skb->head + skb->mac_header;
2169}
2170
2171static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2172{
Cong Wang35d04612013-05-29 15:16:05 +08002173 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002174}
2175
2176static inline void skb_reset_mac_header(struct sk_buff *skb)
2177{
2178 skb->mac_header = skb->data - skb->head;
2179}
2180
2181static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2182{
2183 skb_reset_mac_header(skb);
2184 skb->mac_header += offset;
2185}
2186
Timo Teräs0e3da5b2013-12-16 11:02:09 +02002187static inline void skb_pop_mac_header(struct sk_buff *skb)
2188{
2189 skb->mac_header = skb->network_header;
2190}
2191
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002192static inline void skb_probe_transport_header(struct sk_buff *skb,
2193 const int offset_hint)
2194{
2195 struct flow_keys keys;
2196
2197 if (skb_transport_header_was_set(skb))
2198 return;
Tom Herbertcd79a232015-09-01 09:24:27 -07002199 else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
Tom Herbert42aecaa2015-06-04 09:16:39 -07002200 skb_set_transport_header(skb, keys.control.thoff);
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002201 else
2202 skb_set_transport_header(skb, offset_hint);
2203}
2204
Eric Dumazet03606892012-02-23 10:55:02 +00002205static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2206{
2207 if (skb_mac_header_was_set(skb)) {
2208 const unsigned char *old_mac = skb_mac_header(skb);
2209
2210 skb_set_mac_header(skb, -skb->mac_len);
2211 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2212 }
2213}
2214
Michał Mirosław04fb4512010-12-14 15:24:08 +00002215static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2216{
2217 return skb->csum_start - skb_headroom(skb);
2218}
2219
Alexander Duyck08b64fc2016-02-05 15:27:49 -08002220static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2221{
2222 return skb->head + skb->csum_start;
2223}
2224
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002225static inline int skb_transport_offset(const struct sk_buff *skb)
2226{
2227 return skb_transport_header(skb) - skb->data;
2228}
2229
2230static inline u32 skb_network_header_len(const struct sk_buff *skb)
2231{
2232 return skb->transport_header - skb->network_header;
2233}
2234
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002235static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2236{
2237 return skb->inner_transport_header - skb->inner_network_header;
2238}
2239
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002240static inline int skb_network_offset(const struct sk_buff *skb)
2241{
2242 return skb_network_header(skb) - skb->data;
2243}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03002244
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002245static inline int skb_inner_network_offset(const struct sk_buff *skb)
2246{
2247 return skb_inner_network_header(skb) - skb->data;
2248}
2249
Changli Gaof9599ce2010-08-04 04:43:44 +00002250static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2251{
2252 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2253}
2254
Linus Torvalds1da177e2005-04-16 15:20:36 -07002255/*
2256 * CPUs often take a performance hit when accessing unaligned memory
2257 * locations. The actual performance hit varies, it can be small if the
2258 * hardware handles it or large if we have to take an exception and fix it
2259 * in software.
2260 *
2261 * Since an ethernet header is 14 bytes network drivers often end up with
2262 * the IP header at an unaligned offset. The IP header can be aligned by
2263 * shifting the start of the packet by 2 bytes. Drivers should do this
2264 * with:
2265 *
Tobias Klauser8660c122009-07-13 22:48:16 +00002266 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002267 *
2268 * The downside to this alignment of the IP header is that the DMA is now
2269 * unaligned. On some architectures the cost of an unaligned DMA is high
2270 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00002271 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002272 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2273 * to be overridden.
2274 */
2275#ifndef NET_IP_ALIGN
2276#define NET_IP_ALIGN 2
2277#endif
2278
Anton Blanchard025be812006-03-31 02:27:06 -08002279/*
2280 * The networking layer reserves some headroom in skb data (via
2281 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2282 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08002283 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08002284 *
2285 * Unfortunately this headroom changes the DMA alignment of the resulting
2286 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2287 * on some architectures. An architecture can override this value,
2288 * perhaps setting it to a cacheline in size (since that will maintain
2289 * cacheline alignment of the DMA). It must be a power of 2.
2290 *
David S. Millerd6301d32009-02-08 19:24:13 -08002291 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08002292 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07002293 *
2294 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2295 * to reduce average number of cache lines per packet.
2296 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07002297 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08002298 */
2299#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07002300#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08002301#endif
2302
Joe Perches7965bd42013-09-26 14:48:15 -07002303int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002304
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002305static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002306{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00002307 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07002308 WARN_ON(1);
2309 return;
2310 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002311 skb->len = len;
2312 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002313}
2314
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002315static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2316{
2317 __skb_set_length(skb, len);
2318}
2319
Joe Perches7965bd42013-09-26 14:48:15 -07002320void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002321
2322static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2323{
Herbert Xu3cc0e872006-06-09 16:13:38 -07002324 if (skb->data_len)
2325 return ___pskb_trim(skb, len);
2326 __skb_trim(skb, len);
2327 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002328}
2329
2330static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2331{
2332 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2333}
2334
2335/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07002336 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2337 * @skb: buffer to alter
2338 * @len: new length
2339 *
2340 * This is identical to pskb_trim except that the caller knows that
2341 * the skb is not cloned so we should never get an error due to out-
2342 * of-memory.
2343 */
2344static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2345{
2346 int err = pskb_trim(skb, len);
2347 BUG_ON(err);
2348}
2349
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002350static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2351{
2352 unsigned int diff = len - skb->len;
2353
2354 if (skb_tailroom(skb) < diff) {
2355 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2356 GFP_ATOMIC);
2357 if (ret)
2358 return ret;
2359 }
2360 __skb_set_length(skb, len);
2361 return 0;
2362}
2363
Herbert Xue9fa4f72006-08-13 20:12:58 -07002364/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002365 * skb_orphan - orphan a buffer
2366 * @skb: buffer to orphan
2367 *
2368 * If a buffer currently has an owner then we call the owner's
2369 * destructor function and make the @skb unowned. The buffer continues
2370 * to exist but is no longer charged to its former owner.
2371 */
2372static inline void skb_orphan(struct sk_buff *skb)
2373{
Eric Dumazetc34a7612013-07-30 16:11:15 -07002374 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002375 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002376 skb->destructor = NULL;
2377 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07002378 } else {
2379 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002380 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002381}
2382
2383/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002384 * skb_orphan_frags - orphan the frags contained in a buffer
2385 * @skb: buffer to orphan frags from
2386 * @gfp_mask: allocation mask for replacement pages
2387 *
2388 * For each frag in the SKB which needs a destructor (i.e. has an
2389 * owner) create a copy of that frag and release the original
2390 * page by calling the destructor.
2391 */
2392static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2393{
2394 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2395 return 0;
2396 return skb_copy_ubufs(skb, gfp_mask);
2397}
2398
2399/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002400 * __skb_queue_purge - empty a list
2401 * @list: list to empty
2402 *
2403 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2404 * the list and one reference dropped. This function does not take the
2405 * list lock and the caller must hold the relevant locks to use it.
2406 */
Joe Perches7965bd42013-09-26 14:48:15 -07002407void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002408static inline void __skb_queue_purge(struct sk_buff_head *list)
2409{
2410 struct sk_buff *skb;
2411 while ((skb = __skb_dequeue(list)) != NULL)
2412 kfree_skb(skb);
2413}
2414
Yaogong Wang9f5afea2016-09-07 14:49:28 -07002415void skb_rbtree_purge(struct rb_root *root);
2416
Joe Perches7965bd42013-09-26 14:48:15 -07002417void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002418
Joe Perches7965bd42013-09-26 14:48:15 -07002419struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2420 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07002421
2422/**
2423 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2424 * @dev: network device to receive on
2425 * @length: length to allocate
2426 *
2427 * Allocate a new &sk_buff and assign it a usage count of one. The
2428 * buffer has unspecified headroom built in. Users should allocate
2429 * the headroom they think they need without accounting for the
2430 * built in space. The built in space is used for optimisations.
2431 *
2432 * %NULL is returned if there is no free memory. Although this function
2433 * allocates memory it can be called from an interrupt.
2434 */
2435static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002436 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002437{
2438 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2439}
2440
Eric Dumazet6f532612012-05-18 05:12:12 +00002441/* legacy helper around __netdev_alloc_skb() */
2442static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2443 gfp_t gfp_mask)
2444{
2445 return __netdev_alloc_skb(NULL, length, gfp_mask);
2446}
2447
2448/* legacy helper around netdev_alloc_skb() */
2449static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2450{
2451 return netdev_alloc_skb(NULL, length);
2452}
2453
2454
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002455static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2456 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002457{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002458 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002459
2460 if (NET_IP_ALIGN && skb)
2461 skb_reserve(skb, NET_IP_ALIGN);
2462 return skb;
2463}
2464
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002465static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2466 unsigned int length)
2467{
2468 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2469}
2470
Alexander Duyck181edb22015-05-06 21:12:03 -07002471static inline void skb_free_frag(void *addr)
2472{
2473 __free_page_frag(addr);
2474}
2475
Alexander Duyckffde7322014-12-09 19:40:42 -08002476void *napi_alloc_frag(unsigned int fragsz);
Alexander Duyckfd11a832014-12-09 19:40:49 -08002477struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2478 unsigned int length, gfp_t gfp_mask);
2479static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2480 unsigned int length)
2481{
2482 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2483}
Jesper Dangaard Brouer795bb1c2016-02-08 13:14:59 +01002484void napi_consume_skb(struct sk_buff *skb, int budget);
2485
2486void __kfree_skb_flush(void);
Jesper Dangaard Brouer15fad712016-02-08 13:15:04 +01002487void __kfree_skb_defer(struct sk_buff *skb);
Alexander Duyckffde7322014-12-09 19:40:42 -08002488
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002489/**
Alexander Duyck71dfda52014-11-11 09:26:34 -08002490 * __dev_alloc_pages - allocate page for network Rx
2491 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2492 * @order: size of the allocation
2493 *
2494 * Allocate a new page.
2495 *
2496 * %NULL is returned if there is no free memory.
2497*/
2498static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2499 unsigned int order)
2500{
2501 /* This piece of code contains several assumptions.
2502 * 1. This is for device Rx, therefor a cold page is preferred.
2503 * 2. The expectation is the user wants a compound page.
2504 * 3. If requesting a order 0 page it will not be compound
2505 * due to the check to see if order has a value in prep_new_page
2506 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2507 * code in gfp_to_alloc_flags that should be enforcing this.
2508 */
2509 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2510
2511 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2512}
2513
2514static inline struct page *dev_alloc_pages(unsigned int order)
2515{
Neil Horman95829b32016-05-19 11:30:54 -04002516 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002517}
2518
2519/**
2520 * __dev_alloc_page - allocate a page for network Rx
2521 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2522 *
2523 * Allocate a new page.
2524 *
2525 * %NULL is returned if there is no free memory.
2526 */
2527static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2528{
2529 return __dev_alloc_pages(gfp_mask, 0);
2530}
2531
2532static inline struct page *dev_alloc_page(void)
2533{
Neil Horman95829b32016-05-19 11:30:54 -04002534 return dev_alloc_pages(0);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002535}
2536
2537/**
Mel Gorman06140022012-07-31 16:44:24 -07002538 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2539 * @page: The page that was allocated from skb_alloc_page
2540 * @skb: The skb that may need pfmemalloc set
2541 */
2542static inline void skb_propagate_pfmemalloc(struct page *page,
2543 struct sk_buff *skb)
2544{
Michal Hocko2f064f32015-08-21 14:11:51 -07002545 if (page_is_pfmemalloc(page))
Mel Gorman06140022012-07-31 16:44:24 -07002546 skb->pfmemalloc = true;
2547}
2548
Eric Dumazet564824b2010-10-11 19:05:25 +00002549/**
Masanari Iidae2278672014-02-18 22:54:36 +09002550 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002551 * @frag: the paged fragment
2552 *
2553 * Returns the &struct page associated with @frag.
2554 */
2555static inline struct page *skb_frag_page(const skb_frag_t *frag)
2556{
Ian Campbella8605c62011-10-19 23:01:49 +00002557 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00002558}
2559
2560/**
2561 * __skb_frag_ref - take an addition reference on a paged fragment.
2562 * @frag: the paged fragment
2563 *
2564 * Takes an additional reference on the paged fragment @frag.
2565 */
2566static inline void __skb_frag_ref(skb_frag_t *frag)
2567{
2568 get_page(skb_frag_page(frag));
2569}
2570
2571/**
2572 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2573 * @skb: the buffer
2574 * @f: the fragment offset.
2575 *
2576 * Takes an additional reference on the @f'th paged fragment of @skb.
2577 */
2578static inline void skb_frag_ref(struct sk_buff *skb, int f)
2579{
2580 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2581}
2582
2583/**
2584 * __skb_frag_unref - release a reference on a paged fragment.
2585 * @frag: the paged fragment
2586 *
2587 * Releases a reference on the paged fragment @frag.
2588 */
2589static inline void __skb_frag_unref(skb_frag_t *frag)
2590{
2591 put_page(skb_frag_page(frag));
2592}
2593
2594/**
2595 * skb_frag_unref - release a reference on a paged fragment of an skb.
2596 * @skb: the buffer
2597 * @f: the fragment offset
2598 *
2599 * Releases a reference on the @f'th paged fragment of @skb.
2600 */
2601static inline void skb_frag_unref(struct sk_buff *skb, int f)
2602{
2603 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2604}
2605
2606/**
2607 * skb_frag_address - gets the address of the data contained in a paged fragment
2608 * @frag: the paged fragment buffer
2609 *
2610 * Returns the address of the data within @frag. The page must already
2611 * be mapped.
2612 */
2613static inline void *skb_frag_address(const skb_frag_t *frag)
2614{
2615 return page_address(skb_frag_page(frag)) + frag->page_offset;
2616}
2617
2618/**
2619 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2620 * @frag: the paged fragment buffer
2621 *
2622 * Returns the address of the data within @frag. Checks that the page
2623 * is mapped and returns %NULL otherwise.
2624 */
2625static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2626{
2627 void *ptr = page_address(skb_frag_page(frag));
2628 if (unlikely(!ptr))
2629 return NULL;
2630
2631 return ptr + frag->page_offset;
2632}
2633
2634/**
2635 * __skb_frag_set_page - sets the page contained in a paged fragment
2636 * @frag: the paged fragment
2637 * @page: the page to set
2638 *
2639 * Sets the fragment @frag to contain @page.
2640 */
2641static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2642{
Ian Campbella8605c62011-10-19 23:01:49 +00002643 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002644}
2645
2646/**
2647 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2648 * @skb: the buffer
2649 * @f: the fragment offset
2650 * @page: the page to set
2651 *
2652 * Sets the @f'th fragment of @skb to contain @page.
2653 */
2654static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2655 struct page *page)
2656{
2657 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2658}
2659
Eric Dumazet400dfd32013-10-17 16:27:07 -07002660bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2661
Ian Campbell131ea662011-08-19 06:25:00 +00002662/**
2663 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002664 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002665 * @frag: the paged fragment to map
2666 * @offset: the offset within the fragment (starting at the
2667 * fragment's own offset)
2668 * @size: the number of bytes to map
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002669 * @dir: the direction of the mapping (%PCI_DMA_*)
Ian Campbell131ea662011-08-19 06:25:00 +00002670 *
2671 * Maps the page associated with @frag to @device.
2672 */
2673static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2674 const skb_frag_t *frag,
2675 size_t offset, size_t size,
2676 enum dma_data_direction dir)
2677{
2678 return dma_map_page(dev, skb_frag_page(frag),
2679 frag->page_offset + offset, size, dir);
2680}
2681
Eric Dumazet117632e2011-12-03 21:39:53 +00002682static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2683 gfp_t gfp_mask)
2684{
2685 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2686}
2687
Octavian Purdilabad93e92014-06-12 01:36:26 +03002688
2689static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2690 gfp_t gfp_mask)
2691{
2692 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2693}
2694
2695
Ian Campbell131ea662011-08-19 06:25:00 +00002696/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002697 * skb_clone_writable - is the header of a clone writable
2698 * @skb: buffer to check
2699 * @len: length up to which to write
2700 *
2701 * Returns true if modifying the header part of the cloned buffer
2702 * does not requires the data to be copied.
2703 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002704static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002705{
2706 return !skb_header_cloned(skb) &&
2707 skb_headroom(skb) + len <= skb->hdr_len;
2708}
2709
Daniel Borkmann36976492016-02-19 23:05:25 +01002710static inline int skb_try_make_writable(struct sk_buff *skb,
2711 unsigned int write_len)
2712{
2713 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
2714 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
2715}
2716
Herbert Xud9cc2042007-09-16 16:21:16 -07002717static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2718 int cloned)
2719{
2720 int delta = 0;
2721
Herbert Xud9cc2042007-09-16 16:21:16 -07002722 if (headroom > skb_headroom(skb))
2723 delta = headroom - skb_headroom(skb);
2724
2725 if (delta || cloned)
2726 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2727 GFP_ATOMIC);
2728 return 0;
2729}
2730
Patrick McHardy334a8132007-06-25 04:35:20 -07002731/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002732 * skb_cow - copy header of skb when it is required
2733 * @skb: buffer to cow
2734 * @headroom: needed headroom
2735 *
2736 * If the skb passed lacks sufficient headroom or its data part
2737 * is shared, data is reallocated. If reallocation fails, an error
2738 * is returned and original skb is not changed.
2739 *
2740 * The result is skb with writable area skb->head...skb->tail
2741 * and at least @headroom of space at head.
2742 */
2743static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2744{
Herbert Xud9cc2042007-09-16 16:21:16 -07002745 return __skb_cow(skb, headroom, skb_cloned(skb));
2746}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002747
Herbert Xud9cc2042007-09-16 16:21:16 -07002748/**
2749 * skb_cow_head - skb_cow but only making the head writable
2750 * @skb: buffer to cow
2751 * @headroom: needed headroom
2752 *
2753 * This function is identical to skb_cow except that we replace the
2754 * skb_cloned check by skb_header_cloned. It should be used when
2755 * you only need to push on some header and do not need to modify
2756 * the data.
2757 */
2758static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2759{
2760 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002761}
2762
2763/**
2764 * skb_padto - pad an skbuff up to a minimal size
2765 * @skb: buffer to pad
2766 * @len: minimal length
2767 *
2768 * Pads up a buffer to ensure the trailing bytes exist and are
2769 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002770 * is untouched. Otherwise it is extended. Returns zero on
2771 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002772 */
Herbert Xu5b057c62006-06-23 02:06:41 -07002773static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002774{
2775 unsigned int size = skb->len;
2776 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002777 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002778 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002779}
2780
Alexander Duyck9c0c1122014-12-03 08:17:33 -08002781/**
2782 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2783 * @skb: buffer to pad
2784 * @len: minimal length
2785 *
2786 * Pads up a buffer to ensure the trailing bytes exist and are
2787 * blanked. If the buffer already contains sufficient data it
2788 * is untouched. Otherwise it is extended. Returns zero on
2789 * success. The skb is freed on error.
2790 */
2791static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2792{
2793 unsigned int size = skb->len;
2794
2795 if (unlikely(size < len)) {
2796 len -= size;
2797 if (skb_pad(skb, len))
2798 return -ENOMEM;
2799 __skb_put(skb, len);
2800 }
2801 return 0;
2802}
2803
Linus Torvalds1da177e2005-04-16 15:20:36 -07002804static inline int skb_add_data(struct sk_buff *skb,
Al Viroaf2b0402014-11-27 21:44:24 -05002805 struct iov_iter *from, int copy)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002806{
2807 const int off = skb->len;
2808
2809 if (skb->ip_summed == CHECKSUM_NONE) {
Al Viroaf2b0402014-11-27 21:44:24 -05002810 __wsum csum = 0;
2811 if (csum_and_copy_from_iter(skb_put(skb, copy), copy,
2812 &csum, from) == copy) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002813 skb->csum = csum_block_add(skb->csum, csum, off);
2814 return 0;
2815 }
Al Viroaf2b0402014-11-27 21:44:24 -05002816 } else if (copy_from_iter(skb_put(skb, copy), copy, from) == copy)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002817 return 0;
2818
2819 __skb_trim(skb, off);
2820 return -EFAULT;
2821}
2822
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002823static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2824 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002825{
2826 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002827 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002828
Ian Campbellea2ab692011-08-22 23:44:58 +00002829 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002830 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002831 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002832 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002833}
2834
Herbert Xu364c6ba2006-06-09 16:10:40 -07002835static inline int __skb_linearize(struct sk_buff *skb)
2836{
2837 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2838}
2839
Linus Torvalds1da177e2005-04-16 15:20:36 -07002840/**
2841 * skb_linearize - convert paged skb to linear one
2842 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002843 *
2844 * If there is no free memory -ENOMEM is returned, otherwise zero
2845 * is returned and the old skb data released.
2846 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002847static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002848{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002849 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2850}
2851
2852/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00002853 * skb_has_shared_frag - can any frag be overwritten
2854 * @skb: buffer to test
2855 *
2856 * Return true if the skb has at least one frag that might be modified
2857 * by an external entity (as in vmsplice()/sendfile())
2858 */
2859static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2860{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00002861 return skb_is_nonlinear(skb) &&
2862 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00002863}
2864
2865/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07002866 * skb_linearize_cow - make sure skb is linear and writable
2867 * @skb: buffer to process
2868 *
2869 * If there is no free memory -ENOMEM is returned, otherwise zero
2870 * is returned and the old skb data released.
2871 */
2872static inline int skb_linearize_cow(struct sk_buff *skb)
2873{
2874 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2875 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002876}
2877
Daniel Borkmann479ffcc2016-08-05 00:11:12 +02002878static __always_inline void
2879__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
2880 unsigned int off)
2881{
2882 if (skb->ip_summed == CHECKSUM_COMPLETE)
2883 skb->csum = csum_block_sub(skb->csum,
2884 csum_partial(start, len, 0), off);
2885 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
2886 skb_checksum_start_offset(skb) < 0)
2887 skb->ip_summed = CHECKSUM_NONE;
2888}
2889
Linus Torvalds1da177e2005-04-16 15:20:36 -07002890/**
2891 * skb_postpull_rcsum - update checksum for received skb after pull
2892 * @skb: buffer to update
2893 * @start: start of data before pull
2894 * @len: length of data pulled
2895 *
2896 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002897 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2898 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002899 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002900static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002901 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002902{
Daniel Borkmann479ffcc2016-08-05 00:11:12 +02002903 __skb_postpull_rcsum(skb, start, len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002904}
2905
Daniel Borkmann479ffcc2016-08-05 00:11:12 +02002906static __always_inline void
2907__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
2908 unsigned int off)
2909{
2910 if (skb->ip_summed == CHECKSUM_COMPLETE)
2911 skb->csum = csum_block_add(skb->csum,
2912 csum_partial(start, len, 0), off);
2913}
Herbert Xucbb042f2006-03-20 22:43:56 -08002914
Daniel Borkmann479ffcc2016-08-05 00:11:12 +02002915/**
2916 * skb_postpush_rcsum - update checksum for received skb after push
2917 * @skb: buffer to update
2918 * @start: start of data after push
2919 * @len: length of data pushed
2920 *
2921 * After doing a push on a received packet, you need to call this to
2922 * update the CHECKSUM_COMPLETE checksum.
2923 */
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01002924static inline void skb_postpush_rcsum(struct sk_buff *skb,
2925 const void *start, unsigned int len)
2926{
Daniel Borkmann479ffcc2016-08-05 00:11:12 +02002927 __skb_postpush_rcsum(skb, start, len, 0);
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01002928}
2929
Daniel Borkmann479ffcc2016-08-05 00:11:12 +02002930unsigned char *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
2931
David S. Miller7ce5a272013-12-02 17:26:05 -05002932/**
WANG Cong82a31b92016-06-30 10:15:22 -07002933 * skb_push_rcsum - push skb and update receive checksum
2934 * @skb: buffer to update
2935 * @len: length of data pulled
2936 *
2937 * This function performs an skb_push on the packet and updates
2938 * the CHECKSUM_COMPLETE checksum. It should be used on
2939 * receive path processing instead of skb_push unless you know
2940 * that the checksum difference is zero (e.g., a valid IP header)
2941 * or you are setting ip_summed to CHECKSUM_NONE.
2942 */
2943static inline unsigned char *skb_push_rcsum(struct sk_buff *skb,
2944 unsigned int len)
2945{
2946 skb_push(skb, len);
2947 skb_postpush_rcsum(skb, skb->data, len);
2948 return skb->data;
2949}
2950
2951/**
David S. Miller7ce5a272013-12-02 17:26:05 -05002952 * pskb_trim_rcsum - trim received skb and update checksum
2953 * @skb: buffer to trim
2954 * @len: new length
2955 *
2956 * This is exactly the same as pskb_trim except that it ensures the
2957 * checksum of received packets are still valid after the operation.
2958 */
2959
2960static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2961{
2962 if (likely(len >= skb->len))
2963 return 0;
2964 if (skb->ip_summed == CHECKSUM_COMPLETE)
2965 skb->ip_summed = CHECKSUM_NONE;
2966 return __pskb_trim(skb, len);
2967}
2968
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002969static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
2970{
2971 if (skb->ip_summed == CHECKSUM_COMPLETE)
2972 skb->ip_summed = CHECKSUM_NONE;
2973 __skb_trim(skb, len);
2974 return 0;
2975}
2976
2977static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
2978{
2979 if (skb->ip_summed == CHECKSUM_COMPLETE)
2980 skb->ip_summed = CHECKSUM_NONE;
2981 return __skb_grow(skb, len);
2982}
2983
Linus Torvalds1da177e2005-04-16 15:20:36 -07002984#define skb_queue_walk(queue, skb) \
2985 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002986 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07002987 skb = skb->next)
2988
James Chapman46f89142007-04-30 00:07:31 -07002989#define skb_queue_walk_safe(queue, skb, tmp) \
2990 for (skb = (queue)->next, tmp = skb->next; \
2991 skb != (struct sk_buff *)(queue); \
2992 skb = tmp, tmp = skb->next)
2993
David S. Miller1164f522008-09-23 00:49:44 -07002994#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07002995 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07002996 skb = skb->next)
2997
2998#define skb_queue_walk_from_safe(queue, skb, tmp) \
2999 for (tmp = skb->next; \
3000 skb != (struct sk_buff *)(queue); \
3001 skb = tmp, tmp = skb->next)
3002
Stephen Hemminger300ce172005-10-30 13:47:34 -08003003#define skb_queue_reverse_walk(queue, skb) \
3004 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003005 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08003006 skb = skb->prev)
3007
David S. Miller686a2952011-01-20 22:47:32 -08003008#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3009 for (skb = (queue)->prev, tmp = skb->prev; \
3010 skb != (struct sk_buff *)(queue); \
3011 skb = tmp, tmp = skb->prev)
3012
3013#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3014 for (tmp = skb->prev; \
3015 skb != (struct sk_buff *)(queue); \
3016 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003017
David S. Miller21dc3302010-08-23 00:13:46 -07003018static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07003019{
3020 return skb_shinfo(skb)->frag_list != NULL;
3021}
3022
3023static inline void skb_frag_list_init(struct sk_buff *skb)
3024{
3025 skb_shinfo(skb)->frag_list = NULL;
3026}
3027
David S. Milleree039872009-06-09 00:17:13 -07003028#define skb_walk_frags(skb, iter) \
3029 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3030
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003031
3032int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
3033 const struct sk_buff *skb);
3034struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
3035 int *peeked, int *off, int *err,
3036 struct sk_buff **last);
Joe Perches7965bd42013-09-26 14:48:15 -07003037struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
3038 int *peeked, int *off, int *err);
3039struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3040 int *err);
3041unsigned int datagram_poll(struct file *file, struct socket *sock,
3042 struct poll_table_struct *wait);
Al Viroc0371da2014-11-24 10:42:55 -05003043int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3044 struct iov_iter *to, int size);
David S. Miller51f3d022014-11-05 16:46:40 -05003045static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3046 struct msghdr *msg, int size)
3047{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003048 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
David S. Miller51f3d022014-11-05 16:46:40 -05003049}
Al Viroe5a4b0b2014-11-24 18:17:55 -05003050int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3051 struct msghdr *msg);
Al Viro3a654f92014-06-19 14:15:22 -04003052int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3053 struct iov_iter *from, int len);
Al Viro3a654f92014-06-19 14:15:22 -04003054int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
Joe Perches7965bd42013-09-26 14:48:15 -07003055void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
samanthakumar627d2d62016-04-05 12:41:16 -04003056void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3057static inline void skb_free_datagram_locked(struct sock *sk,
3058 struct sk_buff *skb)
3059{
3060 __skb_free_datagram_locked(sk, skb, 0);
3061}
Joe Perches7965bd42013-09-26 14:48:15 -07003062int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07003063int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3064int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3065__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3066 int len, __wsum csum);
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +02003067int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
Joe Perches7965bd42013-09-26 14:48:15 -07003068 struct pipe_inode_info *pipe, unsigned int len,
Al Viro25869262016-09-17 21:02:10 -04003069 unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07003070void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01003071unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00003072int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3073 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07003074void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3075int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3076void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Florian Westphalde960aa2014-01-26 10:58:16 +01003077unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
Marcelo Ricardo Leitnerae7ef812016-06-02 15:05:41 -03003078bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu);
Joe Perches7965bd42013-09-26 14:48:15 -07003079struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Vlad Yasevich0d5501c2014-08-08 14:42:13 -04003080struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
Jiri Pirkoe2195122014-11-19 14:05:01 +01003081int skb_ensure_writable(struct sk_buff *skb, int write_len);
Shmulik Ladkanibfca4c52016-09-19 19:11:09 +03003082int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
Jiri Pirko93515d52014-11-19 14:05:02 +01003083int skb_vlan_pop(struct sk_buff *skb);
3084int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
Sowmini Varadhan6fa01cc2016-04-22 18:36:35 -07003085struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3086 gfp_t gfp);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003087
Al Viro6ce8e9c2014-04-06 21:25:44 -04003088static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3089{
Al Viro21226ab2014-11-28 15:48:29 -05003090 return copy_from_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro6ce8e9c2014-04-06 21:25:44 -04003091}
3092
Al Viro7eab8d92014-04-06 21:51:23 -04003093static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3094{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003095 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro7eab8d92014-04-06 21:51:23 -04003096}
3097
Daniel Borkmann2817a332013-10-30 11:50:51 +01003098struct skb_checksum_ops {
3099 __wsum (*update)(const void *mem, int len, __wsum wsum);
3100 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3101};
3102
3103__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3104 __wsum csum, const struct skb_checksum_ops *ops);
3105__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3106 __wsum csum);
3107
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003108static inline void * __must_check
3109__skb_header_pointer(const struct sk_buff *skb, int offset,
3110 int len, void *data, int hlen, void *buffer)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003111{
Patrick McHardy55820ee2005-07-05 14:08:10 -07003112 if (hlen - offset >= len)
David S. Miller690e36e2014-08-23 12:13:41 -07003113 return data + offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003114
David S. Miller690e36e2014-08-23 12:13:41 -07003115 if (!skb ||
3116 skb_copy_bits(skb, offset, buffer, len) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003117 return NULL;
3118
3119 return buffer;
3120}
3121
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003122static inline void * __must_check
3123skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
David S. Miller690e36e2014-08-23 12:13:41 -07003124{
3125 return __skb_header_pointer(skb, offset, len, skb->data,
3126 skb_headlen(skb), buffer);
3127}
3128
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01003129/**
3130 * skb_needs_linearize - check if we need to linearize a given skb
3131 * depending on the given device features.
3132 * @skb: socket buffer to check
3133 * @features: net device features
3134 *
3135 * Returns true if either:
3136 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3137 * 2. skb is fragmented and the device does not support SG.
3138 */
3139static inline bool skb_needs_linearize(struct sk_buff *skb,
3140 netdev_features_t features)
3141{
3142 return skb_is_nonlinear(skb) &&
3143 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3144 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3145}
3146
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03003147static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3148 void *to,
3149 const unsigned int len)
3150{
3151 memcpy(to, skb->data, len);
3152}
3153
3154static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3155 const int offset, void *to,
3156 const unsigned int len)
3157{
3158 memcpy(to, skb->data + offset, len);
3159}
3160
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03003161static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3162 const void *from,
3163 const unsigned int len)
3164{
3165 memcpy(skb->data, from, len);
3166}
3167
3168static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3169 const int offset,
3170 const void *from,
3171 const unsigned int len)
3172{
3173 memcpy(skb->data + offset, from, len);
3174}
3175
Joe Perches7965bd42013-09-26 14:48:15 -07003176void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003177
Patrick Ohlyac45f602009-02-12 05:03:37 +00003178static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3179{
3180 return skb->tstamp;
3181}
3182
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003183/**
3184 * skb_get_timestamp - get timestamp from a skb
3185 * @skb: skb to get stamp from
3186 * @stamp: pointer to struct timeval to store stamp in
3187 *
3188 * Timestamps are stored in the skb as offsets to a base timestamp.
3189 * This function converts the offset back to a struct timeval and stores
3190 * it in stamp.
3191 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00003192static inline void skb_get_timestamp(const struct sk_buff *skb,
3193 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003194{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003195 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003196}
3197
Patrick Ohlyac45f602009-02-12 05:03:37 +00003198static inline void skb_get_timestampns(const struct sk_buff *skb,
3199 struct timespec *stamp)
3200{
3201 *stamp = ktime_to_timespec(skb->tstamp);
3202}
3203
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003204static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003205{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003206 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003207}
3208
Stephen Hemminger164891a2007-04-23 22:26:16 -07003209static inline ktime_t net_timedelta(ktime_t t)
3210{
3211 return ktime_sub(ktime_get_real(), t);
3212}
3213
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003214static inline ktime_t net_invalid_timestamp(void)
3215{
3216 return ktime_set(0, 0);
3217}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003218
Alexander Duyck62bccb82014-09-04 13:31:35 -04003219struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3220
Richard Cochranc1f19b52010-07-17 08:49:36 +00003221#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3222
Joe Perches7965bd42013-09-26 14:48:15 -07003223void skb_clone_tx_timestamp(struct sk_buff *skb);
3224bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00003225
3226#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3227
3228static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3229{
3230}
3231
3232static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3233{
3234 return false;
3235}
3236
3237#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3238
3239/**
3240 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3241 *
Richard Cochranda92b192011-10-21 00:49:15 +00003242 * PHY drivers may accept clones of transmitted packets for
3243 * timestamping via their phy_driver.txtstamp method. These drivers
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003244 * must call this function to return the skb back to the stack with a
3245 * timestamp.
Richard Cochranda92b192011-10-21 00:49:15 +00003246 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00003247 * @skb: clone of the the original outgoing packet
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003248 * @hwtstamps: hardware time stamps
Richard Cochranc1f19b52010-07-17 08:49:36 +00003249 *
3250 */
3251void skb_complete_tx_timestamp(struct sk_buff *skb,
3252 struct skb_shared_hwtstamps *hwtstamps);
3253
Willem de Bruijne7fd2882014-08-04 22:11:48 -04003254void __skb_tstamp_tx(struct sk_buff *orig_skb,
3255 struct skb_shared_hwtstamps *hwtstamps,
3256 struct sock *sk, int tstype);
3257
Patrick Ohlyac45f602009-02-12 05:03:37 +00003258/**
3259 * skb_tstamp_tx - queue clone of skb with send time stamps
3260 * @orig_skb: the original outgoing packet
3261 * @hwtstamps: hardware time stamps, may be NULL if not available
3262 *
3263 * If the skb has a socket associated, then this function clones the
3264 * skb (thus sharing the actual data and optional structures), stores
3265 * the optional hardware time stamping information (if non NULL) or
3266 * generates a software time stamp (otherwise), then queues the clone
3267 * to the error queue of the socket. Errors are silently ignored.
3268 */
Joe Perches7965bd42013-09-26 14:48:15 -07003269void skb_tstamp_tx(struct sk_buff *orig_skb,
3270 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00003271
Richard Cochran4507a712010-07-17 08:48:28 +00003272static inline void sw_tx_timestamp(struct sk_buff *skb)
3273{
Oliver Hartkopp2244d072010-08-17 08:59:14 +00003274 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP &&
3275 !(skb_shinfo(skb)->tx_flags & SKBTX_IN_PROGRESS))
Richard Cochran4507a712010-07-17 08:48:28 +00003276 skb_tstamp_tx(skb, NULL);
3277}
3278
3279/**
3280 * skb_tx_timestamp() - Driver hook for transmit timestamping
3281 *
3282 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00003283 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00003284 *
David S. Miller73409f32013-12-27 13:04:33 -05003285 * Specifically, one should make absolutely sure that this function is
3286 * called before TX completion of this packet can trigger. Otherwise
3287 * the packet could potentially already be freed.
3288 *
Richard Cochran4507a712010-07-17 08:48:28 +00003289 * @skb: A socket buffer.
3290 */
3291static inline void skb_tx_timestamp(struct sk_buff *skb)
3292{
Richard Cochranc1f19b52010-07-17 08:49:36 +00003293 skb_clone_tx_timestamp(skb);
Richard Cochran4507a712010-07-17 08:48:28 +00003294 sw_tx_timestamp(skb);
3295}
3296
Johannes Berg6e3e9392011-11-09 10:15:42 +01003297/**
3298 * skb_complete_wifi_ack - deliver skb with wifi status
3299 *
3300 * @skb: the original outgoing packet
3301 * @acked: ack status
3302 *
3303 */
3304void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3305
Joe Perches7965bd42013-09-26 14:48:15 -07003306__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3307__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003308
Herbert Xu60476372007-04-09 11:59:39 -07003309static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3310{
Tom Herbert6edec0e2015-02-10 16:30:28 -08003311 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3312 skb->csum_valid ||
3313 (skb->ip_summed == CHECKSUM_PARTIAL &&
3314 skb_checksum_start_offset(skb) >= 0));
Herbert Xu60476372007-04-09 11:59:39 -07003315}
3316
Herbert Xufb286bb2005-11-10 13:01:24 -08003317/**
3318 * skb_checksum_complete - Calculate checksum of an entire packet
3319 * @skb: packet to process
3320 *
3321 * This function calculates the checksum over the entire packet plus
3322 * the value of skb->csum. The latter can be used to supply the
3323 * checksum of a pseudo header as used by TCP/UDP. It returns the
3324 * checksum.
3325 *
3326 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3327 * this function can be used to verify that checksum on received
3328 * packets. In that case the function should return zero if the
3329 * checksum is correct. In particular, this function will return zero
3330 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3331 * hardware has already verified the correctness of the checksum.
3332 */
Al Viro4381ca32007-07-15 21:00:11 +01003333static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08003334{
Herbert Xu60476372007-04-09 11:59:39 -07003335 return skb_csum_unnecessary(skb) ?
3336 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003337}
3338
Tom Herbert77cffe22014-08-27 21:26:46 -07003339static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3340{
3341 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3342 if (skb->csum_level == 0)
3343 skb->ip_summed = CHECKSUM_NONE;
3344 else
3345 skb->csum_level--;
3346 }
3347}
3348
3349static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3350{
3351 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3352 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3353 skb->csum_level++;
3354 } else if (skb->ip_summed == CHECKSUM_NONE) {
3355 skb->ip_summed = CHECKSUM_UNNECESSARY;
3356 skb->csum_level = 0;
3357 }
3358}
3359
Tom Herbert5a212322014-08-31 15:12:41 -07003360static inline void __skb_mark_checksum_bad(struct sk_buff *skb)
3361{
3362 /* Mark current checksum as bad (typically called from GRO
3363 * path). In the case that ip_summed is CHECKSUM_NONE
3364 * this must be the first checksum encountered in the packet.
3365 * When ip_summed is CHECKSUM_UNNECESSARY, this is the first
3366 * checksum after the last one validated. For UDP, a zero
3367 * checksum can not be marked as bad.
3368 */
3369
3370 if (skb->ip_summed == CHECKSUM_NONE ||
3371 skb->ip_summed == CHECKSUM_UNNECESSARY)
3372 skb->csum_bad = 1;
3373}
3374
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003375/* Check if we need to perform checksum complete validation.
3376 *
3377 * Returns true if checksum complete is needed, false otherwise
3378 * (either checksum is unnecessary or zero checksum is allowed).
3379 */
3380static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3381 bool zero_okay,
3382 __sum16 check)
3383{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003384 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3385 skb->csum_valid = 1;
Tom Herbert77cffe22014-08-27 21:26:46 -07003386 __skb_decr_checksum_unnecessary(skb);
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003387 return false;
3388 }
3389
3390 return true;
3391}
3392
3393/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3394 * in checksum_init.
3395 */
3396#define CHECKSUM_BREAK 76
3397
Tom Herbert4e18b9a2015-04-20 14:10:04 -07003398/* Unset checksum-complete
3399 *
3400 * Unset checksum complete can be done when packet is being modified
3401 * (uncompressed for instance) and checksum-complete value is
3402 * invalidated.
3403 */
3404static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3405{
3406 if (skb->ip_summed == CHECKSUM_COMPLETE)
3407 skb->ip_summed = CHECKSUM_NONE;
3408}
3409
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003410/* Validate (init) checksum based on checksum complete.
3411 *
3412 * Return values:
3413 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3414 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3415 * checksum is stored in skb->csum for use in __skb_checksum_complete
3416 * non-zero: value of invalid checksum
3417 *
3418 */
3419static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3420 bool complete,
3421 __wsum psum)
3422{
3423 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3424 if (!csum_fold(csum_add(psum, skb->csum))) {
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003425 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003426 return 0;
3427 }
Tom Herbert5a212322014-08-31 15:12:41 -07003428 } else if (skb->csum_bad) {
3429 /* ip_summed == CHECKSUM_NONE in this case */
Eric Dumazetc91d4602015-05-15 05:48:07 -07003430 return (__force __sum16)1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003431 }
3432
3433 skb->csum = psum;
3434
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003435 if (complete || skb->len <= CHECKSUM_BREAK) {
3436 __sum16 csum;
3437
3438 csum = __skb_checksum_complete(skb);
3439 skb->csum_valid = !csum;
3440 return csum;
3441 }
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003442
3443 return 0;
3444}
3445
3446static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3447{
3448 return 0;
3449}
3450
3451/* Perform checksum validate (init). Note that this is a macro since we only
3452 * want to calculate the pseudo header which is an input function if necessary.
3453 * First we try to validate without any computation (checksum unnecessary) and
3454 * then calculate based on checksum complete calling the function to compute
3455 * pseudo header.
3456 *
3457 * Return values:
3458 * 0: checksum is validated or try to in skb_checksum_complete
3459 * non-zero: value of invalid checksum
3460 */
3461#define __skb_checksum_validate(skb, proto, complete, \
3462 zero_okay, check, compute_pseudo) \
3463({ \
3464 __sum16 __ret = 0; \
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003465 skb->csum_valid = 0; \
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003466 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3467 __ret = __skb_checksum_validate_complete(skb, \
3468 complete, compute_pseudo(skb, proto)); \
3469 __ret; \
3470})
3471
3472#define skb_checksum_init(skb, proto, compute_pseudo) \
3473 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3474
3475#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3476 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3477
3478#define skb_checksum_validate(skb, proto, compute_pseudo) \
3479 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3480
3481#define skb_checksum_validate_zero_check(skb, proto, check, \
3482 compute_pseudo) \
Sabrina Dubroca096a4cf2015-02-06 18:54:19 +01003483 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003484
3485#define skb_checksum_simple_validate(skb) \
3486 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3487
Tom Herbertd96535a2014-08-31 15:12:42 -07003488static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3489{
3490 return (skb->ip_summed == CHECKSUM_NONE &&
3491 skb->csum_valid && !skb->csum_bad);
3492}
3493
3494static inline void __skb_checksum_convert(struct sk_buff *skb,
3495 __sum16 check, __wsum pseudo)
3496{
3497 skb->csum = ~pseudo;
3498 skb->ip_summed = CHECKSUM_COMPLETE;
3499}
3500
3501#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3502do { \
3503 if (__skb_checksum_convert_check(skb)) \
3504 __skb_checksum_convert(skb, check, \
3505 compute_pseudo(skb, proto)); \
3506} while (0)
3507
Tom Herbert15e23962015-02-10 16:30:31 -08003508static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3509 u16 start, u16 offset)
3510{
3511 skb->ip_summed = CHECKSUM_PARTIAL;
3512 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3513 skb->csum_offset = offset - start;
3514}
3515
Tom Herbertdcdc8992015-02-02 16:07:34 -08003516/* Update skbuf and packet to reflect the remote checksum offload operation.
3517 * When called, ptr indicates the starting point for skb->csum when
3518 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3519 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3520 */
3521static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
Tom Herbert15e23962015-02-10 16:30:31 -08003522 int start, int offset, bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08003523{
3524 __wsum delta;
3525
Tom Herbert15e23962015-02-10 16:30:31 -08003526 if (!nopartial) {
3527 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3528 return;
3529 }
3530
Tom Herbertdcdc8992015-02-02 16:07:34 -08003531 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3532 __skb_checksum_complete(skb);
3533 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3534 }
3535
3536 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3537
3538 /* Adjust skb->csum since we changed the packet */
3539 skb->csum = csum_add(skb->csum, delta);
3540}
3541
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003542#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07003543void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003544static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3545{
3546 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07003547 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003548}
3549static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3550{
3551 if (nfct)
3552 atomic_inc(&nfct->use);
3553}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003554#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003555#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003556static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3557{
3558 if (nf_bridge && atomic_dec_and_test(&nf_bridge->use))
3559 kfree(nf_bridge);
3560}
3561static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3562{
3563 if (nf_bridge)
3564 atomic_inc(&nf_bridge->use);
3565}
3566#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08003567static inline void nf_reset(struct sk_buff *skb)
3568{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003569#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Patrick McHardya193a4a2006-03-20 19:23:05 -08003570 nf_conntrack_put(skb->nfct);
3571 skb->nfct = NULL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003572#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003573#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Patrick McHardya193a4a2006-03-20 19:23:05 -08003574 nf_bridge_put(skb->nf_bridge);
3575 skb->nf_bridge = NULL;
3576#endif
3577}
3578
Patrick McHardy124dff02013-04-05 20:42:05 +02003579static inline void nf_reset_trace(struct sk_buff *skb)
3580{
Florian Westphal478b3602014-02-15 23:48:45 +01003581#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00003582 skb->nf_trace = 0;
3583#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003584}
3585
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003586/* Note: This doesn't put any conntrack and bridge info in dst. */
Eric Dumazetb1937222014-09-28 22:18:47 -07003587static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3588 bool copy)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003589{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003590#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003591 dst->nfct = src->nfct;
3592 nf_conntrack_get(src->nfct);
Eric Dumazetb1937222014-09-28 22:18:47 -07003593 if (copy)
3594 dst->nfctinfo = src->nfctinfo;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003595#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003596#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003597 dst->nf_bridge = src->nf_bridge;
3598 nf_bridge_get(src->nf_bridge);
3599#endif
Florian Westphal478b3602014-02-15 23:48:45 +01003600#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Eric Dumazetb1937222014-09-28 22:18:47 -07003601 if (copy)
3602 dst->nf_trace = src->nf_trace;
Florian Westphal478b3602014-02-15 23:48:45 +01003603#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003604}
3605
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003606static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3607{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003608#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003609 nf_conntrack_put(dst->nfct);
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003610#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003611#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003612 nf_bridge_put(dst->nf_bridge);
3613#endif
Eric Dumazetb1937222014-09-28 22:18:47 -07003614 __nf_copy(dst, src, true);
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003615}
3616
James Morris984bc162006-06-09 00:29:17 -07003617#ifdef CONFIG_NETWORK_SECMARK
3618static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3619{
3620 to->secmark = from->secmark;
3621}
3622
3623static inline void skb_init_secmark(struct sk_buff *skb)
3624{
3625 skb->secmark = 0;
3626}
3627#else
3628static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3629{ }
3630
3631static inline void skb_init_secmark(struct sk_buff *skb)
3632{ }
3633#endif
3634
Eric W. Biederman574f7192014-04-01 12:20:24 -07003635static inline bool skb_irq_freeable(const struct sk_buff *skb)
3636{
3637 return !skb->destructor &&
3638#if IS_ENABLED(CONFIG_XFRM)
3639 !skb->sp &&
3640#endif
3641#if IS_ENABLED(CONFIG_NF_CONNTRACK)
3642 !skb->nfct &&
3643#endif
3644 !skb->_skb_refdst &&
3645 !skb_has_frag_list(skb);
3646}
3647
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003648static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3649{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003650 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003651}
3652
Stephen Hemminger92477442009-03-21 13:39:26 -07003653static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003654{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003655 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003656}
3657
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003658static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3659{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003660 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003661}
3662
David S. Millerd5a9e242009-01-27 16:22:11 -08003663static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3664{
3665 skb->queue_mapping = rx_queue + 1;
3666}
3667
Stephen Hemminger92477442009-03-21 13:39:26 -07003668static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003669{
3670 return skb->queue_mapping - 1;
3671}
3672
Stephen Hemminger92477442009-03-21 13:39:26 -07003673static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003674{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003675 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08003676}
3677
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003678static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3679{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003680#ifdef CONFIG_XFRM
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003681 return skb->sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003682#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003683 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003684#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003685}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003686
Pravin B Shelar68c33162013-02-14 14:02:41 +00003687/* Keeps track of mac header offset relative to skb->head.
3688 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3689 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07003690 * tunnel skb it points to outer mac header.
3691 * Keeps track of level of encapsulation of network headers.
3692 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00003693struct skb_gso_cb {
Alexander Duyck802ab552016-04-10 21:45:03 -04003694 union {
3695 int mac_offset;
3696 int data_offset;
3697 };
Eric Dumazet3347c962013-10-19 11:42:56 -07003698 int encap_level;
Alexander Duyck76443452016-02-05 15:27:37 -08003699 __wsum csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003700 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00003701};
Konstantin Khlebnikov9207f9d2016-01-08 15:21:46 +03003702#define SKB_SGO_CB_OFFSET 32
3703#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
Pravin B Shelar68c33162013-02-14 14:02:41 +00003704
3705static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3706{
3707 return (skb_mac_header(inner_skb) - inner_skb->head) -
3708 SKB_GSO_CB(inner_skb)->mac_offset;
3709}
3710
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00003711static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3712{
3713 int new_headroom, headroom;
3714 int ret;
3715
3716 headroom = skb_headroom(skb);
3717 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3718 if (ret)
3719 return ret;
3720
3721 new_headroom = skb_headroom(skb);
3722 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3723 return 0;
3724}
3725
Alexander Duyck08b64fc2016-02-05 15:27:49 -08003726static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
3727{
3728 /* Do not update partial checksums if remote checksum is enabled. */
3729 if (skb->remcsum_offload)
3730 return;
3731
3732 SKB_GSO_CB(skb)->csum = res;
3733 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
3734}
3735
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003736/* Compute the checksum for a gso segment. First compute the checksum value
3737 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3738 * then add in skb->csum (checksum from csum_start to end of packet).
3739 * skb->csum and csum_start are then updated to reflect the checksum of the
3740 * resultant packet starting from the transport header-- the resultant checksum
3741 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3742 * header.
3743 */
3744static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3745{
Alexander Duyck76443452016-02-05 15:27:37 -08003746 unsigned char *csum_start = skb_transport_header(skb);
3747 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
3748 __wsum partial = SKB_GSO_CB(skb)->csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003749
Alexander Duyck76443452016-02-05 15:27:37 -08003750 SKB_GSO_CB(skb)->csum = res;
3751 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003752
Alexander Duyck76443452016-02-05 15:27:37 -08003753 return csum_fold(csum_partial(csum_start, plen, partial));
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003754}
3755
David S. Millerbdcc0922012-03-07 20:53:36 -05003756static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07003757{
3758 return skb_shinfo(skb)->gso_size;
3759}
3760
Eric Dumazet36a8f392013-09-29 01:21:32 -07003761/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05003762static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02003763{
3764 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3765}
3766
Daniel Borkmann5293efe2016-08-18 01:00:39 +02003767static inline void skb_gso_reset(struct sk_buff *skb)
3768{
3769 skb_shinfo(skb)->gso_size = 0;
3770 skb_shinfo(skb)->gso_segs = 0;
3771 skb_shinfo(skb)->gso_type = 0;
3772}
3773
Joe Perches7965bd42013-09-26 14:48:15 -07003774void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07003775
3776static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3777{
3778 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3779 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003780 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3781
Alexander Duyckb78462e2010-06-02 12:24:37 +00003782 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3783 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07003784 __skb_warn_lro_forwarding(skb);
3785 return true;
3786 }
3787 return false;
3788}
3789
Herbert Xu35fc92a2007-03-26 23:22:20 -07003790static inline void skb_forward_csum(struct sk_buff *skb)
3791{
3792 /* Unfortunately we don't support this one. Any brave souls? */
3793 if (skb->ip_summed == CHECKSUM_COMPLETE)
3794 skb->ip_summed = CHECKSUM_NONE;
3795}
3796
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003797/**
3798 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3799 * @skb: skb to check
3800 *
3801 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3802 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3803 * use this helper, to document places where we make this assertion.
3804 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003805static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003806{
3807#ifdef DEBUG
3808 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3809#endif
3810}
3811
Rusty Russellf35d9d82008-02-04 23:49:54 -05003812bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00003813
Paul Durranted1f50c2014-01-09 10:02:46 +00003814int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
Linus Lüssing9afd85c2015-05-02 14:01:07 +02003815struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
3816 unsigned int transport_len,
3817 __sum16(*skb_chkf)(struct sk_buff *skb));
Paul Durranted1f50c2014-01-09 10:02:46 +00003818
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00003819/**
3820 * skb_head_is_locked - Determine if the skb->head is locked down
3821 * @skb: skb to check
3822 *
3823 * The head on skbs build around a head frag can be removed if they are
3824 * not cloned. This function returns true if the skb head is locked down
3825 * due to either being allocated via kmalloc, or by being a clone with
3826 * multiple references to the head.
3827 */
3828static inline bool skb_head_is_locked(const struct sk_buff *skb)
3829{
3830 return !skb->head_frag || skb_cloned(skb);
3831}
Florian Westphalfe6cc552014-02-13 23:09:12 +01003832
3833/**
3834 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3835 *
3836 * @skb: GSO skb
3837 *
3838 * skb_gso_network_seglen is used to determine the real size of the
3839 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3840 *
3841 * The MAC/L2 header is not accounted for.
3842 */
3843static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3844{
3845 unsigned int hdr_len = skb_transport_header(skb) -
3846 skb_network_header(skb);
3847 return hdr_len + skb_gso_transport_seglen(skb);
3848}
Thomas Grafee122c72015-07-21 10:43:58 +02003849
Edward Cree179bc672016-02-11 20:48:04 +00003850/* Local Checksum Offload.
3851 * Compute outer checksum based on the assumption that the
3852 * inner checksum will be offloaded later.
Edward Creee8ae7b02016-02-11 21:03:37 +00003853 * See Documentation/networking/checksum-offloads.txt for
3854 * explanation of how this works.
Edward Cree179bc672016-02-11 20:48:04 +00003855 * Fill in outer checksum adjustment (e.g. with sum of outer
3856 * pseudo-header) before calling.
3857 * Also ensure that inner checksum is in linear data area.
3858 */
3859static inline __wsum lco_csum(struct sk_buff *skb)
3860{
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003861 unsigned char *csum_start = skb_checksum_start(skb);
3862 unsigned char *l4_hdr = skb_transport_header(skb);
3863 __wsum partial;
Edward Cree179bc672016-02-11 20:48:04 +00003864
3865 /* Start with complement of inner checksum adjustment */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003866 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
3867 skb->csum_offset));
3868
Edward Cree179bc672016-02-11 20:48:04 +00003869 /* Add in checksum of our headers (incl. outer checksum
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003870 * adjustment filled in by caller) and return result.
Edward Cree179bc672016-02-11 20:48:04 +00003871 */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003872 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
Edward Cree179bc672016-02-11 20:48:04 +00003873}
3874
Linus Torvalds1da177e2005-04-16 15:20:36 -07003875#endif /* __KERNEL__ */
3876#endif /* _LINUX_SKBUFF_H */