blob: 2f64e2bbb59219560951177fb5def4e1fc69902b [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * Definitions for the 'struct sk_buff' memory handlers.
3 *
4 * Authors:
5 * Alan Cox, <gw4pts@gw4pts.ampr.org>
6 * Florian La Roche, <rzsfl@rz.uni-sb.de>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; either version
11 * 2 of the License, or (at your option) any later version.
12 */
13
14#ifndef _LINUX_SKBUFF_H
15#define _LINUX_SKBUFF_H
16
Linus Torvalds1da177e2005-04-16 15:20:36 -070017#include <linux/kernel.h>
Vegard Nossumfe55f6d2008-08-30 12:16:35 +020018#include <linux/kmemcheck.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070019#include <linux/compiler.h>
20#include <linux/time.h>
Paul Gortmaker187f1882011-11-23 20:12:59 -050021#include <linux/bug.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070022#include <linux/cache.h>
Eric Dumazet56b17422014-11-03 08:19:53 -080023#include <linux/rbtree.h>
David S. Miller51f3d022014-11-05 16:46:40 -050024#include <linux/socket.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070025
Arun Sharma600634972011-07-26 16:09:06 -070026#include <linux/atomic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include <asm/types.h>
28#include <linux/spinlock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070029#include <linux/net.h>
Thomas Graf3fc7e8a2005-06-23 21:00:17 -070030#include <linux/textsearch.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070031#include <net/checksum.h>
Al Viroa80958f2006-12-04 20:41:19 +000032#include <linux/rcupdate.h>
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -070033#include <linux/hrtimer.h>
Ian Campbell131ea662011-08-19 06:25:00 +000034#include <linux/dma-mapping.h>
Michał Mirosławc8f44af2011-11-15 15:29:55 +000035#include <linux/netdev_features.h>
Eric Dumazet363ec39232014-02-26 14:02:11 -080036#include <linux/sched.h>
Ingo Molnare6017572017-02-01 16:36:40 +010037#include <linux/sched/clock.h>
Jiri Pirko1bd758e2015-05-12 14:56:07 +020038#include <net/flow_dissector.h>
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +020039#include <linux/splice.h>
Bernhard Thaler72b31f72015-05-30 15:27:40 +020040#include <linux/in6.h>
Jamal Hadi Salim8b10cab2016-07-02 06:43:14 -040041#include <linux/if_packet.h>
Tom Herbertf70ea012015-07-31 16:52:10 -070042#include <net/flow.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070043
Tom Herbert7a6ae712015-12-14 11:19:47 -080044/* The interface for checksum offload between the stack and networking drivers
45 * is as follows...
46 *
47 * A. IP checksum related features
48 *
49 * Drivers advertise checksum offload capabilities in the features of a device.
50 * From the stack's point of view these are capabilities offered by the driver,
51 * a driver typically only advertises features that it is capable of offloading
52 * to its device.
53 *
54 * The checksum related features are:
55 *
56 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
57 * IP (one's complement) checksum for any combination
58 * of protocols or protocol layering. The checksum is
59 * computed and set in a packet per the CHECKSUM_PARTIAL
60 * interface (see below).
61 *
62 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
63 * TCP or UDP packets over IPv4. These are specifically
64 * unencapsulated packets of the form IPv4|TCP or
65 * IPv4|UDP where the Protocol field in the IPv4 header
66 * is TCP or UDP. The IPv4 header may contain IP options
67 * This feature cannot be set in features for a device
68 * with NETIF_F_HW_CSUM also set. This feature is being
69 * DEPRECATED (see below).
70 *
71 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
72 * TCP or UDP packets over IPv6. These are specifically
73 * unencapsulated packets of the form IPv6|TCP or
74 * IPv4|UDP where the Next Header field in the IPv6
75 * header is either TCP or UDP. IPv6 extension headers
76 * are not supported with this feature. This feature
77 * cannot be set in features for a device with
78 * NETIF_F_HW_CSUM also set. This feature is being
79 * DEPRECATED (see below).
80 *
81 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
82 * This flag is used only used to disable the RX checksum
83 * feature for a device. The stack will accept receive
84 * checksum indication in packets received on a device
85 * regardless of whether NETIF_F_RXCSUM is set.
86 *
87 * B. Checksumming of received packets by device. Indication of checksum
88 * verification is in set skb->ip_summed. Possible values are:
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010089 *
90 * CHECKSUM_NONE:
91 *
Tom Herbert7a6ae712015-12-14 11:19:47 -080092 * Device did not checksum this packet e.g. due to lack of capabilities.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010093 * The packet contains full (though not verified) checksum in packet but
94 * not in skb->csum. Thus, skb->csum is undefined in this case.
95 *
96 * CHECKSUM_UNNECESSARY:
97 *
98 * The hardware you're dealing with doesn't calculate the full checksum
99 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
Tom Herbert77cffe22014-08-27 21:26:46 -0700100 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
101 * if their checksums are okay. skb->csum is still undefined in this case
Tom Herbert7a6ae712015-12-14 11:19:47 -0800102 * though. A driver or device must never modify the checksum field in the
103 * packet even if checksum is verified.
Tom Herbert77cffe22014-08-27 21:26:46 -0700104 *
105 * CHECKSUM_UNNECESSARY is applicable to following protocols:
106 * TCP: IPv6 and IPv4.
107 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
108 * zero UDP checksum for either IPv4 or IPv6, the networking stack
109 * may perform further validation in this case.
110 * GRE: only if the checksum is present in the header.
111 * SCTP: indicates the CRC in SCTP header has been validated.
Davide Carattib4759dc2017-05-18 15:44:43 +0200112 * FCOE: indicates the CRC in FC frame has been validated.
Tom Herbert77cffe22014-08-27 21:26:46 -0700113 *
114 * skb->csum_level indicates the number of consecutive checksums found in
115 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
116 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
117 * and a device is able to verify the checksums for UDP (possibly zero),
118 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
119 * two. If the device were only able to verify the UDP checksum and not
120 * GRE, either because it doesn't support GRE checksum of because GRE
121 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
122 * not considered in this case).
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100123 *
124 * CHECKSUM_COMPLETE:
125 *
126 * This is the most generic way. The device supplied checksum of the _whole_
127 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
128 * hardware doesn't need to parse L3/L4 headers to implement this.
129 *
Davide Carattib4759dc2017-05-18 15:44:43 +0200130 * Notes:
131 * - Even if device supports only some protocols, but is able to produce
132 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
133 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100134 *
135 * CHECKSUM_PARTIAL:
136 *
Tom Herbert6edec0e2015-02-10 16:30:28 -0800137 * A checksum is set up to be offloaded to a device as described in the
138 * output description for CHECKSUM_PARTIAL. This may occur on a packet
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100139 * received directly from another Linux OS, e.g., a virtualized Linux kernel
Tom Herbert6edec0e2015-02-10 16:30:28 -0800140 * on the same host, or it may be set in the input path in GRO or remote
141 * checksum offload. For the purposes of checksum verification, the checksum
142 * referred to by skb->csum_start + skb->csum_offset and any preceding
143 * checksums in the packet are considered verified. Any checksums in the
144 * packet that are after the checksum being offloaded are not considered to
145 * be verified.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100146 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800147 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
148 * in the skb->ip_summed for a packet. Values are:
149 *
150 * CHECKSUM_PARTIAL:
151 *
152 * The driver is required to checksum the packet as seen by hard_start_xmit()
153 * from skb->csum_start up to the end, and to record/write the checksum at
154 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
155 * csum_start and csum_offset values are valid values given the length and
156 * offset of the packet, however they should not attempt to validate that the
157 * checksum refers to a legitimate transport layer checksum-- it is the
158 * purview of the stack to validate that csum_start and csum_offset are set
159 * correctly.
160 *
161 * When the stack requests checksum offload for a packet, the driver MUST
162 * ensure that the checksum is set correctly. A driver can either offload the
163 * checksum calculation to the device, or call skb_checksum_help (in the case
164 * that the device does not support offload for a particular checksum).
165 *
166 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
167 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
Davide Caratti43c26a12017-05-18 15:44:41 +0200168 * checksum offload capability.
169 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based
170 * on network device checksumming capabilities: if a packet does not match
171 * them, skb_checksum_help or skb_crc32c_help (depending on the value of
172 * csum_not_inet, see item D.) is called to resolve the checksum.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100173 *
174 * CHECKSUM_NONE:
175 *
176 * The skb was already checksummed by the protocol, or a checksum is not
177 * required.
178 *
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100179 * CHECKSUM_UNNECESSARY:
180 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800181 * This has the same meaning on as CHECKSUM_NONE for checksum offload on
182 * output.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100183 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800184 * CHECKSUM_COMPLETE:
185 * Not used in checksum output. If a driver observes a packet with this value
186 * set in skbuff, if should treat as CHECKSUM_NONE being set.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100187 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800188 * D. Non-IP checksum (CRC) offloads
189 *
190 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
191 * offloading the SCTP CRC in a packet. To perform this offload the stack
Davide Carattidba00302017-05-18 15:44:40 +0200192 * will set set csum_start and csum_offset accordingly, set ip_summed to
193 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
194 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
195 * A driver that supports both IP checksum offload and SCTP CRC32c offload
196 * must verify which offload is configured for a packet by testing the
197 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
198 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
Tom Herbert7a6ae712015-12-14 11:19:47 -0800199 *
200 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
201 * offloading the FCOE CRC in a packet. To perform this offload the stack
202 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
203 * accordingly. Note the there is no indication in the skbuff that the
204 * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
205 * both IP checksum offload and FCOE CRC offload must verify which offload
206 * is configured for a packet presumably by inspecting packet headers.
207 *
208 * E. Checksumming on output with GSO.
209 *
210 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
211 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
212 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
213 * part of the GSO operation is implied. If a checksum is being offloaded
214 * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
215 * are set to refer to the outermost checksum being offload (two offloaded
216 * checksums are possible with UDP encapsulation).
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100217 */
218
Herbert Xu60476372007-04-09 11:59:39 -0700219/* Don't change this without changing skb_csum_unnecessary! */
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100220#define CHECKSUM_NONE 0
221#define CHECKSUM_UNNECESSARY 1
222#define CHECKSUM_COMPLETE 2
223#define CHECKSUM_PARTIAL 3
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224
Tom Herbert77cffe22014-08-27 21:26:46 -0700225/* Maximum value in skb->csum_level */
226#define SKB_MAX_CSUM_LEVEL 3
227
Tobias Klauser0bec8c82014-07-22 12:06:23 +0200228#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
David S. Millerfc910a22007-03-25 20:27:59 -0700229#define SKB_WITH_OVERHEAD(X) \
Herbert Xudeea84b2007-10-21 16:27:46 -0700230 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
David S. Millerfc910a22007-03-25 20:27:59 -0700231#define SKB_MAX_ORDER(X, ORDER) \
232 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
234#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
235
Eric Dumazet87fb4b72011-10-13 07:28:54 +0000236/* return minimum truesize of one skb containing X bytes of data */
237#define SKB_TRUESIZE(X) ((X) + \
238 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
239 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
240
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241struct net_device;
David Howells716ea3a2007-04-02 20:19:53 -0700242struct scatterlist;
Jens Axboe9c55e012007-11-06 23:30:13 -0800243struct pipe_inode_info;
Herbert Xua8f820aa2014-11-07 21:22:22 +0800244struct iov_iter;
Alexander Duyckfd11a832014-12-09 19:40:49 -0800245struct napi_struct;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700246
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700247#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248struct nf_conntrack {
249 atomic_t use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700250};
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700251#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +0200253#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254struct nf_bridge_info {
Reshetova, Elena53869ce2017-06-30 13:07:57 +0300255 refcount_t use;
Florian Westphal3eaf4022015-04-02 14:31:44 +0200256 enum {
257 BRNF_PROTO_UNCHANGED,
258 BRNF_PROTO_8021Q,
259 BRNF_PROTO_PPPOE
Florian Westphal7fb48c52015-05-03 22:05:28 +0200260 } orig_proto:8;
Florian Westphal72b1e5e2015-07-23 16:21:30 +0200261 u8 pkt_otherhost:1;
262 u8 in_prerouting:1;
263 u8 bridged_dnat:1;
Bernhard Thaler411ffb42015-05-30 15:28:28 +0200264 __u16 frag_max_size;
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000265 struct net_device *physindev;
Florian Westphal63cdbc02015-09-14 17:06:27 +0200266
267 /* always valid & non-NULL from FORWARD on, for physdev match */
268 struct net_device *physoutdev;
Florian Westphal7fb48c52015-05-03 22:05:28 +0200269 union {
Florian Westphal72b1e5e2015-07-23 16:21:30 +0200270 /* prerouting: detect dnat in orig/reply direction */
Bernhard Thaler72b31f72015-05-30 15:27:40 +0200271 __be32 ipv4_daddr;
272 struct in6_addr ipv6_daddr;
Florian Westphal72b1e5e2015-07-23 16:21:30 +0200273
274 /* after prerouting + nat detected: store original source
275 * mac since neigh resolution overwrites it, only used while
276 * skb is out in neigh layer.
277 */
278 char neigh_header[8];
Bernhard Thaler72b31f72015-05-30 15:27:40 +0200279 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700280};
281#endif
282
Linus Torvalds1da177e2005-04-16 15:20:36 -0700283struct sk_buff_head {
284 /* These two members must be first. */
285 struct sk_buff *next;
286 struct sk_buff *prev;
287
288 __u32 qlen;
289 spinlock_t lock;
290};
291
292struct sk_buff;
293
Ian Campbell9d4dde52011-12-22 23:39:14 +0000294/* To allow 64K frame to be packed as single skb without frag_list we
295 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
296 * buffers which do not start on a page boundary.
297 *
298 * Since GRO uses frags we allocate at least 16 regardless of page
299 * size.
Anton Blancharda715dea2011-03-27 14:57:26 +0000300 */
Ian Campbell9d4dde52011-12-22 23:39:14 +0000301#if (65536/PAGE_SIZE + 1) < 16
David S. Millereec00952011-03-29 23:34:08 -0700302#define MAX_SKB_FRAGS 16UL
Anton Blancharda715dea2011-03-27 14:57:26 +0000303#else
Ian Campbell9d4dde52011-12-22 23:39:14 +0000304#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
Anton Blancharda715dea2011-03-27 14:57:26 +0000305#endif
Hans Westgaard Ry5f74f82e2016-02-03 09:26:57 +0100306extern int sysctl_max_skb_frags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700307
Marcelo Ricardo Leitner3953c462016-06-02 15:05:40 -0300308/* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
309 * segment using its current segmentation instead.
310 */
311#define GSO_BY_FRAGS 0xFFFF
312
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313typedef struct skb_frag_struct skb_frag_t;
314
315struct skb_frag_struct {
Ian Campbella8605c62011-10-19 23:01:49 +0000316 struct {
317 struct page *p;
318 } page;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000319#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
David S. Millera309bb02007-07-30 18:47:03 -0700320 __u32 page_offset;
321 __u32 size;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000322#else
323 __u16 page_offset;
324 __u16 size;
325#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700326};
327
Eric Dumazet9e903e02011-10-18 21:00:24 +0000328static inline unsigned int skb_frag_size(const skb_frag_t *frag)
329{
330 return frag->size;
331}
332
333static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
334{
335 frag->size = size;
336}
337
338static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
339{
340 frag->size += delta;
341}
342
343static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
344{
345 frag->size -= delta;
346}
347
Willem de Bruijnc613c202017-07-31 08:15:47 -0400348static inline bool skb_frag_must_loop(struct page *p)
349{
350#if defined(CONFIG_HIGHMEM)
351 if (PageHighMem(p))
352 return true;
353#endif
354 return false;
355}
356
357/**
358 * skb_frag_foreach_page - loop over pages in a fragment
359 *
360 * @f: skb frag to operate on
361 * @f_off: offset from start of f->page.p
362 * @f_len: length from f_off to loop over
363 * @p: (temp var) current page
364 * @p_off: (temp var) offset from start of current page,
365 * non-zero only on first page.
366 * @p_len: (temp var) length in current page,
367 * < PAGE_SIZE only on first and last page.
368 * @copied: (temp var) length so far, excluding current p_len.
369 *
370 * A fragment can hold a compound page, in which case per-page
371 * operations, notably kmap_atomic, must be called for each
372 * regular page.
373 */
374#define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
375 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
376 p_off = (f_off) & (PAGE_SIZE - 1), \
377 p_len = skb_frag_must_loop(p) ? \
378 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
379 copied = 0; \
380 copied < f_len; \
381 copied += p_len, p++, p_off = 0, \
382 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
383
Patrick Ohlyac45f602009-02-12 05:03:37 +0000384#define HAVE_HW_TIME_STAMP
385
386/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800387 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000388 * @hwtstamp: hardware time stamp transformed into duration
389 * since arbitrary point in time
Patrick Ohlyac45f602009-02-12 05:03:37 +0000390 *
391 * Software time stamps generated by ktime_get_real() are stored in
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400392 * skb->tstamp.
Patrick Ohlyac45f602009-02-12 05:03:37 +0000393 *
394 * hwtstamps can only be compared against other hwtstamps from
395 * the same device.
396 *
397 * This structure is attached to packets as part of the
398 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
399 */
400struct skb_shared_hwtstamps {
401 ktime_t hwtstamp;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000402};
403
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000404/* Definitions for tx_flags in struct skb_shared_info */
405enum {
406 /* generate hardware time stamp */
407 SKBTX_HW_TSTAMP = 1 << 0,
408
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400409 /* generate software time stamp when queueing packet to NIC */
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000410 SKBTX_SW_TSTAMP = 1 << 1,
411
412 /* device driver is going to provide hardware time stamp */
413 SKBTX_IN_PROGRESS = 1 << 2,
414
Shirley Maa6686f22011-07-06 12:22:12 +0000415 /* device driver supports TX zero-copy buffers */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000416 SKBTX_DEV_ZEROCOPY = 1 << 3,
Johannes Berg6e3e9392011-11-09 10:15:42 +0100417
418 /* generate wifi status information (where possible) */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000419 SKBTX_WIFI_STATUS = 1 << 4,
Pravin B Shelarc9af6db2013-02-11 09:27:41 +0000420
421 /* This indicates at least one fragment might be overwritten
422 * (as in vmsplice(), sendfile() ...)
423 * If we need to compute a TX checksum, we'll need to copy
424 * all frags to avoid possible bad checksum
425 */
426 SKBTX_SHARED_FRAG = 1 << 5,
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400427
428 /* generate software time stamp when entering packet scheduling */
429 SKBTX_SCHED_TSTAMP = 1 << 6,
Shirley Maa6686f22011-07-06 12:22:12 +0000430};
431
Willem de Bruijne1c8a602014-08-04 22:11:50 -0400432#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
Soheil Hassas Yeganeh0a2cf202016-04-27 23:39:01 -0400433 SKBTX_SCHED_TSTAMP)
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400434#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
435
Shirley Maa6686f22011-07-06 12:22:12 +0000436/*
437 * The callback notifies userspace to release buffers when skb DMA is done in
438 * lower device, the skb last reference should be 0 when calling this.
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000439 * The zerocopy_success argument is true if zero copy transmit occurred,
440 * false on data copy or out of memory error caused by data copy attempt.
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000441 * The ctx field is used to track device context.
442 * The desc field is used to track userspace buffer index.
Shirley Maa6686f22011-07-06 12:22:12 +0000443 */
444struct ubuf_info {
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000445 void (*callback)(struct ubuf_info *, bool zerocopy_success);
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000446 void *ctx;
Shirley Maa6686f22011-07-06 12:22:12 +0000447 unsigned long desc;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000448};
449
Linus Torvalds1da177e2005-04-16 15:20:36 -0700450/* This data is invariant across clones and lives at
451 * the end of the header data, ie. at skb->end.
452 */
453struct skb_shared_info {
Steffen Klassert7f564522017-04-08 20:36:24 +0200454 unsigned short _unused;
Ian Campbell9f42f122012-01-05 07:13:39 +0000455 unsigned char nr_frags;
456 __u8 tx_flags;
Herbert Xu79671682006-06-22 02:40:14 -0700457 unsigned short gso_size;
458 /* Warning: this field is not always filled in (UFO)! */
459 unsigned short gso_segs;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700460 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000461 struct skb_shared_hwtstamps hwtstamps;
Steffen Klassert7f564522017-04-08 20:36:24 +0200462 unsigned int gso_type;
Willem de Bruijn09c2d252014-08-04 22:11:47 -0400463 u32 tskey;
Ian Campbell9f42f122012-01-05 07:13:39 +0000464 __be32 ip6_frag_id;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700465
466 /*
467 * Warning : all fields before dataref are cleared in __alloc_skb()
468 */
469 atomic_t dataref;
470
Johann Baudy69e3c752009-05-18 22:11:22 -0700471 /* Intermediate layers must ensure that destructor_arg
472 * remains valid until skb destructor */
473 void * destructor_arg;
Shirley Maa6686f22011-07-06 12:22:12 +0000474
Eric Dumazetfed66382010-07-22 19:09:08 +0000475 /* must be last field, see pskb_expand_head() */
476 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700477};
478
479/* We divide dataref into two halves. The higher 16 bits hold references
480 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700481 * the entire skb->data. A clone of a headerless skb holds the length of
482 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700483 *
484 * All users must obey the rule that the skb->data reference count must be
485 * greater than or equal to the payload reference count.
486 *
487 * Holding a reference to the payload part means that the user does not
488 * care about modifications to the header part of skb->data.
489 */
490#define SKB_DATAREF_SHIFT 16
491#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
492
David S. Millerd179cd12005-08-17 14:57:30 -0700493
494enum {
Vijay Subramanianc8753d52014-10-02 10:00:43 -0700495 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
496 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
497 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
David S. Millerd179cd12005-08-17 14:57:30 -0700498};
499
Herbert Xu79671682006-06-22 02:40:14 -0700500enum {
501 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xu576a30e2006-06-27 13:22:38 -0700502
503 /* This indicates the skb is from an untrusted source. */
David S. Millerd9d30ad2017-07-03 07:31:57 -0700504 SKB_GSO_DODGY = 1 << 1,
Michael Chanb0da85372006-06-29 12:30:00 -0700505
506 /* This indicates the tcp segment has CWR set. */
David S. Millerd9d30ad2017-07-03 07:31:57 -0700507 SKB_GSO_TCP_ECN = 1 << 2,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700508
David S. Millerd9d30ad2017-07-03 07:31:57 -0700509 SKB_GSO_TCP_FIXEDID = 1 << 3,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800510
David S. Millerd9d30ad2017-07-03 07:31:57 -0700511 SKB_GSO_TCPV6 = 1 << 4,
Pravin B Shelar68c33162013-02-14 14:02:41 +0000512
David S. Millerd9d30ad2017-07-03 07:31:57 -0700513 SKB_GSO_FCOE = 1 << 5,
Pravin B Shelar73136262013-03-07 13:21:51 +0000514
David S. Millerd9d30ad2017-07-03 07:31:57 -0700515 SKB_GSO_GRE = 1 << 6,
Simon Horman0d89d202013-05-23 21:02:52 +0000516
David S. Millerd9d30ad2017-07-03 07:31:57 -0700517 SKB_GSO_GRE_CSUM = 1 << 7,
Eric Dumazetcb32f512013-10-19 11:42:57 -0700518
David S. Millerd9d30ad2017-07-03 07:31:57 -0700519 SKB_GSO_IPXIP4 = 1 << 8,
Eric Dumazet61c1db72013-10-20 20:47:30 -0700520
David S. Millerd9d30ad2017-07-03 07:31:57 -0700521 SKB_GSO_IPXIP6 = 1 << 9,
Tom Herbert0f4f4ff2014-06-04 17:20:16 -0700522
David S. Millerd9d30ad2017-07-03 07:31:57 -0700523 SKB_GSO_UDP_TUNNEL = 1 << 10,
Tom Herbert4749c092014-06-04 17:20:23 -0700524
David S. Millerd9d30ad2017-07-03 07:31:57 -0700525 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400526
David S. Millerd9d30ad2017-07-03 07:31:57 -0700527 SKB_GSO_PARTIAL = 1 << 12,
Alexander Duyck802ab552016-04-10 21:45:03 -0400528
David S. Millerd9d30ad2017-07-03 07:31:57 -0700529 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
Marcelo Ricardo Leitner90017ac2016-06-02 15:05:43 -0300530
David S. Millerd9d30ad2017-07-03 07:31:57 -0700531 SKB_GSO_SCTP = 1 << 14,
Steffen Klassertc7ef8f02017-04-14 10:05:36 +0200532
David S. Millerd9d30ad2017-07-03 07:31:57 -0700533 SKB_GSO_ESP = 1 << 15,
Herbert Xu79671682006-06-22 02:40:14 -0700534};
535
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700536#if BITS_PER_LONG > 32
537#define NET_SKBUFF_DATA_USES_OFFSET 1
538#endif
539
540#ifdef NET_SKBUFF_DATA_USES_OFFSET
541typedef unsigned int sk_buff_data_t;
542#else
543typedef unsigned char *sk_buff_data_t;
544#endif
545
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546/**
547 * struct sk_buff - socket buffer
548 * @next: Next buffer in list
549 * @prev: Previous buffer in list
Eric Dumazet363ec39232014-02-26 14:02:11 -0800550 * @tstamp: Time we arrived/left
Eric Dumazet56b17422014-11-03 08:19:53 -0800551 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700552 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700553 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700554 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000555 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700556 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700557 * @len: Length of actual data
558 * @data_len: Data length
559 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700560 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700561 * @csum: Checksum (must include start/offset pair)
562 * @csum_start: Offset from skb->head where checksumming should start
563 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700564 * @priority: Packet queueing priority
WANG Cong60ff7462014-05-04 16:39:18 -0700565 * @ignore_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700566 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700567 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700568 * @nohdr: Payload reference only, must not modify header
569 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700570 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700571 * @ipvs_property: skbuff is owned by ipvs
Willem de Bruijne7246e12017-01-07 17:06:35 -0500572 * @tc_skip_classify: do not classify packet. set by IFB device
Willem de Bruijn8dc07fd2017-01-07 17:06:37 -0500573 * @tc_at_ingress: used within tc_classify to distinguish in/egress
Willem de Bruijnbc31c902017-01-07 17:06:38 -0500574 * @tc_redirected: packet was redirected by a tc action
575 * @tc_from_ingress: if tc_redirected, tc_at_ingress at time of redirect
Randy Dunlap31729362008-02-18 20:52:13 -0800576 * @peeked: this packet has been seen already, so stats have been
577 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700578 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700579 * @protocol: Packet protocol from driver
580 * @destructor: Destruct function
Florian Westphala9e419d2017-01-23 18:21:59 +0100581 * @_nfct: Associated connection, if any (with nfctinfo bits)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700582 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800583 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700584 * @tc_index: Traffic control index
Tom Herbert61b905d2014-03-24 15:34:47 -0700585 * @hash: the packet hash
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700586 * @queue_mapping: Queue mapping for multiqueue devices
David S. Miller0b725a22014-08-25 15:51:53 -0700587 * @xmit_more: More SKBs are pending for this queue
Randy Dunlap553a5672008-04-20 10:51:01 -0700588 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700589 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Tom Herbert61b905d2014-03-24 15:34:47 -0700590 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
Changli Gao4ca24622011-08-19 07:26:44 -0700591 * ports.
Tom Herberta3b18dd2014-07-01 21:33:17 -0700592 * @sw_hash: indicates hash was computed in software stack
Johannes Berg6e3e9392011-11-09 10:15:42 +0100593 * @wifi_acked_valid: wifi_acked was set
594 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000595 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Davide Carattidba00302017-05-18 15:44:40 +0200596 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
Julian Anastasov4ff06202017-02-06 23:14:12 +0200597 * @dst_pending_confirm: need to confirm neighbour
Eliezer Tamir06021292013-06-10 11:39:50 +0300598 * @napi_id: id of the NAPI struct this skb came from
James Morris984bc162006-06-09 00:29:17 -0700599 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700600 * @mark: Generic packet mark
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000601 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b2008-07-14 22:49:06 -0700602 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000603 * @inner_protocol: Protocol (encapsulation)
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000604 * @inner_transport_header: Inner transport layer header (encapsulation)
605 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000606 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700607 * @transport_header: Transport layer header
608 * @network_header: Network layer header
609 * @mac_header: Link layer header
610 * @tail: Tail pointer
611 * @end: End pointer
612 * @head: Head of buffer
613 * @data: Data head pointer
614 * @truesize: Buffer size
615 * @users: User count - see {datagram,tcp}.c
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616 */
617
618struct sk_buff {
Eric Dumazet363ec39232014-02-26 14:02:11 -0800619 union {
Eric Dumazet56b17422014-11-03 08:19:53 -0800620 struct {
621 /* These two members must be first. */
622 struct sk_buff *next;
623 struct sk_buff *prev;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000624
Eric Dumazet56b17422014-11-03 08:19:53 -0800625 union {
626 ktime_t tstamp;
Eric Dumazet9a568de2017-05-16 14:00:14 -0700627 u64 skb_mstamp;
Eric Dumazet56b17422014-11-03 08:19:53 -0800628 };
629 };
630 struct rb_node rbnode; /* used in netem & tcp stack */
631 };
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000632 struct sock *sk;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700633
Eric Dumazetc84d9492016-12-08 11:41:55 -0800634 union {
635 struct net_device *dev;
636 /* Some protocols might use this space to store information,
637 * while device pointer would be NULL.
638 * UDP receive path is one user.
639 */
640 unsigned long dev_scratch;
641 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700642 /*
643 * This is the control buffer. It is free to use for every
644 * layer. Please put your private variables there. If you
645 * want to keep them across layers you have to do a skb_clone()
646 * first. This is owned by whoever has the skb queued ATM.
647 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000648 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700649
Eric Dumazet7fee2262010-05-11 23:19:48 +0000650 unsigned long _skb_refdst;
Eric Dumazetb1937222014-09-28 22:18:47 -0700651 void (*destructor)(struct sk_buff *skb);
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000652#ifdef CONFIG_XFRM
653 struct sec_path *sp;
654#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700655#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +0100656 unsigned long _nfct;
Eric Dumazetb1937222014-09-28 22:18:47 -0700657#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +0200658#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Eric Dumazetb1937222014-09-28 22:18:47 -0700659 struct nf_bridge_info *nf_bridge;
660#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700661 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700662 data_len;
663 __u16 mac_len,
664 hdr_len;
Eric Dumazetb1937222014-09-28 22:18:47 -0700665
666 /* Following fields are _not_ copied in __copy_skb_header()
667 * Note that queue_mapping is here mostly to fill a hole.
668 */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200669 kmemcheck_bitfield_begin(flags1);
Eric Dumazetb1937222014-09-28 22:18:47 -0700670 __u16 queue_mapping;
Daniel Borkmann36bbef52016-09-20 00:26:13 +0200671
672/* if you move cloned around you also must adapt those constants */
673#ifdef __BIG_ENDIAN_BITFIELD
674#define CLONED_MASK (1 << 7)
675#else
676#define CLONED_MASK 1
677#endif
678#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
679
680 __u8 __cloned_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700681 __u8 cloned:1,
Harald Welte6869c4d2005-08-09 19:24:19 -0700682 nohdr:1,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800683 fclone:2,
Herbert Xua59322b2007-12-05 01:53:40 -0800684 peeked:1,
Eric Dumazetb1937222014-09-28 22:18:47 -0700685 head_frag:1,
Daniel Borkmann36bbef52016-09-20 00:26:13 +0200686 xmit_more:1,
687 __unused:1; /* one bit hole */
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200688 kmemcheck_bitfield_end(flags1);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700689
Eric Dumazetb1937222014-09-28 22:18:47 -0700690 /* fields enclosed in headers_start/headers_end are copied
691 * using a single memcpy() in __copy_skb_header()
692 */
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700693 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700694 __u32 headers_start[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700695 /* public: */
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200696
697/* if you move pkt_type around you also must adapt those constants */
698#ifdef __BIG_ENDIAN_BITFIELD
699#define PKT_TYPE_MAX (7 << 5)
700#else
701#define PKT_TYPE_MAX 7
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100702#endif
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200703#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
704
705 __u8 __pkt_type_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700706 __u8 pkt_type:3;
707 __u8 pfmemalloc:1;
708 __u8 ignore_df:1;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709
Eric Dumazetb1937222014-09-28 22:18:47 -0700710 __u8 nf_trace:1;
711 __u8 ip_summed:2;
712 __u8 ooo_okay:1;
713 __u8 l4_hash:1;
714 __u8 sw_hash:1;
715 __u8 wifi_acked_valid:1;
716 __u8 wifi_acked:1;
717
718 __u8 no_fcs:1;
719 /* Indicates the inner headers are valid in the skbuff. */
720 __u8 encapsulation:1;
721 __u8 encap_hdr_csum:1;
722 __u8 csum_valid:1;
723 __u8 csum_complete_sw:1;
724 __u8 csum_level:2;
Davide Carattidba00302017-05-18 15:44:40 +0200725 __u8 csum_not_inet:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700726
Julian Anastasov4ff06202017-02-06 23:14:12 +0200727 __u8 dst_pending_confirm:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700728#ifdef CONFIG_IPV6_NDISC_NODETYPE
729 __u8 ndisc_nodetype:2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700731 __u8 ipvs_property:1;
Tom Herbert8bce6d72014-09-29 20:22:29 -0700732 __u8 inner_protocol_type:1;
Tom Herberte585f232014-11-04 09:06:54 -0800733 __u8 remcsum_offload:1;
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200734#ifdef CONFIG_NET_SWITCHDEV
735 __u8 offload_fwd_mark:1;
736#endif
Willem de Bruijne7246e12017-01-07 17:06:35 -0500737#ifdef CONFIG_NET_CLS_ACT
738 __u8 tc_skip_classify:1;
Willem de Bruijn8dc07fd2017-01-07 17:06:37 -0500739 __u8 tc_at_ingress:1;
Willem de Bruijnbc31c902017-01-07 17:06:38 -0500740 __u8 tc_redirected:1;
741 __u8 tc_from_ingress:1;
Willem de Bruijne7246e12017-01-07 17:06:35 -0500742#endif
Alexander Duyck4031ae62012-01-27 06:22:53 +0000743
Linus Torvalds1da177e2005-04-16 15:20:36 -0700744#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700745 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200747
Eric Dumazetb1937222014-09-28 22:18:47 -0700748 union {
749 __wsum csum;
750 struct {
751 __u16 csum_start;
752 __u16 csum_offset;
753 };
754 };
755 __u32 priority;
756 int skb_iif;
757 __u32 hash;
758 __be16 vlan_proto;
759 __u16 vlan_tci;
Eric Dumazet2bd82482015-02-03 23:48:24 -0800760#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
761 union {
762 unsigned int napi_id;
763 unsigned int sender_cpu;
764 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700765#endif
James Morris984bc162006-06-09 00:29:17 -0700766#ifdef CONFIG_NETWORK_SECMARK
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200767 __u32 secmark;
James Morris984bc162006-06-09 00:29:17 -0700768#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -0700769
Neil Horman3b885782009-10-12 13:26:31 -0700770 union {
771 __u32 mark;
Eric Dumazet16fad692013-03-14 05:40:32 +0000772 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700773 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700774
Tom Herbert8bce6d72014-09-29 20:22:29 -0700775 union {
776 __be16 inner_protocol;
777 __u8 inner_ipproto;
778 };
779
Simon Horman1a37e412013-05-23 21:02:51 +0000780 __u16 inner_transport_header;
781 __u16 inner_network_header;
782 __u16 inner_mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700783
784 __be16 protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000785 __u16 transport_header;
786 __u16 network_header;
787 __u16 mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700788
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700789 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700790 __u32 headers_end[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700791 /* public: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700792
Linus Torvalds1da177e2005-04-16 15:20:36 -0700793 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700794 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700795 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700796 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700797 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700798 unsigned int truesize;
Reshetova, Elena63354792017-06-30 13:07:58 +0300799 refcount_t users;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700800};
801
802#ifdef __KERNEL__
803/*
804 * Handling routines are only of interest to the kernel
805 */
806#include <linux/slab.h>
807
Linus Torvalds1da177e2005-04-16 15:20:36 -0700808
Mel Gormanc93bdd02012-07-31 16:44:19 -0700809#define SKB_ALLOC_FCLONE 0x01
810#define SKB_ALLOC_RX 0x02
Alexander Duyckfd11a832014-12-09 19:40:49 -0800811#define SKB_ALLOC_NAPI 0x04
Mel Gormanc93bdd02012-07-31 16:44:19 -0700812
813/* Returns true if the skb was allocated from PFMEMALLOC reserves */
814static inline bool skb_pfmemalloc(const struct sk_buff *skb)
815{
816 return unlikely(skb->pfmemalloc);
817}
818
Eric Dumazet7fee2262010-05-11 23:19:48 +0000819/*
820 * skb might have a dst pointer attached, refcounted or not.
821 * _skb_refdst low order bit is set if refcount was _not_ taken
822 */
823#define SKB_DST_NOREF 1UL
824#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
825
Florian Westphala9e419d2017-01-23 18:21:59 +0100826#define SKB_NFCT_PTRMASK ~(7UL)
Eric Dumazet7fee2262010-05-11 23:19:48 +0000827/**
828 * skb_dst - returns skb dst_entry
829 * @skb: buffer
830 *
831 * Returns skb dst_entry, regardless of reference taken or not.
832 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000833static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
834{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000835 /* If refdst was not refcounted, check we still are in a
836 * rcu_read_lock section
837 */
838 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
839 !rcu_read_lock_held() &&
840 !rcu_read_lock_bh_held());
841 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000842}
843
Eric Dumazet7fee2262010-05-11 23:19:48 +0000844/**
845 * skb_dst_set - sets skb dst
846 * @skb: buffer
847 * @dst: dst entry
848 *
849 * Sets skb dst, assuming a reference was taken on dst and should
850 * be released by skb_dst_drop()
851 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000852static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
853{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000854 skb->_skb_refdst = (unsigned long)dst;
855}
856
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200857/**
858 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
859 * @skb: buffer
860 * @dst: dst entry
861 *
862 * Sets skb dst, assuming a reference was not taken on dst.
863 * If dst entry is cached, we do not take reference and dst_release
864 * will be avoided by refdst_drop. If dst entry is not cached, we take
865 * reference, so that last dst_release can destroy the dst immediately.
866 */
867static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
868{
Hannes Frederic Sowadbfc4fb2014-12-06 19:19:42 +0100869 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
870 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200871}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000872
873/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300874 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000875 * @skb: buffer
876 */
877static inline bool skb_dst_is_noref(const struct sk_buff *skb)
878{
879 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000880}
881
Eric Dumazet511c3f92009-06-02 05:14:27 +0000882static inline struct rtable *skb_rtable(const struct sk_buff *skb)
883{
Eric Dumazetadf30902009-06-02 05:19:30 +0000884 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000885}
886
Jamal Hadi Salim8b10cab2016-07-02 06:43:14 -0400887/* For mangling skb->pkt_type from user space side from applications
888 * such as nft, tc, etc, we only allow a conservative subset of
889 * possible pkt_types to be set.
890*/
891static inline bool skb_pkt_type_ok(u32 ptype)
892{
893 return ptype <= PACKET_OTHERHOST;
894}
895
Miroslav Lichvar90b602f2017-05-19 17:52:37 +0200896static inline unsigned int skb_napi_id(const struct sk_buff *skb)
897{
898#ifdef CONFIG_NET_RX_BUSY_POLL
899 return skb->napi_id;
900#else
901 return 0;
902#endif
903}
904
Paolo Abeni3889a8032017-06-12 11:23:41 +0200905/* decrement the reference count and return true if we can free the skb */
906static inline bool skb_unref(struct sk_buff *skb)
907{
908 if (unlikely(!skb))
909 return false;
Reshetova, Elena63354792017-06-30 13:07:58 +0300910 if (likely(refcount_read(&skb->users) == 1))
Paolo Abeni3889a8032017-06-12 11:23:41 +0200911 smp_rmb();
Reshetova, Elena63354792017-06-30 13:07:58 +0300912 else if (likely(!refcount_dec_and_test(&skb->users)))
Paolo Abeni3889a8032017-06-12 11:23:41 +0200913 return false;
914
915 return true;
916}
917
Paolo Abeni0a463c72017-06-12 11:23:42 +0200918void skb_release_head_state(struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -0700919void kfree_skb(struct sk_buff *skb);
920void kfree_skb_list(struct sk_buff *segs);
921void skb_tx_error(struct sk_buff *skb);
922void consume_skb(struct sk_buff *skb);
Paolo Abeni0a463c72017-06-12 11:23:42 +0200923void consume_stateless_skb(struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -0700924void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +0000925extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000926
Joe Perches7965bd42013-09-26 14:48:15 -0700927void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
928bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
929 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +0000930
Joe Perches7965bd42013-09-26 14:48:15 -0700931struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
932 int node);
Eric Dumazet2ea2f622015-04-24 16:05:01 -0700933struct sk_buff *__build_skb(void *data, unsigned int frag_size);
Joe Perches7965bd42013-09-26 14:48:15 -0700934struct sk_buff *build_skb(void *data, unsigned int frag_size);
David S. Millerd179cd12005-08-17 14:57:30 -0700935static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100936 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700937{
Eric Dumazet564824b2010-10-11 19:05:25 +0000938 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700939}
940
Eric Dumazet2e4e4412014-09-17 04:49:49 -0700941struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
942 unsigned long data_len,
943 int max_page_order,
944 int *errcode,
945 gfp_t gfp_mask);
946
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700947/* Layout of fast clones : [skb1][skb2][fclone_ref] */
948struct sk_buff_fclones {
949 struct sk_buff skb1;
950
951 struct sk_buff skb2;
952
Reshetova, Elena26385952017-06-30 13:07:59 +0300953 refcount_t fclone_ref;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700954};
955
956/**
957 * skb_fclone_busy - check if fclone is busy
Stephen Hemminger293de7d2016-10-23 09:28:29 -0700958 * @sk: socket
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700959 * @skb: buffer
960 *
Masanari Iidabda13fe2015-12-13 16:53:02 +0900961 * Returns true if skb is a fast clone, and its clone is not freed.
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700962 * Some drivers call skb_orphan() in their ndo_start_xmit(),
963 * so we also check that this didnt happen.
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700964 */
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700965static inline bool skb_fclone_busy(const struct sock *sk,
966 const struct sk_buff *skb)
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700967{
968 const struct sk_buff_fclones *fclones;
969
970 fclones = container_of(skb, struct sk_buff_fclones, skb1);
971
972 return skb->fclone == SKB_FCLONE_ORIG &&
Reshetova, Elena26385952017-06-30 13:07:59 +0300973 refcount_read(&fclones->fclone_ref) > 1 &&
Eric Dumazet39bb5e62014-10-30 10:32:34 -0700974 fclones->skb2.sk == sk;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -0700975}
976
David S. Millerd179cd12005-08-17 14:57:30 -0700977static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +0100978 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -0700979{
Mel Gormanc93bdd02012-07-31 16:44:19 -0700980 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -0700981}
982
Joe Perches7965bd42013-09-26 14:48:15 -0700983struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
984int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
985struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
986struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
Octavian Purdilabad93e92014-06-12 01:36:26 +0300987struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
988 gfp_t gfp_mask, bool fclone);
989static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
990 gfp_t gfp_mask)
991{
992 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
993}
Eric Dumazet117632e2011-12-03 21:39:53 +0000994
Joe Perches7965bd42013-09-26 14:48:15 -0700995int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
996struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
997 unsigned int headroom);
998struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
999 int newtailroom, gfp_t priority);
Jason A. Donenfeld48a1df62017-06-04 04:16:22 +02001000int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1001 int offset, int len);
1002int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1003 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001004int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
1005int skb_pad(struct sk_buff *skb, int pad);
Neil Hormanead2ceb2009-03-11 09:49:55 +00001006#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001007
Joe Perches7965bd42013-09-26 14:48:15 -07001008int skb_append_datato_frags(struct sock *sk, struct sk_buff *skb,
1009 int getfrag(void *from, char *to, int offset,
1010 int len, int odd, struct sk_buff *skb),
1011 void *from, int length);
Ananda Rajue89e9cf2005-10-18 15:46:41 -07001012
Hannes Frederic Sowabe12a1f2015-05-21 16:59:58 +02001013int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1014 int offset, size_t size);
1015
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08001016struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -07001017 __u32 lower_offset;
1018 __u32 upper_offset;
1019 __u32 frag_idx;
1020 __u32 stepped_offset;
1021 struct sk_buff *root_skb;
1022 struct sk_buff *cur_skb;
1023 __u8 *frag_data;
1024};
1025
Joe Perches7965bd42013-09-26 14:48:15 -07001026void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1027 unsigned int to, struct skb_seq_state *st);
1028unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1029 struct skb_seq_state *st);
1030void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -07001031
Joe Perches7965bd42013-09-26 14:48:15 -07001032unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
Bojan Prtvar059a2442015-02-22 11:46:35 +01001033 unsigned int to, struct ts_config *config);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -07001034
Tom Herbert09323cc2013-12-15 22:16:19 -08001035/*
1036 * Packet hash types specify the type of hash in skb_set_hash.
1037 *
1038 * Hash types refer to the protocol layer addresses which are used to
1039 * construct a packet's hash. The hashes are used to differentiate or identify
1040 * flows of the protocol layer for the hash type. Hash types are either
1041 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1042 *
1043 * Properties of hashes:
1044 *
1045 * 1) Two packets in different flows have different hash values
1046 * 2) Two packets in the same flow should have the same hash value
1047 *
1048 * A hash at a higher layer is considered to be more specific. A driver should
1049 * set the most specific hash possible.
1050 *
1051 * A driver cannot indicate a more specific hash than the layer at which a hash
1052 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1053 *
1054 * A driver may indicate a hash level which is less specific than the
1055 * actual layer the hash was computed on. For instance, a hash computed
1056 * at L4 may be considered an L3 hash. This should only be done if the
1057 * driver can't unambiguously determine that the HW computed the hash at
1058 * the higher layer. Note that the "should" in the second property above
1059 * permits this.
1060 */
1061enum pkt_hash_types {
1062 PKT_HASH_TYPE_NONE, /* Undefined type */
1063 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1064 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1065 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1066};
1067
Tom Herbertbcc83832015-09-01 09:24:24 -07001068static inline void skb_clear_hash(struct sk_buff *skb)
1069{
1070 skb->hash = 0;
1071 skb->sw_hash = 0;
1072 skb->l4_hash = 0;
1073}
1074
1075static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1076{
1077 if (!skb->l4_hash)
1078 skb_clear_hash(skb);
1079}
1080
1081static inline void
1082__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1083{
1084 skb->l4_hash = is_l4;
1085 skb->sw_hash = is_sw;
1086 skb->hash = hash;
1087}
1088
Tom Herbert09323cc2013-12-15 22:16:19 -08001089static inline void
1090skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1091{
Tom Herbertbcc83832015-09-01 09:24:24 -07001092 /* Used by drivers to set hash from HW */
1093 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1094}
1095
1096static inline void
1097__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1098{
1099 __skb_set_hash(skb, hash, true, is_l4);
Tom Herbert09323cc2013-12-15 22:16:19 -08001100}
1101
Tom Herberte5276932015-09-01 09:24:23 -07001102void __skb_get_hash(struct sk_buff *skb);
Florian Westphalb9177832016-10-26 18:49:46 +02001103u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
Tom Herberte5276932015-09-01 09:24:23 -07001104u32 skb_get_poff(const struct sk_buff *skb);
1105u32 __skb_get_poff(const struct sk_buff *skb, void *data,
1106 const struct flow_keys *keys, int hlen);
1107__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1108 void *data, int hlen_proto);
1109
1110static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1111 int thoff, u8 ip_proto)
1112{
1113 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1114}
1115
1116void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1117 const struct flow_dissector_key *key,
1118 unsigned int key_count);
1119
1120bool __skb_flow_dissect(const struct sk_buff *skb,
1121 struct flow_dissector *flow_dissector,
1122 void *target_container,
Tom Herbertcd79a232015-09-01 09:24:27 -07001123 void *data, __be16 proto, int nhoff, int hlen,
1124 unsigned int flags);
Tom Herberte5276932015-09-01 09:24:23 -07001125
1126static inline bool skb_flow_dissect(const struct sk_buff *skb,
1127 struct flow_dissector *flow_dissector,
Tom Herbertcd79a232015-09-01 09:24:27 -07001128 void *target_container, unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001129{
1130 return __skb_flow_dissect(skb, flow_dissector, target_container,
Tom Herbertcd79a232015-09-01 09:24:27 -07001131 NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001132}
1133
1134static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
Tom Herbertcd79a232015-09-01 09:24:27 -07001135 struct flow_keys *flow,
1136 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001137{
1138 memset(flow, 0, sizeof(*flow));
1139 return __skb_flow_dissect(skb, &flow_keys_dissector, flow,
Tom Herbertcd79a232015-09-01 09:24:27 -07001140 NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001141}
1142
1143static inline bool skb_flow_dissect_flow_keys_buf(struct flow_keys *flow,
1144 void *data, __be16 proto,
Tom Herbertcd79a232015-09-01 09:24:27 -07001145 int nhoff, int hlen,
1146 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001147{
1148 memset(flow, 0, sizeof(*flow));
1149 return __skb_flow_dissect(NULL, &flow_keys_buf_dissector, flow,
Tom Herbertcd79a232015-09-01 09:24:27 -07001150 data, proto, nhoff, hlen, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001151}
1152
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001153static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001154{
Tom Herberta3b18dd2014-07-01 21:33:17 -07001155 if (!skb->l4_hash && !skb->sw_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001156 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001157
Tom Herbert61b905d2014-03-24 15:34:47 -07001158 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001159}
1160
David S. Miller20a17bf2015-09-01 21:19:17 -07001161static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
Tom Herbertf70ea012015-07-31 16:52:10 -07001162{
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001163 if (!skb->l4_hash && !skb->sw_hash) {
1164 struct flow_keys keys;
Tom Herbertde4c1f82015-09-01 18:11:04 -07001165 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001166
Tom Herbertde4c1f82015-09-01 18:11:04 -07001167 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001168 }
Tom Herbertf70ea012015-07-31 16:52:10 -07001169
1170 return skb->hash;
1171}
1172
Tom Herbert50fb7992015-05-01 11:30:12 -07001173__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
1174
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001175static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1176{
Tom Herbert61b905d2014-03-24 15:34:47 -07001177 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001178}
1179
Tom Herbert3df7a742013-12-15 22:16:29 -08001180static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1181{
Tom Herbert61b905d2014-03-24 15:34:47 -07001182 to->hash = from->hash;
Tom Herberta3b18dd2014-07-01 21:33:17 -07001183 to->sw_hash = from->sw_hash;
Tom Herbert61b905d2014-03-24 15:34:47 -07001184 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -08001185};
1186
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001187#ifdef NET_SKBUFF_DATA_USES_OFFSET
1188static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1189{
1190 return skb->head + skb->end;
1191}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001192
1193static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1194{
1195 return skb->end;
1196}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001197#else
1198static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1199{
1200 return skb->end;
1201}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001202
1203static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1204{
1205 return skb->end - skb->head;
1206}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001207#endif
1208
Linus Torvalds1da177e2005-04-16 15:20:36 -07001209/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001210#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001211
Patrick Ohlyac45f602009-02-12 05:03:37 +00001212static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1213{
1214 return &skb_shinfo(skb)->hwtstamps;
1215}
1216
Linus Torvalds1da177e2005-04-16 15:20:36 -07001217/**
1218 * skb_queue_empty - check if a queue is empty
1219 * @list: queue head
1220 *
1221 * Returns true if the queue is empty, false otherwise.
1222 */
1223static inline int skb_queue_empty(const struct sk_buff_head *list)
1224{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001225 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001226}
1227
1228/**
David S. Millerfc7ebb22008-09-23 00:34:07 -07001229 * skb_queue_is_last - check if skb is the last entry in the queue
1230 * @list: queue head
1231 * @skb: buffer
1232 *
1233 * Returns true if @skb is the last buffer on the list.
1234 */
1235static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1236 const struct sk_buff *skb)
1237{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001238 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -07001239}
1240
1241/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001242 * skb_queue_is_first - check if skb is the first entry in the queue
1243 * @list: queue head
1244 * @skb: buffer
1245 *
1246 * Returns true if @skb is the first buffer on the list.
1247 */
1248static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1249 const struct sk_buff *skb)
1250{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001251 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001252}
1253
1254/**
David S. Miller249c8b42008-09-23 00:44:42 -07001255 * skb_queue_next - return the next packet in the queue
1256 * @list: queue head
1257 * @skb: current buffer
1258 *
1259 * Return the next packet in @list after @skb. It is only valid to
1260 * call this if skb_queue_is_last() evaluates to false.
1261 */
1262static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1263 const struct sk_buff *skb)
1264{
1265 /* This BUG_ON may seem severe, but if we just return then we
1266 * are going to dereference garbage.
1267 */
1268 BUG_ON(skb_queue_is_last(list, skb));
1269 return skb->next;
1270}
1271
1272/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001273 * skb_queue_prev - return the prev packet in the queue
1274 * @list: queue head
1275 * @skb: current buffer
1276 *
1277 * Return the prev packet in @list before @skb. It is only valid to
1278 * call this if skb_queue_is_first() evaluates to false.
1279 */
1280static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1281 const struct sk_buff *skb)
1282{
1283 /* This BUG_ON may seem severe, but if we just return then we
1284 * are going to dereference garbage.
1285 */
1286 BUG_ON(skb_queue_is_first(list, skb));
1287 return skb->prev;
1288}
1289
1290/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001291 * skb_get - reference buffer
1292 * @skb: buffer to reference
1293 *
1294 * Makes another reference to a socket buffer and returns a pointer
1295 * to the buffer.
1296 */
1297static inline struct sk_buff *skb_get(struct sk_buff *skb)
1298{
Reshetova, Elena63354792017-06-30 13:07:58 +03001299 refcount_inc(&skb->users);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001300 return skb;
1301}
1302
1303/*
1304 * If users == 1, we are the only owner and are can avoid redundant
1305 * atomic change.
1306 */
1307
1308/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001309 * skb_cloned - is the buffer a clone
1310 * @skb: buffer to check
1311 *
1312 * Returns true if the buffer was generated with skb_clone() and is
1313 * one of multiple shared copies of the buffer. Cloned buffers are
1314 * shared data so must not be written to under normal circumstances.
1315 */
1316static inline int skb_cloned(const struct sk_buff *skb)
1317{
1318 return skb->cloned &&
1319 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1320}
1321
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001322static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1323{
Mel Gormand0164ad2015-11-06 16:28:21 -08001324 might_sleep_if(gfpflags_allow_blocking(pri));
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001325
1326 if (skb_cloned(skb))
1327 return pskb_expand_head(skb, 0, 0, pri);
1328
1329 return 0;
1330}
1331
Linus Torvalds1da177e2005-04-16 15:20:36 -07001332/**
1333 * skb_header_cloned - is the header a clone
1334 * @skb: buffer to check
1335 *
1336 * Returns true if modifying the header part of the buffer requires
1337 * the data to be copied.
1338 */
1339static inline int skb_header_cloned(const struct sk_buff *skb)
1340{
1341 int dataref;
1342
1343 if (!skb->cloned)
1344 return 0;
1345
1346 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1347 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1348 return dataref != 1;
1349}
1350
Eric Dumazet9580bf22016-04-30 10:19:29 -07001351static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1352{
1353 might_sleep_if(gfpflags_allow_blocking(pri));
1354
1355 if (skb_header_cloned(skb))
1356 return pskb_expand_head(skb, 0, 0, pri);
1357
1358 return 0;
1359}
1360
Linus Torvalds1da177e2005-04-16 15:20:36 -07001361/**
1362 * skb_header_release - release reference to header
1363 * @skb: buffer to operate on
1364 *
1365 * Drop a reference to the header part of the buffer. This is done
1366 * by acquiring a payload reference. You must not read from the header
1367 * part of skb->data after this.
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001368 * Note : Check if you can use __skb_header_release() instead.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001369 */
1370static inline void skb_header_release(struct sk_buff *skb)
1371{
1372 BUG_ON(skb->nohdr);
1373 skb->nohdr = 1;
1374 atomic_add(1 << SKB_DATAREF_SHIFT, &skb_shinfo(skb)->dataref);
1375}
1376
1377/**
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001378 * __skb_header_release - release reference to header
1379 * @skb: buffer to operate on
1380 *
1381 * Variant of skb_header_release() assuming skb is private to caller.
1382 * We can avoid one atomic operation.
1383 */
1384static inline void __skb_header_release(struct sk_buff *skb)
1385{
1386 skb->nohdr = 1;
1387 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1388}
1389
1390
1391/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001392 * skb_shared - is the buffer shared
1393 * @skb: buffer to check
1394 *
1395 * Returns true if more than one person has a reference to this
1396 * buffer.
1397 */
1398static inline int skb_shared(const struct sk_buff *skb)
1399{
Reshetova, Elena63354792017-06-30 13:07:58 +03001400 return refcount_read(&skb->users) != 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001401}
1402
1403/**
1404 * skb_share_check - check if buffer is shared and if so clone it
1405 * @skb: buffer to check
1406 * @pri: priority for memory allocation
1407 *
1408 * If the buffer is shared the buffer is cloned and the old copy
1409 * drops a reference. A new clone with a single reference is returned.
1410 * If the buffer is not shared the original buffer is returned. When
1411 * being called from interrupt status or with spinlocks held pri must
1412 * be GFP_ATOMIC.
1413 *
1414 * NULL is returned on a memory allocation failure.
1415 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001416static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001417{
Mel Gormand0164ad2015-11-06 16:28:21 -08001418 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001419 if (skb_shared(skb)) {
1420 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001421
1422 if (likely(nskb))
1423 consume_skb(skb);
1424 else
1425 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001426 skb = nskb;
1427 }
1428 return skb;
1429}
1430
1431/*
1432 * Copy shared buffers into a new sk_buff. We effectively do COW on
1433 * packets to handle cases where we have a local reader and forward
1434 * and a couple of other messy ones. The normal one is tcpdumping
1435 * a packet thats being forwarded.
1436 */
1437
1438/**
1439 * skb_unshare - make a copy of a shared buffer
1440 * @skb: buffer to check
1441 * @pri: priority for memory allocation
1442 *
1443 * If the socket buffer is a clone then this function creates a new
1444 * copy of the data, drops a reference count on the old copy and returns
1445 * the new copy with the reference count at 1. If the buffer is not a clone
1446 * the original buffer is returned. When called with a spinlock held or
1447 * from interrupt state @pri must be %GFP_ATOMIC
1448 *
1449 * %NULL is returned on a memory allocation failure.
1450 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001451static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001452 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001453{
Mel Gormand0164ad2015-11-06 16:28:21 -08001454 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001455 if (skb_cloned(skb)) {
1456 struct sk_buff *nskb = skb_copy(skb, pri);
Alexander Aring31eff812014-10-10 23:10:47 +02001457
1458 /* Free our shared copy */
1459 if (likely(nskb))
1460 consume_skb(skb);
1461 else
1462 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001463 skb = nskb;
1464 }
1465 return skb;
1466}
1467
1468/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001469 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001470 * @list_: list to peek at
1471 *
1472 * Peek an &sk_buff. Unlike most other operations you _MUST_
1473 * be careful with this one. A peek leaves the buffer on the
1474 * list and someone else may run off with it. You must hold
1475 * the appropriate locks or have a private queue to do this.
1476 *
1477 * Returns %NULL for an empty list or a pointer to the head element.
1478 * The reference count is not incremented and the reference is therefore
1479 * volatile. Use with caution.
1480 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001481static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001482{
Eric Dumazet18d07002012-04-30 16:31:46 +00001483 struct sk_buff *skb = list_->next;
1484
1485 if (skb == (struct sk_buff *)list_)
1486 skb = NULL;
1487 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001488}
1489
1490/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001491 * skb_peek_next - peek skb following the given one from a queue
1492 * @skb: skb to start from
1493 * @list_: list to peek at
1494 *
1495 * Returns %NULL when the end of the list is met or a pointer to the
1496 * next element. The reference count is not incremented and the
1497 * reference is therefore volatile. Use with caution.
1498 */
1499static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1500 const struct sk_buff_head *list_)
1501{
1502 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001503
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001504 if (next == (struct sk_buff *)list_)
1505 next = NULL;
1506 return next;
1507}
1508
1509/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001510 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001511 * @list_: list to peek at
1512 *
1513 * Peek an &sk_buff. Unlike most other operations you _MUST_
1514 * be careful with this one. A peek leaves the buffer on the
1515 * list and someone else may run off with it. You must hold
1516 * the appropriate locks or have a private queue to do this.
1517 *
1518 * Returns %NULL for an empty list or a pointer to the tail element.
1519 * The reference count is not incremented and the reference is therefore
1520 * volatile. Use with caution.
1521 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001522static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001523{
Eric Dumazet18d07002012-04-30 16:31:46 +00001524 struct sk_buff *skb = list_->prev;
1525
1526 if (skb == (struct sk_buff *)list_)
1527 skb = NULL;
1528 return skb;
1529
Linus Torvalds1da177e2005-04-16 15:20:36 -07001530}
1531
1532/**
1533 * skb_queue_len - get queue length
1534 * @list_: list to measure
1535 *
1536 * Return the length of an &sk_buff queue.
1537 */
1538static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1539{
1540 return list_->qlen;
1541}
1542
David S. Miller67fed452008-09-21 22:36:24 -07001543/**
1544 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1545 * @list: queue to initialize
1546 *
1547 * This initializes only the list and queue length aspects of
1548 * an sk_buff_head object. This allows to initialize the list
1549 * aspects of an sk_buff_head without reinitializing things like
1550 * the spinlock. It can also be used for on-stack sk_buff_head
1551 * objects where the spinlock is known to not be used.
1552 */
1553static inline void __skb_queue_head_init(struct sk_buff_head *list)
1554{
1555 list->prev = list->next = (struct sk_buff *)list;
1556 list->qlen = 0;
1557}
1558
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001559/*
1560 * This function creates a split out lock class for each invocation;
1561 * this is needed for now since a whole lot of users of the skb-queue
1562 * infrastructure in drivers have different locking usage (in hardirq)
1563 * than the networking core (in softirq only). In the long run either the
1564 * network layer or drivers should need annotation to consolidate the
1565 * main types of usage into 3 classes.
1566 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001567static inline void skb_queue_head_init(struct sk_buff_head *list)
1568{
1569 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001570 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001571}
1572
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001573static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1574 struct lock_class_key *class)
1575{
1576 skb_queue_head_init(list);
1577 lockdep_set_class(&list->lock, class);
1578}
1579
Linus Torvalds1da177e2005-04-16 15:20:36 -07001580/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001581 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001582 *
1583 * The "__skb_xxxx()" functions are the non-atomic ones that
1584 * can only be called with interrupts disabled.
1585 */
Joe Perches7965bd42013-09-26 14:48:15 -07001586void skb_insert(struct sk_buff *old, struct sk_buff *newsk,
1587 struct sk_buff_head *list);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001588static inline void __skb_insert(struct sk_buff *newsk,
1589 struct sk_buff *prev, struct sk_buff *next,
1590 struct sk_buff_head *list)
1591{
1592 newsk->next = next;
1593 newsk->prev = prev;
1594 next->prev = prev->next = newsk;
1595 list->qlen++;
1596}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001597
David S. Miller67fed452008-09-21 22:36:24 -07001598static inline void __skb_queue_splice(const struct sk_buff_head *list,
1599 struct sk_buff *prev,
1600 struct sk_buff *next)
1601{
1602 struct sk_buff *first = list->next;
1603 struct sk_buff *last = list->prev;
1604
1605 first->prev = prev;
1606 prev->next = first;
1607
1608 last->next = next;
1609 next->prev = last;
1610}
1611
1612/**
1613 * skb_queue_splice - join two skb lists, this is designed for stacks
1614 * @list: the new list to add
1615 * @head: the place to add it in the first list
1616 */
1617static inline void skb_queue_splice(const struct sk_buff_head *list,
1618 struct sk_buff_head *head)
1619{
1620 if (!skb_queue_empty(list)) {
1621 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001622 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001623 }
1624}
1625
1626/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001627 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001628 * @list: the new list to add
1629 * @head: the place to add it in the first list
1630 *
1631 * The list at @list is reinitialised
1632 */
1633static inline void skb_queue_splice_init(struct sk_buff_head *list,
1634 struct sk_buff_head *head)
1635{
1636 if (!skb_queue_empty(list)) {
1637 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001638 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001639 __skb_queue_head_init(list);
1640 }
1641}
1642
1643/**
1644 * skb_queue_splice_tail - join two skb lists, each list being a queue
1645 * @list: the new list to add
1646 * @head: the place to add it in the first list
1647 */
1648static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1649 struct sk_buff_head *head)
1650{
1651 if (!skb_queue_empty(list)) {
1652 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001653 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001654 }
1655}
1656
1657/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001658 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001659 * @list: the new list to add
1660 * @head: the place to add it in the first list
1661 *
1662 * Each of the lists is a queue.
1663 * The list at @list is reinitialised
1664 */
1665static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1666 struct sk_buff_head *head)
1667{
1668 if (!skb_queue_empty(list)) {
1669 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001670 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001671 __skb_queue_head_init(list);
1672 }
1673}
1674
Linus Torvalds1da177e2005-04-16 15:20:36 -07001675/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001676 * __skb_queue_after - queue a buffer at the list head
1677 * @list: list to use
1678 * @prev: place after this buffer
1679 * @newsk: buffer to queue
1680 *
1681 * Queue a buffer int the middle of a list. This function takes no locks
1682 * and you must therefore hold required locks before calling it.
1683 *
1684 * A buffer cannot be placed on two lists at the same time.
1685 */
1686static inline void __skb_queue_after(struct sk_buff_head *list,
1687 struct sk_buff *prev,
1688 struct sk_buff *newsk)
1689{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001690 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001691}
1692
Joe Perches7965bd42013-09-26 14:48:15 -07001693void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1694 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001695
Gerrit Renkerf5572852008-04-14 00:05:28 -07001696static inline void __skb_queue_before(struct sk_buff_head *list,
1697 struct sk_buff *next,
1698 struct sk_buff *newsk)
1699{
1700 __skb_insert(newsk, next->prev, next, list);
1701}
1702
Stephen Hemminger300ce172005-10-30 13:47:34 -08001703/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001704 * __skb_queue_head - queue a buffer at the list head
1705 * @list: list to use
1706 * @newsk: buffer to queue
1707 *
1708 * Queue a buffer at the start of a list. This function takes no locks
1709 * and you must therefore hold required locks before calling it.
1710 *
1711 * A buffer cannot be placed on two lists at the same time.
1712 */
Joe Perches7965bd42013-09-26 14:48:15 -07001713void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001714static inline void __skb_queue_head(struct sk_buff_head *list,
1715 struct sk_buff *newsk)
1716{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001717 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001718}
1719
1720/**
1721 * __skb_queue_tail - queue a buffer at the list tail
1722 * @list: list to use
1723 * @newsk: buffer to queue
1724 *
1725 * Queue a buffer at the end of a list. This function takes no locks
1726 * and you must therefore hold required locks before calling it.
1727 *
1728 * A buffer cannot be placed on two lists at the same time.
1729 */
Joe Perches7965bd42013-09-26 14:48:15 -07001730void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001731static inline void __skb_queue_tail(struct sk_buff_head *list,
1732 struct sk_buff *newsk)
1733{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001734 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001735}
1736
Linus Torvalds1da177e2005-04-16 15:20:36 -07001737/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001738 * remove sk_buff from list. _Must_ be called atomically, and with
1739 * the list known..
1740 */
Joe Perches7965bd42013-09-26 14:48:15 -07001741void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001742static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1743{
1744 struct sk_buff *next, *prev;
1745
1746 list->qlen--;
1747 next = skb->next;
1748 prev = skb->prev;
1749 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001750 next->prev = prev;
1751 prev->next = next;
1752}
1753
Gerrit Renkerf525c062008-04-14 00:04:12 -07001754/**
1755 * __skb_dequeue - remove from the head of the queue
1756 * @list: list to dequeue from
1757 *
1758 * Remove the head of the list. This function does not take any locks
1759 * so must be used with appropriate locks held only. The head item is
1760 * returned or %NULL if the list is empty.
1761 */
Joe Perches7965bd42013-09-26 14:48:15 -07001762struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Gerrit Renkerf525c062008-04-14 00:04:12 -07001763static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
1764{
1765 struct sk_buff *skb = skb_peek(list);
1766 if (skb)
1767 __skb_unlink(skb, list);
1768 return skb;
1769}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001770
1771/**
1772 * __skb_dequeue_tail - remove from the tail of the queue
1773 * @list: list to dequeue from
1774 *
1775 * Remove the tail of the list. This function does not take any locks
1776 * so must be used with appropriate locks held only. The tail item is
1777 * returned or %NULL if the list is empty.
1778 */
Joe Perches7965bd42013-09-26 14:48:15 -07001779struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001780static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
1781{
1782 struct sk_buff *skb = skb_peek_tail(list);
1783 if (skb)
1784 __skb_unlink(skb, list);
1785 return skb;
1786}
1787
1788
David S. Millerbdcc0922012-03-07 20:53:36 -05001789static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001790{
1791 return skb->data_len;
1792}
1793
1794static inline unsigned int skb_headlen(const struct sk_buff *skb)
1795{
1796 return skb->len - skb->data_len;
1797}
1798
Willem de Bruijn3ece7822017-08-03 16:29:38 -04001799static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001800{
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03001801 unsigned int i, len = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001802
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03001803 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00001804 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Willem de Bruijn3ece7822017-08-03 16:29:38 -04001805 return len;
1806}
1807
1808static inline unsigned int skb_pagelen(const struct sk_buff *skb)
1809{
1810 return skb_headlen(skb) + __skb_pagelen(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001811}
1812
Ian Campbell131ea662011-08-19 06:25:00 +00001813/**
1814 * __skb_fill_page_desc - initialise a paged fragment in an skb
1815 * @skb: buffer containing fragment to be initialised
1816 * @i: paged fragment index to initialise
1817 * @page: the page to use for this fragment
1818 * @off: the offset to the data with @page
1819 * @size: the length of the data
1820 *
1821 * Initialises the @i'th fragment of @skb to point to &size bytes at
1822 * offset @off within @page.
1823 *
1824 * Does not take any additional reference on the fragment.
1825 */
1826static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
1827 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001828{
1829 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1830
Mel Gormanc48a11c2012-07-31 16:44:23 -07001831 /*
Michal Hocko2f064f32015-08-21 14:11:51 -07001832 * Propagate page pfmemalloc to the skb if we can. The problem is
1833 * that not all callers have unique ownership of the page but rely
1834 * on page_is_pfmemalloc doing the right thing(tm).
Mel Gormanc48a11c2012-07-31 16:44:23 -07001835 */
Ian Campbella8605c62011-10-19 23:01:49 +00001836 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001837 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00001838 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001839
1840 page = compound_head(page);
Michal Hocko2f064f32015-08-21 14:11:51 -07001841 if (page_is_pfmemalloc(page))
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00001842 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00001843}
1844
1845/**
1846 * skb_fill_page_desc - initialise a paged fragment in an skb
1847 * @skb: buffer containing fragment to be initialised
1848 * @i: paged fragment index to initialise
1849 * @page: the page to use for this fragment
1850 * @off: the offset to the data with @page
1851 * @size: the length of the data
1852 *
1853 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01001854 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00001855 * addition updates @skb such that @i is the last fragment.
1856 *
1857 * Does not take any additional reference on the fragment.
1858 */
1859static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
1860 struct page *page, int off, int size)
1861{
1862 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001863 skb_shinfo(skb)->nr_frags = i + 1;
1864}
1865
Joe Perches7965bd42013-09-26 14:48:15 -07001866void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
1867 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07001868
Jason Wangf8e617e2013-11-01 14:07:47 +08001869void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
1870 unsigned int truesize);
1871
Linus Torvalds1da177e2005-04-16 15:20:36 -07001872#define SKB_PAGE_ASSERT(skb) BUG_ON(skb_shinfo(skb)->nr_frags)
David S. Miller21dc3302010-08-23 00:13:46 -07001873#define SKB_FRAG_ASSERT(skb) BUG_ON(skb_has_frag_list(skb))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001874#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
1875
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001876#ifdef NET_SKBUFF_DATA_USES_OFFSET
1877static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1878{
1879 return skb->head + skb->tail;
1880}
1881
1882static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1883{
1884 skb->tail = skb->data - skb->head;
1885}
1886
1887static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1888{
1889 skb_reset_tail_pointer(skb);
1890 skb->tail += offset;
1891}
Simon Horman7cc46192013-05-28 20:34:29 +00001892
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001893#else /* NET_SKBUFF_DATA_USES_OFFSET */
1894static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
1895{
1896 return skb->tail;
1897}
1898
1899static inline void skb_reset_tail_pointer(struct sk_buff *skb)
1900{
1901 skb->tail = skb->data;
1902}
1903
1904static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
1905{
1906 skb->tail = skb->data + offset;
1907}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001908
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07001909#endif /* NET_SKBUFF_DATA_USES_OFFSET */
1910
Linus Torvalds1da177e2005-04-16 15:20:36 -07001911/*
1912 * Add data to an sk_buff
1913 */
Johannes Berg4df864c2017-06-16 14:29:21 +02001914void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
1915void *skb_put(struct sk_buff *skb, unsigned int len);
1916static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001917{
Johannes Berg4df864c2017-06-16 14:29:21 +02001918 void *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001919 SKB_LINEAR_ASSERT(skb);
1920 skb->tail += len;
1921 skb->len += len;
1922 return tmp;
1923}
1924
yuan linyude77b962017-06-18 22:48:17 +08001925static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
1926{
1927 void *tmp = __skb_put(skb, len);
1928
1929 memset(tmp, 0, len);
1930 return tmp;
1931}
1932
1933static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
1934 unsigned int len)
1935{
1936 void *tmp = __skb_put(skb, len);
1937
1938 memcpy(tmp, data, len);
1939 return tmp;
1940}
1941
1942static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
1943{
1944 *(u8 *)__skb_put(skb, 1) = val;
1945}
1946
Johannes Berg83ad3572017-06-14 22:17:20 +02001947static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
Johannes Berge45a79d2017-05-24 09:07:47 +02001948{
Johannes Berg83ad3572017-06-14 22:17:20 +02001949 void *tmp = skb_put(skb, len);
Johannes Berge45a79d2017-05-24 09:07:47 +02001950
1951 memset(tmp, 0, len);
1952
1953 return tmp;
1954}
1955
Johannes Berg59ae1d12017-06-16 14:29:20 +02001956static inline void *skb_put_data(struct sk_buff *skb, const void *data,
1957 unsigned int len)
1958{
1959 void *tmp = skb_put(skb, len);
1960
1961 memcpy(tmp, data, len);
1962
1963 return tmp;
1964}
1965
Johannes Berg634fef62017-06-16 14:29:24 +02001966static inline void skb_put_u8(struct sk_buff *skb, u8 val)
1967{
1968 *(u8 *)skb_put(skb, 1) = val;
1969}
1970
Johannes Bergd58ff352017-06-16 14:29:23 +02001971void *skb_push(struct sk_buff *skb, unsigned int len);
1972static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001973{
1974 skb->data -= len;
1975 skb->len += len;
1976 return skb->data;
1977}
1978
Johannes Bergaf728682017-06-16 14:29:22 +02001979void *skb_pull(struct sk_buff *skb, unsigned int len);
1980static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001981{
1982 skb->len -= len;
1983 BUG_ON(skb->len < skb->data_len);
1984 return skb->data += len;
1985}
1986
Johannes Bergaf728682017-06-16 14:29:22 +02001987static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
David S. Miller47d29642010-05-02 02:21:44 -07001988{
1989 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
1990}
1991
Johannes Bergaf728682017-06-16 14:29:22 +02001992void *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001993
Johannes Bergaf728682017-06-16 14:29:22 +02001994static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001995{
1996 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07001997 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001998 return NULL;
1999 skb->len -= len;
2000 return skb->data += len;
2001}
2002
Johannes Bergaf728682017-06-16 14:29:22 +02002003static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002004{
2005 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2006}
2007
2008static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
2009{
2010 if (likely(len <= skb_headlen(skb)))
2011 return 1;
2012 if (unlikely(len > skb->len))
2013 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002014 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002015}
2016
Eric Dumazetc8c8b122016-12-07 09:19:33 -08002017void skb_condense(struct sk_buff *skb);
2018
Linus Torvalds1da177e2005-04-16 15:20:36 -07002019/**
2020 * skb_headroom - bytes at buffer head
2021 * @skb: buffer to check
2022 *
2023 * Return the number of bytes of free space at the head of an &sk_buff.
2024 */
Chuck Leverc2636b42007-10-23 21:07:32 -07002025static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002026{
2027 return skb->data - skb->head;
2028}
2029
2030/**
2031 * skb_tailroom - bytes at buffer end
2032 * @skb: buffer to check
2033 *
2034 * Return the number of bytes of free space at the tail of an sk_buff
2035 */
2036static inline int skb_tailroom(const struct sk_buff *skb)
2037{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07002038 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002039}
2040
2041/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00002042 * skb_availroom - bytes at buffer end
2043 * @skb: buffer to check
2044 *
2045 * Return the number of bytes of free space at the tail of an sk_buff
2046 * allocated by sk_stream_alloc()
2047 */
2048static inline int skb_availroom(const struct sk_buff *skb)
2049{
Eric Dumazet16fad692013-03-14 05:40:32 +00002050 if (skb_is_nonlinear(skb))
2051 return 0;
2052
2053 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00002054}
2055
2056/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002057 * skb_reserve - adjust headroom
2058 * @skb: buffer to alter
2059 * @len: bytes to move
2060 *
2061 * Increase the headroom of an empty &sk_buff by reducing the tail
2062 * room. This is only allowed for an empty buffer.
2063 */
David S. Miller8243126c2006-01-17 02:54:21 -08002064static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002065{
2066 skb->data += len;
2067 skb->tail += len;
2068}
2069
Benjamin Poirier1837b2e2016-02-29 15:03:33 -08002070/**
2071 * skb_tailroom_reserve - adjust reserved_tailroom
2072 * @skb: buffer to alter
2073 * @mtu: maximum amount of headlen permitted
2074 * @needed_tailroom: minimum amount of reserved_tailroom
2075 *
2076 * Set reserved_tailroom so that headlen can be as large as possible but
2077 * not larger than mtu and tailroom cannot be smaller than
2078 * needed_tailroom.
2079 * The required headroom should already have been reserved before using
2080 * this function.
2081 */
2082static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2083 unsigned int needed_tailroom)
2084{
2085 SKB_LINEAR_ASSERT(skb);
2086 if (mtu < skb_tailroom(skb) - needed_tailroom)
2087 /* use at most mtu */
2088 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2089 else
2090 /* use up to all available space */
2091 skb->reserved_tailroom = needed_tailroom;
2092}
2093
Tom Herbert8bce6d72014-09-29 20:22:29 -07002094#define ENCAP_TYPE_ETHER 0
2095#define ENCAP_TYPE_IPPROTO 1
2096
2097static inline void skb_set_inner_protocol(struct sk_buff *skb,
2098 __be16 protocol)
2099{
2100 skb->inner_protocol = protocol;
2101 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2102}
2103
2104static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2105 __u8 ipproto)
2106{
2107 skb->inner_ipproto = ipproto;
2108 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2109}
2110
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002111static inline void skb_reset_inner_headers(struct sk_buff *skb)
2112{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002113 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002114 skb->inner_network_header = skb->network_header;
2115 skb->inner_transport_header = skb->transport_header;
2116}
2117
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00002118static inline void skb_reset_mac_len(struct sk_buff *skb)
2119{
2120 skb->mac_len = skb->network_header - skb->mac_header;
2121}
2122
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002123static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2124 *skb)
2125{
2126 return skb->head + skb->inner_transport_header;
2127}
2128
Tom Herbert55dc5a92015-12-14 11:19:40 -08002129static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2130{
2131 return skb_inner_transport_header(skb) - skb->data;
2132}
2133
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002134static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2135{
2136 skb->inner_transport_header = skb->data - skb->head;
2137}
2138
2139static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2140 const int offset)
2141{
2142 skb_reset_inner_transport_header(skb);
2143 skb->inner_transport_header += offset;
2144}
2145
2146static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2147{
2148 return skb->head + skb->inner_network_header;
2149}
2150
2151static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2152{
2153 skb->inner_network_header = skb->data - skb->head;
2154}
2155
2156static inline void skb_set_inner_network_header(struct sk_buff *skb,
2157 const int offset)
2158{
2159 skb_reset_inner_network_header(skb);
2160 skb->inner_network_header += offset;
2161}
2162
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002163static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2164{
2165 return skb->head + skb->inner_mac_header;
2166}
2167
2168static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2169{
2170 skb->inner_mac_header = skb->data - skb->head;
2171}
2172
2173static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2174 const int offset)
2175{
2176 skb_reset_inner_mac_header(skb);
2177 skb->inner_mac_header += offset;
2178}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002179static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2180{
Cong Wang35d04612013-05-29 15:16:05 +08002181 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002182}
2183
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002184static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2185{
2186 return skb->head + skb->transport_header;
2187}
2188
2189static inline void skb_reset_transport_header(struct sk_buff *skb)
2190{
2191 skb->transport_header = skb->data - skb->head;
2192}
2193
2194static inline void skb_set_transport_header(struct sk_buff *skb,
2195 const int offset)
2196{
2197 skb_reset_transport_header(skb);
2198 skb->transport_header += offset;
2199}
2200
2201static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2202{
2203 return skb->head + skb->network_header;
2204}
2205
2206static inline void skb_reset_network_header(struct sk_buff *skb)
2207{
2208 skb->network_header = skb->data - skb->head;
2209}
2210
2211static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2212{
2213 skb_reset_network_header(skb);
2214 skb->network_header += offset;
2215}
2216
2217static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2218{
2219 return skb->head + skb->mac_header;
2220}
2221
Amir Vadaiea6da4f2017-02-07 09:56:06 +02002222static inline int skb_mac_offset(const struct sk_buff *skb)
2223{
2224 return skb_mac_header(skb) - skb->data;
2225}
2226
Daniel Borkmann0daf4342017-07-02 02:13:25 +02002227static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2228{
2229 return skb->network_header - skb->mac_header;
2230}
2231
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002232static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2233{
Cong Wang35d04612013-05-29 15:16:05 +08002234 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002235}
2236
2237static inline void skb_reset_mac_header(struct sk_buff *skb)
2238{
2239 skb->mac_header = skb->data - skb->head;
2240}
2241
2242static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2243{
2244 skb_reset_mac_header(skb);
2245 skb->mac_header += offset;
2246}
2247
Timo Teräs0e3da5b2013-12-16 11:02:09 +02002248static inline void skb_pop_mac_header(struct sk_buff *skb)
2249{
2250 skb->mac_header = skb->network_header;
2251}
2252
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002253static inline void skb_probe_transport_header(struct sk_buff *skb,
2254 const int offset_hint)
2255{
2256 struct flow_keys keys;
2257
2258 if (skb_transport_header_was_set(skb))
2259 return;
Tom Herbertcd79a232015-09-01 09:24:27 -07002260 else if (skb_flow_dissect_flow_keys(skb, &keys, 0))
Tom Herbert42aecaa2015-06-04 09:16:39 -07002261 skb_set_transport_header(skb, keys.control.thoff);
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002262 else
2263 skb_set_transport_header(skb, offset_hint);
2264}
2265
Eric Dumazet03606892012-02-23 10:55:02 +00002266static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2267{
2268 if (skb_mac_header_was_set(skb)) {
2269 const unsigned char *old_mac = skb_mac_header(skb);
2270
2271 skb_set_mac_header(skb, -skb->mac_len);
2272 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2273 }
2274}
2275
Michał Mirosław04fb4512010-12-14 15:24:08 +00002276static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2277{
2278 return skb->csum_start - skb_headroom(skb);
2279}
2280
Alexander Duyck08b64fc2016-02-05 15:27:49 -08002281static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2282{
2283 return skb->head + skb->csum_start;
2284}
2285
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002286static inline int skb_transport_offset(const struct sk_buff *skb)
2287{
2288 return skb_transport_header(skb) - skb->data;
2289}
2290
2291static inline u32 skb_network_header_len(const struct sk_buff *skb)
2292{
2293 return skb->transport_header - skb->network_header;
2294}
2295
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002296static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2297{
2298 return skb->inner_transport_header - skb->inner_network_header;
2299}
2300
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002301static inline int skb_network_offset(const struct sk_buff *skb)
2302{
2303 return skb_network_header(skb) - skb->data;
2304}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03002305
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002306static inline int skb_inner_network_offset(const struct sk_buff *skb)
2307{
2308 return skb_inner_network_header(skb) - skb->data;
2309}
2310
Changli Gaof9599ce2010-08-04 04:43:44 +00002311static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2312{
2313 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2314}
2315
Linus Torvalds1da177e2005-04-16 15:20:36 -07002316/*
2317 * CPUs often take a performance hit when accessing unaligned memory
2318 * locations. The actual performance hit varies, it can be small if the
2319 * hardware handles it or large if we have to take an exception and fix it
2320 * in software.
2321 *
2322 * Since an ethernet header is 14 bytes network drivers often end up with
2323 * the IP header at an unaligned offset. The IP header can be aligned by
2324 * shifting the start of the packet by 2 bytes. Drivers should do this
2325 * with:
2326 *
Tobias Klauser8660c122009-07-13 22:48:16 +00002327 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002328 *
2329 * The downside to this alignment of the IP header is that the DMA is now
2330 * unaligned. On some architectures the cost of an unaligned DMA is high
2331 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00002332 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002333 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2334 * to be overridden.
2335 */
2336#ifndef NET_IP_ALIGN
2337#define NET_IP_ALIGN 2
2338#endif
2339
Anton Blanchard025be812006-03-31 02:27:06 -08002340/*
2341 * The networking layer reserves some headroom in skb data (via
2342 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2343 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08002344 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08002345 *
2346 * Unfortunately this headroom changes the DMA alignment of the resulting
2347 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2348 * on some architectures. An architecture can override this value,
2349 * perhaps setting it to a cacheline in size (since that will maintain
2350 * cacheline alignment of the DMA). It must be a power of 2.
2351 *
David S. Millerd6301d32009-02-08 19:24:13 -08002352 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08002353 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07002354 *
2355 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2356 * to reduce average number of cache lines per packet.
2357 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07002358 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08002359 */
2360#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07002361#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08002362#endif
2363
Joe Perches7965bd42013-09-26 14:48:15 -07002364int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002365
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002366static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002367{
Emmanuel Grumbachc4264f22011-05-21 19:46:09 +00002368 if (unlikely(skb_is_nonlinear(skb))) {
Herbert Xu3cc0e872006-06-09 16:13:38 -07002369 WARN_ON(1);
2370 return;
2371 }
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002372 skb->len = len;
2373 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002374}
2375
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002376static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2377{
2378 __skb_set_length(skb, len);
2379}
2380
Joe Perches7965bd42013-09-26 14:48:15 -07002381void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002382
2383static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2384{
Herbert Xu3cc0e872006-06-09 16:13:38 -07002385 if (skb->data_len)
2386 return ___pskb_trim(skb, len);
2387 __skb_trim(skb, len);
2388 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002389}
2390
2391static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2392{
2393 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2394}
2395
2396/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07002397 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2398 * @skb: buffer to alter
2399 * @len: new length
2400 *
2401 * This is identical to pskb_trim except that the caller knows that
2402 * the skb is not cloned so we should never get an error due to out-
2403 * of-memory.
2404 */
2405static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2406{
2407 int err = pskb_trim(skb, len);
2408 BUG_ON(err);
2409}
2410
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002411static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2412{
2413 unsigned int diff = len - skb->len;
2414
2415 if (skb_tailroom(skb) < diff) {
2416 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2417 GFP_ATOMIC);
2418 if (ret)
2419 return ret;
2420 }
2421 __skb_set_length(skb, len);
2422 return 0;
2423}
2424
Herbert Xue9fa4f72006-08-13 20:12:58 -07002425/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002426 * skb_orphan - orphan a buffer
2427 * @skb: buffer to orphan
2428 *
2429 * If a buffer currently has an owner then we call the owner's
2430 * destructor function and make the @skb unowned. The buffer continues
2431 * to exist but is no longer charged to its former owner.
2432 */
2433static inline void skb_orphan(struct sk_buff *skb)
2434{
Eric Dumazetc34a7612013-07-30 16:11:15 -07002435 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002436 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002437 skb->destructor = NULL;
2438 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07002439 } else {
2440 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002441 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002442}
2443
2444/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002445 * skb_orphan_frags - orphan the frags contained in a buffer
2446 * @skb: buffer to orphan frags from
2447 * @gfp_mask: allocation mask for replacement pages
2448 *
2449 * For each frag in the SKB which needs a destructor (i.e. has an
2450 * owner) create a copy of that frag and release the original
2451 * page by calling the destructor.
2452 */
2453static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2454{
2455 if (likely(!(skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY)))
2456 return 0;
2457 return skb_copy_ubufs(skb, gfp_mask);
2458}
2459
2460/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002461 * __skb_queue_purge - empty a list
2462 * @list: list to empty
2463 *
2464 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2465 * the list and one reference dropped. This function does not take the
2466 * list lock and the caller must hold the relevant locks to use it.
2467 */
Joe Perches7965bd42013-09-26 14:48:15 -07002468void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002469static inline void __skb_queue_purge(struct sk_buff_head *list)
2470{
2471 struct sk_buff *skb;
2472 while ((skb = __skb_dequeue(list)) != NULL)
2473 kfree_skb(skb);
2474}
2475
Yaogong Wang9f5afea2016-09-07 14:49:28 -07002476void skb_rbtree_purge(struct rb_root *root);
2477
Joe Perches7965bd42013-09-26 14:48:15 -07002478void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002479
Joe Perches7965bd42013-09-26 14:48:15 -07002480struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2481 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07002482
2483/**
2484 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2485 * @dev: network device to receive on
2486 * @length: length to allocate
2487 *
2488 * Allocate a new &sk_buff and assign it a usage count of one. The
2489 * buffer has unspecified headroom built in. Users should allocate
2490 * the headroom they think they need without accounting for the
2491 * built in space. The built in space is used for optimisations.
2492 *
2493 * %NULL is returned if there is no free memory. Although this function
2494 * allocates memory it can be called from an interrupt.
2495 */
2496static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002497 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002498{
2499 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2500}
2501
Eric Dumazet6f532612012-05-18 05:12:12 +00002502/* legacy helper around __netdev_alloc_skb() */
2503static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2504 gfp_t gfp_mask)
2505{
2506 return __netdev_alloc_skb(NULL, length, gfp_mask);
2507}
2508
2509/* legacy helper around netdev_alloc_skb() */
2510static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2511{
2512 return netdev_alloc_skb(NULL, length);
2513}
2514
2515
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002516static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2517 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002518{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002519 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002520
2521 if (NET_IP_ALIGN && skb)
2522 skb_reserve(skb, NET_IP_ALIGN);
2523 return skb;
2524}
2525
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002526static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2527 unsigned int length)
2528{
2529 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2530}
2531
Alexander Duyck181edb22015-05-06 21:12:03 -07002532static inline void skb_free_frag(void *addr)
2533{
Alexander Duyck8c2dd3e2017-01-10 16:58:06 -08002534 page_frag_free(addr);
Alexander Duyck181edb22015-05-06 21:12:03 -07002535}
2536
Alexander Duyckffde7322014-12-09 19:40:42 -08002537void *napi_alloc_frag(unsigned int fragsz);
Alexander Duyckfd11a832014-12-09 19:40:49 -08002538struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2539 unsigned int length, gfp_t gfp_mask);
2540static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2541 unsigned int length)
2542{
2543 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2544}
Jesper Dangaard Brouer795bb1c2016-02-08 13:14:59 +01002545void napi_consume_skb(struct sk_buff *skb, int budget);
2546
2547void __kfree_skb_flush(void);
Jesper Dangaard Brouer15fad712016-02-08 13:15:04 +01002548void __kfree_skb_defer(struct sk_buff *skb);
Alexander Duyckffde7322014-12-09 19:40:42 -08002549
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002550/**
Alexander Duyck71dfda52014-11-11 09:26:34 -08002551 * __dev_alloc_pages - allocate page for network Rx
2552 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2553 * @order: size of the allocation
2554 *
2555 * Allocate a new page.
2556 *
2557 * %NULL is returned if there is no free memory.
2558*/
2559static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2560 unsigned int order)
2561{
2562 /* This piece of code contains several assumptions.
2563 * 1. This is for device Rx, therefor a cold page is preferred.
2564 * 2. The expectation is the user wants a compound page.
2565 * 3. If requesting a order 0 page it will not be compound
2566 * due to the check to see if order has a value in prep_new_page
2567 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2568 * code in gfp_to_alloc_flags that should be enforcing this.
2569 */
2570 gfp_mask |= __GFP_COLD | __GFP_COMP | __GFP_MEMALLOC;
2571
2572 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2573}
2574
2575static inline struct page *dev_alloc_pages(unsigned int order)
2576{
Neil Horman95829b32016-05-19 11:30:54 -04002577 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002578}
2579
2580/**
2581 * __dev_alloc_page - allocate a page for network Rx
2582 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2583 *
2584 * Allocate a new page.
2585 *
2586 * %NULL is returned if there is no free memory.
2587 */
2588static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2589{
2590 return __dev_alloc_pages(gfp_mask, 0);
2591}
2592
2593static inline struct page *dev_alloc_page(void)
2594{
Neil Horman95829b32016-05-19 11:30:54 -04002595 return dev_alloc_pages(0);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002596}
2597
2598/**
Mel Gorman06140022012-07-31 16:44:24 -07002599 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2600 * @page: The page that was allocated from skb_alloc_page
2601 * @skb: The skb that may need pfmemalloc set
2602 */
2603static inline void skb_propagate_pfmemalloc(struct page *page,
2604 struct sk_buff *skb)
2605{
Michal Hocko2f064f32015-08-21 14:11:51 -07002606 if (page_is_pfmemalloc(page))
Mel Gorman06140022012-07-31 16:44:24 -07002607 skb->pfmemalloc = true;
2608}
2609
Eric Dumazet564824b2010-10-11 19:05:25 +00002610/**
Masanari Iidae2278672014-02-18 22:54:36 +09002611 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002612 * @frag: the paged fragment
2613 *
2614 * Returns the &struct page associated with @frag.
2615 */
2616static inline struct page *skb_frag_page(const skb_frag_t *frag)
2617{
Ian Campbella8605c62011-10-19 23:01:49 +00002618 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00002619}
2620
2621/**
2622 * __skb_frag_ref - take an addition reference on a paged fragment.
2623 * @frag: the paged fragment
2624 *
2625 * Takes an additional reference on the paged fragment @frag.
2626 */
2627static inline void __skb_frag_ref(skb_frag_t *frag)
2628{
2629 get_page(skb_frag_page(frag));
2630}
2631
2632/**
2633 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2634 * @skb: the buffer
2635 * @f: the fragment offset.
2636 *
2637 * Takes an additional reference on the @f'th paged fragment of @skb.
2638 */
2639static inline void skb_frag_ref(struct sk_buff *skb, int f)
2640{
2641 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2642}
2643
2644/**
2645 * __skb_frag_unref - release a reference on a paged fragment.
2646 * @frag: the paged fragment
2647 *
2648 * Releases a reference on the paged fragment @frag.
2649 */
2650static inline void __skb_frag_unref(skb_frag_t *frag)
2651{
2652 put_page(skb_frag_page(frag));
2653}
2654
2655/**
2656 * skb_frag_unref - release a reference on a paged fragment of an skb.
2657 * @skb: the buffer
2658 * @f: the fragment offset
2659 *
2660 * Releases a reference on the @f'th paged fragment of @skb.
2661 */
2662static inline void skb_frag_unref(struct sk_buff *skb, int f)
2663{
2664 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2665}
2666
2667/**
2668 * skb_frag_address - gets the address of the data contained in a paged fragment
2669 * @frag: the paged fragment buffer
2670 *
2671 * Returns the address of the data within @frag. The page must already
2672 * be mapped.
2673 */
2674static inline void *skb_frag_address(const skb_frag_t *frag)
2675{
2676 return page_address(skb_frag_page(frag)) + frag->page_offset;
2677}
2678
2679/**
2680 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2681 * @frag: the paged fragment buffer
2682 *
2683 * Returns the address of the data within @frag. Checks that the page
2684 * is mapped and returns %NULL otherwise.
2685 */
2686static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2687{
2688 void *ptr = page_address(skb_frag_page(frag));
2689 if (unlikely(!ptr))
2690 return NULL;
2691
2692 return ptr + frag->page_offset;
2693}
2694
2695/**
2696 * __skb_frag_set_page - sets the page contained in a paged fragment
2697 * @frag: the paged fragment
2698 * @page: the page to set
2699 *
2700 * Sets the fragment @frag to contain @page.
2701 */
2702static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2703{
Ian Campbella8605c62011-10-19 23:01:49 +00002704 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002705}
2706
2707/**
2708 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2709 * @skb: the buffer
2710 * @f: the fragment offset
2711 * @page: the page to set
2712 *
2713 * Sets the @f'th fragment of @skb to contain @page.
2714 */
2715static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2716 struct page *page)
2717{
2718 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2719}
2720
Eric Dumazet400dfd32013-10-17 16:27:07 -07002721bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2722
Ian Campbell131ea662011-08-19 06:25:00 +00002723/**
2724 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002725 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002726 * @frag: the paged fragment to map
2727 * @offset: the offset within the fragment (starting at the
2728 * fragment's own offset)
2729 * @size: the number of bytes to map
Mauro Carvalho Chehab771b00a2017-05-12 09:19:29 -03002730 * @dir: the direction of the mapping (``PCI_DMA_*``)
Ian Campbell131ea662011-08-19 06:25:00 +00002731 *
2732 * Maps the page associated with @frag to @device.
2733 */
2734static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2735 const skb_frag_t *frag,
2736 size_t offset, size_t size,
2737 enum dma_data_direction dir)
2738{
2739 return dma_map_page(dev, skb_frag_page(frag),
2740 frag->page_offset + offset, size, dir);
2741}
2742
Eric Dumazet117632e2011-12-03 21:39:53 +00002743static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2744 gfp_t gfp_mask)
2745{
2746 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2747}
2748
Octavian Purdilabad93e92014-06-12 01:36:26 +03002749
2750static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2751 gfp_t gfp_mask)
2752{
2753 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2754}
2755
2756
Ian Campbell131ea662011-08-19 06:25:00 +00002757/**
Patrick McHardy334a8132007-06-25 04:35:20 -07002758 * skb_clone_writable - is the header of a clone writable
2759 * @skb: buffer to check
2760 * @len: length up to which to write
2761 *
2762 * Returns true if modifying the header part of the cloned buffer
2763 * does not requires the data to be copied.
2764 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04002765static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07002766{
2767 return !skb_header_cloned(skb) &&
2768 skb_headroom(skb) + len <= skb->hdr_len;
2769}
2770
Daniel Borkmann36976492016-02-19 23:05:25 +01002771static inline int skb_try_make_writable(struct sk_buff *skb,
2772 unsigned int write_len)
2773{
2774 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
2775 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
2776}
2777
Herbert Xud9cc2042007-09-16 16:21:16 -07002778static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
2779 int cloned)
2780{
2781 int delta = 0;
2782
Herbert Xud9cc2042007-09-16 16:21:16 -07002783 if (headroom > skb_headroom(skb))
2784 delta = headroom - skb_headroom(skb);
2785
2786 if (delta || cloned)
2787 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
2788 GFP_ATOMIC);
2789 return 0;
2790}
2791
Patrick McHardy334a8132007-06-25 04:35:20 -07002792/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002793 * skb_cow - copy header of skb when it is required
2794 * @skb: buffer to cow
2795 * @headroom: needed headroom
2796 *
2797 * If the skb passed lacks sufficient headroom or its data part
2798 * is shared, data is reallocated. If reallocation fails, an error
2799 * is returned and original skb is not changed.
2800 *
2801 * The result is skb with writable area skb->head...skb->tail
2802 * and at least @headroom of space at head.
2803 */
2804static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
2805{
Herbert Xud9cc2042007-09-16 16:21:16 -07002806 return __skb_cow(skb, headroom, skb_cloned(skb));
2807}
Linus Torvalds1da177e2005-04-16 15:20:36 -07002808
Herbert Xud9cc2042007-09-16 16:21:16 -07002809/**
2810 * skb_cow_head - skb_cow but only making the head writable
2811 * @skb: buffer to cow
2812 * @headroom: needed headroom
2813 *
2814 * This function is identical to skb_cow except that we replace the
2815 * skb_cloned check by skb_header_cloned. It should be used when
2816 * you only need to push on some header and do not need to modify
2817 * the data.
2818 */
2819static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
2820{
2821 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07002822}
2823
2824/**
2825 * skb_padto - pad an skbuff up to a minimal size
2826 * @skb: buffer to pad
2827 * @len: minimal length
2828 *
2829 * Pads up a buffer to ensure the trailing bytes exist and are
2830 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07002831 * is untouched. Otherwise it is extended. Returns zero on
2832 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002833 */
Herbert Xu5b057c62006-06-23 02:06:41 -07002834static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002835{
2836 unsigned int size = skb->len;
2837 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07002838 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002839 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002840}
2841
Alexander Duyck9c0c1122014-12-03 08:17:33 -08002842/**
2843 * skb_put_padto - increase size and pad an skbuff up to a minimal size
2844 * @skb: buffer to pad
2845 * @len: minimal length
2846 *
2847 * Pads up a buffer to ensure the trailing bytes exist and are
2848 * blanked. If the buffer already contains sufficient data it
2849 * is untouched. Otherwise it is extended. Returns zero on
2850 * success. The skb is freed on error.
2851 */
2852static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
2853{
2854 unsigned int size = skb->len;
2855
2856 if (unlikely(size < len)) {
2857 len -= size;
2858 if (skb_pad(skb, len))
2859 return -ENOMEM;
2860 __skb_put(skb, len);
2861 }
2862 return 0;
2863}
2864
Linus Torvalds1da177e2005-04-16 15:20:36 -07002865static inline int skb_add_data(struct sk_buff *skb,
Al Viroaf2b0402014-11-27 21:44:24 -05002866 struct iov_iter *from, int copy)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002867{
2868 const int off = skb->len;
2869
2870 if (skb->ip_summed == CHECKSUM_NONE) {
Al Viroaf2b0402014-11-27 21:44:24 -05002871 __wsum csum = 0;
Al Viro15e6cb42016-11-01 22:42:45 -04002872 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
2873 &csum, from)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002874 skb->csum = csum_block_add(skb->csum, csum, off);
2875 return 0;
2876 }
Al Viro15e6cb42016-11-01 22:42:45 -04002877 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002878 return 0;
2879
2880 __skb_trim(skb, off);
2881 return -EFAULT;
2882}
2883
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002884static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
2885 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002886{
2887 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00002888 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07002889
Ian Campbellea2ab692011-08-22 23:44:58 +00002890 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00002891 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002892 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00002893 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002894}
2895
Herbert Xu364c6ba2006-06-09 16:10:40 -07002896static inline int __skb_linearize(struct sk_buff *skb)
2897{
2898 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
2899}
2900
Linus Torvalds1da177e2005-04-16 15:20:36 -07002901/**
2902 * skb_linearize - convert paged skb to linear one
2903 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07002904 *
2905 * If there is no free memory -ENOMEM is returned, otherwise zero
2906 * is returned and the old skb data released.
2907 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07002908static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002909{
Herbert Xu364c6ba2006-06-09 16:10:40 -07002910 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
2911}
2912
2913/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00002914 * skb_has_shared_frag - can any frag be overwritten
2915 * @skb: buffer to test
2916 *
2917 * Return true if the skb has at least one frag that might be modified
2918 * by an external entity (as in vmsplice()/sendfile())
2919 */
2920static inline bool skb_has_shared_frag(const struct sk_buff *skb)
2921{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00002922 return skb_is_nonlinear(skb) &&
2923 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00002924}
2925
2926/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07002927 * skb_linearize_cow - make sure skb is linear and writable
2928 * @skb: buffer to process
2929 *
2930 * If there is no free memory -ENOMEM is returned, otherwise zero
2931 * is returned and the old skb data released.
2932 */
2933static inline int skb_linearize_cow(struct sk_buff *skb)
2934{
2935 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
2936 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002937}
2938
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02002939static __always_inline void
2940__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
2941 unsigned int off)
2942{
2943 if (skb->ip_summed == CHECKSUM_COMPLETE)
2944 skb->csum = csum_block_sub(skb->csum,
2945 csum_partial(start, len, 0), off);
2946 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
2947 skb_checksum_start_offset(skb) < 0)
2948 skb->ip_summed = CHECKSUM_NONE;
2949}
2950
Linus Torvalds1da177e2005-04-16 15:20:36 -07002951/**
2952 * skb_postpull_rcsum - update checksum for received skb after pull
2953 * @skb: buffer to update
2954 * @start: start of data before pull
2955 * @len: length of data pulled
2956 *
2957 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07002958 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
2959 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07002960 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002961static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08002962 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002963{
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02002964 __skb_postpull_rcsum(skb, start, len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002965}
2966
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02002967static __always_inline void
2968__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
2969 unsigned int off)
2970{
2971 if (skb->ip_summed == CHECKSUM_COMPLETE)
2972 skb->csum = csum_block_add(skb->csum,
2973 csum_partial(start, len, 0), off);
2974}
Herbert Xucbb042f2006-03-20 22:43:56 -08002975
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02002976/**
2977 * skb_postpush_rcsum - update checksum for received skb after push
2978 * @skb: buffer to update
2979 * @start: start of data after push
2980 * @len: length of data pushed
2981 *
2982 * After doing a push on a received packet, you need to call this to
2983 * update the CHECKSUM_COMPLETE checksum.
2984 */
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01002985static inline void skb_postpush_rcsum(struct sk_buff *skb,
2986 const void *start, unsigned int len)
2987{
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02002988 __skb_postpush_rcsum(skb, start, len, 0);
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01002989}
2990
Johannes Bergaf728682017-06-16 14:29:22 +02002991void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02002992
David S. Miller7ce5a272013-12-02 17:26:05 -05002993/**
WANG Cong82a31b92016-06-30 10:15:22 -07002994 * skb_push_rcsum - push skb and update receive checksum
2995 * @skb: buffer to update
2996 * @len: length of data pulled
2997 *
2998 * This function performs an skb_push on the packet and updates
2999 * the CHECKSUM_COMPLETE checksum. It should be used on
3000 * receive path processing instead of skb_push unless you know
3001 * that the checksum difference is zero (e.g., a valid IP header)
3002 * or you are setting ip_summed to CHECKSUM_NONE.
3003 */
Johannes Bergd58ff352017-06-16 14:29:23 +02003004static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
WANG Cong82a31b92016-06-30 10:15:22 -07003005{
3006 skb_push(skb, len);
3007 skb_postpush_rcsum(skb, skb->data, len);
3008 return skb->data;
3009}
3010
3011/**
David S. Miller7ce5a272013-12-02 17:26:05 -05003012 * pskb_trim_rcsum - trim received skb and update checksum
3013 * @skb: buffer to trim
3014 * @len: new length
3015 *
3016 * This is exactly the same as pskb_trim except that it ensures the
3017 * checksum of received packets are still valid after the operation.
3018 */
3019
3020static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3021{
3022 if (likely(len >= skb->len))
3023 return 0;
3024 if (skb->ip_summed == CHECKSUM_COMPLETE)
3025 skb->ip_summed = CHECKSUM_NONE;
3026 return __pskb_trim(skb, len);
3027}
3028
Daniel Borkmann5293efe2016-08-18 01:00:39 +02003029static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3030{
3031 if (skb->ip_summed == CHECKSUM_COMPLETE)
3032 skb->ip_summed = CHECKSUM_NONE;
3033 __skb_trim(skb, len);
3034 return 0;
3035}
3036
3037static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3038{
3039 if (skb->ip_summed == CHECKSUM_COMPLETE)
3040 skb->ip_summed = CHECKSUM_NONE;
3041 return __skb_grow(skb, len);
3042}
3043
Linus Torvalds1da177e2005-04-16 15:20:36 -07003044#define skb_queue_walk(queue, skb) \
3045 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003046 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07003047 skb = skb->next)
3048
James Chapman46f89142007-04-30 00:07:31 -07003049#define skb_queue_walk_safe(queue, skb, tmp) \
3050 for (skb = (queue)->next, tmp = skb->next; \
3051 skb != (struct sk_buff *)(queue); \
3052 skb = tmp, tmp = skb->next)
3053
David S. Miller1164f522008-09-23 00:49:44 -07003054#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003055 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07003056 skb = skb->next)
3057
3058#define skb_queue_walk_from_safe(queue, skb, tmp) \
3059 for (tmp = skb->next; \
3060 skb != (struct sk_buff *)(queue); \
3061 skb = tmp, tmp = skb->next)
3062
Stephen Hemminger300ce172005-10-30 13:47:34 -08003063#define skb_queue_reverse_walk(queue, skb) \
3064 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003065 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08003066 skb = skb->prev)
3067
David S. Miller686a2952011-01-20 22:47:32 -08003068#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3069 for (skb = (queue)->prev, tmp = skb->prev; \
3070 skb != (struct sk_buff *)(queue); \
3071 skb = tmp, tmp = skb->prev)
3072
3073#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3074 for (tmp = skb->prev; \
3075 skb != (struct sk_buff *)(queue); \
3076 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003077
David S. Miller21dc3302010-08-23 00:13:46 -07003078static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07003079{
3080 return skb_shinfo(skb)->frag_list != NULL;
3081}
3082
3083static inline void skb_frag_list_init(struct sk_buff *skb)
3084{
3085 skb_shinfo(skb)->frag_list = NULL;
3086}
3087
David S. Milleree039872009-06-09 00:17:13 -07003088#define skb_walk_frags(skb, iter) \
3089 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3090
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003091
3092int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
3093 const struct sk_buff *skb);
Paolo Abeni65101ae2017-05-16 11:20:13 +02003094struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3095 struct sk_buff_head *queue,
3096 unsigned int flags,
3097 void (*destructor)(struct sock *sk,
3098 struct sk_buff *skb),
3099 int *peeked, int *off, int *err,
3100 struct sk_buff **last);
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003101struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
Paolo Abeni7c13f972016-11-04 11:28:59 +01003102 void (*destructor)(struct sock *sk,
3103 struct sk_buff *skb),
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003104 int *peeked, int *off, int *err,
3105 struct sk_buff **last);
Joe Perches7965bd42013-09-26 14:48:15 -07003106struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
Paolo Abeni7c13f972016-11-04 11:28:59 +01003107 void (*destructor)(struct sock *sk,
3108 struct sk_buff *skb),
Joe Perches7965bd42013-09-26 14:48:15 -07003109 int *peeked, int *off, int *err);
3110struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3111 int *err);
3112unsigned int datagram_poll(struct file *file, struct socket *sock,
3113 struct poll_table_struct *wait);
Al Viroc0371da2014-11-24 10:42:55 -05003114int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3115 struct iov_iter *to, int size);
David S. Miller51f3d022014-11-05 16:46:40 -05003116static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3117 struct msghdr *msg, int size)
3118{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003119 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
David S. Miller51f3d022014-11-05 16:46:40 -05003120}
Al Viroe5a4b0b2014-11-24 18:17:55 -05003121int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3122 struct msghdr *msg);
Al Viro3a654f92014-06-19 14:15:22 -04003123int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3124 struct iov_iter *from, int len);
Al Viro3a654f92014-06-19 14:15:22 -04003125int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
Joe Perches7965bd42013-09-26 14:48:15 -07003126void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
samanthakumar627d2d62016-04-05 12:41:16 -04003127void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3128static inline void skb_free_datagram_locked(struct sock *sk,
3129 struct sk_buff *skb)
3130{
3131 __skb_free_datagram_locked(sk, skb, 0);
3132}
Joe Perches7965bd42013-09-26 14:48:15 -07003133int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07003134int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3135int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3136__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3137 int len, __wsum csum);
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +02003138int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
Joe Perches7965bd42013-09-26 14:48:15 -07003139 struct pipe_inode_info *pipe, unsigned int len,
Al Viro25869262016-09-17 21:02:10 -04003140 unsigned int flags);
Tom Herbert20bf50d2017-07-28 16:22:42 -07003141int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3142 int len);
3143int skb_send_sock(struct sock *sk, struct sk_buff *skb, int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07003144void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01003145unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00003146int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3147 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07003148void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3149int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3150void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Florian Westphalde960aa2014-01-26 10:58:16 +01003151unsigned int skb_gso_transport_seglen(const struct sk_buff *skb);
Marcelo Ricardo Leitnerae7ef812016-06-02 15:05:41 -03003152bool skb_gso_validate_mtu(const struct sk_buff *skb, unsigned int mtu);
Joe Perches7965bd42013-09-26 14:48:15 -07003153struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Vlad Yasevich0d5501c2014-08-08 14:42:13 -04003154struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
Jiri Pirkoe2195122014-11-19 14:05:01 +01003155int skb_ensure_writable(struct sk_buff *skb, int write_len);
Shmulik Ladkanibfca4c52016-09-19 19:11:09 +03003156int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
Jiri Pirko93515d52014-11-19 14:05:02 +01003157int skb_vlan_pop(struct sk_buff *skb);
3158int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
Sowmini Varadhan6fa01cc2016-04-22 18:36:35 -07003159struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3160 gfp_t gfp);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003161
Al Viro6ce8e9c2014-04-06 21:25:44 -04003162static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3163{
Al Viro3073f072017-02-17 23:13:25 -05003164 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
Al Viro6ce8e9c2014-04-06 21:25:44 -04003165}
3166
Al Viro7eab8d92014-04-06 21:51:23 -04003167static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3168{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003169 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro7eab8d92014-04-06 21:51:23 -04003170}
3171
Daniel Borkmann2817a332013-10-30 11:50:51 +01003172struct skb_checksum_ops {
3173 __wsum (*update)(const void *mem, int len, __wsum wsum);
3174 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3175};
3176
Davide Caratti96178132017-05-18 15:44:37 +02003177extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3178
Daniel Borkmann2817a332013-10-30 11:50:51 +01003179__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3180 __wsum csum, const struct skb_checksum_ops *ops);
3181__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3182 __wsum csum);
3183
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003184static inline void * __must_check
3185__skb_header_pointer(const struct sk_buff *skb, int offset,
3186 int len, void *data, int hlen, void *buffer)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003187{
Patrick McHardy55820ee2005-07-05 14:08:10 -07003188 if (hlen - offset >= len)
David S. Miller690e36e2014-08-23 12:13:41 -07003189 return data + offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003190
David S. Miller690e36e2014-08-23 12:13:41 -07003191 if (!skb ||
3192 skb_copy_bits(skb, offset, buffer, len) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003193 return NULL;
3194
3195 return buffer;
3196}
3197
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003198static inline void * __must_check
3199skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
David S. Miller690e36e2014-08-23 12:13:41 -07003200{
3201 return __skb_header_pointer(skb, offset, len, skb->data,
3202 skb_headlen(skb), buffer);
3203}
3204
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01003205/**
3206 * skb_needs_linearize - check if we need to linearize a given skb
3207 * depending on the given device features.
3208 * @skb: socket buffer to check
3209 * @features: net device features
3210 *
3211 * Returns true if either:
3212 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3213 * 2. skb is fragmented and the device does not support SG.
3214 */
3215static inline bool skb_needs_linearize(struct sk_buff *skb,
3216 netdev_features_t features)
3217{
3218 return skb_is_nonlinear(skb) &&
3219 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3220 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3221}
3222
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03003223static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3224 void *to,
3225 const unsigned int len)
3226{
3227 memcpy(to, skb->data, len);
3228}
3229
3230static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3231 const int offset, void *to,
3232 const unsigned int len)
3233{
3234 memcpy(to, skb->data + offset, len);
3235}
3236
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03003237static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3238 const void *from,
3239 const unsigned int len)
3240{
3241 memcpy(skb->data, from, len);
3242}
3243
3244static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3245 const int offset,
3246 const void *from,
3247 const unsigned int len)
3248{
3249 memcpy(skb->data + offset, from, len);
3250}
3251
Joe Perches7965bd42013-09-26 14:48:15 -07003252void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003253
Patrick Ohlyac45f602009-02-12 05:03:37 +00003254static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3255{
3256 return skb->tstamp;
3257}
3258
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003259/**
3260 * skb_get_timestamp - get timestamp from a skb
3261 * @skb: skb to get stamp from
3262 * @stamp: pointer to struct timeval to store stamp in
3263 *
3264 * Timestamps are stored in the skb as offsets to a base timestamp.
3265 * This function converts the offset back to a struct timeval and stores
3266 * it in stamp.
3267 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00003268static inline void skb_get_timestamp(const struct sk_buff *skb,
3269 struct timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003270{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003271 *stamp = ktime_to_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003272}
3273
Patrick Ohlyac45f602009-02-12 05:03:37 +00003274static inline void skb_get_timestampns(const struct sk_buff *skb,
3275 struct timespec *stamp)
3276{
3277 *stamp = ktime_to_timespec(skb->tstamp);
3278}
3279
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003280static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003281{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003282 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003283}
3284
Stephen Hemminger164891a2007-04-23 22:26:16 -07003285static inline ktime_t net_timedelta(ktime_t t)
3286{
3287 return ktime_sub(ktime_get_real(), t);
3288}
3289
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003290static inline ktime_t net_invalid_timestamp(void)
3291{
Thomas Gleixner8b0e1952016-12-25 12:30:41 +01003292 return 0;
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003293}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003294
Alexander Duyck62bccb82014-09-04 13:31:35 -04003295struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3296
Richard Cochranc1f19b52010-07-17 08:49:36 +00003297#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3298
Joe Perches7965bd42013-09-26 14:48:15 -07003299void skb_clone_tx_timestamp(struct sk_buff *skb);
3300bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00003301
3302#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3303
3304static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3305{
3306}
3307
3308static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3309{
3310 return false;
3311}
3312
3313#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3314
3315/**
3316 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3317 *
Richard Cochranda92b192011-10-21 00:49:15 +00003318 * PHY drivers may accept clones of transmitted packets for
3319 * timestamping via their phy_driver.txtstamp method. These drivers
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003320 * must call this function to return the skb back to the stack with a
3321 * timestamp.
Richard Cochranda92b192011-10-21 00:49:15 +00003322 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00003323 * @skb: clone of the the original outgoing packet
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003324 * @hwtstamps: hardware time stamps
Richard Cochranc1f19b52010-07-17 08:49:36 +00003325 *
3326 */
3327void skb_complete_tx_timestamp(struct sk_buff *skb,
3328 struct skb_shared_hwtstamps *hwtstamps);
3329
Willem de Bruijne7fd2882014-08-04 22:11:48 -04003330void __skb_tstamp_tx(struct sk_buff *orig_skb,
3331 struct skb_shared_hwtstamps *hwtstamps,
3332 struct sock *sk, int tstype);
3333
Patrick Ohlyac45f602009-02-12 05:03:37 +00003334/**
3335 * skb_tstamp_tx - queue clone of skb with send time stamps
3336 * @orig_skb: the original outgoing packet
3337 * @hwtstamps: hardware time stamps, may be NULL if not available
3338 *
3339 * If the skb has a socket associated, then this function clones the
3340 * skb (thus sharing the actual data and optional structures), stores
3341 * the optional hardware time stamping information (if non NULL) or
3342 * generates a software time stamp (otherwise), then queues the clone
3343 * to the error queue of the socket. Errors are silently ignored.
3344 */
Joe Perches7965bd42013-09-26 14:48:15 -07003345void skb_tstamp_tx(struct sk_buff *orig_skb,
3346 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00003347
Richard Cochran4507a712010-07-17 08:48:28 +00003348/**
3349 * skb_tx_timestamp() - Driver hook for transmit timestamping
3350 *
3351 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00003352 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00003353 *
David S. Miller73409f32013-12-27 13:04:33 -05003354 * Specifically, one should make absolutely sure that this function is
3355 * called before TX completion of this packet can trigger. Otherwise
3356 * the packet could potentially already be freed.
3357 *
Richard Cochran4507a712010-07-17 08:48:28 +00003358 * @skb: A socket buffer.
3359 */
3360static inline void skb_tx_timestamp(struct sk_buff *skb)
3361{
Richard Cochranc1f19b52010-07-17 08:49:36 +00003362 skb_clone_tx_timestamp(skb);
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +02003363 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3364 skb_tstamp_tx(skb, NULL);
Richard Cochran4507a712010-07-17 08:48:28 +00003365}
3366
Johannes Berg6e3e9392011-11-09 10:15:42 +01003367/**
3368 * skb_complete_wifi_ack - deliver skb with wifi status
3369 *
3370 * @skb: the original outgoing packet
3371 * @acked: ack status
3372 *
3373 */
3374void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3375
Joe Perches7965bd42013-09-26 14:48:15 -07003376__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3377__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003378
Herbert Xu60476372007-04-09 11:59:39 -07003379static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3380{
Tom Herbert6edec0e2015-02-10 16:30:28 -08003381 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3382 skb->csum_valid ||
3383 (skb->ip_summed == CHECKSUM_PARTIAL &&
3384 skb_checksum_start_offset(skb) >= 0));
Herbert Xu60476372007-04-09 11:59:39 -07003385}
3386
Herbert Xufb286bb2005-11-10 13:01:24 -08003387/**
3388 * skb_checksum_complete - Calculate checksum of an entire packet
3389 * @skb: packet to process
3390 *
3391 * This function calculates the checksum over the entire packet plus
3392 * the value of skb->csum. The latter can be used to supply the
3393 * checksum of a pseudo header as used by TCP/UDP. It returns the
3394 * checksum.
3395 *
3396 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3397 * this function can be used to verify that checksum on received
3398 * packets. In that case the function should return zero if the
3399 * checksum is correct. In particular, this function will return zero
3400 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3401 * hardware has already verified the correctness of the checksum.
3402 */
Al Viro4381ca32007-07-15 21:00:11 +01003403static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08003404{
Herbert Xu60476372007-04-09 11:59:39 -07003405 return skb_csum_unnecessary(skb) ?
3406 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003407}
3408
Tom Herbert77cffe22014-08-27 21:26:46 -07003409static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3410{
3411 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3412 if (skb->csum_level == 0)
3413 skb->ip_summed = CHECKSUM_NONE;
3414 else
3415 skb->csum_level--;
3416 }
3417}
3418
3419static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3420{
3421 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3422 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3423 skb->csum_level++;
3424 } else if (skb->ip_summed == CHECKSUM_NONE) {
3425 skb->ip_summed = CHECKSUM_UNNECESSARY;
3426 skb->csum_level = 0;
3427 }
3428}
3429
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003430/* Check if we need to perform checksum complete validation.
3431 *
3432 * Returns true if checksum complete is needed, false otherwise
3433 * (either checksum is unnecessary or zero checksum is allowed).
3434 */
3435static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3436 bool zero_okay,
3437 __sum16 check)
3438{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003439 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3440 skb->csum_valid = 1;
Tom Herbert77cffe22014-08-27 21:26:46 -07003441 __skb_decr_checksum_unnecessary(skb);
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003442 return false;
3443 }
3444
3445 return true;
3446}
3447
3448/* For small packets <= CHECKSUM_BREAK peform checksum complete directly
3449 * in checksum_init.
3450 */
3451#define CHECKSUM_BREAK 76
3452
Tom Herbert4e18b9a2015-04-20 14:10:04 -07003453/* Unset checksum-complete
3454 *
3455 * Unset checksum complete can be done when packet is being modified
3456 * (uncompressed for instance) and checksum-complete value is
3457 * invalidated.
3458 */
3459static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3460{
3461 if (skb->ip_summed == CHECKSUM_COMPLETE)
3462 skb->ip_summed = CHECKSUM_NONE;
3463}
3464
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003465/* Validate (init) checksum based on checksum complete.
3466 *
3467 * Return values:
3468 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3469 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3470 * checksum is stored in skb->csum for use in __skb_checksum_complete
3471 * non-zero: value of invalid checksum
3472 *
3473 */
3474static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3475 bool complete,
3476 __wsum psum)
3477{
3478 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3479 if (!csum_fold(csum_add(psum, skb->csum))) {
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003480 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003481 return 0;
3482 }
3483 }
3484
3485 skb->csum = psum;
3486
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003487 if (complete || skb->len <= CHECKSUM_BREAK) {
3488 __sum16 csum;
3489
3490 csum = __skb_checksum_complete(skb);
3491 skb->csum_valid = !csum;
3492 return csum;
3493 }
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003494
3495 return 0;
3496}
3497
3498static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3499{
3500 return 0;
3501}
3502
3503/* Perform checksum validate (init). Note that this is a macro since we only
3504 * want to calculate the pseudo header which is an input function if necessary.
3505 * First we try to validate without any computation (checksum unnecessary) and
3506 * then calculate based on checksum complete calling the function to compute
3507 * pseudo header.
3508 *
3509 * Return values:
3510 * 0: checksum is validated or try to in skb_checksum_complete
3511 * non-zero: value of invalid checksum
3512 */
3513#define __skb_checksum_validate(skb, proto, complete, \
3514 zero_okay, check, compute_pseudo) \
3515({ \
3516 __sum16 __ret = 0; \
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003517 skb->csum_valid = 0; \
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003518 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3519 __ret = __skb_checksum_validate_complete(skb, \
3520 complete, compute_pseudo(skb, proto)); \
3521 __ret; \
3522})
3523
3524#define skb_checksum_init(skb, proto, compute_pseudo) \
3525 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3526
3527#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3528 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3529
3530#define skb_checksum_validate(skb, proto, compute_pseudo) \
3531 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3532
3533#define skb_checksum_validate_zero_check(skb, proto, check, \
3534 compute_pseudo) \
Sabrina Dubroca096a4cf2015-02-06 18:54:19 +01003535 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003536
3537#define skb_checksum_simple_validate(skb) \
3538 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3539
Tom Herbertd96535a2014-08-31 15:12:42 -07003540static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3541{
Davide Caratti219f1d792017-05-18 15:44:39 +02003542 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
Tom Herbertd96535a2014-08-31 15:12:42 -07003543}
3544
3545static inline void __skb_checksum_convert(struct sk_buff *skb,
3546 __sum16 check, __wsum pseudo)
3547{
3548 skb->csum = ~pseudo;
3549 skb->ip_summed = CHECKSUM_COMPLETE;
3550}
3551
3552#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3553do { \
3554 if (__skb_checksum_convert_check(skb)) \
3555 __skb_checksum_convert(skb, check, \
3556 compute_pseudo(skb, proto)); \
3557} while (0)
3558
Tom Herbert15e23962015-02-10 16:30:31 -08003559static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3560 u16 start, u16 offset)
3561{
3562 skb->ip_summed = CHECKSUM_PARTIAL;
3563 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3564 skb->csum_offset = offset - start;
3565}
3566
Tom Herbertdcdc8992015-02-02 16:07:34 -08003567/* Update skbuf and packet to reflect the remote checksum offload operation.
3568 * When called, ptr indicates the starting point for skb->csum when
3569 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3570 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3571 */
3572static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
Tom Herbert15e23962015-02-10 16:30:31 -08003573 int start, int offset, bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08003574{
3575 __wsum delta;
3576
Tom Herbert15e23962015-02-10 16:30:31 -08003577 if (!nopartial) {
3578 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3579 return;
3580 }
3581
Tom Herbertdcdc8992015-02-02 16:07:34 -08003582 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3583 __skb_checksum_complete(skb);
3584 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3585 }
3586
3587 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3588
3589 /* Adjust skb->csum since we changed the packet */
3590 skb->csum = csum_add(skb->csum, delta);
3591}
3592
Florian Westphalcb9c6832017-01-23 18:21:56 +01003593static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
3594{
3595#if IS_ENABLED(CONFIG_NF_CONNTRACK)
Florian Westphala9e419d2017-01-23 18:21:59 +01003596 return (void *)(skb->_nfct & SKB_NFCT_PTRMASK);
Florian Westphalcb9c6832017-01-23 18:21:56 +01003597#else
3598 return NULL;
3599#endif
3600}
3601
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003602#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07003603void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003604static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3605{
3606 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07003607 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003608}
3609static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3610{
3611 if (nfct)
3612 atomic_inc(&nfct->use);
3613}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003614#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003615#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003616static inline void nf_bridge_put(struct nf_bridge_info *nf_bridge)
3617{
Reshetova, Elena53869ce2017-06-30 13:07:57 +03003618 if (nf_bridge && refcount_dec_and_test(&nf_bridge->use))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003619 kfree(nf_bridge);
3620}
3621static inline void nf_bridge_get(struct nf_bridge_info *nf_bridge)
3622{
3623 if (nf_bridge)
Reshetova, Elena53869ce2017-06-30 13:07:57 +03003624 refcount_inc(&nf_bridge->use);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003625}
3626#endif /* CONFIG_BRIDGE_NETFILTER */
Patrick McHardya193a4a2006-03-20 19:23:05 -08003627static inline void nf_reset(struct sk_buff *skb)
3628{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003629#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01003630 nf_conntrack_put(skb_nfct(skb));
3631 skb->_nfct = 0;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003632#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003633#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Patrick McHardya193a4a2006-03-20 19:23:05 -08003634 nf_bridge_put(skb->nf_bridge);
3635 skb->nf_bridge = NULL;
3636#endif
3637}
3638
Patrick McHardy124dff02013-04-05 20:42:05 +02003639static inline void nf_reset_trace(struct sk_buff *skb)
3640{
Florian Westphal478b3602014-02-15 23:48:45 +01003641#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00003642 skb->nf_trace = 0;
3643#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07003644}
3645
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003646/* Note: This doesn't put any conntrack and bridge info in dst. */
Eric Dumazetb1937222014-09-28 22:18:47 -07003647static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
3648 bool copy)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003649{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003650#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01003651 dst->_nfct = src->_nfct;
3652 nf_conntrack_get(skb_nfct(src));
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003653#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003654#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003655 dst->nf_bridge = src->nf_bridge;
3656 nf_bridge_get(src->nf_bridge);
3657#endif
Florian Westphal478b3602014-02-15 23:48:45 +01003658#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Eric Dumazetb1937222014-09-28 22:18:47 -07003659 if (copy)
3660 dst->nf_trace = src->nf_trace;
Florian Westphal478b3602014-02-15 23:48:45 +01003661#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07003662}
3663
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003664static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
3665{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003666#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01003667 nf_conntrack_put(skb_nfct(dst));
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003668#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02003669#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003670 nf_bridge_put(dst->nf_bridge);
3671#endif
Eric Dumazetb1937222014-09-28 22:18:47 -07003672 __nf_copy(dst, src, true);
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07003673}
3674
James Morris984bc162006-06-09 00:29:17 -07003675#ifdef CONFIG_NETWORK_SECMARK
3676static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3677{
3678 to->secmark = from->secmark;
3679}
3680
3681static inline void skb_init_secmark(struct sk_buff *skb)
3682{
3683 skb->secmark = 0;
3684}
3685#else
3686static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
3687{ }
3688
3689static inline void skb_init_secmark(struct sk_buff *skb)
3690{ }
3691#endif
3692
Eric W. Biederman574f7192014-04-01 12:20:24 -07003693static inline bool skb_irq_freeable(const struct sk_buff *skb)
3694{
3695 return !skb->destructor &&
3696#if IS_ENABLED(CONFIG_XFRM)
3697 !skb->sp &&
3698#endif
Florian Westphalcb9c6832017-01-23 18:21:56 +01003699 !skb_nfct(skb) &&
Eric W. Biederman574f7192014-04-01 12:20:24 -07003700 !skb->_skb_refdst &&
3701 !skb_has_frag_list(skb);
3702}
3703
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003704static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
3705{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003706 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003707}
3708
Stephen Hemminger92477442009-03-21 13:39:26 -07003709static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003710{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003711 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07003712}
3713
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003714static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
3715{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003716 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07003717}
3718
David S. Millerd5a9e242009-01-27 16:22:11 -08003719static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
3720{
3721 skb->queue_mapping = rx_queue + 1;
3722}
3723
Stephen Hemminger92477442009-03-21 13:39:26 -07003724static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003725{
3726 return skb->queue_mapping - 1;
3727}
3728
Stephen Hemminger92477442009-03-21 13:39:26 -07003729static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08003730{
Eric Dumazeta02cec22010-09-22 20:43:57 +00003731 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08003732}
3733
Julian Anastasov4ff06202017-02-06 23:14:12 +02003734static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
3735{
3736 skb->dst_pending_confirm = val;
3737}
3738
3739static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
3740{
3741 return skb->dst_pending_confirm != 0;
3742}
3743
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003744static inline struct sec_path *skb_sec_path(struct sk_buff *skb)
3745{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003746#ifdef CONFIG_XFRM
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003747 return skb->sp;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003748#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003749 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003750#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04003751}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07003752
Pravin B Shelar68c33162013-02-14 14:02:41 +00003753/* Keeps track of mac header offset relative to skb->head.
3754 * It is useful for TSO of Tunneling protocol. e.g. GRE.
3755 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07003756 * tunnel skb it points to outer mac header.
3757 * Keeps track of level of encapsulation of network headers.
3758 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00003759struct skb_gso_cb {
Alexander Duyck802ab552016-04-10 21:45:03 -04003760 union {
3761 int mac_offset;
3762 int data_offset;
3763 };
Eric Dumazet3347c962013-10-19 11:42:56 -07003764 int encap_level;
Alexander Duyck76443452016-02-05 15:27:37 -08003765 __wsum csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003766 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00003767};
Konstantin Khlebnikov9207f9d2016-01-08 15:21:46 +03003768#define SKB_SGO_CB_OFFSET 32
3769#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
Pravin B Shelar68c33162013-02-14 14:02:41 +00003770
3771static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
3772{
3773 return (skb_mac_header(inner_skb) - inner_skb->head) -
3774 SKB_GSO_CB(inner_skb)->mac_offset;
3775}
3776
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00003777static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
3778{
3779 int new_headroom, headroom;
3780 int ret;
3781
3782 headroom = skb_headroom(skb);
3783 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
3784 if (ret)
3785 return ret;
3786
3787 new_headroom = skb_headroom(skb);
3788 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
3789 return 0;
3790}
3791
Alexander Duyck08b64fc2016-02-05 15:27:49 -08003792static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
3793{
3794 /* Do not update partial checksums if remote checksum is enabled. */
3795 if (skb->remcsum_offload)
3796 return;
3797
3798 SKB_GSO_CB(skb)->csum = res;
3799 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
3800}
3801
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003802/* Compute the checksum for a gso segment. First compute the checksum value
3803 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
3804 * then add in skb->csum (checksum from csum_start to end of packet).
3805 * skb->csum and csum_start are then updated to reflect the checksum of the
3806 * resultant packet starting from the transport header-- the resultant checksum
3807 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
3808 * header.
3809 */
3810static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
3811{
Alexander Duyck76443452016-02-05 15:27:37 -08003812 unsigned char *csum_start = skb_transport_header(skb);
3813 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
3814 __wsum partial = SKB_GSO_CB(skb)->csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003815
Alexander Duyck76443452016-02-05 15:27:37 -08003816 SKB_GSO_CB(skb)->csum = res;
3817 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003818
Alexander Duyck76443452016-02-05 15:27:37 -08003819 return csum_fold(csum_partial(csum_start, plen, partial));
Tom Herbert7e2b10c2014-06-04 17:20:02 -07003820}
3821
David S. Millerbdcc0922012-03-07 20:53:36 -05003822static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07003823{
3824 return skb_shinfo(skb)->gso_size;
3825}
3826
Eric Dumazet36a8f392013-09-29 01:21:32 -07003827/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05003828static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02003829{
3830 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
3831}
3832
Daniel Borkmann5293efe2016-08-18 01:00:39 +02003833static inline void skb_gso_reset(struct sk_buff *skb)
3834{
3835 skb_shinfo(skb)->gso_size = 0;
3836 skb_shinfo(skb)->gso_segs = 0;
3837 skb_shinfo(skb)->gso_type = 0;
3838}
3839
Joe Perches7965bd42013-09-26 14:48:15 -07003840void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07003841
3842static inline bool skb_warn_if_lro(const struct sk_buff *skb)
3843{
3844 /* LRO sets gso_size but not gso_type, whereas if GSO is really
3845 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003846 const struct skb_shared_info *shinfo = skb_shinfo(skb);
3847
Alexander Duyckb78462e2010-06-02 12:24:37 +00003848 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
3849 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07003850 __skb_warn_lro_forwarding(skb);
3851 return true;
3852 }
3853 return false;
3854}
3855
Herbert Xu35fc92a2007-03-26 23:22:20 -07003856static inline void skb_forward_csum(struct sk_buff *skb)
3857{
3858 /* Unfortunately we don't support this one. Any brave souls? */
3859 if (skb->ip_summed == CHECKSUM_COMPLETE)
3860 skb->ip_summed = CHECKSUM_NONE;
3861}
3862
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003863/**
3864 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
3865 * @skb: skb to check
3866 *
3867 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
3868 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
3869 * use this helper, to document places where we make this assertion.
3870 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003871static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07003872{
3873#ifdef DEBUG
3874 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
3875#endif
3876}
3877
Rusty Russellf35d9d82008-02-04 23:49:54 -05003878bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00003879
Paul Durranted1f50c2014-01-09 10:02:46 +00003880int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
Linus Lüssing9afd85c2015-05-02 14:01:07 +02003881struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
3882 unsigned int transport_len,
3883 __sum16(*skb_chkf)(struct sk_buff *skb));
Paul Durranted1f50c2014-01-09 10:02:46 +00003884
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00003885/**
3886 * skb_head_is_locked - Determine if the skb->head is locked down
3887 * @skb: skb to check
3888 *
3889 * The head on skbs build around a head frag can be removed if they are
3890 * not cloned. This function returns true if the skb head is locked down
3891 * due to either being allocated via kmalloc, or by being a clone with
3892 * multiple references to the head.
3893 */
3894static inline bool skb_head_is_locked(const struct sk_buff *skb)
3895{
3896 return !skb->head_frag || skb_cloned(skb);
3897}
Florian Westphalfe6cc552014-02-13 23:09:12 +01003898
3899/**
3900 * skb_gso_network_seglen - Return length of individual segments of a gso packet
3901 *
3902 * @skb: GSO skb
3903 *
3904 * skb_gso_network_seglen is used to determine the real size of the
3905 * individual segments, including Layer3 (IP, IPv6) and L4 headers (TCP/UDP).
3906 *
3907 * The MAC/L2 header is not accounted for.
3908 */
3909static inline unsigned int skb_gso_network_seglen(const struct sk_buff *skb)
3910{
3911 unsigned int hdr_len = skb_transport_header(skb) -
3912 skb_network_header(skb);
3913 return hdr_len + skb_gso_transport_seglen(skb);
3914}
Thomas Grafee122c72015-07-21 10:43:58 +02003915
Edward Cree179bc672016-02-11 20:48:04 +00003916/* Local Checksum Offload.
3917 * Compute outer checksum based on the assumption that the
3918 * inner checksum will be offloaded later.
Edward Creee8ae7b02016-02-11 21:03:37 +00003919 * See Documentation/networking/checksum-offloads.txt for
3920 * explanation of how this works.
Edward Cree179bc672016-02-11 20:48:04 +00003921 * Fill in outer checksum adjustment (e.g. with sum of outer
3922 * pseudo-header) before calling.
3923 * Also ensure that inner checksum is in linear data area.
3924 */
3925static inline __wsum lco_csum(struct sk_buff *skb)
3926{
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003927 unsigned char *csum_start = skb_checksum_start(skb);
3928 unsigned char *l4_hdr = skb_transport_header(skb);
3929 __wsum partial;
Edward Cree179bc672016-02-11 20:48:04 +00003930
3931 /* Start with complement of inner checksum adjustment */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003932 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
3933 skb->csum_offset));
3934
Edward Cree179bc672016-02-11 20:48:04 +00003935 /* Add in checksum of our headers (incl. outer checksum
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003936 * adjustment filled in by caller) and return result.
Edward Cree179bc672016-02-11 20:48:04 +00003937 */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08003938 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
Edward Cree179bc672016-02-11 20:48:04 +00003939}
3940
Linus Torvalds1da177e2005-04-16 15:20:36 -07003941#endif /* __KERNEL__ */
3942#endif /* _LINUX_SKBUFF_H */