blob: b466fbface2ec555c6e6bbdd01cf61df4caf16ed [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>
18#include <linux/compiler.h>
19#include <linux/time.h>
Paul Gortmaker187f1882011-11-23 20:12:59 -050020#include <linux/bug.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070021#include <linux/cache.h>
Eric Dumazet56b17422014-11-03 08:19:53 -080022#include <linux/rbtree.h>
David S. Miller51f3d022014-11-05 16:46:40 -050023#include <linux/socket.h>
Eric Dumazetc1d1b432017-08-31 16:48:22 -070024#include <linux/refcount.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;
Petar Penkovd58e4682018-09-14 07:46:18 -0700246struct bpf_prog;
247union bpf_attr;
Florian Westphaldf5042f2018-12-18 17:15:16 +0100248struct skb_ext;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700250#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251struct nf_conntrack {
252 atomic_t use;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253};
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -0700254#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700255
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +0200256#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700257struct nf_bridge_info {
Florian Westphal3eaf4022015-04-02 14:31:44 +0200258 enum {
259 BRNF_PROTO_UNCHANGED,
260 BRNF_PROTO_8021Q,
261 BRNF_PROTO_PPPOE
Florian Westphal7fb48c52015-05-03 22:05:28 +0200262 } orig_proto:8;
Florian Westphal72b1e5e2015-07-23 16:21:30 +0200263 u8 pkt_otherhost:1;
264 u8 in_prerouting:1;
265 u8 bridged_dnat:1;
Bernhard Thaler411ffb42015-05-30 15:28:28 +0200266 __u16 frag_max_size;
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000267 struct net_device *physindev;
Florian Westphal63cdbc02015-09-14 17:06:27 +0200268
269 /* always valid & non-NULL from FORWARD on, for physdev match */
270 struct net_device *physoutdev;
Florian Westphal7fb48c52015-05-03 22:05:28 +0200271 union {
Florian Westphal72b1e5e2015-07-23 16:21:30 +0200272 /* prerouting: detect dnat in orig/reply direction */
Bernhard Thaler72b31f72015-05-30 15:27:40 +0200273 __be32 ipv4_daddr;
274 struct in6_addr ipv6_daddr;
Florian Westphal72b1e5e2015-07-23 16:21:30 +0200275
276 /* after prerouting + nat detected: store original source
277 * mac since neigh resolution overwrites it, only used while
278 * skb is out in neigh layer.
279 */
280 char neigh_header[8];
Bernhard Thaler72b31f72015-05-30 15:27:40 +0200281 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700282};
283#endif
284
Linus Torvalds1da177e2005-04-16 15:20:36 -0700285struct sk_buff_head {
286 /* These two members must be first. */
287 struct sk_buff *next;
288 struct sk_buff *prev;
289
290 __u32 qlen;
291 spinlock_t lock;
292};
293
294struct sk_buff;
295
Ian Campbell9d4dde52011-12-22 23:39:14 +0000296/* To allow 64K frame to be packed as single skb without frag_list we
297 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
298 * buffers which do not start on a page boundary.
299 *
300 * Since GRO uses frags we allocate at least 16 regardless of page
301 * size.
Anton Blancharda715dea2011-03-27 14:57:26 +0000302 */
Ian Campbell9d4dde52011-12-22 23:39:14 +0000303#if (65536/PAGE_SIZE + 1) < 16
David S. Millereec00952011-03-29 23:34:08 -0700304#define MAX_SKB_FRAGS 16UL
Anton Blancharda715dea2011-03-27 14:57:26 +0000305#else
Ian Campbell9d4dde52011-12-22 23:39:14 +0000306#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
Anton Blancharda715dea2011-03-27 14:57:26 +0000307#endif
Hans Westgaard Ry5f74f82e2016-02-03 09:26:57 +0100308extern int sysctl_max_skb_frags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700309
Marcelo Ricardo Leitner3953c462016-06-02 15:05:40 -0300310/* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
311 * segment using its current segmentation instead.
312 */
313#define GSO_BY_FRAGS 0xFFFF
314
Linus Torvalds1da177e2005-04-16 15:20:36 -0700315typedef struct skb_frag_struct skb_frag_t;
316
317struct skb_frag_struct {
Ian Campbella8605c62011-10-19 23:01:49 +0000318 struct {
319 struct page *p;
320 } page;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000321#if (BITS_PER_LONG > 32) || (PAGE_SIZE >= 65536)
David S. Millera309bb02007-07-30 18:47:03 -0700322 __u32 page_offset;
323 __u32 size;
Eric Dumazetcb4dfe52010-09-23 05:06:54 +0000324#else
325 __u16 page_offset;
326 __u16 size;
327#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700328};
329
Pedro Tammela161e6132019-03-05 11:35:54 -0300330/**
331 * skb_frag_size - Returns the size of a skb fragment
332 * @frag: skb fragment
333 */
Eric Dumazet9e903e02011-10-18 21:00:24 +0000334static inline unsigned int skb_frag_size(const skb_frag_t *frag)
335{
336 return frag->size;
337}
338
Pedro Tammela161e6132019-03-05 11:35:54 -0300339/**
340 * skb_frag_size_set - Sets the size of a skb fragment
341 * @frag: skb fragment
342 * @size: size of fragment
343 */
Eric Dumazet9e903e02011-10-18 21:00:24 +0000344static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
345{
346 frag->size = size;
347}
348
Pedro Tammela161e6132019-03-05 11:35:54 -0300349/**
350 * skb_frag_size_add - Incrementes the size of a skb fragment by %delta
351 * @frag: skb fragment
352 * @delta: value to add
353 */
Eric Dumazet9e903e02011-10-18 21:00:24 +0000354static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
355{
356 frag->size += delta;
357}
358
Pedro Tammela161e6132019-03-05 11:35:54 -0300359/**
360 * skb_frag_size_sub - Decrements the size of a skb fragment by %delta
361 * @frag: skb fragment
362 * @delta: value to subtract
363 */
Eric Dumazet9e903e02011-10-18 21:00:24 +0000364static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
365{
366 frag->size -= delta;
367}
368
Pedro Tammela161e6132019-03-05 11:35:54 -0300369/**
370 * skb_frag_must_loop - Test if %p is a high memory page
371 * @p: fragment's page
372 */
Willem de Bruijnc613c202017-07-31 08:15:47 -0400373static inline bool skb_frag_must_loop(struct page *p)
374{
375#if defined(CONFIG_HIGHMEM)
376 if (PageHighMem(p))
377 return true;
378#endif
379 return false;
380}
381
382/**
383 * skb_frag_foreach_page - loop over pages in a fragment
384 *
385 * @f: skb frag to operate on
386 * @f_off: offset from start of f->page.p
387 * @f_len: length from f_off to loop over
388 * @p: (temp var) current page
389 * @p_off: (temp var) offset from start of current page,
390 * non-zero only on first page.
391 * @p_len: (temp var) length in current page,
392 * < PAGE_SIZE only on first and last page.
393 * @copied: (temp var) length so far, excluding current p_len.
394 *
395 * A fragment can hold a compound page, in which case per-page
396 * operations, notably kmap_atomic, must be called for each
397 * regular page.
398 */
399#define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
400 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
401 p_off = (f_off) & (PAGE_SIZE - 1), \
402 p_len = skb_frag_must_loop(p) ? \
403 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
404 copied = 0; \
405 copied < f_len; \
406 copied += p_len, p++, p_off = 0, \
407 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
408
Patrick Ohlyac45f602009-02-12 05:03:37 +0000409#define HAVE_HW_TIME_STAMP
410
411/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800412 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000413 * @hwtstamp: hardware time stamp transformed into duration
414 * since arbitrary point in time
Patrick Ohlyac45f602009-02-12 05:03:37 +0000415 *
416 * Software time stamps generated by ktime_get_real() are stored in
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400417 * skb->tstamp.
Patrick Ohlyac45f602009-02-12 05:03:37 +0000418 *
419 * hwtstamps can only be compared against other hwtstamps from
420 * the same device.
421 *
422 * This structure is attached to packets as part of the
423 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
424 */
425struct skb_shared_hwtstamps {
426 ktime_t hwtstamp;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000427};
428
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000429/* Definitions for tx_flags in struct skb_shared_info */
430enum {
431 /* generate hardware time stamp */
432 SKBTX_HW_TSTAMP = 1 << 0,
433
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400434 /* generate software time stamp when queueing packet to NIC */
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000435 SKBTX_SW_TSTAMP = 1 << 1,
436
437 /* device driver is going to provide hardware time stamp */
438 SKBTX_IN_PROGRESS = 1 << 2,
439
Shirley Maa6686f22011-07-06 12:22:12 +0000440 /* device driver supports TX zero-copy buffers */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000441 SKBTX_DEV_ZEROCOPY = 1 << 3,
Johannes Berg6e3e9392011-11-09 10:15:42 +0100442
443 /* generate wifi status information (where possible) */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000444 SKBTX_WIFI_STATUS = 1 << 4,
Pravin B Shelarc9af6db2013-02-11 09:27:41 +0000445
446 /* This indicates at least one fragment might be overwritten
447 * (as in vmsplice(), sendfile() ...)
448 * If we need to compute a TX checksum, we'll need to copy
449 * all frags to avoid possible bad checksum
450 */
451 SKBTX_SHARED_FRAG = 1 << 5,
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400452
453 /* generate software time stamp when entering packet scheduling */
454 SKBTX_SCHED_TSTAMP = 1 << 6,
Shirley Maa6686f22011-07-06 12:22:12 +0000455};
456
Willem de Bruijn52267792017-08-03 16:29:39 -0400457#define SKBTX_ZEROCOPY_FRAG (SKBTX_DEV_ZEROCOPY | SKBTX_SHARED_FRAG)
Willem de Bruijne1c8a602014-08-04 22:11:50 -0400458#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
Soheil Hassas Yeganeh0a2cf202016-04-27 23:39:01 -0400459 SKBTX_SCHED_TSTAMP)
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400460#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
461
Shirley Maa6686f22011-07-06 12:22:12 +0000462/*
463 * The callback notifies userspace to release buffers when skb DMA is done in
464 * lower device, the skb last reference should be 0 when calling this.
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000465 * The zerocopy_success argument is true if zero copy transmit occurred,
466 * false on data copy or out of memory error caused by data copy attempt.
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000467 * The ctx field is used to track device context.
468 * The desc field is used to track userspace buffer index.
Shirley Maa6686f22011-07-06 12:22:12 +0000469 */
470struct ubuf_info {
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000471 void (*callback)(struct ubuf_info *, bool zerocopy_success);
Willem de Bruijn4ab6c992017-08-03 16:29:42 -0400472 union {
473 struct {
474 unsigned long desc;
475 void *ctx;
476 };
477 struct {
478 u32 id;
479 u16 len;
480 u16 zerocopy:1;
481 u32 bytelen;
482 };
483 };
Eric Dumazetc1d1b432017-08-31 16:48:22 -0700484 refcount_t refcnt;
Willem de Bruijna91dbff2017-08-03 16:29:43 -0400485
486 struct mmpin {
487 struct user_struct *user;
488 unsigned int num_pg;
489 } mmp;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000490};
491
Willem de Bruijn52267792017-08-03 16:29:39 -0400492#define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
493
Sowmini Varadhan6f89dbc2018-02-15 10:49:32 -0800494int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
495void mm_unaccount_pinned_pages(struct mmpin *mmp);
496
Willem de Bruijn52267792017-08-03 16:29:39 -0400497struct ubuf_info *sock_zerocopy_alloc(struct sock *sk, size_t size);
Willem de Bruijn4ab6c992017-08-03 16:29:42 -0400498struct ubuf_info *sock_zerocopy_realloc(struct sock *sk, size_t size,
499 struct ubuf_info *uarg);
Willem de Bruijn52267792017-08-03 16:29:39 -0400500
501static inline void sock_zerocopy_get(struct ubuf_info *uarg)
502{
Eric Dumazetc1d1b432017-08-31 16:48:22 -0700503 refcount_inc(&uarg->refcnt);
Willem de Bruijn52267792017-08-03 16:29:39 -0400504}
505
506void sock_zerocopy_put(struct ubuf_info *uarg);
Willem de Bruijn52900d22018-11-30 15:32:40 -0500507void sock_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
Willem de Bruijn52267792017-08-03 16:29:39 -0400508
509void sock_zerocopy_callback(struct ubuf_info *uarg, bool success);
510
Willem de Bruijnb5947e52018-11-30 15:32:39 -0500511int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len);
Willem de Bruijn52267792017-08-03 16:29:39 -0400512int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
513 struct msghdr *msg, int len,
514 struct ubuf_info *uarg);
515
Linus Torvalds1da177e2005-04-16 15:20:36 -0700516/* This data is invariant across clones and lives at
517 * the end of the header data, ie. at skb->end.
518 */
519struct skb_shared_info {
Daniel Borkmannde8f3a82017-09-25 02:25:51 +0200520 __u8 __unused;
521 __u8 meta_len;
522 __u8 nr_frags;
Ian Campbell9f42f122012-01-05 07:13:39 +0000523 __u8 tx_flags;
Herbert Xu79671682006-06-22 02:40:14 -0700524 unsigned short gso_size;
525 /* Warning: this field is not always filled in (UFO)! */
526 unsigned short gso_segs;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700527 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000528 struct skb_shared_hwtstamps hwtstamps;
Steffen Klassert7f564522017-04-08 20:36:24 +0200529 unsigned int gso_type;
Willem de Bruijn09c2d252014-08-04 22:11:47 -0400530 u32 tskey;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700531
532 /*
533 * Warning : all fields before dataref are cleared in __alloc_skb()
534 */
535 atomic_t dataref;
536
Johann Baudy69e3c752009-05-18 22:11:22 -0700537 /* Intermediate layers must ensure that destructor_arg
538 * remains valid until skb destructor */
539 void * destructor_arg;
Shirley Maa6686f22011-07-06 12:22:12 +0000540
Eric Dumazetfed66382010-07-22 19:09:08 +0000541 /* must be last field, see pskb_expand_head() */
542 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700543};
544
545/* We divide dataref into two halves. The higher 16 bits hold references
546 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700547 * the entire skb->data. A clone of a headerless skb holds the length of
548 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700549 *
550 * All users must obey the rule that the skb->data reference count must be
551 * greater than or equal to the payload reference count.
552 *
553 * Holding a reference to the payload part means that the user does not
554 * care about modifications to the header part of skb->data.
555 */
556#define SKB_DATAREF_SHIFT 16
557#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
558
David S. Millerd179cd12005-08-17 14:57:30 -0700559
560enum {
Vijay Subramanianc8753d52014-10-02 10:00:43 -0700561 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
562 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
563 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
David S. Millerd179cd12005-08-17 14:57:30 -0700564};
565
Herbert Xu79671682006-06-22 02:40:14 -0700566enum {
567 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xu576a30e2006-06-27 13:22:38 -0700568
569 /* This indicates the skb is from an untrusted source. */
David S. Millerd9d30ad2017-07-03 07:31:57 -0700570 SKB_GSO_DODGY = 1 << 1,
Michael Chanb0da85372006-06-29 12:30:00 -0700571
572 /* This indicates the tcp segment has CWR set. */
David S. Millerd9d30ad2017-07-03 07:31:57 -0700573 SKB_GSO_TCP_ECN = 1 << 2,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700574
David S. Millerd9d30ad2017-07-03 07:31:57 -0700575 SKB_GSO_TCP_FIXEDID = 1 << 3,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800576
David S. Millerd9d30ad2017-07-03 07:31:57 -0700577 SKB_GSO_TCPV6 = 1 << 4,
Pravin B Shelar68c33162013-02-14 14:02:41 +0000578
David S. Millerd9d30ad2017-07-03 07:31:57 -0700579 SKB_GSO_FCOE = 1 << 5,
Pravin B Shelar73136262013-03-07 13:21:51 +0000580
David S. Millerd9d30ad2017-07-03 07:31:57 -0700581 SKB_GSO_GRE = 1 << 6,
Simon Horman0d89d202013-05-23 21:02:52 +0000582
David S. Millerd9d30ad2017-07-03 07:31:57 -0700583 SKB_GSO_GRE_CSUM = 1 << 7,
Eric Dumazetcb32f512013-10-19 11:42:57 -0700584
David S. Millerd9d30ad2017-07-03 07:31:57 -0700585 SKB_GSO_IPXIP4 = 1 << 8,
Eric Dumazet61c1db72013-10-20 20:47:30 -0700586
David S. Millerd9d30ad2017-07-03 07:31:57 -0700587 SKB_GSO_IPXIP6 = 1 << 9,
Tom Herbert0f4f4ff2014-06-04 17:20:16 -0700588
David S. Millerd9d30ad2017-07-03 07:31:57 -0700589 SKB_GSO_UDP_TUNNEL = 1 << 10,
Tom Herbert4749c092014-06-04 17:20:23 -0700590
David S. Millerd9d30ad2017-07-03 07:31:57 -0700591 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400592
David S. Millerd9d30ad2017-07-03 07:31:57 -0700593 SKB_GSO_PARTIAL = 1 << 12,
Alexander Duyck802ab552016-04-10 21:45:03 -0400594
David S. Millerd9d30ad2017-07-03 07:31:57 -0700595 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
Marcelo Ricardo Leitner90017ac2016-06-02 15:05:43 -0300596
David S. Millerd9d30ad2017-07-03 07:31:57 -0700597 SKB_GSO_SCTP = 1 << 14,
Steffen Klassertc7ef8f02017-04-14 10:05:36 +0200598
David S. Millerd9d30ad2017-07-03 07:31:57 -0700599 SKB_GSO_ESP = 1 << 15,
Willem de Bruijn0c19f8462017-11-21 10:22:25 -0500600
601 SKB_GSO_UDP = 1 << 16,
Willem de Bruijnee80d1e2018-04-26 13:42:16 -0400602
603 SKB_GSO_UDP_L4 = 1 << 17,
Herbert Xu79671682006-06-22 02:40:14 -0700604};
605
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700606#if BITS_PER_LONG > 32
607#define NET_SKBUFF_DATA_USES_OFFSET 1
608#endif
609
610#ifdef NET_SKBUFF_DATA_USES_OFFSET
611typedef unsigned int sk_buff_data_t;
612#else
613typedef unsigned char *sk_buff_data_t;
614#endif
615
Pedro Tammela161e6132019-03-05 11:35:54 -0300616/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700617 * struct sk_buff - socket buffer
618 * @next: Next buffer in list
619 * @prev: Previous buffer in list
Eric Dumazet363ec39232014-02-26 14:02:11 -0800620 * @tstamp: Time we arrived/left
Eric Dumazet56b17422014-11-03 08:19:53 -0800621 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700622 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700623 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700624 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000625 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700626 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627 * @len: Length of actual data
628 * @data_len: Data length
629 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700630 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700631 * @csum: Checksum (must include start/offset pair)
632 * @csum_start: Offset from skb->head where checksumming should start
633 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700634 * @priority: Packet queueing priority
WANG Cong60ff7462014-05-04 16:39:18 -0700635 * @ignore_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700636 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700637 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700638 * @nohdr: Payload reference only, must not modify header
639 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700640 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700641 * @ipvs_property: skbuff is owned by ipvs
Ido Schimmel875e8932018-12-04 08:15:10 +0000642 * @offload_fwd_mark: Packet was L2-forwarded in hardware
643 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
Willem de Bruijne7246e12017-01-07 17:06:35 -0500644 * @tc_skip_classify: do not classify packet. set by IFB device
Willem de Bruijn8dc07fd2017-01-07 17:06:37 -0500645 * @tc_at_ingress: used within tc_classify to distinguish in/egress
Willem de Bruijnbc31c902017-01-07 17:06:38 -0500646 * @tc_redirected: packet was redirected by a tc action
647 * @tc_from_ingress: if tc_redirected, tc_at_ingress at time of redirect
Randy Dunlap31729362008-02-18 20:52:13 -0800648 * @peeked: this packet has been seen already, so stats have been
649 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700650 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700651 * @protocol: Packet protocol from driver
652 * @destructor: Destruct function
Eric Dumazete2080072017-10-04 12:59:58 -0700653 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
Florian Westphala9e419d2017-01-23 18:21:59 +0100654 * @_nfct: Associated connection, if any (with nfctinfo bits)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700655 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800656 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657 * @tc_index: Traffic control index
Tom Herbert61b905d2014-03-24 15:34:47 -0700658 * @hash: the packet hash
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700659 * @queue_mapping: Queue mapping for multiqueue devices
Stefano Brivio8b700862018-07-11 14:39:42 +0200660 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
Florian Westphaldf5042f2018-12-18 17:15:16 +0100661 * @active_extensions: active extensions (skb_ext_id types)
Randy Dunlap553a5672008-04-20 10:51:01 -0700662 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700663 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Tom Herbert61b905d2014-03-24 15:34:47 -0700664 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
Changli Gao4ca24622011-08-19 07:26:44 -0700665 * ports.
Tom Herberta3b18dd2014-07-01 21:33:17 -0700666 * @sw_hash: indicates hash was computed in software stack
Johannes Berg6e3e9392011-11-09 10:15:42 +0100667 * @wifi_acked_valid: wifi_acked was set
668 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000669 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Davide Carattidba00302017-05-18 15:44:40 +0200670 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
Julian Anastasov4ff06202017-02-06 23:14:12 +0200671 * @dst_pending_confirm: need to confirm neighbour
Stefano Brivioa48d1892018-07-17 11:52:57 +0200672 * @decrypted: Decrypted SKB
Pedro Tammela161e6132019-03-05 11:35:54 -0300673 * @napi_id: id of the NAPI struct this skb came from
James Morris984bc162006-06-09 00:29:17 -0700674 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700675 * @mark: Generic packet mark
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000676 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b2008-07-14 22:49:06 -0700677 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000678 * @inner_protocol: Protocol (encapsulation)
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000679 * @inner_transport_header: Inner transport layer header (encapsulation)
680 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000681 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700682 * @transport_header: Transport layer header
683 * @network_header: Network layer header
684 * @mac_header: Link layer header
685 * @tail: Tail pointer
686 * @end: End pointer
687 * @head: Head of buffer
688 * @data: Data head pointer
689 * @truesize: Buffer size
690 * @users: User count - see {datagram,tcp}.c
Florian Westphaldf5042f2018-12-18 17:15:16 +0100691 * @extensions: allocated extensions, valid if active_extensions is nonzero
Linus Torvalds1da177e2005-04-16 15:20:36 -0700692 */
693
694struct sk_buff {
Eric Dumazet363ec39232014-02-26 14:02:11 -0800695 union {
Eric Dumazet56b17422014-11-03 08:19:53 -0800696 struct {
697 /* These two members must be first. */
698 struct sk_buff *next;
699 struct sk_buff *prev;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000700
Eric Dumazet56b17422014-11-03 08:19:53 -0800701 union {
Eric Dumazetbffa72c2017-09-19 05:14:24 -0700702 struct net_device *dev;
703 /* Some protocols might use this space to store information,
704 * while device pointer would be NULL.
705 * UDP receive path is one user.
706 */
707 unsigned long dev_scratch;
Eric Dumazet56b17422014-11-03 08:19:53 -0800708 };
709 };
Peter Oskolkovfa0f5272018-08-02 23:34:39 +0000710 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
David Millerd4546c22018-06-24 14:13:49 +0900711 struct list_head list;
Eric Dumazet56b17422014-11-03 08:19:53 -0800712 };
Peter Oskolkovfa0f5272018-08-02 23:34:39 +0000713
714 union {
715 struct sock *sk;
716 int ip_defrag_offset;
717 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700718
Eric Dumazetc84d9492016-12-08 11:41:55 -0800719 union {
Eric Dumazetbffa72c2017-09-19 05:14:24 -0700720 ktime_t tstamp;
Eric Dumazetd3edd062018-09-21 08:51:50 -0700721 u64 skb_mstamp_ns; /* earliest departure time */
Eric Dumazetc84d9492016-12-08 11:41:55 -0800722 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723 /*
724 * This is the control buffer. It is free to use for every
725 * layer. Please put your private variables there. If you
726 * want to keep them across layers you have to do a skb_clone()
727 * first. This is owned by whoever has the skb queued ATM.
728 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000729 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700730
Eric Dumazete2080072017-10-04 12:59:58 -0700731 union {
732 struct {
733 unsigned long _skb_refdst;
734 void (*destructor)(struct sk_buff *skb);
735 };
736 struct list_head tcp_tsorted_anchor;
737 };
738
Eric Dumazetb1937222014-09-28 22:18:47 -0700739#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +0100740 unsigned long _nfct;
Eric Dumazetb1937222014-09-28 22:18:47 -0700741#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700742 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700743 data_len;
744 __u16 mac_len,
745 hdr_len;
Eric Dumazetb1937222014-09-28 22:18:47 -0700746
747 /* Following fields are _not_ copied in __copy_skb_header()
748 * Note that queue_mapping is here mostly to fill a hole.
749 */
Eric Dumazetb1937222014-09-28 22:18:47 -0700750 __u16 queue_mapping;
Daniel Borkmann36bbef52016-09-20 00:26:13 +0200751
752/* if you move cloned around you also must adapt those constants */
753#ifdef __BIG_ENDIAN_BITFIELD
754#define CLONED_MASK (1 << 7)
755#else
756#define CLONED_MASK 1
757#endif
758#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
759
760 __u8 __cloned_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700761 __u8 cloned:1,
Harald Welte6869c4d2005-08-09 19:24:19 -0700762 nohdr:1,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800763 fclone:2,
Herbert Xua59322b2007-12-05 01:53:40 -0800764 peeked:1,
Eric Dumazetb1937222014-09-28 22:18:47 -0700765 head_frag:1,
Stefano Brivio8b700862018-07-11 14:39:42 +0200766 pfmemalloc:1;
Florian Westphaldf5042f2018-12-18 17:15:16 +0100767#ifdef CONFIG_SKB_EXTENSIONS
768 __u8 active_extensions;
769#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700770 /* fields enclosed in headers_start/headers_end are copied
771 * using a single memcpy() in __copy_skb_header()
772 */
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700773 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700774 __u32 headers_start[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700775 /* public: */
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200776
777/* if you move pkt_type around you also must adapt those constants */
778#ifdef __BIG_ENDIAN_BITFIELD
779#define PKT_TYPE_MAX (7 << 5)
780#else
781#define PKT_TYPE_MAX 7
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100782#endif
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200783#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
784
785 __u8 __pkt_type_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700786 __u8 pkt_type:3;
Eric Dumazetb1937222014-09-28 22:18:47 -0700787 __u8 ignore_df:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700788 __u8 nf_trace:1;
789 __u8 ip_summed:2;
790 __u8 ooo_okay:1;
Stefano Brivio8b700862018-07-11 14:39:42 +0200791
Eric Dumazetb1937222014-09-28 22:18:47 -0700792 __u8 l4_hash:1;
793 __u8 sw_hash:1;
794 __u8 wifi_acked_valid:1;
795 __u8 wifi_acked:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700796 __u8 no_fcs:1;
797 /* Indicates the inner headers are valid in the skbuff. */
798 __u8 encapsulation:1;
799 __u8 encap_hdr_csum:1;
800 __u8 csum_valid:1;
Stefano Brivio8b700862018-07-11 14:39:42 +0200801
Michał Mirosław0c4b2d32018-11-10 19:58:36 +0100802#ifdef __BIG_ENDIAN_BITFIELD
803#define PKT_VLAN_PRESENT_BIT 7
804#else
805#define PKT_VLAN_PRESENT_BIT 0
806#endif
807#define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset)
808 __u8 __pkt_vlan_present_offset[0];
809 __u8 vlan_present:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700810 __u8 csum_complete_sw:1;
811 __u8 csum_level:2;
Davide Carattidba00302017-05-18 15:44:40 +0200812 __u8 csum_not_inet:1;
Julian Anastasov4ff06202017-02-06 23:14:12 +0200813 __u8 dst_pending_confirm:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700814#ifdef CONFIG_IPV6_NDISC_NODETYPE
815 __u8 ndisc_nodetype:2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700816#endif
Stefano Brivio8b700862018-07-11 14:39:42 +0200817
Michał Mirosław0c4b2d32018-11-10 19:58:36 +0100818 __u8 ipvs_property:1;
Tom Herbert8bce6d72014-09-29 20:22:29 -0700819 __u8 inner_protocol_type:1;
Tom Herberte585f232014-11-04 09:06:54 -0800820 __u8 remcsum_offload:1;
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200821#ifdef CONFIG_NET_SWITCHDEV
822 __u8 offload_fwd_mark:1;
Ido Schimmel875e8932018-12-04 08:15:10 +0000823 __u8 offload_l3_fwd_mark:1;
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200824#endif
Willem de Bruijne7246e12017-01-07 17:06:35 -0500825#ifdef CONFIG_NET_CLS_ACT
826 __u8 tc_skip_classify:1;
Willem de Bruijn8dc07fd2017-01-07 17:06:37 -0500827 __u8 tc_at_ingress:1;
Willem de Bruijnbc31c902017-01-07 17:06:38 -0500828 __u8 tc_redirected:1;
829 __u8 tc_from_ingress:1;
Willem de Bruijne7246e12017-01-07 17:06:35 -0500830#endif
Stefano Brivioa48d1892018-07-17 11:52:57 +0200831#ifdef CONFIG_TLS_DEVICE
832 __u8 decrypted:1;
833#endif
Alexander Duyck4031ae62012-01-27 06:22:53 +0000834
Linus Torvalds1da177e2005-04-16 15:20:36 -0700835#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700836 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700837#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200838
Eric Dumazetb1937222014-09-28 22:18:47 -0700839 union {
840 __wsum csum;
841 struct {
842 __u16 csum_start;
843 __u16 csum_offset;
844 };
845 };
846 __u32 priority;
847 int skb_iif;
848 __u32 hash;
849 __be16 vlan_proto;
850 __u16 vlan_tci;
Eric Dumazet2bd82482015-02-03 23:48:24 -0800851#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
852 union {
853 unsigned int napi_id;
854 unsigned int sender_cpu;
855 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700856#endif
James Morris984bc162006-06-09 00:29:17 -0700857#ifdef CONFIG_NETWORK_SECMARK
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200858 __u32 secmark;
James Morris984bc162006-06-09 00:29:17 -0700859#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -0700860
Neil Horman3b885782009-10-12 13:26:31 -0700861 union {
862 __u32 mark;
Eric Dumazet16fad692013-03-14 05:40:32 +0000863 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700864 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700865
Tom Herbert8bce6d72014-09-29 20:22:29 -0700866 union {
867 __be16 inner_protocol;
868 __u8 inner_ipproto;
869 };
870
Simon Horman1a37e412013-05-23 21:02:51 +0000871 __u16 inner_transport_header;
872 __u16 inner_network_header;
873 __u16 inner_mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700874
875 __be16 protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000876 __u16 transport_header;
877 __u16 network_header;
878 __u16 mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700879
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700880 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700881 __u32 headers_end[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700882 /* public: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700883
Linus Torvalds1da177e2005-04-16 15:20:36 -0700884 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700885 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700886 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700887 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700888 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700889 unsigned int truesize;
Reshetova, Elena63354792017-06-30 13:07:58 +0300890 refcount_t users;
Florian Westphaldf5042f2018-12-18 17:15:16 +0100891
892#ifdef CONFIG_SKB_EXTENSIONS
893 /* only useable after checking ->active_extensions != 0 */
894 struct skb_ext *extensions;
895#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896};
897
898#ifdef __KERNEL__
899/*
900 * Handling routines are only of interest to the kernel
901 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700902
Mel Gormanc93bdd02012-07-31 16:44:19 -0700903#define SKB_ALLOC_FCLONE 0x01
904#define SKB_ALLOC_RX 0x02
Alexander Duyckfd11a832014-12-09 19:40:49 -0800905#define SKB_ALLOC_NAPI 0x04
Mel Gormanc93bdd02012-07-31 16:44:19 -0700906
Pedro Tammela161e6132019-03-05 11:35:54 -0300907/**
908 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
909 * @skb: buffer
910 */
Mel Gormanc93bdd02012-07-31 16:44:19 -0700911static inline bool skb_pfmemalloc(const struct sk_buff *skb)
912{
913 return unlikely(skb->pfmemalloc);
914}
915
Eric Dumazet7fee2262010-05-11 23:19:48 +0000916/*
917 * skb might have a dst pointer attached, refcounted or not.
918 * _skb_refdst low order bit is set if refcount was _not_ taken
919 */
920#define SKB_DST_NOREF 1UL
921#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
922
Florian Westphala9e419d2017-01-23 18:21:59 +0100923#define SKB_NFCT_PTRMASK ~(7UL)
Eric Dumazet7fee2262010-05-11 23:19:48 +0000924/**
925 * skb_dst - returns skb dst_entry
926 * @skb: buffer
927 *
928 * Returns skb dst_entry, regardless of reference taken or not.
929 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000930static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
931{
Pedro Tammela161e6132019-03-05 11:35:54 -0300932 /* If refdst was not refcounted, check we still are in a
Eric Dumazet7fee2262010-05-11 23:19:48 +0000933 * rcu_read_lock section
934 */
935 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
936 !rcu_read_lock_held() &&
937 !rcu_read_lock_bh_held());
938 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000939}
940
Eric Dumazet7fee2262010-05-11 23:19:48 +0000941/**
942 * skb_dst_set - sets skb dst
943 * @skb: buffer
944 * @dst: dst entry
945 *
946 * Sets skb dst, assuming a reference was taken on dst and should
947 * be released by skb_dst_drop()
948 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000949static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
950{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000951 skb->_skb_refdst = (unsigned long)dst;
952}
953
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200954/**
955 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
956 * @skb: buffer
957 * @dst: dst entry
958 *
959 * Sets skb dst, assuming a reference was not taken on dst.
960 * If dst entry is cached, we do not take reference and dst_release
961 * will be avoided by refdst_drop. If dst entry is not cached, we take
962 * reference, so that last dst_release can destroy the dst immediately.
963 */
964static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
965{
Hannes Frederic Sowadbfc4fb2014-12-06 19:19:42 +0100966 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
967 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200968}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000969
970/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300971 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000972 * @skb: buffer
973 */
974static inline bool skb_dst_is_noref(const struct sk_buff *skb)
975{
976 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000977}
978
Pedro Tammela161e6132019-03-05 11:35:54 -0300979/**
980 * skb_rtable - Returns the skb &rtable
981 * @skb: buffer
982 */
Eric Dumazet511c3f92009-06-02 05:14:27 +0000983static inline struct rtable *skb_rtable(const struct sk_buff *skb)
984{
Eric Dumazetadf30902009-06-02 05:19:30 +0000985 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000986}
987
Jamal Hadi Salim8b10cab2016-07-02 06:43:14 -0400988/* For mangling skb->pkt_type from user space side from applications
989 * such as nft, tc, etc, we only allow a conservative subset of
990 * possible pkt_types to be set.
991*/
992static inline bool skb_pkt_type_ok(u32 ptype)
993{
994 return ptype <= PACKET_OTHERHOST;
995}
996
Pedro Tammela161e6132019-03-05 11:35:54 -0300997/**
998 * skb_napi_id - Returns the skb's NAPI id
999 * @skb: buffer
1000 */
Miroslav Lichvar90b602f2017-05-19 17:52:37 +02001001static inline unsigned int skb_napi_id(const struct sk_buff *skb)
1002{
1003#ifdef CONFIG_NET_RX_BUSY_POLL
1004 return skb->napi_id;
1005#else
1006 return 0;
1007#endif
1008}
1009
Pedro Tammela161e6132019-03-05 11:35:54 -03001010/**
1011 * skb_unref - decrement the skb's reference count
1012 * @skb: buffer
1013 *
1014 * Returns true if we can free the skb.
1015 */
Paolo Abeni3889a8032017-06-12 11:23:41 +02001016static inline bool skb_unref(struct sk_buff *skb)
1017{
1018 if (unlikely(!skb))
1019 return false;
Reshetova, Elena63354792017-06-30 13:07:58 +03001020 if (likely(refcount_read(&skb->users) == 1))
Paolo Abeni3889a8032017-06-12 11:23:41 +02001021 smp_rmb();
Reshetova, Elena63354792017-06-30 13:07:58 +03001022 else if (likely(!refcount_dec_and_test(&skb->users)))
Paolo Abeni3889a8032017-06-12 11:23:41 +02001023 return false;
1024
1025 return true;
1026}
1027
Paolo Abeni0a463c72017-06-12 11:23:42 +02001028void skb_release_head_state(struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -07001029void kfree_skb(struct sk_buff *skb);
1030void kfree_skb_list(struct sk_buff *segs);
1031void skb_tx_error(struct sk_buff *skb);
1032void consume_skb(struct sk_buff *skb);
Paolo Abenica2c1412017-09-06 14:44:36 +02001033void __consume_stateless_skb(struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -07001034void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +00001035extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +00001036
Joe Perches7965bd42013-09-26 14:48:15 -07001037void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1038bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1039 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +00001040
Joe Perches7965bd42013-09-26 14:48:15 -07001041struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1042 int node);
Eric Dumazet2ea2f622015-04-24 16:05:01 -07001043struct sk_buff *__build_skb(void *data, unsigned int frag_size);
Joe Perches7965bd42013-09-26 14:48:15 -07001044struct sk_buff *build_skb(void *data, unsigned int frag_size);
Jesper Dangaard Brouerba0509b2019-04-12 17:07:37 +02001045struct sk_buff *build_skb_around(struct sk_buff *skb,
1046 void *data, unsigned int frag_size);
Pedro Tammela161e6132019-03-05 11:35:54 -03001047
1048/**
1049 * alloc_skb - allocate a network buffer
1050 * @size: size to allocate
1051 * @priority: allocation mask
1052 *
1053 * This function is a convenient wrapper around __alloc_skb().
1054 */
David S. Millerd179cd12005-08-17 14:57:30 -07001055static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +01001056 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -07001057{
Eric Dumazet564824b2010-10-11 19:05:25 +00001058 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -07001059}
1060
Eric Dumazet2e4e4412014-09-17 04:49:49 -07001061struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1062 unsigned long data_len,
1063 int max_page_order,
1064 int *errcode,
1065 gfp_t gfp_mask);
1066
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001067/* Layout of fast clones : [skb1][skb2][fclone_ref] */
1068struct sk_buff_fclones {
1069 struct sk_buff skb1;
1070
1071 struct sk_buff skb2;
1072
Reshetova, Elena26385952017-06-30 13:07:59 +03001073 refcount_t fclone_ref;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001074};
1075
1076/**
1077 * skb_fclone_busy - check if fclone is busy
Stephen Hemminger293de7d2016-10-23 09:28:29 -07001078 * @sk: socket
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001079 * @skb: buffer
1080 *
Masanari Iidabda13fe2015-12-13 16:53:02 +09001081 * Returns true if skb is a fast clone, and its clone is not freed.
Eric Dumazet39bb5e62014-10-30 10:32:34 -07001082 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1083 * so we also check that this didnt happen.
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001084 */
Eric Dumazet39bb5e62014-10-30 10:32:34 -07001085static inline bool skb_fclone_busy(const struct sock *sk,
1086 const struct sk_buff *skb)
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001087{
1088 const struct sk_buff_fclones *fclones;
1089
1090 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1091
1092 return skb->fclone == SKB_FCLONE_ORIG &&
Reshetova, Elena26385952017-06-30 13:07:59 +03001093 refcount_read(&fclones->fclone_ref) > 1 &&
Eric Dumazet39bb5e62014-10-30 10:32:34 -07001094 fclones->skb2.sk == sk;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001095}
1096
Pedro Tammela161e6132019-03-05 11:35:54 -03001097/**
1098 * alloc_skb_fclone - allocate a network buffer from fclone cache
1099 * @size: size to allocate
1100 * @priority: allocation mask
1101 *
1102 * This function is a convenient wrapper around __alloc_skb().
1103 */
David S. Millerd179cd12005-08-17 14:57:30 -07001104static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +01001105 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -07001106{
Mel Gormanc93bdd02012-07-31 16:44:19 -07001107 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -07001108}
1109
Joe Perches7965bd42013-09-26 14:48:15 -07001110struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
Toshiaki Makitab0768a82018-08-03 16:58:09 +09001111void skb_headers_offset_update(struct sk_buff *skb, int off);
Joe Perches7965bd42013-09-26 14:48:15 -07001112int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1113struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
Ilya Lesokhin08303c12018-04-30 10:16:11 +03001114void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
Joe Perches7965bd42013-09-26 14:48:15 -07001115struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
Octavian Purdilabad93e92014-06-12 01:36:26 +03001116struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1117 gfp_t gfp_mask, bool fclone);
1118static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1119 gfp_t gfp_mask)
1120{
1121 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1122}
Eric Dumazet117632e2011-12-03 21:39:53 +00001123
Joe Perches7965bd42013-09-26 14:48:15 -07001124int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1125struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1126 unsigned int headroom);
1127struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1128 int newtailroom, gfp_t priority);
Jason A. Donenfeld48a1df62017-06-04 04:16:22 +02001129int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1130 int offset, int len);
1131int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1132 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001133int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
Florian Fainellicd0a1372017-08-22 15:12:14 -07001134int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1135
1136/**
1137 * skb_pad - zero pad the tail of an skb
1138 * @skb: buffer to pad
1139 * @pad: space to pad
1140 *
1141 * Ensure that a buffer is followed by a padding area that is zero
1142 * filled. Used by network drivers which may DMA or transfer data
1143 * beyond the buffer end onto the wire.
1144 *
1145 * May return error in out of memory cases. The skb is freed on error.
1146 */
1147static inline int skb_pad(struct sk_buff *skb, int pad)
1148{
1149 return __skb_pad(skb, pad, true);
1150}
Neil Hormanead2ceb2009-03-11 09:49:55 +00001151#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001152
Hannes Frederic Sowabe12a1f2015-05-21 16:59:58 +02001153int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1154 int offset, size_t size);
1155
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08001156struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -07001157 __u32 lower_offset;
1158 __u32 upper_offset;
1159 __u32 frag_idx;
1160 __u32 stepped_offset;
1161 struct sk_buff *root_skb;
1162 struct sk_buff *cur_skb;
1163 __u8 *frag_data;
1164};
1165
Joe Perches7965bd42013-09-26 14:48:15 -07001166void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1167 unsigned int to, struct skb_seq_state *st);
1168unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1169 struct skb_seq_state *st);
1170void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -07001171
Joe Perches7965bd42013-09-26 14:48:15 -07001172unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
Bojan Prtvar059a2442015-02-22 11:46:35 +01001173 unsigned int to, struct ts_config *config);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -07001174
Tom Herbert09323cc2013-12-15 22:16:19 -08001175/*
1176 * Packet hash types specify the type of hash in skb_set_hash.
1177 *
1178 * Hash types refer to the protocol layer addresses which are used to
1179 * construct a packet's hash. The hashes are used to differentiate or identify
1180 * flows of the protocol layer for the hash type. Hash types are either
1181 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1182 *
1183 * Properties of hashes:
1184 *
1185 * 1) Two packets in different flows have different hash values
1186 * 2) Two packets in the same flow should have the same hash value
1187 *
1188 * A hash at a higher layer is considered to be more specific. A driver should
1189 * set the most specific hash possible.
1190 *
1191 * A driver cannot indicate a more specific hash than the layer at which a hash
1192 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1193 *
1194 * A driver may indicate a hash level which is less specific than the
1195 * actual layer the hash was computed on. For instance, a hash computed
1196 * at L4 may be considered an L3 hash. This should only be done if the
1197 * driver can't unambiguously determine that the HW computed the hash at
1198 * the higher layer. Note that the "should" in the second property above
1199 * permits this.
1200 */
1201enum pkt_hash_types {
1202 PKT_HASH_TYPE_NONE, /* Undefined type */
1203 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1204 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1205 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1206};
1207
Tom Herbertbcc83832015-09-01 09:24:24 -07001208static inline void skb_clear_hash(struct sk_buff *skb)
1209{
1210 skb->hash = 0;
1211 skb->sw_hash = 0;
1212 skb->l4_hash = 0;
1213}
1214
1215static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1216{
1217 if (!skb->l4_hash)
1218 skb_clear_hash(skb);
1219}
1220
1221static inline void
1222__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1223{
1224 skb->l4_hash = is_l4;
1225 skb->sw_hash = is_sw;
1226 skb->hash = hash;
1227}
1228
Tom Herbert09323cc2013-12-15 22:16:19 -08001229static inline void
1230skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1231{
Tom Herbertbcc83832015-09-01 09:24:24 -07001232 /* Used by drivers to set hash from HW */
1233 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1234}
1235
1236static inline void
1237__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1238{
1239 __skb_set_hash(skb, hash, true, is_l4);
Tom Herbert09323cc2013-12-15 22:16:19 -08001240}
1241
Tom Herberte5276932015-09-01 09:24:23 -07001242void __skb_get_hash(struct sk_buff *skb);
Florian Westphalb9177832016-10-26 18:49:46 +02001243u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
Tom Herberte5276932015-09-01 09:24:23 -07001244u32 skb_get_poff(const struct sk_buff *skb);
1245u32 __skb_get_poff(const struct sk_buff *skb, void *data,
Paolo Abeni72a338b2018-05-04 11:32:59 +02001246 const struct flow_keys_basic *keys, int hlen);
Tom Herberte5276932015-09-01 09:24:23 -07001247__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1248 void *data, int hlen_proto);
1249
1250static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1251 int thoff, u8 ip_proto)
1252{
1253 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1254}
1255
1256void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1257 const struct flow_dissector_key *key,
1258 unsigned int key_count);
1259
Willem de Bruijn2dfd1842018-09-18 16:20:18 -04001260#ifdef CONFIG_NET
Petar Penkovd58e4682018-09-14 07:46:18 -07001261int skb_flow_dissector_bpf_prog_attach(const union bpf_attr *attr,
1262 struct bpf_prog *prog);
1263
1264int skb_flow_dissector_bpf_prog_detach(const union bpf_attr *attr);
Willem de Bruijn2dfd1842018-09-18 16:20:18 -04001265#else
1266static inline int skb_flow_dissector_bpf_prog_attach(const union bpf_attr *attr,
1267 struct bpf_prog *prog)
1268{
1269 return -EOPNOTSUPP;
1270}
1271
1272static inline int skb_flow_dissector_bpf_prog_detach(const union bpf_attr *attr)
1273{
1274 return -EOPNOTSUPP;
1275}
1276#endif
Petar Penkovd58e4682018-09-14 07:46:18 -07001277
Stanislav Fomichev089b19a2019-04-22 08:55:44 -07001278struct bpf_flow_dissector;
1279bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
1280 __be16 proto, int nhoff, int hlen);
1281
Stanislav Fomichevc8aa7032019-01-28 08:53:53 -08001282struct bpf_flow_keys;
1283bool __skb_flow_bpf_dissect(struct bpf_prog *prog,
1284 const struct sk_buff *skb,
1285 struct flow_dissector *flow_dissector,
1286 struct bpf_flow_keys *flow_keys);
Stanislav Fomichev3cbf4ff2019-04-22 08:55:46 -07001287bool __skb_flow_dissect(const struct net *net,
1288 const struct sk_buff *skb,
Tom Herberte5276932015-09-01 09:24:23 -07001289 struct flow_dissector *flow_dissector,
1290 void *target_container,
Tom Herbertcd79a232015-09-01 09:24:27 -07001291 void *data, __be16 proto, int nhoff, int hlen,
1292 unsigned int flags);
Tom Herberte5276932015-09-01 09:24:23 -07001293
1294static inline bool skb_flow_dissect(const struct sk_buff *skb,
1295 struct flow_dissector *flow_dissector,
Tom Herbertcd79a232015-09-01 09:24:27 -07001296 void *target_container, unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001297{
Stanislav Fomichev3cbf4ff2019-04-22 08:55:46 -07001298 return __skb_flow_dissect(NULL, skb, flow_dissector,
1299 target_container, NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001300}
1301
1302static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
Tom Herbertcd79a232015-09-01 09:24:27 -07001303 struct flow_keys *flow,
1304 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001305{
1306 memset(flow, 0, sizeof(*flow));
Stanislav Fomichev3cbf4ff2019-04-22 08:55:46 -07001307 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1308 flow, NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001309}
1310
Paolo Abeni72a338b2018-05-04 11:32:59 +02001311static inline bool
Stanislav Fomichev3cbf4ff2019-04-22 08:55:46 -07001312skb_flow_dissect_flow_keys_basic(const struct net *net,
1313 const struct sk_buff *skb,
Paolo Abeni72a338b2018-05-04 11:32:59 +02001314 struct flow_keys_basic *flow, void *data,
1315 __be16 proto, int nhoff, int hlen,
1316 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001317{
1318 memset(flow, 0, sizeof(*flow));
Stanislav Fomichev3cbf4ff2019-04-22 08:55:46 -07001319 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
Tom Herbertcd79a232015-09-01 09:24:27 -07001320 data, proto, nhoff, hlen, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001321}
1322
Simon Horman62b32372017-12-04 11:31:48 +01001323void
1324skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1325 struct flow_dissector *flow_dissector,
1326 void *target_container);
1327
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001328static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001329{
Tom Herberta3b18dd2014-07-01 21:33:17 -07001330 if (!skb->l4_hash && !skb->sw_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001331 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001332
Tom Herbert61b905d2014-03-24 15:34:47 -07001333 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001334}
1335
David S. Miller20a17bf2015-09-01 21:19:17 -07001336static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
Tom Herbertf70ea012015-07-31 16:52:10 -07001337{
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001338 if (!skb->l4_hash && !skb->sw_hash) {
1339 struct flow_keys keys;
Tom Herbertde4c1f82015-09-01 18:11:04 -07001340 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001341
Tom Herbertde4c1f82015-09-01 18:11:04 -07001342 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001343 }
Tom Herbertf70ea012015-07-31 16:52:10 -07001344
1345 return skb->hash;
1346}
1347
Tom Herbert50fb7992015-05-01 11:30:12 -07001348__u32 skb_get_hash_perturb(const struct sk_buff *skb, u32 perturb);
1349
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001350static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1351{
Tom Herbert61b905d2014-03-24 15:34:47 -07001352 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001353}
1354
Tom Herbert3df7a742013-12-15 22:16:29 -08001355static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1356{
Tom Herbert61b905d2014-03-24 15:34:47 -07001357 to->hash = from->hash;
Tom Herberta3b18dd2014-07-01 21:33:17 -07001358 to->sw_hash = from->sw_hash;
Tom Herbert61b905d2014-03-24 15:34:47 -07001359 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -08001360};
1361
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001362#ifdef NET_SKBUFF_DATA_USES_OFFSET
1363static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1364{
1365 return skb->head + skb->end;
1366}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001367
1368static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1369{
1370 return skb->end;
1371}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001372#else
1373static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1374{
1375 return skb->end;
1376}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001377
1378static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1379{
1380 return skb->end - skb->head;
1381}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001382#endif
1383
Linus Torvalds1da177e2005-04-16 15:20:36 -07001384/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001385#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001386
Patrick Ohlyac45f602009-02-12 05:03:37 +00001387static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1388{
1389 return &skb_shinfo(skb)->hwtstamps;
1390}
1391
Willem de Bruijn52267792017-08-03 16:29:39 -04001392static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1393{
1394 bool is_zcopy = skb && skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY;
1395
1396 return is_zcopy ? skb_uarg(skb) : NULL;
1397}
1398
Willem de Bruijn52900d22018-11-30 15:32:40 -05001399static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1400 bool *have_ref)
Willem de Bruijn52267792017-08-03 16:29:39 -04001401{
1402 if (skb && uarg && !skb_zcopy(skb)) {
Willem de Bruijn52900d22018-11-30 15:32:40 -05001403 if (unlikely(have_ref && *have_ref))
1404 *have_ref = false;
1405 else
1406 sock_zerocopy_get(uarg);
Willem de Bruijn52267792017-08-03 16:29:39 -04001407 skb_shinfo(skb)->destructor_arg = uarg;
1408 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1409 }
1410}
1411
Willem de Bruijn5cd8d462018-11-20 13:00:18 -05001412static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1413{
1414 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1415 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1416}
1417
1418static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1419{
1420 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1421}
1422
1423static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1424{
1425 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1426}
1427
Willem de Bruijn52267792017-08-03 16:29:39 -04001428/* Release a reference on a zerocopy structure */
1429static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy)
1430{
1431 struct ubuf_info *uarg = skb_zcopy(skb);
1432
1433 if (uarg) {
Willem de Bruijn0a4a0602017-08-09 19:09:44 -04001434 if (uarg->callback == sock_zerocopy_callback) {
1435 uarg->zerocopy = uarg->zerocopy && zerocopy;
1436 sock_zerocopy_put(uarg);
Willem de Bruijn5cd8d462018-11-20 13:00:18 -05001437 } else if (!skb_zcopy_is_nouarg(skb)) {
Willem de Bruijn0a4a0602017-08-09 19:09:44 -04001438 uarg->callback(uarg, zerocopy);
1439 }
1440
Willem de Bruijn52267792017-08-03 16:29:39 -04001441 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1442 }
1443}
1444
1445/* Abort a zerocopy operation and revert zckey on error in send syscall */
1446static inline void skb_zcopy_abort(struct sk_buff *skb)
1447{
1448 struct ubuf_info *uarg = skb_zcopy(skb);
1449
1450 if (uarg) {
Willem de Bruijn52900d22018-11-30 15:32:40 -05001451 sock_zerocopy_put_abort(uarg, false);
Willem de Bruijn52267792017-08-03 16:29:39 -04001452 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1453 }
1454}
1455
David S. Millera8305bf2018-07-29 20:42:53 -07001456static inline void skb_mark_not_on_list(struct sk_buff *skb)
1457{
1458 skb->next = NULL;
1459}
1460
David S. Miller992cba72018-07-31 15:27:56 -07001461static inline void skb_list_del_init(struct sk_buff *skb)
1462{
1463 __list_del_entry(&skb->list);
1464 skb_mark_not_on_list(skb);
1465}
1466
Linus Torvalds1da177e2005-04-16 15:20:36 -07001467/**
1468 * skb_queue_empty - check if a queue is empty
1469 * @list: queue head
1470 *
1471 * Returns true if the queue is empty, false otherwise.
1472 */
1473static inline int skb_queue_empty(const struct sk_buff_head *list)
1474{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001475 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001476}
1477
1478/**
David S. Millerfc7ebb22008-09-23 00:34:07 -07001479 * skb_queue_is_last - check if skb is the last entry in the queue
1480 * @list: queue head
1481 * @skb: buffer
1482 *
1483 * Returns true if @skb is the last buffer on the list.
1484 */
1485static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1486 const struct sk_buff *skb)
1487{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001488 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -07001489}
1490
1491/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001492 * skb_queue_is_first - check if skb is the first entry in the queue
1493 * @list: queue head
1494 * @skb: buffer
1495 *
1496 * Returns true if @skb is the first buffer on the list.
1497 */
1498static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1499 const struct sk_buff *skb)
1500{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001501 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001502}
1503
1504/**
David S. Miller249c8b42008-09-23 00:44:42 -07001505 * skb_queue_next - return the next packet in the queue
1506 * @list: queue head
1507 * @skb: current buffer
1508 *
1509 * Return the next packet in @list after @skb. It is only valid to
1510 * call this if skb_queue_is_last() evaluates to false.
1511 */
1512static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1513 const struct sk_buff *skb)
1514{
1515 /* This BUG_ON may seem severe, but if we just return then we
1516 * are going to dereference garbage.
1517 */
1518 BUG_ON(skb_queue_is_last(list, skb));
1519 return skb->next;
1520}
1521
1522/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001523 * skb_queue_prev - return the prev packet in the queue
1524 * @list: queue head
1525 * @skb: current buffer
1526 *
1527 * Return the prev packet in @list before @skb. It is only valid to
1528 * call this if skb_queue_is_first() evaluates to false.
1529 */
1530static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1531 const struct sk_buff *skb)
1532{
1533 /* This BUG_ON may seem severe, but if we just return then we
1534 * are going to dereference garbage.
1535 */
1536 BUG_ON(skb_queue_is_first(list, skb));
1537 return skb->prev;
1538}
1539
1540/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001541 * skb_get - reference buffer
1542 * @skb: buffer to reference
1543 *
1544 * Makes another reference to a socket buffer and returns a pointer
1545 * to the buffer.
1546 */
1547static inline struct sk_buff *skb_get(struct sk_buff *skb)
1548{
Reshetova, Elena63354792017-06-30 13:07:58 +03001549 refcount_inc(&skb->users);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001550 return skb;
1551}
1552
1553/*
Geert Uytterhoevenf8821f92017-11-30 14:33:56 +01001554 * If users == 1, we are the only owner and can avoid redundant atomic changes.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001555 */
1556
1557/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001558 * skb_cloned - is the buffer a clone
1559 * @skb: buffer to check
1560 *
1561 * Returns true if the buffer was generated with skb_clone() and is
1562 * one of multiple shared copies of the buffer. Cloned buffers are
1563 * shared data so must not be written to under normal circumstances.
1564 */
1565static inline int skb_cloned(const struct sk_buff *skb)
1566{
1567 return skb->cloned &&
1568 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1569}
1570
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001571static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1572{
Mel Gormand0164ad2015-11-06 16:28:21 -08001573 might_sleep_if(gfpflags_allow_blocking(pri));
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001574
1575 if (skb_cloned(skb))
1576 return pskb_expand_head(skb, 0, 0, pri);
1577
1578 return 0;
1579}
1580
Linus Torvalds1da177e2005-04-16 15:20:36 -07001581/**
1582 * skb_header_cloned - is the header a clone
1583 * @skb: buffer to check
1584 *
1585 * Returns true if modifying the header part of the buffer requires
1586 * the data to be copied.
1587 */
1588static inline int skb_header_cloned(const struct sk_buff *skb)
1589{
1590 int dataref;
1591
1592 if (!skb->cloned)
1593 return 0;
1594
1595 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1596 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1597 return dataref != 1;
1598}
1599
Eric Dumazet9580bf22016-04-30 10:19:29 -07001600static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1601{
1602 might_sleep_if(gfpflags_allow_blocking(pri));
1603
1604 if (skb_header_cloned(skb))
1605 return pskb_expand_head(skb, 0, 0, pri);
1606
1607 return 0;
1608}
1609
Linus Torvalds1da177e2005-04-16 15:20:36 -07001610/**
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001611 * __skb_header_release - release reference to header
1612 * @skb: buffer to operate on
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001613 */
1614static inline void __skb_header_release(struct sk_buff *skb)
1615{
1616 skb->nohdr = 1;
1617 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1618}
1619
1620
1621/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001622 * skb_shared - is the buffer shared
1623 * @skb: buffer to check
1624 *
1625 * Returns true if more than one person has a reference to this
1626 * buffer.
1627 */
1628static inline int skb_shared(const struct sk_buff *skb)
1629{
Reshetova, Elena63354792017-06-30 13:07:58 +03001630 return refcount_read(&skb->users) != 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001631}
1632
1633/**
1634 * skb_share_check - check if buffer is shared and if so clone it
1635 * @skb: buffer to check
1636 * @pri: priority for memory allocation
1637 *
1638 * If the buffer is shared the buffer is cloned and the old copy
1639 * drops a reference. A new clone with a single reference is returned.
1640 * If the buffer is not shared the original buffer is returned. When
1641 * being called from interrupt status or with spinlocks held pri must
1642 * be GFP_ATOMIC.
1643 *
1644 * NULL is returned on a memory allocation failure.
1645 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001646static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001647{
Mel Gormand0164ad2015-11-06 16:28:21 -08001648 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001649 if (skb_shared(skb)) {
1650 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001651
1652 if (likely(nskb))
1653 consume_skb(skb);
1654 else
1655 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001656 skb = nskb;
1657 }
1658 return skb;
1659}
1660
1661/*
1662 * Copy shared buffers into a new sk_buff. We effectively do COW on
1663 * packets to handle cases where we have a local reader and forward
1664 * and a couple of other messy ones. The normal one is tcpdumping
1665 * a packet thats being forwarded.
1666 */
1667
1668/**
1669 * skb_unshare - make a copy of a shared buffer
1670 * @skb: buffer to check
1671 * @pri: priority for memory allocation
1672 *
1673 * If the socket buffer is a clone then this function creates a new
1674 * copy of the data, drops a reference count on the old copy and returns
1675 * the new copy with the reference count at 1. If the buffer is not a clone
1676 * the original buffer is returned. When called with a spinlock held or
1677 * from interrupt state @pri must be %GFP_ATOMIC
1678 *
1679 * %NULL is returned on a memory allocation failure.
1680 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001681static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001682 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001683{
Mel Gormand0164ad2015-11-06 16:28:21 -08001684 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001685 if (skb_cloned(skb)) {
1686 struct sk_buff *nskb = skb_copy(skb, pri);
Alexander Aring31eff812014-10-10 23:10:47 +02001687
1688 /* Free our shared copy */
1689 if (likely(nskb))
1690 consume_skb(skb);
1691 else
1692 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001693 skb = nskb;
1694 }
1695 return skb;
1696}
1697
1698/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001699 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001700 * @list_: list to peek at
1701 *
1702 * Peek an &sk_buff. Unlike most other operations you _MUST_
1703 * be careful with this one. A peek leaves the buffer on the
1704 * list and someone else may run off with it. You must hold
1705 * the appropriate locks or have a private queue to do this.
1706 *
1707 * Returns %NULL for an empty list or a pointer to the head element.
1708 * The reference count is not incremented and the reference is therefore
1709 * volatile. Use with caution.
1710 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001711static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001712{
Eric Dumazet18d07002012-04-30 16:31:46 +00001713 struct sk_buff *skb = list_->next;
1714
1715 if (skb == (struct sk_buff *)list_)
1716 skb = NULL;
1717 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001718}
1719
1720/**
David S. Miller8b69bd72018-08-11 18:43:38 -07001721 * __skb_peek - peek at the head of a non-empty &sk_buff_head
1722 * @list_: list to peek at
1723 *
1724 * Like skb_peek(), but the caller knows that the list is not empty.
1725 */
1726static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
1727{
1728 return list_->next;
1729}
1730
1731/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001732 * skb_peek_next - peek skb following the given one from a queue
1733 * @skb: skb to start from
1734 * @list_: list to peek at
1735 *
1736 * Returns %NULL when the end of the list is met or a pointer to the
1737 * next element. The reference count is not incremented and the
1738 * reference is therefore volatile. Use with caution.
1739 */
1740static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1741 const struct sk_buff_head *list_)
1742{
1743 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001744
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001745 if (next == (struct sk_buff *)list_)
1746 next = NULL;
1747 return next;
1748}
1749
1750/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001751 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001752 * @list_: list to peek at
1753 *
1754 * Peek an &sk_buff. Unlike most other operations you _MUST_
1755 * be careful with this one. A peek leaves the buffer on the
1756 * list and someone else may run off with it. You must hold
1757 * the appropriate locks or have a private queue to do this.
1758 *
1759 * Returns %NULL for an empty list or a pointer to the tail element.
1760 * The reference count is not incremented and the reference is therefore
1761 * volatile. Use with caution.
1762 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001763static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001764{
Eric Dumazet18d07002012-04-30 16:31:46 +00001765 struct sk_buff *skb = list_->prev;
1766
1767 if (skb == (struct sk_buff *)list_)
1768 skb = NULL;
1769 return skb;
1770
Linus Torvalds1da177e2005-04-16 15:20:36 -07001771}
1772
1773/**
1774 * skb_queue_len - get queue length
1775 * @list_: list to measure
1776 *
1777 * Return the length of an &sk_buff queue.
1778 */
1779static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1780{
1781 return list_->qlen;
1782}
1783
David S. Miller67fed452008-09-21 22:36:24 -07001784/**
1785 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1786 * @list: queue to initialize
1787 *
1788 * This initializes only the list and queue length aspects of
1789 * an sk_buff_head object. This allows to initialize the list
1790 * aspects of an sk_buff_head without reinitializing things like
1791 * the spinlock. It can also be used for on-stack sk_buff_head
1792 * objects where the spinlock is known to not be used.
1793 */
1794static inline void __skb_queue_head_init(struct sk_buff_head *list)
1795{
1796 list->prev = list->next = (struct sk_buff *)list;
1797 list->qlen = 0;
1798}
1799
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001800/*
1801 * This function creates a split out lock class for each invocation;
1802 * this is needed for now since a whole lot of users of the skb-queue
1803 * infrastructure in drivers have different locking usage (in hardirq)
1804 * than the networking core (in softirq only). In the long run either the
1805 * network layer or drivers should need annotation to consolidate the
1806 * main types of usage into 3 classes.
1807 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001808static inline void skb_queue_head_init(struct sk_buff_head *list)
1809{
1810 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001811 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001812}
1813
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001814static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1815 struct lock_class_key *class)
1816{
1817 skb_queue_head_init(list);
1818 lockdep_set_class(&list->lock, class);
1819}
1820
Linus Torvalds1da177e2005-04-16 15:20:36 -07001821/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001822 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001823 *
1824 * The "__skb_xxxx()" functions are the non-atomic ones that
1825 * can only be called with interrupts disabled.
1826 */
Gerrit Renkerbf299272008-04-14 00:04:51 -07001827static inline void __skb_insert(struct sk_buff *newsk,
1828 struct sk_buff *prev, struct sk_buff *next,
1829 struct sk_buff_head *list)
1830{
1831 newsk->next = next;
1832 newsk->prev = prev;
1833 next->prev = prev->next = newsk;
1834 list->qlen++;
1835}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001836
David S. Miller67fed452008-09-21 22:36:24 -07001837static inline void __skb_queue_splice(const struct sk_buff_head *list,
1838 struct sk_buff *prev,
1839 struct sk_buff *next)
1840{
1841 struct sk_buff *first = list->next;
1842 struct sk_buff *last = list->prev;
1843
1844 first->prev = prev;
1845 prev->next = first;
1846
1847 last->next = next;
1848 next->prev = last;
1849}
1850
1851/**
1852 * skb_queue_splice - join two skb lists, this is designed for stacks
1853 * @list: the new list to add
1854 * @head: the place to add it in the first list
1855 */
1856static inline void skb_queue_splice(const struct sk_buff_head *list,
1857 struct sk_buff_head *head)
1858{
1859 if (!skb_queue_empty(list)) {
1860 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001861 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001862 }
1863}
1864
1865/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001866 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001867 * @list: the new list to add
1868 * @head: the place to add it in the first list
1869 *
1870 * The list at @list is reinitialised
1871 */
1872static inline void skb_queue_splice_init(struct sk_buff_head *list,
1873 struct sk_buff_head *head)
1874{
1875 if (!skb_queue_empty(list)) {
1876 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001877 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001878 __skb_queue_head_init(list);
1879 }
1880}
1881
1882/**
1883 * skb_queue_splice_tail - join two skb lists, each list being a queue
1884 * @list: the new list to add
1885 * @head: the place to add it in the first list
1886 */
1887static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1888 struct sk_buff_head *head)
1889{
1890 if (!skb_queue_empty(list)) {
1891 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001892 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001893 }
1894}
1895
1896/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001897 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001898 * @list: the new list to add
1899 * @head: the place to add it in the first list
1900 *
1901 * Each of the lists is a queue.
1902 * The list at @list is reinitialised
1903 */
1904static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1905 struct sk_buff_head *head)
1906{
1907 if (!skb_queue_empty(list)) {
1908 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001909 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001910 __skb_queue_head_init(list);
1911 }
1912}
1913
Linus Torvalds1da177e2005-04-16 15:20:36 -07001914/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001915 * __skb_queue_after - queue a buffer at the list head
1916 * @list: list to use
1917 * @prev: place after this buffer
1918 * @newsk: buffer to queue
1919 *
1920 * Queue a buffer int the middle of a list. This function takes no locks
1921 * and you must therefore hold required locks before calling it.
1922 *
1923 * A buffer cannot be placed on two lists at the same time.
1924 */
1925static inline void __skb_queue_after(struct sk_buff_head *list,
1926 struct sk_buff *prev,
1927 struct sk_buff *newsk)
1928{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001929 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001930}
1931
Joe Perches7965bd42013-09-26 14:48:15 -07001932void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1933 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001934
Gerrit Renkerf5572852008-04-14 00:05:28 -07001935static inline void __skb_queue_before(struct sk_buff_head *list,
1936 struct sk_buff *next,
1937 struct sk_buff *newsk)
1938{
1939 __skb_insert(newsk, next->prev, next, list);
1940}
1941
Stephen Hemminger300ce172005-10-30 13:47:34 -08001942/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001943 * __skb_queue_head - queue a buffer at the list head
1944 * @list: list to use
1945 * @newsk: buffer to queue
1946 *
1947 * Queue a buffer at the start of a list. This function takes no locks
1948 * and you must therefore hold required locks before calling it.
1949 *
1950 * A buffer cannot be placed on two lists at the same time.
1951 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001952static inline void __skb_queue_head(struct sk_buff_head *list,
1953 struct sk_buff *newsk)
1954{
Stephen Hemminger300ce172005-10-30 13:47:34 -08001955 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001956}
Brian Norris4ea7b0c2019-02-11 13:02:25 -08001957void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001958
1959/**
1960 * __skb_queue_tail - queue a buffer at the list tail
1961 * @list: list to use
1962 * @newsk: buffer to queue
1963 *
1964 * Queue a buffer at the end of a list. This function takes no locks
1965 * and you must therefore hold required locks before calling it.
1966 *
1967 * A buffer cannot be placed on two lists at the same time.
1968 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001969static inline void __skb_queue_tail(struct sk_buff_head *list,
1970 struct sk_buff *newsk)
1971{
Gerrit Renkerf5572852008-04-14 00:05:28 -07001972 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001973}
Brian Norris4ea7b0c2019-02-11 13:02:25 -08001974void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001975
Linus Torvalds1da177e2005-04-16 15:20:36 -07001976/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07001977 * remove sk_buff from list. _Must_ be called atomically, and with
1978 * the list known..
1979 */
Joe Perches7965bd42013-09-26 14:48:15 -07001980void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001981static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
1982{
1983 struct sk_buff *next, *prev;
1984
1985 list->qlen--;
1986 next = skb->next;
1987 prev = skb->prev;
1988 skb->next = skb->prev = NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001989 next->prev = prev;
1990 prev->next = next;
1991}
1992
Gerrit Renkerf525c062008-04-14 00:04:12 -07001993/**
1994 * __skb_dequeue - remove from the head of the queue
1995 * @list: list to dequeue from
1996 *
1997 * Remove the head of the list. This function does not take any locks
1998 * so must be used with appropriate locks held only. The head item is
1999 * returned or %NULL if the list is empty.
2000 */
Gerrit Renkerf525c062008-04-14 00:04:12 -07002001static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2002{
2003 struct sk_buff *skb = skb_peek(list);
2004 if (skb)
2005 __skb_unlink(skb, list);
2006 return skb;
2007}
Brian Norris4ea7b0c2019-02-11 13:02:25 -08002008struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002009
2010/**
2011 * __skb_dequeue_tail - remove from the tail of the queue
2012 * @list: list to dequeue from
2013 *
2014 * Remove the tail of the list. This function does not take any locks
2015 * so must be used with appropriate locks held only. The tail item is
2016 * returned or %NULL if the list is empty.
2017 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002018static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2019{
2020 struct sk_buff *skb = skb_peek_tail(list);
2021 if (skb)
2022 __skb_unlink(skb, list);
2023 return skb;
2024}
Brian Norris4ea7b0c2019-02-11 13:02:25 -08002025struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002026
2027
David S. Millerbdcc0922012-03-07 20:53:36 -05002028static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002029{
2030 return skb->data_len;
2031}
2032
2033static inline unsigned int skb_headlen(const struct sk_buff *skb)
2034{
2035 return skb->len - skb->data_len;
2036}
2037
Willem de Bruijn3ece7822017-08-03 16:29:38 -04002038static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002039{
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03002040 unsigned int i, len = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002041
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03002042 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00002043 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Willem de Bruijn3ece7822017-08-03 16:29:38 -04002044 return len;
2045}
2046
2047static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2048{
2049 return skb_headlen(skb) + __skb_pagelen(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002050}
2051
Ian Campbell131ea662011-08-19 06:25:00 +00002052/**
2053 * __skb_fill_page_desc - initialise a paged fragment in an skb
2054 * @skb: buffer containing fragment to be initialised
2055 * @i: paged fragment index to initialise
2056 * @page: the page to use for this fragment
2057 * @off: the offset to the data with @page
2058 * @size: the length of the data
2059 *
2060 * Initialises the @i'th fragment of @skb to point to &size bytes at
2061 * offset @off within @page.
2062 *
2063 * Does not take any additional reference on the fragment.
2064 */
2065static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2066 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002067{
2068 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2069
Mel Gormanc48a11c2012-07-31 16:44:23 -07002070 /*
Michal Hocko2f064f32015-08-21 14:11:51 -07002071 * Propagate page pfmemalloc to the skb if we can. The problem is
2072 * that not all callers have unique ownership of the page but rely
2073 * on page_is_pfmemalloc doing the right thing(tm).
Mel Gormanc48a11c2012-07-31 16:44:23 -07002074 */
Ian Campbella8605c62011-10-19 23:01:49 +00002075 frag->page.p = page;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002076 frag->page_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00002077 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00002078
2079 page = compound_head(page);
Michal Hocko2f064f32015-08-21 14:11:51 -07002080 if (page_is_pfmemalloc(page))
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00002081 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00002082}
2083
2084/**
2085 * skb_fill_page_desc - initialise a paged fragment in an skb
2086 * @skb: buffer containing fragment to be initialised
2087 * @i: paged fragment index to initialise
2088 * @page: the page to use for this fragment
2089 * @off: the offset to the data with @page
2090 * @size: the length of the data
2091 *
2092 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01002093 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00002094 * addition updates @skb such that @i is the last fragment.
2095 *
2096 * Does not take any additional reference on the fragment.
2097 */
2098static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2099 struct page *page, int off, int size)
2100{
2101 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002102 skb_shinfo(skb)->nr_frags = i + 1;
2103}
2104
Joe Perches7965bd42013-09-26 14:48:15 -07002105void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2106 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07002107
Jason Wangf8e617e2013-11-01 14:07:47 +08002108void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2109 unsigned int truesize);
2110
Linus Torvalds1da177e2005-04-16 15:20:36 -07002111#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2112
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002113#ifdef NET_SKBUFF_DATA_USES_OFFSET
2114static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2115{
2116 return skb->head + skb->tail;
2117}
2118
2119static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2120{
2121 skb->tail = skb->data - skb->head;
2122}
2123
2124static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2125{
2126 skb_reset_tail_pointer(skb);
2127 skb->tail += offset;
2128}
Simon Horman7cc46192013-05-28 20:34:29 +00002129
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002130#else /* NET_SKBUFF_DATA_USES_OFFSET */
2131static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2132{
2133 return skb->tail;
2134}
2135
2136static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2137{
2138 skb->tail = skb->data;
2139}
2140
2141static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2142{
2143 skb->tail = skb->data + offset;
2144}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07002145
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002146#endif /* NET_SKBUFF_DATA_USES_OFFSET */
2147
Linus Torvalds1da177e2005-04-16 15:20:36 -07002148/*
2149 * Add data to an sk_buff
2150 */
Johannes Berg4df864c2017-06-16 14:29:21 +02002151void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2152void *skb_put(struct sk_buff *skb, unsigned int len);
2153static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002154{
Johannes Berg4df864c2017-06-16 14:29:21 +02002155 void *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002156 SKB_LINEAR_ASSERT(skb);
2157 skb->tail += len;
2158 skb->len += len;
2159 return tmp;
2160}
2161
yuan linyude77b962017-06-18 22:48:17 +08002162static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2163{
2164 void *tmp = __skb_put(skb, len);
2165
2166 memset(tmp, 0, len);
2167 return tmp;
2168}
2169
2170static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2171 unsigned int len)
2172{
2173 void *tmp = __skb_put(skb, len);
2174
2175 memcpy(tmp, data, len);
2176 return tmp;
2177}
2178
2179static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2180{
2181 *(u8 *)__skb_put(skb, 1) = val;
2182}
2183
Johannes Berg83ad3572017-06-14 22:17:20 +02002184static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
Johannes Berge45a79d2017-05-24 09:07:47 +02002185{
Johannes Berg83ad3572017-06-14 22:17:20 +02002186 void *tmp = skb_put(skb, len);
Johannes Berge45a79d2017-05-24 09:07:47 +02002187
2188 memset(tmp, 0, len);
2189
2190 return tmp;
2191}
2192
Johannes Berg59ae1d12017-06-16 14:29:20 +02002193static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2194 unsigned int len)
2195{
2196 void *tmp = skb_put(skb, len);
2197
2198 memcpy(tmp, data, len);
2199
2200 return tmp;
2201}
2202
Johannes Berg634fef62017-06-16 14:29:24 +02002203static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2204{
2205 *(u8 *)skb_put(skb, 1) = val;
2206}
2207
Johannes Bergd58ff352017-06-16 14:29:23 +02002208void *skb_push(struct sk_buff *skb, unsigned int len);
2209static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002210{
2211 skb->data -= len;
2212 skb->len += len;
2213 return skb->data;
2214}
2215
Johannes Bergaf728682017-06-16 14:29:22 +02002216void *skb_pull(struct sk_buff *skb, unsigned int len);
2217static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002218{
2219 skb->len -= len;
2220 BUG_ON(skb->len < skb->data_len);
2221 return skb->data += len;
2222}
2223
Johannes Bergaf728682017-06-16 14:29:22 +02002224static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
David S. Miller47d29642010-05-02 02:21:44 -07002225{
2226 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2227}
2228
Johannes Bergaf728682017-06-16 14:29:22 +02002229void *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002230
Johannes Bergaf728682017-06-16 14:29:22 +02002231static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002232{
2233 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07002234 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002235 return NULL;
2236 skb->len -= len;
2237 return skb->data += len;
2238}
2239
Johannes Bergaf728682017-06-16 14:29:22 +02002240static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002241{
2242 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2243}
2244
2245static inline int pskb_may_pull(struct sk_buff *skb, unsigned int len)
2246{
2247 if (likely(len <= skb_headlen(skb)))
2248 return 1;
2249 if (unlikely(len > skb->len))
2250 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07002251 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002252}
2253
Eric Dumazetc8c8b122016-12-07 09:19:33 -08002254void skb_condense(struct sk_buff *skb);
2255
Linus Torvalds1da177e2005-04-16 15:20:36 -07002256/**
2257 * skb_headroom - bytes at buffer head
2258 * @skb: buffer to check
2259 *
2260 * Return the number of bytes of free space at the head of an &sk_buff.
2261 */
Chuck Leverc2636b42007-10-23 21:07:32 -07002262static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002263{
2264 return skb->data - skb->head;
2265}
2266
2267/**
2268 * skb_tailroom - bytes at buffer end
2269 * @skb: buffer to check
2270 *
2271 * Return the number of bytes of free space at the tail of an sk_buff
2272 */
2273static inline int skb_tailroom(const struct sk_buff *skb)
2274{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07002275 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002276}
2277
2278/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00002279 * skb_availroom - bytes at buffer end
2280 * @skb: buffer to check
2281 *
2282 * Return the number of bytes of free space at the tail of an sk_buff
2283 * allocated by sk_stream_alloc()
2284 */
2285static inline int skb_availroom(const struct sk_buff *skb)
2286{
Eric Dumazet16fad692013-03-14 05:40:32 +00002287 if (skb_is_nonlinear(skb))
2288 return 0;
2289
2290 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00002291}
2292
2293/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002294 * skb_reserve - adjust headroom
2295 * @skb: buffer to alter
2296 * @len: bytes to move
2297 *
2298 * Increase the headroom of an empty &sk_buff by reducing the tail
2299 * room. This is only allowed for an empty buffer.
2300 */
David S. Miller8243126c2006-01-17 02:54:21 -08002301static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002302{
2303 skb->data += len;
2304 skb->tail += len;
2305}
2306
Benjamin Poirier1837b2e2016-02-29 15:03:33 -08002307/**
2308 * skb_tailroom_reserve - adjust reserved_tailroom
2309 * @skb: buffer to alter
2310 * @mtu: maximum amount of headlen permitted
2311 * @needed_tailroom: minimum amount of reserved_tailroom
2312 *
2313 * Set reserved_tailroom so that headlen can be as large as possible but
2314 * not larger than mtu and tailroom cannot be smaller than
2315 * needed_tailroom.
2316 * The required headroom should already have been reserved before using
2317 * this function.
2318 */
2319static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2320 unsigned int needed_tailroom)
2321{
2322 SKB_LINEAR_ASSERT(skb);
2323 if (mtu < skb_tailroom(skb) - needed_tailroom)
2324 /* use at most mtu */
2325 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2326 else
2327 /* use up to all available space */
2328 skb->reserved_tailroom = needed_tailroom;
2329}
2330
Tom Herbert8bce6d72014-09-29 20:22:29 -07002331#define ENCAP_TYPE_ETHER 0
2332#define ENCAP_TYPE_IPPROTO 1
2333
2334static inline void skb_set_inner_protocol(struct sk_buff *skb,
2335 __be16 protocol)
2336{
2337 skb->inner_protocol = protocol;
2338 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2339}
2340
2341static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2342 __u8 ipproto)
2343{
2344 skb->inner_ipproto = ipproto;
2345 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2346}
2347
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002348static inline void skb_reset_inner_headers(struct sk_buff *skb)
2349{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002350 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002351 skb->inner_network_header = skb->network_header;
2352 skb->inner_transport_header = skb->transport_header;
2353}
2354
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00002355static inline void skb_reset_mac_len(struct sk_buff *skb)
2356{
2357 skb->mac_len = skb->network_header - skb->mac_header;
2358}
2359
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002360static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2361 *skb)
2362{
2363 return skb->head + skb->inner_transport_header;
2364}
2365
Tom Herbert55dc5a92015-12-14 11:19:40 -08002366static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2367{
2368 return skb_inner_transport_header(skb) - skb->data;
2369}
2370
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002371static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2372{
2373 skb->inner_transport_header = skb->data - skb->head;
2374}
2375
2376static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2377 const int offset)
2378{
2379 skb_reset_inner_transport_header(skb);
2380 skb->inner_transport_header += offset;
2381}
2382
2383static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2384{
2385 return skb->head + skb->inner_network_header;
2386}
2387
2388static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2389{
2390 skb->inner_network_header = skb->data - skb->head;
2391}
2392
2393static inline void skb_set_inner_network_header(struct sk_buff *skb,
2394 const int offset)
2395{
2396 skb_reset_inner_network_header(skb);
2397 skb->inner_network_header += offset;
2398}
2399
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002400static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2401{
2402 return skb->head + skb->inner_mac_header;
2403}
2404
2405static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2406{
2407 skb->inner_mac_header = skb->data - skb->head;
2408}
2409
2410static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2411 const int offset)
2412{
2413 skb_reset_inner_mac_header(skb);
2414 skb->inner_mac_header += offset;
2415}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002416static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2417{
Cong Wang35d04612013-05-29 15:16:05 +08002418 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002419}
2420
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002421static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2422{
2423 return skb->head + skb->transport_header;
2424}
2425
2426static inline void skb_reset_transport_header(struct sk_buff *skb)
2427{
2428 skb->transport_header = skb->data - skb->head;
2429}
2430
2431static inline void skb_set_transport_header(struct sk_buff *skb,
2432 const int offset)
2433{
2434 skb_reset_transport_header(skb);
2435 skb->transport_header += offset;
2436}
2437
2438static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2439{
2440 return skb->head + skb->network_header;
2441}
2442
2443static inline void skb_reset_network_header(struct sk_buff *skb)
2444{
2445 skb->network_header = skb->data - skb->head;
2446}
2447
2448static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2449{
2450 skb_reset_network_header(skb);
2451 skb->network_header += offset;
2452}
2453
2454static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2455{
2456 return skb->head + skb->mac_header;
2457}
2458
Amir Vadaiea6da4f2017-02-07 09:56:06 +02002459static inline int skb_mac_offset(const struct sk_buff *skb)
2460{
2461 return skb_mac_header(skb) - skb->data;
2462}
2463
Daniel Borkmann0daf4342017-07-02 02:13:25 +02002464static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2465{
2466 return skb->network_header - skb->mac_header;
2467}
2468
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002469static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2470{
Cong Wang35d04612013-05-29 15:16:05 +08002471 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002472}
2473
2474static inline void skb_reset_mac_header(struct sk_buff *skb)
2475{
2476 skb->mac_header = skb->data - skb->head;
2477}
2478
2479static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2480{
2481 skb_reset_mac_header(skb);
2482 skb->mac_header += offset;
2483}
2484
Timo Teräs0e3da5b2013-12-16 11:02:09 +02002485static inline void skb_pop_mac_header(struct sk_buff *skb)
2486{
2487 skb->mac_header = skb->network_header;
2488}
2489
Maxim Mikityanskiyd2aa1252019-02-21 12:39:57 +00002490static inline void skb_probe_transport_header(struct sk_buff *skb)
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002491{
Paolo Abeni72a338b2018-05-04 11:32:59 +02002492 struct flow_keys_basic keys;
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002493
2494 if (skb_transport_header_was_set(skb))
2495 return;
Paolo Abeni72a338b2018-05-04 11:32:59 +02002496
Stanislav Fomichev3cbf4ff2019-04-22 08:55:46 -07002497 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2498 NULL, 0, 0, 0, 0))
Tom Herbert42aecaa2015-06-04 09:16:39 -07002499 skb_set_transport_header(skb, keys.control.thoff);
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002500}
2501
Eric Dumazet03606892012-02-23 10:55:02 +00002502static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2503{
2504 if (skb_mac_header_was_set(skb)) {
2505 const unsigned char *old_mac = skb_mac_header(skb);
2506
2507 skb_set_mac_header(skb, -skb->mac_len);
2508 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2509 }
2510}
2511
Michał Mirosław04fb4512010-12-14 15:24:08 +00002512static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2513{
2514 return skb->csum_start - skb_headroom(skb);
2515}
2516
Alexander Duyck08b64fc2016-02-05 15:27:49 -08002517static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2518{
2519 return skb->head + skb->csum_start;
2520}
2521
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002522static inline int skb_transport_offset(const struct sk_buff *skb)
2523{
2524 return skb_transport_header(skb) - skb->data;
2525}
2526
2527static inline u32 skb_network_header_len(const struct sk_buff *skb)
2528{
2529 return skb->transport_header - skb->network_header;
2530}
2531
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002532static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2533{
2534 return skb->inner_transport_header - skb->inner_network_header;
2535}
2536
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002537static inline int skb_network_offset(const struct sk_buff *skb)
2538{
2539 return skb_network_header(skb) - skb->data;
2540}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03002541
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002542static inline int skb_inner_network_offset(const struct sk_buff *skb)
2543{
2544 return skb_inner_network_header(skb) - skb->data;
2545}
2546
Changli Gaof9599ce2010-08-04 04:43:44 +00002547static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2548{
2549 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2550}
2551
Linus Torvalds1da177e2005-04-16 15:20:36 -07002552/*
2553 * CPUs often take a performance hit when accessing unaligned memory
2554 * locations. The actual performance hit varies, it can be small if the
2555 * hardware handles it or large if we have to take an exception and fix it
2556 * in software.
2557 *
2558 * Since an ethernet header is 14 bytes network drivers often end up with
2559 * the IP header at an unaligned offset. The IP header can be aligned by
2560 * shifting the start of the packet by 2 bytes. Drivers should do this
2561 * with:
2562 *
Tobias Klauser8660c122009-07-13 22:48:16 +00002563 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002564 *
2565 * The downside to this alignment of the IP header is that the DMA is now
2566 * unaligned. On some architectures the cost of an unaligned DMA is high
2567 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00002568 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002569 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2570 * to be overridden.
2571 */
2572#ifndef NET_IP_ALIGN
2573#define NET_IP_ALIGN 2
2574#endif
2575
Anton Blanchard025be812006-03-31 02:27:06 -08002576/*
2577 * The networking layer reserves some headroom in skb data (via
2578 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2579 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08002580 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08002581 *
2582 * Unfortunately this headroom changes the DMA alignment of the resulting
2583 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2584 * on some architectures. An architecture can override this value,
2585 * perhaps setting it to a cacheline in size (since that will maintain
2586 * cacheline alignment of the DMA). It must be a power of 2.
2587 *
David S. Millerd6301d32009-02-08 19:24:13 -08002588 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08002589 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07002590 *
2591 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2592 * to reduce average number of cache lines per packet.
2593 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07002594 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08002595 */
2596#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07002597#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08002598#endif
2599
Joe Perches7965bd42013-09-26 14:48:15 -07002600int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002601
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002602static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002603{
Yangtao Li5e1abdc2018-11-06 10:45:36 -05002604 if (WARN_ON(skb_is_nonlinear(skb)))
Herbert Xu3cc0e872006-06-09 16:13:38 -07002605 return;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002606 skb->len = len;
2607 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002608}
2609
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002610static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2611{
2612 __skb_set_length(skb, len);
2613}
2614
Joe Perches7965bd42013-09-26 14:48:15 -07002615void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002616
2617static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2618{
Herbert Xu3cc0e872006-06-09 16:13:38 -07002619 if (skb->data_len)
2620 return ___pskb_trim(skb, len);
2621 __skb_trim(skb, len);
2622 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002623}
2624
2625static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2626{
2627 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2628}
2629
2630/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07002631 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2632 * @skb: buffer to alter
2633 * @len: new length
2634 *
2635 * This is identical to pskb_trim except that the caller knows that
2636 * the skb is not cloned so we should never get an error due to out-
2637 * of-memory.
2638 */
2639static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2640{
2641 int err = pskb_trim(skb, len);
2642 BUG_ON(err);
2643}
2644
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002645static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2646{
2647 unsigned int diff = len - skb->len;
2648
2649 if (skb_tailroom(skb) < diff) {
2650 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2651 GFP_ATOMIC);
2652 if (ret)
2653 return ret;
2654 }
2655 __skb_set_length(skb, len);
2656 return 0;
2657}
2658
Herbert Xue9fa4f72006-08-13 20:12:58 -07002659/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002660 * skb_orphan - orphan a buffer
2661 * @skb: buffer to orphan
2662 *
2663 * If a buffer currently has an owner then we call the owner's
2664 * destructor function and make the @skb unowned. The buffer continues
2665 * to exist but is no longer charged to its former owner.
2666 */
2667static inline void skb_orphan(struct sk_buff *skb)
2668{
Eric Dumazetc34a7612013-07-30 16:11:15 -07002669 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002670 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002671 skb->destructor = NULL;
2672 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07002673 } else {
2674 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002675 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002676}
2677
2678/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002679 * skb_orphan_frags - orphan the frags contained in a buffer
2680 * @skb: buffer to orphan frags from
2681 * @gfp_mask: allocation mask for replacement pages
2682 *
2683 * For each frag in the SKB which needs a destructor (i.e. has an
2684 * owner) create a copy of that frag and release the original
2685 * page by calling the destructor.
2686 */
2687static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2688{
Willem de Bruijn1f8b9772017-08-03 16:29:41 -04002689 if (likely(!skb_zcopy(skb)))
2690 return 0;
2691 if (skb_uarg(skb)->callback == sock_zerocopy_callback)
2692 return 0;
2693 return skb_copy_ubufs(skb, gfp_mask);
2694}
2695
2696/* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2697static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2698{
2699 if (likely(!skb_zcopy(skb)))
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002700 return 0;
2701 return skb_copy_ubufs(skb, gfp_mask);
2702}
2703
2704/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002705 * __skb_queue_purge - empty a list
2706 * @list: list to empty
2707 *
2708 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2709 * the list and one reference dropped. This function does not take the
2710 * list lock and the caller must hold the relevant locks to use it.
2711 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002712static inline void __skb_queue_purge(struct sk_buff_head *list)
2713{
2714 struct sk_buff *skb;
2715 while ((skb = __skb_dequeue(list)) != NULL)
2716 kfree_skb(skb);
2717}
Brian Norris4ea7b0c2019-02-11 13:02:25 -08002718void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002719
Peter Oskolkov385114d2018-08-02 23:34:38 +00002720unsigned int skb_rbtree_purge(struct rb_root *root);
Yaogong Wang9f5afea2016-09-07 14:49:28 -07002721
Joe Perches7965bd42013-09-26 14:48:15 -07002722void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002723
Joe Perches7965bd42013-09-26 14:48:15 -07002724struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2725 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07002726
2727/**
2728 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2729 * @dev: network device to receive on
2730 * @length: length to allocate
2731 *
2732 * Allocate a new &sk_buff and assign it a usage count of one. The
2733 * buffer has unspecified headroom built in. Users should allocate
2734 * the headroom they think they need without accounting for the
2735 * built in space. The built in space is used for optimisations.
2736 *
2737 * %NULL is returned if there is no free memory. Although this function
2738 * allocates memory it can be called from an interrupt.
2739 */
2740static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002741 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002742{
2743 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2744}
2745
Eric Dumazet6f532612012-05-18 05:12:12 +00002746/* legacy helper around __netdev_alloc_skb() */
2747static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2748 gfp_t gfp_mask)
2749{
2750 return __netdev_alloc_skb(NULL, length, gfp_mask);
2751}
2752
2753/* legacy helper around netdev_alloc_skb() */
2754static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2755{
2756 return netdev_alloc_skb(NULL, length);
2757}
2758
2759
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002760static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2761 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002762{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002763 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002764
2765 if (NET_IP_ALIGN && skb)
2766 skb_reserve(skb, NET_IP_ALIGN);
2767 return skb;
2768}
2769
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002770static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2771 unsigned int length)
2772{
2773 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2774}
2775
Alexander Duyck181edb22015-05-06 21:12:03 -07002776static inline void skb_free_frag(void *addr)
2777{
Alexander Duyck8c2dd3e2017-01-10 16:58:06 -08002778 page_frag_free(addr);
Alexander Duyck181edb22015-05-06 21:12:03 -07002779}
2780
Alexander Duyckffde7322014-12-09 19:40:42 -08002781void *napi_alloc_frag(unsigned int fragsz);
Alexander Duyckfd11a832014-12-09 19:40:49 -08002782struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2783 unsigned int length, gfp_t gfp_mask);
2784static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2785 unsigned int length)
2786{
2787 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2788}
Jesper Dangaard Brouer795bb1c2016-02-08 13:14:59 +01002789void napi_consume_skb(struct sk_buff *skb, int budget);
2790
2791void __kfree_skb_flush(void);
Jesper Dangaard Brouer15fad712016-02-08 13:15:04 +01002792void __kfree_skb_defer(struct sk_buff *skb);
Alexander Duyckffde7322014-12-09 19:40:42 -08002793
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002794/**
Alexander Duyck71dfda52014-11-11 09:26:34 -08002795 * __dev_alloc_pages - allocate page for network Rx
2796 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2797 * @order: size of the allocation
2798 *
2799 * Allocate a new page.
2800 *
2801 * %NULL is returned if there is no free memory.
2802*/
2803static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2804 unsigned int order)
2805{
2806 /* This piece of code contains several assumptions.
2807 * 1. This is for device Rx, therefor a cold page is preferred.
2808 * 2. The expectation is the user wants a compound page.
2809 * 3. If requesting a order 0 page it will not be compound
2810 * due to the check to see if order has a value in prep_new_page
2811 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2812 * code in gfp_to_alloc_flags that should be enforcing this.
2813 */
Mel Gorman453f85d2017-11-15 17:38:03 -08002814 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
Alexander Duyck71dfda52014-11-11 09:26:34 -08002815
2816 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2817}
2818
2819static inline struct page *dev_alloc_pages(unsigned int order)
2820{
Neil Horman95829b32016-05-19 11:30:54 -04002821 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002822}
2823
2824/**
2825 * __dev_alloc_page - allocate a page for network Rx
2826 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2827 *
2828 * Allocate a new page.
2829 *
2830 * %NULL is returned if there is no free memory.
2831 */
2832static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2833{
2834 return __dev_alloc_pages(gfp_mask, 0);
2835}
2836
2837static inline struct page *dev_alloc_page(void)
2838{
Neil Horman95829b32016-05-19 11:30:54 -04002839 return dev_alloc_pages(0);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002840}
2841
2842/**
Mel Gorman06140022012-07-31 16:44:24 -07002843 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2844 * @page: The page that was allocated from skb_alloc_page
2845 * @skb: The skb that may need pfmemalloc set
2846 */
2847static inline void skb_propagate_pfmemalloc(struct page *page,
2848 struct sk_buff *skb)
2849{
Michal Hocko2f064f32015-08-21 14:11:51 -07002850 if (page_is_pfmemalloc(page))
Mel Gorman06140022012-07-31 16:44:24 -07002851 skb->pfmemalloc = true;
2852}
2853
Eric Dumazet564824b2010-10-11 19:05:25 +00002854/**
Masanari Iidae2278672014-02-18 22:54:36 +09002855 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002856 * @frag: the paged fragment
2857 *
2858 * Returns the &struct page associated with @frag.
2859 */
2860static inline struct page *skb_frag_page(const skb_frag_t *frag)
2861{
Ian Campbella8605c62011-10-19 23:01:49 +00002862 return frag->page.p;
Ian Campbell131ea662011-08-19 06:25:00 +00002863}
2864
2865/**
2866 * __skb_frag_ref - take an addition reference on a paged fragment.
2867 * @frag: the paged fragment
2868 *
2869 * Takes an additional reference on the paged fragment @frag.
2870 */
2871static inline void __skb_frag_ref(skb_frag_t *frag)
2872{
2873 get_page(skb_frag_page(frag));
2874}
2875
2876/**
2877 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2878 * @skb: the buffer
2879 * @f: the fragment offset.
2880 *
2881 * Takes an additional reference on the @f'th paged fragment of @skb.
2882 */
2883static inline void skb_frag_ref(struct sk_buff *skb, int f)
2884{
2885 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2886}
2887
2888/**
2889 * __skb_frag_unref - release a reference on a paged fragment.
2890 * @frag: the paged fragment
2891 *
2892 * Releases a reference on the paged fragment @frag.
2893 */
2894static inline void __skb_frag_unref(skb_frag_t *frag)
2895{
2896 put_page(skb_frag_page(frag));
2897}
2898
2899/**
2900 * skb_frag_unref - release a reference on a paged fragment of an skb.
2901 * @skb: the buffer
2902 * @f: the fragment offset
2903 *
2904 * Releases a reference on the @f'th paged fragment of @skb.
2905 */
2906static inline void skb_frag_unref(struct sk_buff *skb, int f)
2907{
2908 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
2909}
2910
2911/**
2912 * skb_frag_address - gets the address of the data contained in a paged fragment
2913 * @frag: the paged fragment buffer
2914 *
2915 * Returns the address of the data within @frag. The page must already
2916 * be mapped.
2917 */
2918static inline void *skb_frag_address(const skb_frag_t *frag)
2919{
2920 return page_address(skb_frag_page(frag)) + frag->page_offset;
2921}
2922
2923/**
2924 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
2925 * @frag: the paged fragment buffer
2926 *
2927 * Returns the address of the data within @frag. Checks that the page
2928 * is mapped and returns %NULL otherwise.
2929 */
2930static inline void *skb_frag_address_safe(const skb_frag_t *frag)
2931{
2932 void *ptr = page_address(skb_frag_page(frag));
2933 if (unlikely(!ptr))
2934 return NULL;
2935
2936 return ptr + frag->page_offset;
2937}
2938
2939/**
2940 * __skb_frag_set_page - sets the page contained in a paged fragment
2941 * @frag: the paged fragment
2942 * @page: the page to set
2943 *
2944 * Sets the fragment @frag to contain @page.
2945 */
2946static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
2947{
Ian Campbella8605c62011-10-19 23:01:49 +00002948 frag->page.p = page;
Ian Campbell131ea662011-08-19 06:25:00 +00002949}
2950
2951/**
2952 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
2953 * @skb: the buffer
2954 * @f: the fragment offset
2955 * @page: the page to set
2956 *
2957 * Sets the @f'th fragment of @skb to contain @page.
2958 */
2959static inline void skb_frag_set_page(struct sk_buff *skb, int f,
2960 struct page *page)
2961{
2962 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
2963}
2964
Eric Dumazet400dfd32013-10-17 16:27:07 -07002965bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
2966
Ian Campbell131ea662011-08-19 06:25:00 +00002967/**
2968 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00002969 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00002970 * @frag: the paged fragment to map
2971 * @offset: the offset within the fragment (starting at the
2972 * fragment's own offset)
2973 * @size: the number of bytes to map
Mauro Carvalho Chehab771b00a2017-05-12 09:19:29 -03002974 * @dir: the direction of the mapping (``PCI_DMA_*``)
Ian Campbell131ea662011-08-19 06:25:00 +00002975 *
2976 * Maps the page associated with @frag to @device.
2977 */
2978static inline dma_addr_t skb_frag_dma_map(struct device *dev,
2979 const skb_frag_t *frag,
2980 size_t offset, size_t size,
2981 enum dma_data_direction dir)
2982{
2983 return dma_map_page(dev, skb_frag_page(frag),
2984 frag->page_offset + offset, size, dir);
2985}
2986
Eric Dumazet117632e2011-12-03 21:39:53 +00002987static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
2988 gfp_t gfp_mask)
2989{
2990 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
2991}
2992
Octavian Purdilabad93e92014-06-12 01:36:26 +03002993
2994static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
2995 gfp_t gfp_mask)
2996{
2997 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
2998}
2999
3000
Ian Campbell131ea662011-08-19 06:25:00 +00003001/**
Patrick McHardy334a8132007-06-25 04:35:20 -07003002 * skb_clone_writable - is the header of a clone writable
3003 * @skb: buffer to check
3004 * @len: length up to which to write
3005 *
3006 * Returns true if modifying the header part of the cloned buffer
3007 * does not requires the data to be copied.
3008 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003009static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07003010{
3011 return !skb_header_cloned(skb) &&
3012 skb_headroom(skb) + len <= skb->hdr_len;
3013}
3014
Daniel Borkmann36976492016-02-19 23:05:25 +01003015static inline int skb_try_make_writable(struct sk_buff *skb,
3016 unsigned int write_len)
3017{
3018 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3019 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3020}
3021
Herbert Xud9cc2042007-09-16 16:21:16 -07003022static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3023 int cloned)
3024{
3025 int delta = 0;
3026
Herbert Xud9cc2042007-09-16 16:21:16 -07003027 if (headroom > skb_headroom(skb))
3028 delta = headroom - skb_headroom(skb);
3029
3030 if (delta || cloned)
3031 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3032 GFP_ATOMIC);
3033 return 0;
3034}
3035
Patrick McHardy334a8132007-06-25 04:35:20 -07003036/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07003037 * skb_cow - copy header of skb when it is required
3038 * @skb: buffer to cow
3039 * @headroom: needed headroom
3040 *
3041 * If the skb passed lacks sufficient headroom or its data part
3042 * is shared, data is reallocated. If reallocation fails, an error
3043 * is returned and original skb is not changed.
3044 *
3045 * The result is skb with writable area skb->head...skb->tail
3046 * and at least @headroom of space at head.
3047 */
3048static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3049{
Herbert Xud9cc2042007-09-16 16:21:16 -07003050 return __skb_cow(skb, headroom, skb_cloned(skb));
3051}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003052
Herbert Xud9cc2042007-09-16 16:21:16 -07003053/**
3054 * skb_cow_head - skb_cow but only making the head writable
3055 * @skb: buffer to cow
3056 * @headroom: needed headroom
3057 *
3058 * This function is identical to skb_cow except that we replace the
3059 * skb_cloned check by skb_header_cloned. It should be used when
3060 * you only need to push on some header and do not need to modify
3061 * the data.
3062 */
3063static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3064{
3065 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07003066}
3067
3068/**
3069 * skb_padto - pad an skbuff up to a minimal size
3070 * @skb: buffer to pad
3071 * @len: minimal length
3072 *
3073 * Pads up a buffer to ensure the trailing bytes exist and are
3074 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07003075 * is untouched. Otherwise it is extended. Returns zero on
3076 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003077 */
Herbert Xu5b057c62006-06-23 02:06:41 -07003078static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003079{
3080 unsigned int size = skb->len;
3081 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07003082 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07003083 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003084}
3085
Alexander Duyck9c0c1122014-12-03 08:17:33 -08003086/**
Brian Norris4ea7b0c2019-02-11 13:02:25 -08003087 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
Alexander Duyck9c0c1122014-12-03 08:17:33 -08003088 * @skb: buffer to pad
3089 * @len: minimal length
Florian Fainellicd0a1372017-08-22 15:12:14 -07003090 * @free_on_error: free buffer on error
3091 *
3092 * Pads up a buffer to ensure the trailing bytes exist and are
3093 * blanked. If the buffer already contains sufficient data it
3094 * is untouched. Otherwise it is extended. Returns zero on
3095 * success. The skb is freed on error if @free_on_error is true.
3096 */
3097static inline int __skb_put_padto(struct sk_buff *skb, unsigned int len,
3098 bool free_on_error)
3099{
3100 unsigned int size = skb->len;
3101
3102 if (unlikely(size < len)) {
3103 len -= size;
3104 if (__skb_pad(skb, len, free_on_error))
3105 return -ENOMEM;
3106 __skb_put(skb, len);
3107 }
3108 return 0;
3109}
3110
3111/**
3112 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3113 * @skb: buffer to pad
3114 * @len: minimal length
Alexander Duyck9c0c1122014-12-03 08:17:33 -08003115 *
3116 * Pads up a buffer to ensure the trailing bytes exist and are
3117 * blanked. If the buffer already contains sufficient data it
3118 * is untouched. Otherwise it is extended. Returns zero on
3119 * success. The skb is freed on error.
3120 */
3121static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
3122{
Florian Fainellicd0a1372017-08-22 15:12:14 -07003123 return __skb_put_padto(skb, len, true);
Alexander Duyck9c0c1122014-12-03 08:17:33 -08003124}
3125
Linus Torvalds1da177e2005-04-16 15:20:36 -07003126static inline int skb_add_data(struct sk_buff *skb,
Al Viroaf2b0402014-11-27 21:44:24 -05003127 struct iov_iter *from, int copy)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003128{
3129 const int off = skb->len;
3130
3131 if (skb->ip_summed == CHECKSUM_NONE) {
Al Viroaf2b0402014-11-27 21:44:24 -05003132 __wsum csum = 0;
Al Viro15e6cb42016-11-01 22:42:45 -04003133 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3134 &csum, from)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003135 skb->csum = csum_block_add(skb->csum, csum, off);
3136 return 0;
3137 }
Al Viro15e6cb42016-11-01 22:42:45 -04003138 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003139 return 0;
3140
3141 __skb_trim(skb, off);
3142 return -EFAULT;
3143}
3144
Eric Dumazet38ba0a62012-04-23 17:48:27 +00003145static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3146 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003147{
Willem de Bruijn1f8b9772017-08-03 16:29:41 -04003148 if (skb_zcopy(skb))
3149 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003150 if (i) {
Eric Dumazet9e903e02011-10-18 21:00:24 +00003151 const struct skb_frag_struct *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07003152
Ian Campbellea2ab692011-08-22 23:44:58 +00003153 return page == skb_frag_page(frag) &&
Eric Dumazet9e903e02011-10-18 21:00:24 +00003154 off == frag->page_offset + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003155 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00003156 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003157}
3158
Herbert Xu364c6ba2006-06-09 16:10:40 -07003159static inline int __skb_linearize(struct sk_buff *skb)
3160{
3161 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3162}
3163
Linus Torvalds1da177e2005-04-16 15:20:36 -07003164/**
3165 * skb_linearize - convert paged skb to linear one
3166 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07003167 *
3168 * If there is no free memory -ENOMEM is returned, otherwise zero
3169 * is returned and the old skb data released.
3170 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07003171static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003172{
Herbert Xu364c6ba2006-06-09 16:10:40 -07003173 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3174}
3175
3176/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00003177 * skb_has_shared_frag - can any frag be overwritten
3178 * @skb: buffer to test
3179 *
3180 * Return true if the skb has at least one frag that might be modified
3181 * by an external entity (as in vmsplice()/sendfile())
3182 */
3183static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3184{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00003185 return skb_is_nonlinear(skb) &&
3186 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00003187}
3188
3189/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07003190 * skb_linearize_cow - make sure skb is linear and writable
3191 * @skb: buffer to process
3192 *
3193 * If there is no free memory -ENOMEM is returned, otherwise zero
3194 * is returned and the old skb data released.
3195 */
3196static inline int skb_linearize_cow(struct sk_buff *skb)
3197{
3198 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3199 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003200}
3201
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003202static __always_inline void
3203__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3204 unsigned int off)
3205{
3206 if (skb->ip_summed == CHECKSUM_COMPLETE)
3207 skb->csum = csum_block_sub(skb->csum,
3208 csum_partial(start, len, 0), off);
3209 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3210 skb_checksum_start_offset(skb) < 0)
3211 skb->ip_summed = CHECKSUM_NONE;
3212}
3213
Linus Torvalds1da177e2005-04-16 15:20:36 -07003214/**
3215 * skb_postpull_rcsum - update checksum for received skb after pull
3216 * @skb: buffer to update
3217 * @start: start of data before pull
3218 * @len: length of data pulled
3219 *
3220 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07003221 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3222 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003223 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003224static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08003225 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003226{
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003227 __skb_postpull_rcsum(skb, start, len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003228}
3229
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003230static __always_inline void
3231__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3232 unsigned int off)
3233{
3234 if (skb->ip_summed == CHECKSUM_COMPLETE)
3235 skb->csum = csum_block_add(skb->csum,
3236 csum_partial(start, len, 0), off);
3237}
Herbert Xucbb042f2006-03-20 22:43:56 -08003238
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003239/**
3240 * skb_postpush_rcsum - update checksum for received skb after push
3241 * @skb: buffer to update
3242 * @start: start of data after push
3243 * @len: length of data pushed
3244 *
3245 * After doing a push on a received packet, you need to call this to
3246 * update the CHECKSUM_COMPLETE checksum.
3247 */
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01003248static inline void skb_postpush_rcsum(struct sk_buff *skb,
3249 const void *start, unsigned int len)
3250{
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003251 __skb_postpush_rcsum(skb, start, len, 0);
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01003252}
3253
Johannes Bergaf728682017-06-16 14:29:22 +02003254void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003255
David S. Miller7ce5a272013-12-02 17:26:05 -05003256/**
WANG Cong82a31b92016-06-30 10:15:22 -07003257 * skb_push_rcsum - push skb and update receive checksum
3258 * @skb: buffer to update
3259 * @len: length of data pulled
3260 *
3261 * This function performs an skb_push on the packet and updates
3262 * the CHECKSUM_COMPLETE checksum. It should be used on
3263 * receive path processing instead of skb_push unless you know
3264 * that the checksum difference is zero (e.g., a valid IP header)
3265 * or you are setting ip_summed to CHECKSUM_NONE.
3266 */
Johannes Bergd58ff352017-06-16 14:29:23 +02003267static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
WANG Cong82a31b92016-06-30 10:15:22 -07003268{
3269 skb_push(skb, len);
3270 skb_postpush_rcsum(skb, skb->data, len);
3271 return skb->data;
3272}
3273
Eric Dumazet88078d92018-04-18 11:43:15 -07003274int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
WANG Cong82a31b92016-06-30 10:15:22 -07003275/**
David S. Miller7ce5a272013-12-02 17:26:05 -05003276 * pskb_trim_rcsum - trim received skb and update checksum
3277 * @skb: buffer to trim
3278 * @len: new length
3279 *
3280 * This is exactly the same as pskb_trim except that it ensures the
3281 * checksum of received packets are still valid after the operation.
Ross Lagerwall6c57f042019-01-17 15:34:38 +00003282 * It can change skb pointers.
David S. Miller7ce5a272013-12-02 17:26:05 -05003283 */
3284
3285static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3286{
3287 if (likely(len >= skb->len))
3288 return 0;
Eric Dumazet88078d92018-04-18 11:43:15 -07003289 return pskb_trim_rcsum_slow(skb, len);
David S. Miller7ce5a272013-12-02 17:26:05 -05003290}
3291
Daniel Borkmann5293efe2016-08-18 01:00:39 +02003292static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3293{
3294 if (skb->ip_summed == CHECKSUM_COMPLETE)
3295 skb->ip_summed = CHECKSUM_NONE;
3296 __skb_trim(skb, len);
3297 return 0;
3298}
3299
3300static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3301{
3302 if (skb->ip_summed == CHECKSUM_COMPLETE)
3303 skb->ip_summed = CHECKSUM_NONE;
3304 return __skb_grow(skb, len);
3305}
3306
Eric Dumazet18a4c0e2017-10-05 22:21:21 -07003307#define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3308#define skb_rb_first(root) rb_to_skb(rb_first(root))
3309#define skb_rb_last(root) rb_to_skb(rb_last(root))
3310#define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3311#define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3312
Linus Torvalds1da177e2005-04-16 15:20:36 -07003313#define skb_queue_walk(queue, skb) \
3314 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003315 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07003316 skb = skb->next)
3317
James Chapman46f89142007-04-30 00:07:31 -07003318#define skb_queue_walk_safe(queue, skb, tmp) \
3319 for (skb = (queue)->next, tmp = skb->next; \
3320 skb != (struct sk_buff *)(queue); \
3321 skb = tmp, tmp = skb->next)
3322
David S. Miller1164f522008-09-23 00:49:44 -07003323#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003324 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07003325 skb = skb->next)
3326
Eric Dumazet18a4c0e2017-10-05 22:21:21 -07003327#define skb_rbtree_walk(skb, root) \
3328 for (skb = skb_rb_first(root); skb != NULL; \
3329 skb = skb_rb_next(skb))
3330
3331#define skb_rbtree_walk_from(skb) \
3332 for (; skb != NULL; \
3333 skb = skb_rb_next(skb))
3334
3335#define skb_rbtree_walk_from_safe(skb, tmp) \
3336 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3337 skb = tmp)
3338
David S. Miller1164f522008-09-23 00:49:44 -07003339#define skb_queue_walk_from_safe(queue, skb, tmp) \
3340 for (tmp = skb->next; \
3341 skb != (struct sk_buff *)(queue); \
3342 skb = tmp, tmp = skb->next)
3343
Stephen Hemminger300ce172005-10-30 13:47:34 -08003344#define skb_queue_reverse_walk(queue, skb) \
3345 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003346 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08003347 skb = skb->prev)
3348
David S. Miller686a2952011-01-20 22:47:32 -08003349#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3350 for (skb = (queue)->prev, tmp = skb->prev; \
3351 skb != (struct sk_buff *)(queue); \
3352 skb = tmp, tmp = skb->prev)
3353
3354#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3355 for (tmp = skb->prev; \
3356 skb != (struct sk_buff *)(queue); \
3357 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003358
David S. Miller21dc3302010-08-23 00:13:46 -07003359static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07003360{
3361 return skb_shinfo(skb)->frag_list != NULL;
3362}
3363
3364static inline void skb_frag_list_init(struct sk_buff *skb)
3365{
3366 skb_shinfo(skb)->frag_list = NULL;
3367}
3368
David S. Milleree039872009-06-09 00:17:13 -07003369#define skb_walk_frags(skb, iter) \
3370 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3371
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003372
3373int __skb_wait_for_more_packets(struct sock *sk, int *err, long *timeo_p,
3374 const struct sk_buff *skb);
Paolo Abeni65101ae2017-05-16 11:20:13 +02003375struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3376 struct sk_buff_head *queue,
3377 unsigned int flags,
3378 void (*destructor)(struct sock *sk,
3379 struct sk_buff *skb),
Paolo Abenifd69c392019-04-08 10:15:59 +02003380 int *off, int *err,
Paolo Abeni65101ae2017-05-16 11:20:13 +02003381 struct sk_buff **last);
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003382struct sk_buff *__skb_try_recv_datagram(struct sock *sk, unsigned flags,
Paolo Abeni7c13f972016-11-04 11:28:59 +01003383 void (*destructor)(struct sock *sk,
3384 struct sk_buff *skb),
Paolo Abenifd69c392019-04-08 10:15:59 +02003385 int *off, int *err,
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003386 struct sk_buff **last);
Joe Perches7965bd42013-09-26 14:48:15 -07003387struct sk_buff *__skb_recv_datagram(struct sock *sk, unsigned flags,
Paolo Abeni7c13f972016-11-04 11:28:59 +01003388 void (*destructor)(struct sock *sk,
3389 struct sk_buff *skb),
Paolo Abenifd69c392019-04-08 10:15:59 +02003390 int *off, int *err);
Joe Perches7965bd42013-09-26 14:48:15 -07003391struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3392 int *err);
Linus Torvaldsa11e1d42018-06-28 09:43:44 -07003393__poll_t datagram_poll(struct file *file, struct socket *sock,
3394 struct poll_table_struct *wait);
Al Viroc0371da2014-11-24 10:42:55 -05003395int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3396 struct iov_iter *to, int size);
David S. Miller51f3d022014-11-05 16:46:40 -05003397static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3398 struct msghdr *msg, int size)
3399{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003400 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
David S. Miller51f3d022014-11-05 16:46:40 -05003401}
Al Viroe5a4b0b2014-11-24 18:17:55 -05003402int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3403 struct msghdr *msg);
Sagi Grimberg65d69e22018-12-03 17:52:10 -08003404int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3405 struct iov_iter *to, int len,
3406 struct ahash_request *hash);
Al Viro3a654f92014-06-19 14:15:22 -04003407int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3408 struct iov_iter *from, int len);
Al Viro3a654f92014-06-19 14:15:22 -04003409int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
Joe Perches7965bd42013-09-26 14:48:15 -07003410void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
samanthakumar627d2d62016-04-05 12:41:16 -04003411void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3412static inline void skb_free_datagram_locked(struct sock *sk,
3413 struct sk_buff *skb)
3414{
3415 __skb_free_datagram_locked(sk, skb, 0);
3416}
Joe Perches7965bd42013-09-26 14:48:15 -07003417int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07003418int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3419int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3420__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3421 int len, __wsum csum);
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +02003422int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
Joe Perches7965bd42013-09-26 14:48:15 -07003423 struct pipe_inode_info *pipe, unsigned int len,
Al Viro25869262016-09-17 21:02:10 -04003424 unsigned int flags);
Tom Herbert20bf50d2017-07-28 16:22:42 -07003425int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3426 int len);
Joe Perches7965bd42013-09-26 14:48:15 -07003427void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01003428unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00003429int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3430 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07003431void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3432int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3433void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Daniel Axtens779b7932018-03-01 17:13:37 +11003434bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
Daniel Axtens2b16f042018-01-31 14:15:33 +11003435bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
Joe Perches7965bd42013-09-26 14:48:15 -07003436struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Vlad Yasevich0d5501c2014-08-08 14:42:13 -04003437struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
Jiri Pirkoe2195122014-11-19 14:05:01 +01003438int skb_ensure_writable(struct sk_buff *skb, int write_len);
Shmulik Ladkanibfca4c52016-09-19 19:11:09 +03003439int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
Jiri Pirko93515d52014-11-19 14:05:02 +01003440int skb_vlan_pop(struct sk_buff *skb);
3441int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
Sowmini Varadhan6fa01cc2016-04-22 18:36:35 -07003442struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3443 gfp_t gfp);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003444
Al Viro6ce8e9c2014-04-06 21:25:44 -04003445static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3446{
Al Viro3073f072017-02-17 23:13:25 -05003447 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
Al Viro6ce8e9c2014-04-06 21:25:44 -04003448}
3449
Al Viro7eab8d92014-04-06 21:51:23 -04003450static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3451{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003452 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro7eab8d92014-04-06 21:51:23 -04003453}
3454
Daniel Borkmann2817a332013-10-30 11:50:51 +01003455struct skb_checksum_ops {
3456 __wsum (*update)(const void *mem, int len, __wsum wsum);
3457 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3458};
3459
Davide Caratti96178132017-05-18 15:44:37 +02003460extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3461
Daniel Borkmann2817a332013-10-30 11:50:51 +01003462__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3463 __wsum csum, const struct skb_checksum_ops *ops);
3464__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3465 __wsum csum);
3466
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003467static inline void * __must_check
3468__skb_header_pointer(const struct sk_buff *skb, int offset,
3469 int len, void *data, int hlen, void *buffer)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003470{
Patrick McHardy55820ee2005-07-05 14:08:10 -07003471 if (hlen - offset >= len)
David S. Miller690e36e2014-08-23 12:13:41 -07003472 return data + offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003473
David S. Miller690e36e2014-08-23 12:13:41 -07003474 if (!skb ||
3475 skb_copy_bits(skb, offset, buffer, len) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003476 return NULL;
3477
3478 return buffer;
3479}
3480
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003481static inline void * __must_check
3482skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
David S. Miller690e36e2014-08-23 12:13:41 -07003483{
3484 return __skb_header_pointer(skb, offset, len, skb->data,
3485 skb_headlen(skb), buffer);
3486}
3487
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01003488/**
3489 * skb_needs_linearize - check if we need to linearize a given skb
3490 * depending on the given device features.
3491 * @skb: socket buffer to check
3492 * @features: net device features
3493 *
3494 * Returns true if either:
3495 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3496 * 2. skb is fragmented and the device does not support SG.
3497 */
3498static inline bool skb_needs_linearize(struct sk_buff *skb,
3499 netdev_features_t features)
3500{
3501 return skb_is_nonlinear(skb) &&
3502 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3503 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3504}
3505
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03003506static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3507 void *to,
3508 const unsigned int len)
3509{
3510 memcpy(to, skb->data, len);
3511}
3512
3513static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3514 const int offset, void *to,
3515 const unsigned int len)
3516{
3517 memcpy(to, skb->data + offset, len);
3518}
3519
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03003520static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3521 const void *from,
3522 const unsigned int len)
3523{
3524 memcpy(skb->data, from, len);
3525}
3526
3527static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3528 const int offset,
3529 const void *from,
3530 const unsigned int len)
3531{
3532 memcpy(skb->data + offset, from, len);
3533}
3534
Joe Perches7965bd42013-09-26 14:48:15 -07003535void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003536
Patrick Ohlyac45f602009-02-12 05:03:37 +00003537static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3538{
3539 return skb->tstamp;
3540}
3541
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003542/**
3543 * skb_get_timestamp - get timestamp from a skb
3544 * @skb: skb to get stamp from
Deepa Dinamani13c6ee22019-02-02 07:34:48 -08003545 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003546 *
3547 * Timestamps are stored in the skb as offsets to a base timestamp.
3548 * This function converts the offset back to a struct timeval and stores
3549 * it in stamp.
3550 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00003551static inline void skb_get_timestamp(const struct sk_buff *skb,
Deepa Dinamani13c6ee22019-02-02 07:34:48 -08003552 struct __kernel_old_timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003553{
Deepa Dinamani13c6ee22019-02-02 07:34:48 -08003554 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003555}
3556
Deepa Dinamani887feae2019-02-02 07:34:50 -08003557static inline void skb_get_new_timestamp(const struct sk_buff *skb,
3558 struct __kernel_sock_timeval *stamp)
3559{
3560 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3561
3562 stamp->tv_sec = ts.tv_sec;
3563 stamp->tv_usec = ts.tv_nsec / 1000;
3564}
3565
Patrick Ohlyac45f602009-02-12 05:03:37 +00003566static inline void skb_get_timestampns(const struct sk_buff *skb,
3567 struct timespec *stamp)
3568{
3569 *stamp = ktime_to_timespec(skb->tstamp);
3570}
3571
Deepa Dinamani887feae2019-02-02 07:34:50 -08003572static inline void skb_get_new_timestampns(const struct sk_buff *skb,
3573 struct __kernel_timespec *stamp)
3574{
3575 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3576
3577 stamp->tv_sec = ts.tv_sec;
3578 stamp->tv_nsec = ts.tv_nsec;
3579}
3580
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003581static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003582{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003583 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003584}
3585
Stephen Hemminger164891a2007-04-23 22:26:16 -07003586static inline ktime_t net_timedelta(ktime_t t)
3587{
3588 return ktime_sub(ktime_get_real(), t);
3589}
3590
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003591static inline ktime_t net_invalid_timestamp(void)
3592{
Thomas Gleixner8b0e1952016-12-25 12:30:41 +01003593 return 0;
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003594}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003595
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003596static inline u8 skb_metadata_len(const struct sk_buff *skb)
3597{
3598 return skb_shinfo(skb)->meta_len;
3599}
3600
3601static inline void *skb_metadata_end(const struct sk_buff *skb)
3602{
3603 return skb_mac_header(skb);
3604}
3605
3606static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3607 const struct sk_buff *skb_b,
3608 u8 meta_len)
3609{
3610 const void *a = skb_metadata_end(skb_a);
3611 const void *b = skb_metadata_end(skb_b);
3612 /* Using more efficient varaiant than plain call to memcmp(). */
3613#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3614 u64 diffs = 0;
3615
3616 switch (meta_len) {
3617#define __it(x, op) (x -= sizeof(u##op))
3618#define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3619 case 32: diffs |= __it_diff(a, b, 64);
Gustavo A. R. Silva82385b02018-10-17 15:01:37 +02003620 /* fall through */
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003621 case 24: diffs |= __it_diff(a, b, 64);
Gustavo A. R. Silva82385b02018-10-17 15:01:37 +02003622 /* fall through */
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003623 case 16: diffs |= __it_diff(a, b, 64);
Gustavo A. R. Silva82385b02018-10-17 15:01:37 +02003624 /* fall through */
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003625 case 8: diffs |= __it_diff(a, b, 64);
3626 break;
3627 case 28: diffs |= __it_diff(a, b, 64);
Gustavo A. R. Silva82385b02018-10-17 15:01:37 +02003628 /* fall through */
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003629 case 20: diffs |= __it_diff(a, b, 64);
Gustavo A. R. Silva82385b02018-10-17 15:01:37 +02003630 /* fall through */
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003631 case 12: diffs |= __it_diff(a, b, 64);
Gustavo A. R. Silva82385b02018-10-17 15:01:37 +02003632 /* fall through */
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003633 case 4: diffs |= __it_diff(a, b, 32);
3634 break;
3635 }
3636 return diffs;
3637#else
3638 return memcmp(a - meta_len, b - meta_len, meta_len);
3639#endif
3640}
3641
3642static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3643 const struct sk_buff *skb_b)
3644{
3645 u8 len_a = skb_metadata_len(skb_a);
3646 u8 len_b = skb_metadata_len(skb_b);
3647
3648 if (!(len_a | len_b))
3649 return false;
3650
3651 return len_a != len_b ?
3652 true : __skb_metadata_differs(skb_a, skb_b, len_a);
3653}
3654
3655static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3656{
3657 skb_shinfo(skb)->meta_len = meta_len;
3658}
3659
3660static inline void skb_metadata_clear(struct sk_buff *skb)
3661{
3662 skb_metadata_set(skb, 0);
3663}
3664
Alexander Duyck62bccb82014-09-04 13:31:35 -04003665struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3666
Richard Cochranc1f19b52010-07-17 08:49:36 +00003667#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3668
Joe Perches7965bd42013-09-26 14:48:15 -07003669void skb_clone_tx_timestamp(struct sk_buff *skb);
3670bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00003671
3672#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3673
3674static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3675{
3676}
3677
3678static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3679{
3680 return false;
3681}
3682
3683#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3684
3685/**
3686 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3687 *
Richard Cochranda92b192011-10-21 00:49:15 +00003688 * PHY drivers may accept clones of transmitted packets for
3689 * timestamping via their phy_driver.txtstamp method. These drivers
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003690 * must call this function to return the skb back to the stack with a
3691 * timestamp.
Richard Cochranda92b192011-10-21 00:49:15 +00003692 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00003693 * @skb: clone of the the original outgoing packet
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003694 * @hwtstamps: hardware time stamps
Richard Cochranc1f19b52010-07-17 08:49:36 +00003695 *
3696 */
3697void skb_complete_tx_timestamp(struct sk_buff *skb,
3698 struct skb_shared_hwtstamps *hwtstamps);
3699
Willem de Bruijne7fd2882014-08-04 22:11:48 -04003700void __skb_tstamp_tx(struct sk_buff *orig_skb,
3701 struct skb_shared_hwtstamps *hwtstamps,
3702 struct sock *sk, int tstype);
3703
Patrick Ohlyac45f602009-02-12 05:03:37 +00003704/**
3705 * skb_tstamp_tx - queue clone of skb with send time stamps
3706 * @orig_skb: the original outgoing packet
3707 * @hwtstamps: hardware time stamps, may be NULL if not available
3708 *
3709 * If the skb has a socket associated, then this function clones the
3710 * skb (thus sharing the actual data and optional structures), stores
3711 * the optional hardware time stamping information (if non NULL) or
3712 * generates a software time stamp (otherwise), then queues the clone
3713 * to the error queue of the socket. Errors are silently ignored.
3714 */
Joe Perches7965bd42013-09-26 14:48:15 -07003715void skb_tstamp_tx(struct sk_buff *orig_skb,
3716 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00003717
Richard Cochran4507a712010-07-17 08:48:28 +00003718/**
3719 * skb_tx_timestamp() - Driver hook for transmit timestamping
3720 *
3721 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00003722 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00003723 *
David S. Miller73409f32013-12-27 13:04:33 -05003724 * Specifically, one should make absolutely sure that this function is
3725 * called before TX completion of this packet can trigger. Otherwise
3726 * the packet could potentially already be freed.
3727 *
Richard Cochran4507a712010-07-17 08:48:28 +00003728 * @skb: A socket buffer.
3729 */
3730static inline void skb_tx_timestamp(struct sk_buff *skb)
3731{
Richard Cochranc1f19b52010-07-17 08:49:36 +00003732 skb_clone_tx_timestamp(skb);
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +02003733 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3734 skb_tstamp_tx(skb, NULL);
Richard Cochran4507a712010-07-17 08:48:28 +00003735}
3736
Johannes Berg6e3e9392011-11-09 10:15:42 +01003737/**
3738 * skb_complete_wifi_ack - deliver skb with wifi status
3739 *
3740 * @skb: the original outgoing packet
3741 * @acked: ack status
3742 *
3743 */
3744void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3745
Joe Perches7965bd42013-09-26 14:48:15 -07003746__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3747__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003748
Herbert Xu60476372007-04-09 11:59:39 -07003749static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3750{
Tom Herbert6edec0e2015-02-10 16:30:28 -08003751 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3752 skb->csum_valid ||
3753 (skb->ip_summed == CHECKSUM_PARTIAL &&
3754 skb_checksum_start_offset(skb) >= 0));
Herbert Xu60476372007-04-09 11:59:39 -07003755}
3756
Herbert Xufb286bb2005-11-10 13:01:24 -08003757/**
3758 * skb_checksum_complete - Calculate checksum of an entire packet
3759 * @skb: packet to process
3760 *
3761 * This function calculates the checksum over the entire packet plus
3762 * the value of skb->csum. The latter can be used to supply the
3763 * checksum of a pseudo header as used by TCP/UDP. It returns the
3764 * checksum.
3765 *
3766 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3767 * this function can be used to verify that checksum on received
3768 * packets. In that case the function should return zero if the
3769 * checksum is correct. In particular, this function will return zero
3770 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3771 * hardware has already verified the correctness of the checksum.
3772 */
Al Viro4381ca32007-07-15 21:00:11 +01003773static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08003774{
Herbert Xu60476372007-04-09 11:59:39 -07003775 return skb_csum_unnecessary(skb) ?
3776 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003777}
3778
Tom Herbert77cffe22014-08-27 21:26:46 -07003779static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3780{
3781 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3782 if (skb->csum_level == 0)
3783 skb->ip_summed = CHECKSUM_NONE;
3784 else
3785 skb->csum_level--;
3786 }
3787}
3788
3789static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3790{
3791 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3792 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3793 skb->csum_level++;
3794 } else if (skb->ip_summed == CHECKSUM_NONE) {
3795 skb->ip_summed = CHECKSUM_UNNECESSARY;
3796 skb->csum_level = 0;
3797 }
3798}
3799
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003800/* Check if we need to perform checksum complete validation.
3801 *
3802 * Returns true if checksum complete is needed, false otherwise
3803 * (either checksum is unnecessary or zero checksum is allowed).
3804 */
3805static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3806 bool zero_okay,
3807 __sum16 check)
3808{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003809 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3810 skb->csum_valid = 1;
Tom Herbert77cffe22014-08-27 21:26:46 -07003811 __skb_decr_checksum_unnecessary(skb);
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003812 return false;
3813 }
3814
3815 return true;
3816}
3817
David S. Millerda279882018-02-16 15:52:42 -05003818/* For small packets <= CHECKSUM_BREAK perform checksum complete directly
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003819 * in checksum_init.
3820 */
3821#define CHECKSUM_BREAK 76
3822
Tom Herbert4e18b9a2015-04-20 14:10:04 -07003823/* Unset checksum-complete
3824 *
3825 * Unset checksum complete can be done when packet is being modified
3826 * (uncompressed for instance) and checksum-complete value is
3827 * invalidated.
3828 */
3829static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3830{
3831 if (skb->ip_summed == CHECKSUM_COMPLETE)
3832 skb->ip_summed = CHECKSUM_NONE;
3833}
3834
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003835/* Validate (init) checksum based on checksum complete.
3836 *
3837 * Return values:
3838 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3839 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3840 * checksum is stored in skb->csum for use in __skb_checksum_complete
3841 * non-zero: value of invalid checksum
3842 *
3843 */
3844static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3845 bool complete,
3846 __wsum psum)
3847{
3848 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3849 if (!csum_fold(csum_add(psum, skb->csum))) {
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003850 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003851 return 0;
3852 }
3853 }
3854
3855 skb->csum = psum;
3856
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003857 if (complete || skb->len <= CHECKSUM_BREAK) {
3858 __sum16 csum;
3859
3860 csum = __skb_checksum_complete(skb);
3861 skb->csum_valid = !csum;
3862 return csum;
3863 }
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003864
3865 return 0;
3866}
3867
3868static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3869{
3870 return 0;
3871}
3872
3873/* Perform checksum validate (init). Note that this is a macro since we only
3874 * want to calculate the pseudo header which is an input function if necessary.
3875 * First we try to validate without any computation (checksum unnecessary) and
3876 * then calculate based on checksum complete calling the function to compute
3877 * pseudo header.
3878 *
3879 * Return values:
3880 * 0: checksum is validated or try to in skb_checksum_complete
3881 * non-zero: value of invalid checksum
3882 */
3883#define __skb_checksum_validate(skb, proto, complete, \
3884 zero_okay, check, compute_pseudo) \
3885({ \
3886 __sum16 __ret = 0; \
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003887 skb->csum_valid = 0; \
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003888 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
3889 __ret = __skb_checksum_validate_complete(skb, \
3890 complete, compute_pseudo(skb, proto)); \
3891 __ret; \
3892})
3893
3894#define skb_checksum_init(skb, proto, compute_pseudo) \
3895 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
3896
3897#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
3898 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
3899
3900#define skb_checksum_validate(skb, proto, compute_pseudo) \
3901 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
3902
3903#define skb_checksum_validate_zero_check(skb, proto, check, \
3904 compute_pseudo) \
Sabrina Dubroca096a4cf2015-02-06 18:54:19 +01003905 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003906
3907#define skb_checksum_simple_validate(skb) \
3908 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
3909
Tom Herbertd96535a2014-08-31 15:12:42 -07003910static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
3911{
Davide Caratti219f1d792017-05-18 15:44:39 +02003912 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
Tom Herbertd96535a2014-08-31 15:12:42 -07003913}
3914
3915static inline void __skb_checksum_convert(struct sk_buff *skb,
3916 __sum16 check, __wsum pseudo)
3917{
3918 skb->csum = ~pseudo;
3919 skb->ip_summed = CHECKSUM_COMPLETE;
3920}
3921
3922#define skb_checksum_try_convert(skb, proto, check, compute_pseudo) \
3923do { \
3924 if (__skb_checksum_convert_check(skb)) \
3925 __skb_checksum_convert(skb, check, \
3926 compute_pseudo(skb, proto)); \
3927} while (0)
3928
Tom Herbert15e23962015-02-10 16:30:31 -08003929static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
3930 u16 start, u16 offset)
3931{
3932 skb->ip_summed = CHECKSUM_PARTIAL;
3933 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
3934 skb->csum_offset = offset - start;
3935}
3936
Tom Herbertdcdc8992015-02-02 16:07:34 -08003937/* Update skbuf and packet to reflect the remote checksum offload operation.
3938 * When called, ptr indicates the starting point for skb->csum when
3939 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
3940 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
3941 */
3942static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
Tom Herbert15e23962015-02-10 16:30:31 -08003943 int start, int offset, bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08003944{
3945 __wsum delta;
3946
Tom Herbert15e23962015-02-10 16:30:31 -08003947 if (!nopartial) {
3948 skb_remcsum_adjust_partial(skb, ptr, start, offset);
3949 return;
3950 }
3951
Tom Herbertdcdc8992015-02-02 16:07:34 -08003952 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
3953 __skb_checksum_complete(skb);
3954 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
3955 }
3956
3957 delta = remcsum_adjust(ptr, skb->csum, start, offset);
3958
3959 /* Adjust skb->csum since we changed the packet */
3960 skb->csum = csum_add(skb->csum, delta);
3961}
3962
Florian Westphalcb9c6832017-01-23 18:21:56 +01003963static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
3964{
3965#if IS_ENABLED(CONFIG_NF_CONNTRACK)
Florian Westphala9e419d2017-01-23 18:21:59 +01003966 return (void *)(skb->_nfct & SKB_NFCT_PTRMASK);
Florian Westphalcb9c6832017-01-23 18:21:56 +01003967#else
3968 return NULL;
3969#endif
3970}
3971
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07003972#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Joe Perches7965bd42013-09-26 14:48:15 -07003973void nf_conntrack_destroy(struct nf_conntrack *nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003974static inline void nf_conntrack_put(struct nf_conntrack *nfct)
3975{
3976 if (nfct && atomic_dec_and_test(&nfct->use))
Yasuyuki Kozakaide6e05c2007-03-23 11:17:27 -07003977 nf_conntrack_destroy(nfct);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003978}
3979static inline void nf_conntrack_get(struct nf_conntrack *nfct)
3980{
3981 if (nfct)
3982 atomic_inc(&nfct->use);
3983}
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01003984#endif
Florian Westphaldf5042f2018-12-18 17:15:16 +01003985
3986#ifdef CONFIG_SKB_EXTENSIONS
3987enum skb_ext_id {
3988#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
3989 SKB_EXT_BRIDGE_NF,
3990#endif
Florian Westphal41650792018-12-18 17:15:27 +01003991#ifdef CONFIG_XFRM
3992 SKB_EXT_SEC_PATH,
3993#endif
Florian Westphaldf5042f2018-12-18 17:15:16 +01003994 SKB_EXT_NUM, /* must be last */
3995};
3996
3997/**
3998 * struct skb_ext - sk_buff extensions
3999 * @refcnt: 1 on allocation, deallocated on 0
4000 * @offset: offset to add to @data to obtain extension address
4001 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4002 * @data: start of extension data, variable sized
4003 *
4004 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4005 * to use 'u8' types while allowing up to 2kb worth of extension data.
4006 */
4007struct skb_ext {
4008 refcount_t refcnt;
4009 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4010 u8 chunks; /* same */
4011 char data[0] __aligned(8);
4012};
4013
4014void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4015void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4016void __skb_ext_put(struct skb_ext *ext);
4017
4018static inline void skb_ext_put(struct sk_buff *skb)
4019{
4020 if (skb->active_extensions)
4021 __skb_ext_put(skb->extensions);
4022}
4023
Florian Westphaldf5042f2018-12-18 17:15:16 +01004024static inline void __skb_ext_copy(struct sk_buff *dst,
4025 const struct sk_buff *src)
4026{
4027 dst->active_extensions = src->active_extensions;
4028
4029 if (src->active_extensions) {
4030 struct skb_ext *ext = src->extensions;
4031
4032 refcount_inc(&ext->refcnt);
4033 dst->extensions = ext;
4034 }
4035}
4036
4037static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4038{
4039 skb_ext_put(dst);
4040 __skb_ext_copy(dst, src);
4041}
4042
4043static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4044{
4045 return !!ext->offset[i];
4046}
4047
4048static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4049{
4050 return skb->active_extensions & (1 << id);
4051}
4052
4053static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4054{
4055 if (skb_ext_exist(skb, id))
4056 __skb_ext_del(skb, id);
4057}
4058
4059static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4060{
4061 if (skb_ext_exist(skb, id)) {
4062 struct skb_ext *ext = skb->extensions;
4063
4064 return (void *)ext + (ext->offset[id] << 3);
4065 }
4066
4067 return NULL;
4068}
4069#else
4070static inline void skb_ext_put(struct sk_buff *skb) {}
Florian Westphaldf5042f2018-12-18 17:15:16 +01004071static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
4072static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {}
4073static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
4074#endif /* CONFIG_SKB_EXTENSIONS */
4075
Patrick McHardya193a4a2006-03-20 19:23:05 -08004076static inline void nf_reset(struct sk_buff *skb)
4077{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07004078#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01004079 nf_conntrack_put(skb_nfct(skb));
4080 skb->_nfct = 0;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01004081#endif
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +02004082#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Florian Westphalde8bda12018-12-18 17:15:17 +01004083 skb_ext_del(skb, SKB_EXT_BRIDGE_NF);
Patrick McHardya193a4a2006-03-20 19:23:05 -08004084#endif
4085}
4086
Patrick McHardy124dff02013-04-05 20:42:05 +02004087static inline void nf_reset_trace(struct sk_buff *skb)
4088{
Florian Westphal478b3602014-02-15 23:48:45 +01004089#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00004090 skb->nf_trace = 0;
4091#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07004092}
4093
Ye Yin2b5ec1a2017-10-26 16:57:05 +08004094static inline void ipvs_reset(struct sk_buff *skb)
4095{
4096#if IS_ENABLED(CONFIG_IP_VS)
4097 skb->ipvs_property = 0;
4098#endif
4099}
4100
Florian Westphalde8bda12018-12-18 17:15:17 +01004101/* Note: This doesn't put any conntrack info in dst. */
Eric Dumazetb1937222014-09-28 22:18:47 -07004102static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4103 bool copy)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07004104{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07004105#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01004106 dst->_nfct = src->_nfct;
4107 nf_conntrack_get(skb_nfct(src));
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01004108#endif
Florian Westphal478b3602014-02-15 23:48:45 +01004109#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Eric Dumazetb1937222014-09-28 22:18:47 -07004110 if (copy)
4111 dst->nf_trace = src->nf_trace;
Florian Westphal478b3602014-02-15 23:48:45 +01004112#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07004113}
4114
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07004115static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4116{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07004117#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01004118 nf_conntrack_put(skb_nfct(dst));
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01004119#endif
Eric Dumazetb1937222014-09-28 22:18:47 -07004120 __nf_copy(dst, src, true);
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07004121}
4122
James Morris984bc162006-06-09 00:29:17 -07004123#ifdef CONFIG_NETWORK_SECMARK
4124static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4125{
4126 to->secmark = from->secmark;
4127}
4128
4129static inline void skb_init_secmark(struct sk_buff *skb)
4130{
4131 skb->secmark = 0;
4132}
4133#else
4134static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4135{ }
4136
4137static inline void skb_init_secmark(struct sk_buff *skb)
4138{ }
4139#endif
4140
Florian Westphal7af8f4c2018-12-18 17:15:19 +01004141static inline int secpath_exists(const struct sk_buff *skb)
4142{
4143#ifdef CONFIG_XFRM
Florian Westphal41650792018-12-18 17:15:27 +01004144 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
Florian Westphal7af8f4c2018-12-18 17:15:19 +01004145#else
4146 return 0;
4147#endif
4148}
4149
Eric W. Biederman574f7192014-04-01 12:20:24 -07004150static inline bool skb_irq_freeable(const struct sk_buff *skb)
4151{
4152 return !skb->destructor &&
Florian Westphal7af8f4c2018-12-18 17:15:19 +01004153 !secpath_exists(skb) &&
Florian Westphalcb9c6832017-01-23 18:21:56 +01004154 !skb_nfct(skb) &&
Eric W. Biederman574f7192014-04-01 12:20:24 -07004155 !skb->_skb_refdst &&
4156 !skb_has_frag_list(skb);
4157}
4158
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07004159static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4160{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07004161 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07004162}
4163
Stephen Hemminger92477442009-03-21 13:39:26 -07004164static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07004165{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07004166 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07004167}
4168
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07004169static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4170{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07004171 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07004172}
4173
David S. Millerd5a9e242009-01-27 16:22:11 -08004174static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4175{
4176 skb->queue_mapping = rx_queue + 1;
4177}
4178
Stephen Hemminger92477442009-03-21 13:39:26 -07004179static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08004180{
4181 return skb->queue_mapping - 1;
4182}
4183
Stephen Hemminger92477442009-03-21 13:39:26 -07004184static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08004185{
Eric Dumazeta02cec22010-09-22 20:43:57 +00004186 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08004187}
4188
Julian Anastasov4ff06202017-02-06 23:14:12 +02004189static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4190{
4191 skb->dst_pending_confirm = val;
4192}
4193
4194static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4195{
4196 return skb->dst_pending_confirm != 0;
4197}
4198
Florian Westphal2294be0f2018-12-18 17:15:20 +01004199static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04004200{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07004201#ifdef CONFIG_XFRM
Florian Westphal41650792018-12-18 17:15:27 +01004202 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07004203#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07004204 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07004205#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04004206}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07004207
Pravin B Shelar68c33162013-02-14 14:02:41 +00004208/* Keeps track of mac header offset relative to skb->head.
4209 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4210 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07004211 * tunnel skb it points to outer mac header.
4212 * Keeps track of level of encapsulation of network headers.
4213 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00004214struct skb_gso_cb {
Alexander Duyck802ab552016-04-10 21:45:03 -04004215 union {
4216 int mac_offset;
4217 int data_offset;
4218 };
Eric Dumazet3347c962013-10-19 11:42:56 -07004219 int encap_level;
Alexander Duyck76443452016-02-05 15:27:37 -08004220 __wsum csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07004221 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00004222};
Konstantin Khlebnikov9207f9d2016-01-08 15:21:46 +03004223#define SKB_SGO_CB_OFFSET 32
4224#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
Pravin B Shelar68c33162013-02-14 14:02:41 +00004225
4226static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4227{
4228 return (skb_mac_header(inner_skb) - inner_skb->head) -
4229 SKB_GSO_CB(inner_skb)->mac_offset;
4230}
4231
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00004232static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4233{
4234 int new_headroom, headroom;
4235 int ret;
4236
4237 headroom = skb_headroom(skb);
4238 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4239 if (ret)
4240 return ret;
4241
4242 new_headroom = skb_headroom(skb);
4243 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4244 return 0;
4245}
4246
Alexander Duyck08b64fc2016-02-05 15:27:49 -08004247static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4248{
4249 /* Do not update partial checksums if remote checksum is enabled. */
4250 if (skb->remcsum_offload)
4251 return;
4252
4253 SKB_GSO_CB(skb)->csum = res;
4254 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4255}
4256
Tom Herbert7e2b10c2014-06-04 17:20:02 -07004257/* Compute the checksum for a gso segment. First compute the checksum value
4258 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4259 * then add in skb->csum (checksum from csum_start to end of packet).
4260 * skb->csum and csum_start are then updated to reflect the checksum of the
4261 * resultant packet starting from the transport header-- the resultant checksum
4262 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4263 * header.
4264 */
4265static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4266{
Alexander Duyck76443452016-02-05 15:27:37 -08004267 unsigned char *csum_start = skb_transport_header(skb);
4268 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4269 __wsum partial = SKB_GSO_CB(skb)->csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07004270
Alexander Duyck76443452016-02-05 15:27:37 -08004271 SKB_GSO_CB(skb)->csum = res;
4272 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07004273
Alexander Duyck76443452016-02-05 15:27:37 -08004274 return csum_fold(csum_partial(csum_start, plen, partial));
Tom Herbert7e2b10c2014-06-04 17:20:02 -07004275}
4276
David S. Millerbdcc0922012-03-07 20:53:36 -05004277static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07004278{
4279 return skb_shinfo(skb)->gso_size;
4280}
4281
Eric Dumazet36a8f392013-09-29 01:21:32 -07004282/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05004283static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02004284{
4285 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4286}
4287
Daniel Axtensd02f51c2018-03-03 03:03:46 +01004288/* Note: Should be called only if skb_is_gso(skb) is true */
4289static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4290{
4291 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4292}
4293
Willem de Bruijn4c3024d2019-03-06 14:35:15 -05004294/* Note: Should be called only if skb_is_gso(skb) is true */
Willem de Bruijnb90efd22019-02-07 14:54:16 -05004295static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4296{
Willem de Bruijn4c3024d2019-03-06 14:35:15 -05004297 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
Willem de Bruijnb90efd22019-02-07 14:54:16 -05004298}
4299
Daniel Borkmann5293efe2016-08-18 01:00:39 +02004300static inline void skb_gso_reset(struct sk_buff *skb)
4301{
4302 skb_shinfo(skb)->gso_size = 0;
4303 skb_shinfo(skb)->gso_segs = 0;
4304 skb_shinfo(skb)->gso_type = 0;
4305}
4306
Daniel Axtensd02f51c2018-03-03 03:03:46 +01004307static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4308 u16 increment)
4309{
4310 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4311 return;
4312 shinfo->gso_size += increment;
4313}
4314
4315static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4316 u16 decrement)
4317{
4318 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4319 return;
4320 shinfo->gso_size -= decrement;
4321}
4322
Joe Perches7965bd42013-09-26 14:48:15 -07004323void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07004324
4325static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4326{
4327 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4328 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04004329 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4330
Alexander Duyckb78462e2010-06-02 12:24:37 +00004331 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4332 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07004333 __skb_warn_lro_forwarding(skb);
4334 return true;
4335 }
4336 return false;
4337}
4338
Herbert Xu35fc92a2007-03-26 23:22:20 -07004339static inline void skb_forward_csum(struct sk_buff *skb)
4340{
4341 /* Unfortunately we don't support this one. Any brave souls? */
4342 if (skb->ip_summed == CHECKSUM_COMPLETE)
4343 skb->ip_summed = CHECKSUM_NONE;
4344}
4345
Eric Dumazetbc8acf22010-09-02 13:07:41 -07004346/**
4347 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4348 * @skb: skb to check
4349 *
4350 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4351 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4352 * use this helper, to document places where we make this assertion.
4353 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04004354static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07004355{
4356#ifdef DEBUG
4357 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4358#endif
4359}
4360
Rusty Russellf35d9d82008-02-04 23:49:54 -05004361bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00004362
Paul Durranted1f50c2014-01-09 10:02:46 +00004363int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
Linus Lüssing9afd85c2015-05-02 14:01:07 +02004364struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4365 unsigned int transport_len,
4366 __sum16(*skb_chkf)(struct sk_buff *skb));
Paul Durranted1f50c2014-01-09 10:02:46 +00004367
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00004368/**
4369 * skb_head_is_locked - Determine if the skb->head is locked down
4370 * @skb: skb to check
4371 *
4372 * The head on skbs build around a head frag can be removed if they are
4373 * not cloned. This function returns true if the skb head is locked down
4374 * due to either being allocated via kmalloc, or by being a clone with
4375 * multiple references to the head.
4376 */
4377static inline bool skb_head_is_locked(const struct sk_buff *skb)
4378{
4379 return !skb->head_frag || skb_cloned(skb);
4380}
Florian Westphalfe6cc552014-02-13 23:09:12 +01004381
Edward Cree179bc672016-02-11 20:48:04 +00004382/* Local Checksum Offload.
4383 * Compute outer checksum based on the assumption that the
4384 * inner checksum will be offloaded later.
Otto Sabartd0dcde62019-01-06 00:29:15 +01004385 * See Documentation/networking/checksum-offloads.rst for
Edward Creee8ae7b02016-02-11 21:03:37 +00004386 * explanation of how this works.
Edward Cree179bc672016-02-11 20:48:04 +00004387 * Fill in outer checksum adjustment (e.g. with sum of outer
4388 * pseudo-header) before calling.
4389 * Also ensure that inner checksum is in linear data area.
4390 */
4391static inline __wsum lco_csum(struct sk_buff *skb)
4392{
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004393 unsigned char *csum_start = skb_checksum_start(skb);
4394 unsigned char *l4_hdr = skb_transport_header(skb);
4395 __wsum partial;
Edward Cree179bc672016-02-11 20:48:04 +00004396
4397 /* Start with complement of inner checksum adjustment */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004398 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4399 skb->csum_offset));
4400
Edward Cree179bc672016-02-11 20:48:04 +00004401 /* Add in checksum of our headers (incl. outer checksum
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004402 * adjustment filled in by caller) and return result.
Edward Cree179bc672016-02-11 20:48:04 +00004403 */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004404 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
Edward Cree179bc672016-02-11 20:48:04 +00004405}
4406
Linus Torvalds1da177e2005-04-16 15:20:36 -07004407#endif /* __KERNEL__ */
4408#endif /* _LINUX_SKBUFF_H */