blob: aa15c390580a35d2fec8c873c03b42dd4c702436 [file] [log] [blame]
Thomas Gleixner2874c5f2019-05-27 08:55:01 +02001/* SPDX-License-Identifier: GPL-2.0-or-later */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002/*
3 * Definitions for the 'struct sk_buff' memory handlers.
4 *
5 * Authors:
6 * Alan Cox, <gw4pts@gw4pts.ampr.org>
7 * Florian La Roche, <rzsfl@rz.uni-sb.de>
Linus Torvalds1da177e2005-04-16 15:20:36 -07008 */
9
10#ifndef _LINUX_SKBUFF_H
11#define _LINUX_SKBUFF_H
12
Linus Torvalds1da177e2005-04-16 15:20:36 -070013#include <linux/kernel.h>
14#include <linux/compiler.h>
15#include <linux/time.h>
Paul Gortmaker187f1882011-11-23 20:12:59 -050016#include <linux/bug.h>
Matthew Wilcox (Oracle)8842d282019-07-22 20:08:31 -070017#include <linux/bvec.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070018#include <linux/cache.h>
Eric Dumazet56b17422014-11-03 08:19:53 -080019#include <linux/rbtree.h>
David S. Miller51f3d022014-11-05 16:46:40 -050020#include <linux/socket.h>
Eric Dumazetc1d1b432017-08-31 16:48:22 -070021#include <linux/refcount.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070022
Arun Sharma600634972011-07-26 16:09:06 -070023#include <linux/atomic.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070024#include <asm/types.h>
25#include <linux/spinlock.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070026#include <linux/net.h>
Thomas Graf3fc7e8a2005-06-23 21:00:17 -070027#include <linux/textsearch.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070028#include <net/checksum.h>
Al Viroa80958f2006-12-04 20:41:19 +000029#include <linux/rcupdate.h>
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -070030#include <linux/hrtimer.h>
Ian Campbell131ea662011-08-19 06:25:00 +000031#include <linux/dma-mapping.h>
Michał Mirosławc8f44af2011-11-15 15:29:55 +000032#include <linux/netdev_features.h>
Eric Dumazet363ec39232014-02-26 14:02:11 -080033#include <linux/sched.h>
Ingo Molnare6017572017-02-01 16:36:40 +010034#include <linux/sched/clock.h>
Jiri Pirko1bd758e2015-05-12 14:56:07 +020035#include <net/flow_dissector.h>
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +020036#include <linux/splice.h>
Bernhard Thaler72b31f72015-05-30 15:27:40 +020037#include <linux/in6.h>
Jamal Hadi Salim8b10cab2016-07-02 06:43:14 -040038#include <linux/if_packet.h>
Tom Herbertf70ea012015-07-31 16:52:10 -070039#include <net/flow.h>
Jeremy Sowden261db6c2019-09-13 09:13:14 +010040#if IS_ENABLED(CONFIG_NF_CONNTRACK)
41#include <linux/netfilter/nf_conntrack_common.h>
42#endif
Greg Kroah-Hartman73ff58b2020-05-02 09:32:22 +020043#include <linux/android_kabi.h>
Vignesh Saravanaperumal726468d2021-07-08 07:01:33 -070044#include <linux/android_vendor.h>
Linus Torvalds1da177e2005-04-16 15:20:36 -070045
Tom Herbert7a6ae712015-12-14 11:19:47 -080046/* The interface for checksum offload between the stack and networking drivers
47 * is as follows...
48 *
49 * A. IP checksum related features
50 *
51 * Drivers advertise checksum offload capabilities in the features of a device.
Dexuan Cuidb1f00f2020-04-05 18:59:24 -070052 * From the stack's point of view these are capabilities offered by the driver.
53 * A driver typically only advertises features that it is capable of offloading
Tom Herbert7a6ae712015-12-14 11:19:47 -080054 * to its device.
55 *
56 * The checksum related features are:
57 *
58 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
59 * IP (one's complement) checksum for any combination
60 * of protocols or protocol layering. The checksum is
61 * computed and set in a packet per the CHECKSUM_PARTIAL
62 * interface (see below).
63 *
64 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
65 * TCP or UDP packets over IPv4. These are specifically
66 * unencapsulated packets of the form IPv4|TCP or
67 * IPv4|UDP where the Protocol field in the IPv4 header
Dexuan Cuidb1f00f2020-04-05 18:59:24 -070068 * is TCP or UDP. The IPv4 header may contain IP options.
Tom Herbert7a6ae712015-12-14 11:19:47 -080069 * This feature cannot be set in features for a device
70 * with NETIF_F_HW_CSUM also set. This feature is being
71 * DEPRECATED (see below).
72 *
73 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
74 * TCP or UDP packets over IPv6. These are specifically
75 * unencapsulated packets of the form IPv6|TCP or
Miaohe Lin645f0892020-08-27 07:27:49 -040076 * IPv6|UDP where the Next Header field in the IPv6
Tom Herbert7a6ae712015-12-14 11:19:47 -080077 * header is either TCP or UDP. IPv6 extension headers
78 * are not supported with this feature. This feature
79 * cannot be set in features for a device with
80 * NETIF_F_HW_CSUM also set. This feature is being
81 * DEPRECATED (see below).
82 *
83 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
Dexuan Cuidb1f00f2020-04-05 18:59:24 -070084 * This flag is only used to disable the RX checksum
Tom Herbert7a6ae712015-12-14 11:19:47 -080085 * feature for a device. The stack will accept receive
86 * checksum indication in packets received on a device
87 * regardless of whether NETIF_F_RXCSUM is set.
88 *
89 * B. Checksumming of received packets by device. Indication of checksum
Dexuan Cuidb1f00f2020-04-05 18:59:24 -070090 * verification is set in skb->ip_summed. Possible values are:
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010091 *
92 * CHECKSUM_NONE:
93 *
Tom Herbert7a6ae712015-12-14 11:19:47 -080094 * Device did not checksum this packet e.g. due to lack of capabilities.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010095 * The packet contains full (though not verified) checksum in packet but
96 * not in skb->csum. Thus, skb->csum is undefined in this case.
97 *
98 * CHECKSUM_UNNECESSARY:
99 *
100 * The hardware you're dealing with doesn't calculate the full checksum
101 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
Tom Herbert77cffe22014-08-27 21:26:46 -0700102 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
103 * if their checksums are okay. skb->csum is still undefined in this case
Tom Herbert7a6ae712015-12-14 11:19:47 -0800104 * though. A driver or device must never modify the checksum field in the
105 * packet even if checksum is verified.
Tom Herbert77cffe22014-08-27 21:26:46 -0700106 *
107 * CHECKSUM_UNNECESSARY is applicable to following protocols:
108 * TCP: IPv6 and IPv4.
109 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
110 * zero UDP checksum for either IPv4 or IPv6, the networking stack
111 * may perform further validation in this case.
112 * GRE: only if the checksum is present in the header.
113 * SCTP: indicates the CRC in SCTP header has been validated.
Davide Carattib4759dc2017-05-18 15:44:43 +0200114 * FCOE: indicates the CRC in FC frame has been validated.
Tom Herbert77cffe22014-08-27 21:26:46 -0700115 *
116 * skb->csum_level indicates the number of consecutive checksums found in
117 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
118 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
119 * and a device is able to verify the checksums for UDP (possibly zero),
Dexuan Cuidb1f00f2020-04-05 18:59:24 -0700120 * GRE (checksum flag is set) and TCP, skb->csum_level would be set to
Tom Herbert77cffe22014-08-27 21:26:46 -0700121 * two. If the device were only able to verify the UDP checksum and not
Dexuan Cuidb1f00f2020-04-05 18:59:24 -0700122 * GRE, either because it doesn't support GRE checksum or because GRE
Tom Herbert77cffe22014-08-27 21:26:46 -0700123 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
124 * not considered in this case).
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100125 *
126 * CHECKSUM_COMPLETE:
127 *
128 * This is the most generic way. The device supplied checksum of the _whole_
Dexuan Cuidb1f00f2020-04-05 18:59:24 -0700129 * packet as seen by netif_rx() and fills in skb->csum. This means the
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100130 * hardware doesn't need to parse L3/L4 headers to implement this.
131 *
Davide Carattib4759dc2017-05-18 15:44:43 +0200132 * Notes:
133 * - Even if device supports only some protocols, but is able to produce
134 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
135 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100136 *
137 * CHECKSUM_PARTIAL:
138 *
Tom Herbert6edec0e2015-02-10 16:30:28 -0800139 * A checksum is set up to be offloaded to a device as described in the
140 * output description for CHECKSUM_PARTIAL. This may occur on a packet
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100141 * received directly from another Linux OS, e.g., a virtualized Linux kernel
Tom Herbert6edec0e2015-02-10 16:30:28 -0800142 * on the same host, or it may be set in the input path in GRO or remote
143 * checksum offload. For the purposes of checksum verification, the checksum
144 * referred to by skb->csum_start + skb->csum_offset and any preceding
145 * checksums in the packet are considered verified. Any checksums in the
146 * packet that are after the checksum being offloaded are not considered to
147 * be verified.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100148 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800149 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
150 * in the skb->ip_summed for a packet. Values are:
151 *
152 * CHECKSUM_PARTIAL:
153 *
154 * The driver is required to checksum the packet as seen by hard_start_xmit()
155 * from skb->csum_start up to the end, and to record/write the checksum at
156 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
157 * csum_start and csum_offset values are valid values given the length and
Dexuan Cuidb1f00f2020-04-05 18:59:24 -0700158 * offset of the packet, but it should not attempt to validate that the
159 * checksum refers to a legitimate transport layer checksum -- it is the
Tom Herbert7a6ae712015-12-14 11:19:47 -0800160 * purview of the stack to validate that csum_start and csum_offset are set
161 * correctly.
162 *
163 * When the stack requests checksum offload for a packet, the driver MUST
164 * ensure that the checksum is set correctly. A driver can either offload the
165 * checksum calculation to the device, or call skb_checksum_help (in the case
166 * that the device does not support offload for a particular checksum).
167 *
168 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
169 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
Davide Caratti43c26a12017-05-18 15:44:41 +0200170 * checksum offload capability.
171 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based
172 * on network device checksumming capabilities: if a packet does not match
173 * them, skb_checksum_help or skb_crc32c_help (depending on the value of
174 * csum_not_inet, see item D.) is called to resolve the checksum.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100175 *
176 * CHECKSUM_NONE:
177 *
178 * The skb was already checksummed by the protocol, or a checksum is not
179 * required.
180 *
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100181 * CHECKSUM_UNNECESSARY:
182 *
Dexuan Cuidb1f00f2020-04-05 18:59:24 -0700183 * This has the same meaning as CHECKSUM_NONE for checksum offload on
Tom Herbert7a6ae712015-12-14 11:19:47 -0800184 * output.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100185 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800186 * CHECKSUM_COMPLETE:
187 * Not used in checksum output. If a driver observes a packet with this value
Dexuan Cuidb1f00f2020-04-05 18:59:24 -0700188 * set in skbuff, it should treat the packet as if CHECKSUM_NONE were set.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100189 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800190 * D. Non-IP checksum (CRC) offloads
191 *
192 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
193 * offloading the SCTP CRC in a packet. To perform this offload the stack
Dexuan Cuidb1f00f2020-04-05 18:59:24 -0700194 * will set csum_start and csum_offset accordingly, set ip_summed to
Davide Carattidba00302017-05-18 15:44:40 +0200195 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
196 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
197 * A driver that supports both IP checksum offload and SCTP CRC32c offload
198 * must verify which offload is configured for a packet by testing the
199 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
200 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
Tom Herbert7a6ae712015-12-14 11:19:47 -0800201 *
202 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
203 * offloading the FCOE CRC in a packet. To perform this offload the stack
204 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
Dexuan Cuidb1f00f2020-04-05 18:59:24 -0700205 * accordingly. Note that there is no indication in the skbuff that the
206 * CHECKSUM_PARTIAL refers to an FCOE checksum, so a driver that supports
Tom Herbert7a6ae712015-12-14 11:19:47 -0800207 * both IP checksum offload and FCOE CRC offload must verify which offload
Dexuan Cuidb1f00f2020-04-05 18:59:24 -0700208 * is configured for a packet, presumably by inspecting packet headers.
Tom Herbert7a6ae712015-12-14 11:19:47 -0800209 *
210 * E. Checksumming on output with GSO.
211 *
212 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
213 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
214 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
215 * part of the GSO operation is implied. If a checksum is being offloaded
Dexuan Cuidb1f00f2020-04-05 18:59:24 -0700216 * with GSO then ip_summed is CHECKSUM_PARTIAL, and both csum_start and
217 * csum_offset are set to refer to the outermost checksum being offloaded
218 * (two offloaded checksums are possible with UDP encapsulation).
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100219 */
220
Herbert Xu60476372007-04-09 11:59:39 -0700221/* Don't change this without changing skb_csum_unnecessary! */
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100222#define CHECKSUM_NONE 0
223#define CHECKSUM_UNNECESSARY 1
224#define CHECKSUM_COMPLETE 2
225#define CHECKSUM_PARTIAL 3
Linus Torvalds1da177e2005-04-16 15:20:36 -0700226
Tom Herbert77cffe22014-08-27 21:26:46 -0700227/* Maximum value in skb->csum_level */
228#define SKB_MAX_CSUM_LEVEL 3
229
Tobias Klauser0bec8c82014-07-22 12:06:23 +0200230#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
David S. Millerfc910a22007-03-25 20:27:59 -0700231#define SKB_WITH_OVERHEAD(X) \
Herbert Xudeea84b2007-10-21 16:27:46 -0700232 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
David S. Millerfc910a22007-03-25 20:27:59 -0700233#define SKB_MAX_ORDER(X, ORDER) \
234 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700235#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
236#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
237
Eric Dumazet87fb4b72011-10-13 07:28:54 +0000238/* return minimum truesize of one skb containing X bytes of data */
239#define SKB_TRUESIZE(X) ((X) + \
240 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
241 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
242
Herbert Xu79990962020-06-12 16:57:37 +1000243struct ahash_request;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700244struct net_device;
David Howells716ea3a2007-04-02 20:19:53 -0700245struct scatterlist;
Jens Axboe9c55e012007-11-06 23:30:13 -0800246struct pipe_inode_info;
Herbert Xua8f820aa2014-11-07 21:22:22 +0800247struct iov_iter;
Alexander Duyckfd11a832014-12-09 19:40:49 -0800248struct napi_struct;
Petar Penkovd58e4682018-09-14 07:46:18 -0700249struct bpf_prog;
250union bpf_attr;
Florian Westphaldf5042f2018-12-18 17:15:16 +0100251struct skb_ext;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700252
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +0200253#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700254struct nf_bridge_info {
Florian Westphal3eaf4022015-04-02 14:31:44 +0200255 enum {
256 BRNF_PROTO_UNCHANGED,
257 BRNF_PROTO_8021Q,
258 BRNF_PROTO_PPPOE
Florian Westphal7fb48c52015-05-03 22:05:28 +0200259 } orig_proto:8;
Florian Westphal72b1e5e2015-07-23 16:21:30 +0200260 u8 pkt_otherhost:1;
261 u8 in_prerouting:1;
262 u8 bridged_dnat:1;
Bernhard Thaler411ffb42015-05-30 15:28:28 +0200263 __u16 frag_max_size;
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000264 struct net_device *physindev;
Florian Westphal63cdbc02015-09-14 17:06:27 +0200265
266 /* always valid & non-NULL from FORWARD on, for physdev match */
267 struct net_device *physoutdev;
Florian Westphal7fb48c52015-05-03 22:05:28 +0200268 union {
Florian Westphal72b1e5e2015-07-23 16:21:30 +0200269 /* prerouting: detect dnat in orig/reply direction */
Bernhard Thaler72b31f72015-05-30 15:27:40 +0200270 __be32 ipv4_daddr;
271 struct in6_addr ipv6_daddr;
Florian Westphal72b1e5e2015-07-23 16:21:30 +0200272
273 /* after prerouting + nat detected: store original source
274 * mac since neigh resolution overwrites it, only used while
275 * skb is out in neigh layer.
276 */
277 char neigh_header[8];
Bernhard Thaler72b31f72015-05-30 15:27:40 +0200278 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700279};
280#endif
281
Paul Blakey95a72332019-09-04 16:56:37 +0300282#if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
283/* Chain in tc_skb_ext will be used to share the tc chain with
284 * ovs recirc_id. It will be set to the current chain by tc
285 * and read by ovs to recirc_id.
286 */
287struct tc_skb_ext {
288 __u32 chain;
wenxu038ebb12020-07-31 10:45:01 +0800289 __u16 mru;
Paul Blakey95a72332019-09-04 16:56:37 +0300290};
291#endif
292
Linus Torvalds1da177e2005-04-16 15:20:36 -0700293struct sk_buff_head {
294 /* These two members must be first. */
295 struct sk_buff *next;
296 struct sk_buff *prev;
297
298 __u32 qlen;
299 spinlock_t lock;
300};
301
302struct sk_buff;
303
Ian Campbell9d4dde52011-12-22 23:39:14 +0000304/* To allow 64K frame to be packed as single skb without frag_list we
305 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
306 * buffers which do not start on a page boundary.
307 *
308 * Since GRO uses frags we allocate at least 16 regardless of page
309 * size.
Anton Blancharda715dea2011-03-27 14:57:26 +0000310 */
Ian Campbell9d4dde52011-12-22 23:39:14 +0000311#if (65536/PAGE_SIZE + 1) < 16
David S. Millereec00952011-03-29 23:34:08 -0700312#define MAX_SKB_FRAGS 16UL
Anton Blancharda715dea2011-03-27 14:57:26 +0000313#else
Ian Campbell9d4dde52011-12-22 23:39:14 +0000314#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
Anton Blancharda715dea2011-03-27 14:57:26 +0000315#endif
Hans Westgaard Ry5f74f82e2016-02-03 09:26:57 +0100316extern int sysctl_max_skb_frags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700317
Marcelo Ricardo Leitner3953c462016-06-02 15:05:40 -0300318/* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
319 * segment using its current segmentation instead.
320 */
321#define GSO_BY_FRAGS 0xFFFF
322
Matthew Wilcox (Oracle)8842d282019-07-22 20:08:31 -0700323typedef struct bio_vec skb_frag_t;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324
Pedro Tammela161e6132019-03-05 11:35:54 -0300325/**
Jonathan Lemon7240b602019-07-30 07:40:32 -0700326 * skb_frag_size() - Returns the size of a skb fragment
Pedro Tammela161e6132019-03-05 11:35:54 -0300327 * @frag: skb fragment
328 */
Eric Dumazet9e903e02011-10-18 21:00:24 +0000329static inline unsigned int skb_frag_size(const skb_frag_t *frag)
330{
Matthew Wilcox (Oracle)b8b576a2019-07-22 20:08:30 -0700331 return frag->bv_len;
Eric Dumazet9e903e02011-10-18 21:00:24 +0000332}
333
Pedro Tammela161e6132019-03-05 11:35:54 -0300334/**
Jonathan Lemon7240b602019-07-30 07:40:32 -0700335 * skb_frag_size_set() - Sets the size of a skb fragment
Pedro Tammela161e6132019-03-05 11:35:54 -0300336 * @frag: skb fragment
337 * @size: size of fragment
338 */
Eric Dumazet9e903e02011-10-18 21:00:24 +0000339static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
340{
Matthew Wilcox (Oracle)b8b576a2019-07-22 20:08:30 -0700341 frag->bv_len = size;
Eric Dumazet9e903e02011-10-18 21:00:24 +0000342}
343
Pedro Tammela161e6132019-03-05 11:35:54 -0300344/**
Jonathan Lemon7240b602019-07-30 07:40:32 -0700345 * skb_frag_size_add() - Increments the size of a skb fragment by @delta
Pedro Tammela161e6132019-03-05 11:35:54 -0300346 * @frag: skb fragment
347 * @delta: value to add
348 */
Eric Dumazet9e903e02011-10-18 21:00:24 +0000349static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
350{
Matthew Wilcox (Oracle)b8b576a2019-07-22 20:08:30 -0700351 frag->bv_len += delta;
Eric Dumazet9e903e02011-10-18 21:00:24 +0000352}
353
Pedro Tammela161e6132019-03-05 11:35:54 -0300354/**
Jonathan Lemon7240b602019-07-30 07:40:32 -0700355 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta
Pedro Tammela161e6132019-03-05 11:35:54 -0300356 * @frag: skb fragment
357 * @delta: value to subtract
358 */
Eric Dumazet9e903e02011-10-18 21:00:24 +0000359static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
360{
Matthew Wilcox (Oracle)b8b576a2019-07-22 20:08:30 -0700361 frag->bv_len -= delta;
Eric Dumazet9e903e02011-10-18 21:00:24 +0000362}
363
Pedro Tammela161e6132019-03-05 11:35:54 -0300364/**
365 * skb_frag_must_loop - Test if %p is a high memory page
366 * @p: fragment's page
367 */
Willem de Bruijnc613c202017-07-31 08:15:47 -0400368static inline bool skb_frag_must_loop(struct page *p)
369{
370#if defined(CONFIG_HIGHMEM)
371 if (PageHighMem(p))
372 return true;
373#endif
374 return false;
375}
376
377/**
378 * skb_frag_foreach_page - loop over pages in a fragment
379 *
380 * @f: skb frag to operate on
Matthew Wilcox (Oracle)1dfa5bd2019-07-22 20:08:29 -0700381 * @f_off: offset from start of f->bv_page
Willem de Bruijnc613c202017-07-31 08:15:47 -0400382 * @f_len: length from f_off to loop over
383 * @p: (temp var) current page
384 * @p_off: (temp var) offset from start of current page,
385 * non-zero only on first page.
386 * @p_len: (temp var) length in current page,
387 * < PAGE_SIZE only on first and last page.
388 * @copied: (temp var) length so far, excluding current p_len.
389 *
390 * A fragment can hold a compound page, in which case per-page
391 * operations, notably kmap_atomic, must be called for each
392 * regular page.
393 */
394#define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
395 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
396 p_off = (f_off) & (PAGE_SIZE - 1), \
397 p_len = skb_frag_must_loop(p) ? \
398 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
399 copied = 0; \
400 copied < f_len; \
401 copied += p_len, p++, p_off = 0, \
402 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
403
Patrick Ohlyac45f602009-02-12 05:03:37 +0000404#define HAVE_HW_TIME_STAMP
405
406/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800407 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000408 * @hwtstamp: hardware time stamp transformed into duration
409 * since arbitrary point in time
Patrick Ohlyac45f602009-02-12 05:03:37 +0000410 *
411 * Software time stamps generated by ktime_get_real() are stored in
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400412 * skb->tstamp.
Patrick Ohlyac45f602009-02-12 05:03:37 +0000413 *
414 * hwtstamps can only be compared against other hwtstamps from
415 * the same device.
416 *
417 * This structure is attached to packets as part of the
418 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
419 */
420struct skb_shared_hwtstamps {
421 ktime_t hwtstamp;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000422};
423
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000424/* Definitions for tx_flags in struct skb_shared_info */
425enum {
426 /* generate hardware time stamp */
427 SKBTX_HW_TSTAMP = 1 << 0,
428
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400429 /* generate software time stamp when queueing packet to NIC */
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000430 SKBTX_SW_TSTAMP = 1 << 1,
431
432 /* device driver is going to provide hardware time stamp */
433 SKBTX_IN_PROGRESS = 1 << 2,
434
Shirley Maa6686f22011-07-06 12:22:12 +0000435 /* device driver supports TX zero-copy buffers */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000436 SKBTX_DEV_ZEROCOPY = 1 << 3,
Johannes Berg6e3e9392011-11-09 10:15:42 +0100437
438 /* generate wifi status information (where possible) */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000439 SKBTX_WIFI_STATUS = 1 << 4,
Pravin B Shelarc9af6db2013-02-11 09:27:41 +0000440
441 /* This indicates at least one fragment might be overwritten
442 * (as in vmsplice(), sendfile() ...)
443 * If we need to compute a TX checksum, we'll need to copy
444 * all frags to avoid possible bad checksum
445 */
446 SKBTX_SHARED_FRAG = 1 << 5,
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400447
448 /* generate software time stamp when entering packet scheduling */
449 SKBTX_SCHED_TSTAMP = 1 << 6,
Shirley Maa6686f22011-07-06 12:22:12 +0000450};
451
Willem de Bruijn52267792017-08-03 16:29:39 -0400452#define SKBTX_ZEROCOPY_FRAG (SKBTX_DEV_ZEROCOPY | SKBTX_SHARED_FRAG)
Willem de Bruijne1c8a602014-08-04 22:11:50 -0400453#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
Soheil Hassas Yeganeh0a2cf202016-04-27 23:39:01 -0400454 SKBTX_SCHED_TSTAMP)
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400455#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
456
Shirley Maa6686f22011-07-06 12:22:12 +0000457/*
458 * The callback notifies userspace to release buffers when skb DMA is done in
459 * lower device, the skb last reference should be 0 when calling this.
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000460 * The zerocopy_success argument is true if zero copy transmit occurred,
461 * false on data copy or out of memory error caused by data copy attempt.
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000462 * The ctx field is used to track device context.
463 * The desc field is used to track userspace buffer index.
Shirley Maa6686f22011-07-06 12:22:12 +0000464 */
465struct ubuf_info {
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000466 void (*callback)(struct ubuf_info *, bool zerocopy_success);
Willem de Bruijn4ab6c992017-08-03 16:29:42 -0400467 union {
468 struct {
469 unsigned long desc;
470 void *ctx;
471 };
472 struct {
473 u32 id;
474 u16 len;
475 u16 zerocopy:1;
476 u32 bytelen;
477 };
478 };
Eric Dumazetc1d1b432017-08-31 16:48:22 -0700479 refcount_t refcnt;
Willem de Bruijna91dbff2017-08-03 16:29:43 -0400480
481 struct mmpin {
482 struct user_struct *user;
483 unsigned int num_pg;
484 } mmp;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000485};
486
Willem de Bruijn52267792017-08-03 16:29:39 -0400487#define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
488
Sowmini Varadhan6f89dbc2018-02-15 10:49:32 -0800489int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
490void mm_unaccount_pinned_pages(struct mmpin *mmp);
491
Willem de Bruijn52267792017-08-03 16:29:39 -0400492struct ubuf_info *sock_zerocopy_alloc(struct sock *sk, size_t size);
Willem de Bruijn4ab6c992017-08-03 16:29:42 -0400493struct ubuf_info *sock_zerocopy_realloc(struct sock *sk, size_t size,
494 struct ubuf_info *uarg);
Willem de Bruijn52267792017-08-03 16:29:39 -0400495
496static inline void sock_zerocopy_get(struct ubuf_info *uarg)
497{
Eric Dumazetc1d1b432017-08-31 16:48:22 -0700498 refcount_inc(&uarg->refcnt);
Willem de Bruijn52267792017-08-03 16:29:39 -0400499}
500
501void sock_zerocopy_put(struct ubuf_info *uarg);
Willem de Bruijn52900d22018-11-30 15:32:40 -0500502void sock_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
Willem de Bruijn52267792017-08-03 16:29:39 -0400503
504void sock_zerocopy_callback(struct ubuf_info *uarg, bool success);
505
Willem de Bruijnb5947e52018-11-30 15:32:39 -0500506int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len);
Willem de Bruijn52267792017-08-03 16:29:39 -0400507int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
508 struct msghdr *msg, int len,
509 struct ubuf_info *uarg);
510
Linus Torvalds1da177e2005-04-16 15:20:36 -0700511/* This data is invariant across clones and lives at
512 * the end of the header data, ie. at skb->end.
513 */
514struct skb_shared_info {
Daniel Borkmannde8f3a82017-09-25 02:25:51 +0200515 __u8 __unused;
516 __u8 meta_len;
517 __u8 nr_frags;
Ian Campbell9f42f122012-01-05 07:13:39 +0000518 __u8 tx_flags;
Herbert Xu79671682006-06-22 02:40:14 -0700519 unsigned short gso_size;
520 /* Warning: this field is not always filled in (UFO)! */
521 unsigned short gso_segs;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700522 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000523 struct skb_shared_hwtstamps hwtstamps;
Steffen Klassert7f564522017-04-08 20:36:24 +0200524 unsigned int gso_type;
Willem de Bruijn09c2d252014-08-04 22:11:47 -0400525 u32 tskey;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700526
527 /*
528 * Warning : all fields before dataref are cleared in __alloc_skb()
529 */
530 atomic_t dataref;
531
Johann Baudy69e3c752009-05-18 22:11:22 -0700532 /* Intermediate layers must ensure that destructor_arg
533 * remains valid until skb destructor */
534 void * destructor_arg;
Shirley Maa6686f22011-07-06 12:22:12 +0000535
Vignesh Saravanaperumal726468d2021-07-08 07:01:33 -0700536 ANDROID_OEM_DATA_ARRAY(1, 3);
537
Eric Dumazetfed66382010-07-22 19:09:08 +0000538 /* must be last field, see pskb_expand_head() */
539 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540};
541
542/* We divide dataref into two halves. The higher 16 bits hold references
543 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700544 * the entire skb->data. A clone of a headerless skb holds the length of
545 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700546 *
547 * All users must obey the rule that the skb->data reference count must be
548 * greater than or equal to the payload reference count.
549 *
550 * Holding a reference to the payload part means that the user does not
551 * care about modifications to the header part of skb->data.
552 */
553#define SKB_DATAREF_SHIFT 16
554#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
555
David S. Millerd179cd12005-08-17 14:57:30 -0700556
557enum {
Vijay Subramanianc8753d52014-10-02 10:00:43 -0700558 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
559 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
560 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
David S. Millerd179cd12005-08-17 14:57:30 -0700561};
562
Herbert Xu79671682006-06-22 02:40:14 -0700563enum {
564 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xu576a30e2006-06-27 13:22:38 -0700565
566 /* This indicates the skb is from an untrusted source. */
David S. Millerd9d30ad2017-07-03 07:31:57 -0700567 SKB_GSO_DODGY = 1 << 1,
Michael Chanb0da85372006-06-29 12:30:00 -0700568
569 /* This indicates the tcp segment has CWR set. */
David S. Millerd9d30ad2017-07-03 07:31:57 -0700570 SKB_GSO_TCP_ECN = 1 << 2,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700571
David S. Millerd9d30ad2017-07-03 07:31:57 -0700572 SKB_GSO_TCP_FIXEDID = 1 << 3,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800573
David S. Millerd9d30ad2017-07-03 07:31:57 -0700574 SKB_GSO_TCPV6 = 1 << 4,
Pravin B Shelar68c33162013-02-14 14:02:41 +0000575
David S. Millerd9d30ad2017-07-03 07:31:57 -0700576 SKB_GSO_FCOE = 1 << 5,
Pravin B Shelar73136262013-03-07 13:21:51 +0000577
David S. Millerd9d30ad2017-07-03 07:31:57 -0700578 SKB_GSO_GRE = 1 << 6,
Simon Horman0d89d202013-05-23 21:02:52 +0000579
David S. Millerd9d30ad2017-07-03 07:31:57 -0700580 SKB_GSO_GRE_CSUM = 1 << 7,
Eric Dumazetcb32f512013-10-19 11:42:57 -0700581
David S. Millerd9d30ad2017-07-03 07:31:57 -0700582 SKB_GSO_IPXIP4 = 1 << 8,
Eric Dumazet61c1db72013-10-20 20:47:30 -0700583
David S. Millerd9d30ad2017-07-03 07:31:57 -0700584 SKB_GSO_IPXIP6 = 1 << 9,
Tom Herbert0f4f4ff2014-06-04 17:20:16 -0700585
David S. Millerd9d30ad2017-07-03 07:31:57 -0700586 SKB_GSO_UDP_TUNNEL = 1 << 10,
Tom Herbert4749c092014-06-04 17:20:23 -0700587
David S. Millerd9d30ad2017-07-03 07:31:57 -0700588 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400589
David S. Millerd9d30ad2017-07-03 07:31:57 -0700590 SKB_GSO_PARTIAL = 1 << 12,
Alexander Duyck802ab552016-04-10 21:45:03 -0400591
David S. Millerd9d30ad2017-07-03 07:31:57 -0700592 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
Marcelo Ricardo Leitner90017ac2016-06-02 15:05:43 -0300593
David S. Millerd9d30ad2017-07-03 07:31:57 -0700594 SKB_GSO_SCTP = 1 << 14,
Steffen Klassertc7ef8f02017-04-14 10:05:36 +0200595
David S. Millerd9d30ad2017-07-03 07:31:57 -0700596 SKB_GSO_ESP = 1 << 15,
Willem de Bruijn0c19f8462017-11-21 10:22:25 -0500597
598 SKB_GSO_UDP = 1 << 16,
Willem de Bruijnee80d1e2018-04-26 13:42:16 -0400599
600 SKB_GSO_UDP_L4 = 1 << 17,
Steffen Klassert3b335832020-01-25 11:26:42 +0100601
602 SKB_GSO_FRAGLIST = 1 << 18,
Herbert Xu79671682006-06-22 02:40:14 -0700603};
604
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700605#if BITS_PER_LONG > 32
606#define NET_SKBUFF_DATA_USES_OFFSET 1
607#endif
608
609#ifdef NET_SKBUFF_DATA_USES_OFFSET
610typedef unsigned int sk_buff_data_t;
611#else
612typedef unsigned char *sk_buff_data_t;
613#endif
614
Pedro Tammela161e6132019-03-05 11:35:54 -0300615/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616 * struct sk_buff - socket buffer
617 * @next: Next buffer in list
618 * @prev: Previous buffer in list
Eric Dumazet363ec39232014-02-26 14:02:11 -0800619 * @tstamp: Time we arrived/left
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800620 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point
621 * for retransmit timer
Eric Dumazet56b17422014-11-03 08:19:53 -0800622 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800623 * @list: queue head
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700624 * @sk: Socket we are owned by
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800625 * @ip_defrag_offset: (aka @sk) alternate use of @sk, used in
626 * fragmentation management
Linus Torvalds1da177e2005-04-16 15:20:36 -0700627 * @dev: Device we arrived on/are leaving by
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800628 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700629 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000630 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700631 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700632 * @len: Length of actual data
633 * @data_len: Data length
634 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700635 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700636 * @csum: Checksum (must include start/offset pair)
637 * @csum_start: Offset from skb->head where checksumming should start
638 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700639 * @priority: Packet queueing priority
WANG Cong60ff7462014-05-04 16:39:18 -0700640 * @ignore_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700641 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700642 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700643 * @nohdr: Payload reference only, must not modify header
644 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700645 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700646 * @ipvs_property: skbuff is owned by ipvs
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800647 * @inner_protocol_type: whether the inner protocol is
648 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO
649 * @remcsum_offload: remote checksum offload is enabled
Ido Schimmel875e8932018-12-04 08:15:10 +0000650 * @offload_fwd_mark: Packet was L2-forwarded in hardware
651 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
Willem de Bruijne7246e12017-01-07 17:06:35 -0500652 * @tc_skip_classify: do not classify packet. set by IFB device
Willem de Bruijn8dc07fd2017-01-07 17:06:37 -0500653 * @tc_at_ingress: used within tc_classify to distinguish in/egress
Pablo Neira Ayuso2c646052020-03-25 13:47:18 +0100654 * @redirected: packet was redirected by packet classifier
655 * @from_ingress: packet was redirected from the ingress path
Randy Dunlap31729362008-02-18 20:52:13 -0800656 * @peeked: this packet has been seen already, so stats have been
657 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700658 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700659 * @protocol: Packet protocol from driver
660 * @destructor: Destruct function
Eric Dumazete2080072017-10-04 12:59:58 -0700661 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
Florian Westphala9e419d2017-01-23 18:21:59 +0100662 * @_nfct: Associated connection, if any (with nfctinfo bits)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700663 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800664 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700665 * @tc_index: Traffic control index
Tom Herbert61b905d2014-03-24 15:34:47 -0700666 * @hash: the packet hash
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700667 * @queue_mapping: Queue mapping for multiqueue devices
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800668 * @head_frag: skb was allocated from page fragments,
669 * not allocated by kmalloc() or vmalloc().
Stefano Brivio8b700862018-07-11 14:39:42 +0200670 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
Florian Westphaldf5042f2018-12-18 17:15:16 +0100671 * @active_extensions: active extensions (skb_ext_id types)
Randy Dunlap553a5672008-04-20 10:51:01 -0700672 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700673 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Tom Herbert61b905d2014-03-24 15:34:47 -0700674 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
Changli Gao4ca24622011-08-19 07:26:44 -0700675 * ports.
Tom Herberta3b18dd2014-07-01 21:33:17 -0700676 * @sw_hash: indicates hash was computed in software stack
Johannes Berg6e3e9392011-11-09 10:15:42 +0100677 * @wifi_acked_valid: wifi_acked was set
678 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000679 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800680 * @encapsulation: indicates the inner headers in the skbuff are valid
681 * @encap_hdr_csum: software checksum is needed
682 * @csum_valid: checksum is already valid
Davide Carattidba00302017-05-18 15:44:40 +0200683 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800684 * @csum_complete_sw: checksum was completed by software
685 * @csum_level: indicates the number of consecutive checksums found in
686 * the packet minus one that have been verified as
687 * CHECKSUM_UNNECESSARY (max 3)
Julian Anastasov4ff06202017-02-06 23:14:12 +0200688 * @dst_pending_confirm: need to confirm neighbour
Stefano Brivioa48d1892018-07-17 11:52:57 +0200689 * @decrypted: Decrypted SKB
Pedro Tammela161e6132019-03-05 11:35:54 -0300690 * @napi_id: id of the NAPI struct this skb came from
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800691 * @sender_cpu: (aka @napi_id) source CPU in XPS
James Morris984bc162006-06-09 00:29:17 -0700692 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700693 * @mark: Generic packet mark
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800694 * @reserved_tailroom: (aka @mark) number of bytes of free space available
695 * at the tail of an sk_buff
696 * @vlan_present: VLAN tag is present
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000697 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b2008-07-14 22:49:06 -0700698 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000699 * @inner_protocol: Protocol (encapsulation)
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800700 * @inner_ipproto: (aka @inner_protocol) stores ipproto when
701 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000702 * @inner_transport_header: Inner transport layer header (encapsulation)
703 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000704 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700705 * @transport_header: Transport layer header
706 * @network_header: Network layer header
707 * @mac_header: Link layer header
708 * @tail: Tail pointer
709 * @end: End pointer
710 * @head: Head of buffer
711 * @data: Data head pointer
712 * @truesize: Buffer size
713 * @users: User count - see {datagram,tcp}.c
Florian Westphaldf5042f2018-12-18 17:15:16 +0100714 * @extensions: allocated extensions, valid if active_extensions is nonzero
Linus Torvalds1da177e2005-04-16 15:20:36 -0700715 */
716
717struct sk_buff {
Eric Dumazet363ec39232014-02-26 14:02:11 -0800718 union {
Eric Dumazet56b17422014-11-03 08:19:53 -0800719 struct {
720 /* These two members must be first. */
721 struct sk_buff *next;
722 struct sk_buff *prev;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000723
Eric Dumazet56b17422014-11-03 08:19:53 -0800724 union {
Eric Dumazetbffa72c2017-09-19 05:14:24 -0700725 struct net_device *dev;
726 /* Some protocols might use this space to store information,
727 * while device pointer would be NULL.
728 * UDP receive path is one user.
729 */
730 unsigned long dev_scratch;
Eric Dumazet56b17422014-11-03 08:19:53 -0800731 };
732 };
Peter Oskolkovfa0f5272018-08-02 23:34:39 +0000733 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
David Millerd4546c22018-06-24 14:13:49 +0900734 struct list_head list;
Eric Dumazet56b17422014-11-03 08:19:53 -0800735 };
Peter Oskolkovfa0f5272018-08-02 23:34:39 +0000736
737 union {
738 struct sock *sk;
739 int ip_defrag_offset;
740 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700741
Eric Dumazetc84d9492016-12-08 11:41:55 -0800742 union {
Eric Dumazetbffa72c2017-09-19 05:14:24 -0700743 ktime_t tstamp;
Eric Dumazetd3edd062018-09-21 08:51:50 -0700744 u64 skb_mstamp_ns; /* earliest departure time */
Eric Dumazetc84d9492016-12-08 11:41:55 -0800745 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700746 /*
747 * This is the control buffer. It is free to use for every
748 * layer. Please put your private variables there. If you
749 * want to keep them across layers you have to do a skb_clone()
750 * first. This is owned by whoever has the skb queued ATM.
751 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000752 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700753
Eric Dumazete2080072017-10-04 12:59:58 -0700754 union {
755 struct {
756 unsigned long _skb_refdst;
757 void (*destructor)(struct sk_buff *skb);
758 };
759 struct list_head tcp_tsorted_anchor;
760 };
761
Eric Dumazetb1937222014-09-28 22:18:47 -0700762#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +0100763 unsigned long _nfct;
Eric Dumazetb1937222014-09-28 22:18:47 -0700764#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700765 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700766 data_len;
767 __u16 mac_len,
768 hdr_len;
Eric Dumazetb1937222014-09-28 22:18:47 -0700769
770 /* Following fields are _not_ copied in __copy_skb_header()
771 * Note that queue_mapping is here mostly to fill a hole.
772 */
Eric Dumazetb1937222014-09-28 22:18:47 -0700773 __u16 queue_mapping;
Daniel Borkmann36bbef52016-09-20 00:26:13 +0200774
775/* if you move cloned around you also must adapt those constants */
776#ifdef __BIG_ENDIAN_BITFIELD
777#define CLONED_MASK (1 << 7)
778#else
779#define CLONED_MASK 1
780#endif
781#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
782
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800783 /* private: */
Daniel Borkmann36bbef52016-09-20 00:26:13 +0200784 __u8 __cloned_offset[0];
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800785 /* public: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700786 __u8 cloned:1,
Harald Welte6869c4d2005-08-09 19:24:19 -0700787 nohdr:1,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800788 fclone:2,
Herbert Xua59322b2007-12-05 01:53:40 -0800789 peeked:1,
Eric Dumazetb1937222014-09-28 22:18:47 -0700790 head_frag:1,
Stefano Brivio8b700862018-07-11 14:39:42 +0200791 pfmemalloc:1;
Florian Westphaldf5042f2018-12-18 17:15:16 +0100792#ifdef CONFIG_SKB_EXTENSIONS
793 __u8 active_extensions;
794#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700795 /* fields enclosed in headers_start/headers_end are copied
796 * using a single memcpy() in __copy_skb_header()
797 */
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700798 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700799 __u32 headers_start[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700800 /* public: */
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200801
802/* if you move pkt_type around you also must adapt those constants */
803#ifdef __BIG_ENDIAN_BITFIELD
804#define PKT_TYPE_MAX (7 << 5)
805#else
806#define PKT_TYPE_MAX 7
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100807#endif
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200808#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
809
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800810 /* private: */
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200811 __u8 __pkt_type_offset[0];
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800812 /* public: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700813 __u8 pkt_type:3;
Eric Dumazetb1937222014-09-28 22:18:47 -0700814 __u8 ignore_df:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700815 __u8 nf_trace:1;
816 __u8 ip_summed:2;
817 __u8 ooo_okay:1;
Stefano Brivio8b700862018-07-11 14:39:42 +0200818
Eric Dumazetb1937222014-09-28 22:18:47 -0700819 __u8 l4_hash:1;
820 __u8 sw_hash:1;
821 __u8 wifi_acked_valid:1;
822 __u8 wifi_acked:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700823 __u8 no_fcs:1;
824 /* Indicates the inner headers are valid in the skbuff. */
825 __u8 encapsulation:1;
826 __u8 encap_hdr_csum:1;
827 __u8 csum_valid:1;
Stefano Brivio8b700862018-07-11 14:39:42 +0200828
Michał Mirosław0c4b2d32018-11-10 19:58:36 +0100829#ifdef __BIG_ENDIAN_BITFIELD
830#define PKT_VLAN_PRESENT_BIT 7
831#else
832#define PKT_VLAN_PRESENT_BIT 0
833#endif
834#define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset)
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800835 /* private: */
Michał Mirosław0c4b2d32018-11-10 19:58:36 +0100836 __u8 __pkt_vlan_present_offset[0];
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800837 /* public: */
Michał Mirosław0c4b2d32018-11-10 19:58:36 +0100838 __u8 vlan_present:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700839 __u8 csum_complete_sw:1;
840 __u8 csum_level:2;
Davide Carattidba00302017-05-18 15:44:40 +0200841 __u8 csum_not_inet:1;
Julian Anastasov4ff06202017-02-06 23:14:12 +0200842 __u8 dst_pending_confirm:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700843#ifdef CONFIG_IPV6_NDISC_NODETYPE
844 __u8 ndisc_nodetype:2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700845#endif
Stefano Brivio8b700862018-07-11 14:39:42 +0200846
Michał Mirosław0c4b2d32018-11-10 19:58:36 +0100847 __u8 ipvs_property:1;
Tom Herbert8bce6d72014-09-29 20:22:29 -0700848 __u8 inner_protocol_type:1;
Tom Herberte585f232014-11-04 09:06:54 -0800849 __u8 remcsum_offload:1;
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200850#ifdef CONFIG_NET_SWITCHDEV
851 __u8 offload_fwd_mark:1;
Ido Schimmel875e8932018-12-04 08:15:10 +0000852 __u8 offload_l3_fwd_mark:1;
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200853#endif
Willem de Bruijne7246e12017-01-07 17:06:35 -0500854#ifdef CONFIG_NET_CLS_ACT
855 __u8 tc_skip_classify:1;
Willem de Bruijn8dc07fd2017-01-07 17:06:37 -0500856 __u8 tc_at_ingress:1;
Pablo Neira Ayuso2c646052020-03-25 13:47:18 +0100857#endif
858#ifdef CONFIG_NET_REDIRECT
859 __u8 redirected:1;
860 __u8 from_ingress:1;
Willem de Bruijne7246e12017-01-07 17:06:35 -0500861#endif
Stefano Brivioa48d1892018-07-17 11:52:57 +0200862#ifdef CONFIG_TLS_DEVICE
863 __u8 decrypted:1;
864#endif
Alexander Duyck4031ae62012-01-27 06:22:53 +0000865
Linus Torvalds1da177e2005-04-16 15:20:36 -0700866#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700867 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700868#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200869
Eric Dumazetb1937222014-09-28 22:18:47 -0700870 union {
871 __wsum csum;
872 struct {
873 __u16 csum_start;
874 __u16 csum_offset;
875 };
876 };
877 __u32 priority;
878 int skb_iif;
879 __u32 hash;
880 __be16 vlan_proto;
881 __u16 vlan_tci;
Eric Dumazet2bd82482015-02-03 23:48:24 -0800882#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
883 union {
884 unsigned int napi_id;
885 unsigned int sender_cpu;
886 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700887#endif
James Morris984bc162006-06-09 00:29:17 -0700888#ifdef CONFIG_NETWORK_SECMARK
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200889 __u32 secmark;
James Morris984bc162006-06-09 00:29:17 -0700890#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -0700891
Neil Horman3b885782009-10-12 13:26:31 -0700892 union {
893 __u32 mark;
Eric Dumazet16fad692013-03-14 05:40:32 +0000894 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700895 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700896
Tom Herbert8bce6d72014-09-29 20:22:29 -0700897 union {
898 __be16 inner_protocol;
899 __u8 inner_ipproto;
900 };
901
Simon Horman1a37e412013-05-23 21:02:51 +0000902 __u16 inner_transport_header;
903 __u16 inner_network_header;
904 __u16 inner_mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700905
906 __be16 protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000907 __u16 transport_header;
908 __u16 network_header;
909 __u16 mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700910
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700911 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700912 __u32 headers_end[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700913 /* public: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700914
Greg Kroah-Hartman73ff58b2020-05-02 09:32:22 +0200915 ANDROID_KABI_RESERVE(1);
916 ANDROID_KABI_RESERVE(2);
917
Linus Torvalds1da177e2005-04-16 15:20:36 -0700918 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700919 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700920 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700921 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700922 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700923 unsigned int truesize;
Reshetova, Elena63354792017-06-30 13:07:58 +0300924 refcount_t users;
Florian Westphaldf5042f2018-12-18 17:15:16 +0100925
926#ifdef CONFIG_SKB_EXTENSIONS
927 /* only useable after checking ->active_extensions != 0 */
928 struct skb_ext *extensions;
929#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700930};
931
932#ifdef __KERNEL__
933/*
934 * Handling routines are only of interest to the kernel
935 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700936
Mel Gormanc93bdd02012-07-31 16:44:19 -0700937#define SKB_ALLOC_FCLONE 0x01
938#define SKB_ALLOC_RX 0x02
Alexander Duyckfd11a832014-12-09 19:40:49 -0800939#define SKB_ALLOC_NAPI 0x04
Mel Gormanc93bdd02012-07-31 16:44:19 -0700940
Pedro Tammela161e6132019-03-05 11:35:54 -0300941/**
942 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
943 * @skb: buffer
944 */
Mel Gormanc93bdd02012-07-31 16:44:19 -0700945static inline bool skb_pfmemalloc(const struct sk_buff *skb)
946{
947 return unlikely(skb->pfmemalloc);
948}
949
Eric Dumazet7fee2262010-05-11 23:19:48 +0000950/*
951 * skb might have a dst pointer attached, refcounted or not.
952 * _skb_refdst low order bit is set if refcount was _not_ taken
953 */
954#define SKB_DST_NOREF 1UL
955#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
956
957/**
958 * skb_dst - returns skb dst_entry
959 * @skb: buffer
960 *
961 * Returns skb dst_entry, regardless of reference taken or not.
962 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000963static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
964{
Pedro Tammela161e6132019-03-05 11:35:54 -0300965 /* If refdst was not refcounted, check we still are in a
Eric Dumazet7fee2262010-05-11 23:19:48 +0000966 * rcu_read_lock section
967 */
968 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
969 !rcu_read_lock_held() &&
970 !rcu_read_lock_bh_held());
971 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000972}
973
Eric Dumazet7fee2262010-05-11 23:19:48 +0000974/**
975 * skb_dst_set - sets skb dst
976 * @skb: buffer
977 * @dst: dst entry
978 *
979 * Sets skb dst, assuming a reference was taken on dst and should
980 * be released by skb_dst_drop()
981 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000982static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
983{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000984 skb->_skb_refdst = (unsigned long)dst;
985}
986
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200987/**
988 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
989 * @skb: buffer
990 * @dst: dst entry
991 *
992 * Sets skb dst, assuming a reference was not taken on dst.
993 * If dst entry is cached, we do not take reference and dst_release
994 * will be avoided by refdst_drop. If dst entry is not cached, we take
995 * reference, so that last dst_release can destroy the dst immediately.
996 */
997static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
998{
Hannes Frederic Sowadbfc4fb2014-12-06 19:19:42 +0100999 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
1000 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
Julian Anastasov932bc4d2013-03-21 11:57:58 +02001001}
Eric Dumazet7fee2262010-05-11 23:19:48 +00001002
1003/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -03001004 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +00001005 * @skb: buffer
1006 */
1007static inline bool skb_dst_is_noref(const struct sk_buff *skb)
1008{
1009 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +00001010}
1011
Pedro Tammela161e6132019-03-05 11:35:54 -03001012/**
1013 * skb_rtable - Returns the skb &rtable
1014 * @skb: buffer
1015 */
Eric Dumazet511c3f92009-06-02 05:14:27 +00001016static inline struct rtable *skb_rtable(const struct sk_buff *skb)
1017{
Eric Dumazetadf30902009-06-02 05:19:30 +00001018 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +00001019}
1020
Jamal Hadi Salim8b10cab2016-07-02 06:43:14 -04001021/* For mangling skb->pkt_type from user space side from applications
1022 * such as nft, tc, etc, we only allow a conservative subset of
1023 * possible pkt_types to be set.
1024*/
1025static inline bool skb_pkt_type_ok(u32 ptype)
1026{
1027 return ptype <= PACKET_OTHERHOST;
1028}
1029
Pedro Tammela161e6132019-03-05 11:35:54 -03001030/**
1031 * skb_napi_id - Returns the skb's NAPI id
1032 * @skb: buffer
1033 */
Miroslav Lichvar90b602f2017-05-19 17:52:37 +02001034static inline unsigned int skb_napi_id(const struct sk_buff *skb)
1035{
1036#ifdef CONFIG_NET_RX_BUSY_POLL
1037 return skb->napi_id;
1038#else
1039 return 0;
1040#endif
1041}
1042
Pedro Tammela161e6132019-03-05 11:35:54 -03001043/**
1044 * skb_unref - decrement the skb's reference count
1045 * @skb: buffer
1046 *
1047 * Returns true if we can free the skb.
1048 */
Paolo Abeni3889a8032017-06-12 11:23:41 +02001049static inline bool skb_unref(struct sk_buff *skb)
1050{
1051 if (unlikely(!skb))
1052 return false;
Reshetova, Elena63354792017-06-30 13:07:58 +03001053 if (likely(refcount_read(&skb->users) == 1))
Paolo Abeni3889a8032017-06-12 11:23:41 +02001054 smp_rmb();
Reshetova, Elena63354792017-06-30 13:07:58 +03001055 else if (likely(!refcount_dec_and_test(&skb->users)))
Paolo Abeni3889a8032017-06-12 11:23:41 +02001056 return false;
1057
1058 return true;
1059}
1060
Paolo Abeni0a463c72017-06-12 11:23:42 +02001061void skb_release_head_state(struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -07001062void kfree_skb(struct sk_buff *skb);
1063void kfree_skb_list(struct sk_buff *segs);
Willem de Bruijn64131392019-07-07 05:51:55 -04001064void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
Joe Perches7965bd42013-09-26 14:48:15 -07001065void skb_tx_error(struct sk_buff *skb);
Herbert Xube769db2020-08-22 08:23:29 +10001066
1067#ifdef CONFIG_TRACEPOINTS
Joe Perches7965bd42013-09-26 14:48:15 -07001068void consume_skb(struct sk_buff *skb);
Herbert Xube769db2020-08-22 08:23:29 +10001069#else
1070static inline void consume_skb(struct sk_buff *skb)
1071{
1072 return kfree_skb(skb);
1073}
1074#endif
1075
Paolo Abenica2c1412017-09-06 14:44:36 +02001076void __consume_stateless_skb(struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -07001077void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +00001078extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +00001079
Joe Perches7965bd42013-09-26 14:48:15 -07001080void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1081bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1082 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +00001083
Joe Perches7965bd42013-09-26 14:48:15 -07001084struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1085 int node);
Eric Dumazet2ea2f622015-04-24 16:05:01 -07001086struct sk_buff *__build_skb(void *data, unsigned int frag_size);
Joe Perches7965bd42013-09-26 14:48:15 -07001087struct sk_buff *build_skb(void *data, unsigned int frag_size);
Jesper Dangaard Brouerba0509b2019-04-12 17:07:37 +02001088struct sk_buff *build_skb_around(struct sk_buff *skb,
1089 void *data, unsigned int frag_size);
Pedro Tammela161e6132019-03-05 11:35:54 -03001090
1091/**
1092 * alloc_skb - allocate a network buffer
1093 * @size: size to allocate
1094 * @priority: allocation mask
1095 *
1096 * This function is a convenient wrapper around __alloc_skb().
1097 */
David S. Millerd179cd12005-08-17 14:57:30 -07001098static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +01001099 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -07001100{
Eric Dumazet564824b2010-10-11 19:05:25 +00001101 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -07001102}
1103
Eric Dumazet2e4e4412014-09-17 04:49:49 -07001104struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1105 unsigned long data_len,
1106 int max_page_order,
1107 int *errcode,
1108 gfp_t gfp_mask);
Jakub Kicinskida29e4b2019-06-03 15:16:58 -07001109struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
Eric Dumazet2e4e4412014-09-17 04:49:49 -07001110
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001111/* Layout of fast clones : [skb1][skb2][fclone_ref] */
1112struct sk_buff_fclones {
1113 struct sk_buff skb1;
1114
1115 struct sk_buff skb2;
1116
Reshetova, Elena26385952017-06-30 13:07:59 +03001117 refcount_t fclone_ref;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001118};
1119
1120/**
1121 * skb_fclone_busy - check if fclone is busy
Stephen Hemminger293de7d2016-10-23 09:28:29 -07001122 * @sk: socket
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001123 * @skb: buffer
1124 *
Masanari Iidabda13fe2015-12-13 16:53:02 +09001125 * Returns true if skb is a fast clone, and its clone is not freed.
Eric Dumazet39bb5e62014-10-30 10:32:34 -07001126 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1127 * so we also check that this didnt happen.
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001128 */
Eric Dumazet39bb5e62014-10-30 10:32:34 -07001129static inline bool skb_fclone_busy(const struct sock *sk,
1130 const struct sk_buff *skb)
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001131{
1132 const struct sk_buff_fclones *fclones;
1133
1134 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1135
1136 return skb->fclone == SKB_FCLONE_ORIG &&
Reshetova, Elena26385952017-06-30 13:07:59 +03001137 refcount_read(&fclones->fclone_ref) > 1 &&
Eric Dumazet39bb5e62014-10-30 10:32:34 -07001138 fclones->skb2.sk == sk;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001139}
1140
Pedro Tammela161e6132019-03-05 11:35:54 -03001141/**
1142 * alloc_skb_fclone - allocate a network buffer from fclone cache
1143 * @size: size to allocate
1144 * @priority: allocation mask
1145 *
1146 * This function is a convenient wrapper around __alloc_skb().
1147 */
David S. Millerd179cd12005-08-17 14:57:30 -07001148static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +01001149 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -07001150{
Mel Gormanc93bdd02012-07-31 16:44:19 -07001151 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -07001152}
1153
Joe Perches7965bd42013-09-26 14:48:15 -07001154struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
Toshiaki Makitab0768a82018-08-03 16:58:09 +09001155void skb_headers_offset_update(struct sk_buff *skb, int off);
Joe Perches7965bd42013-09-26 14:48:15 -07001156int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1157struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
Ilya Lesokhin08303c12018-04-30 10:16:11 +03001158void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
Joe Perches7965bd42013-09-26 14:48:15 -07001159struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
Octavian Purdilabad93e92014-06-12 01:36:26 +03001160struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1161 gfp_t gfp_mask, bool fclone);
1162static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1163 gfp_t gfp_mask)
1164{
1165 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1166}
Eric Dumazet117632e2011-12-03 21:39:53 +00001167
Joe Perches7965bd42013-09-26 14:48:15 -07001168int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1169struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1170 unsigned int headroom);
1171struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1172 int newtailroom, gfp_t priority);
Jason A. Donenfeld48a1df62017-06-04 04:16:22 +02001173int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1174 int offset, int len);
1175int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1176 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001177int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
Florian Fainellicd0a1372017-08-22 15:12:14 -07001178int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1179
1180/**
1181 * skb_pad - zero pad the tail of an skb
1182 * @skb: buffer to pad
1183 * @pad: space to pad
1184 *
1185 * Ensure that a buffer is followed by a padding area that is zero
1186 * filled. Used by network drivers which may DMA or transfer data
1187 * beyond the buffer end onto the wire.
1188 *
1189 * May return error in out of memory cases. The skb is freed on error.
1190 */
1191static inline int skb_pad(struct sk_buff *skb, int pad)
1192{
1193 return __skb_pad(skb, pad, true);
1194}
Neil Hormanead2ceb2009-03-11 09:49:55 +00001195#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001196
Hannes Frederic Sowabe12a1f2015-05-21 16:59:58 +02001197int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1198 int offset, size_t size);
1199
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08001200struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -07001201 __u32 lower_offset;
1202 __u32 upper_offset;
1203 __u32 frag_idx;
1204 __u32 stepped_offset;
1205 struct sk_buff *root_skb;
1206 struct sk_buff *cur_skb;
1207 __u8 *frag_data;
1208};
1209
Joe Perches7965bd42013-09-26 14:48:15 -07001210void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1211 unsigned int to, struct skb_seq_state *st);
1212unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1213 struct skb_seq_state *st);
1214void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -07001215
Joe Perches7965bd42013-09-26 14:48:15 -07001216unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
Bojan Prtvar059a2442015-02-22 11:46:35 +01001217 unsigned int to, struct ts_config *config);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -07001218
Tom Herbert09323cc2013-12-15 22:16:19 -08001219/*
1220 * Packet hash types specify the type of hash in skb_set_hash.
1221 *
1222 * Hash types refer to the protocol layer addresses which are used to
1223 * construct a packet's hash. The hashes are used to differentiate or identify
1224 * flows of the protocol layer for the hash type. Hash types are either
1225 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1226 *
1227 * Properties of hashes:
1228 *
1229 * 1) Two packets in different flows have different hash values
1230 * 2) Two packets in the same flow should have the same hash value
1231 *
1232 * A hash at a higher layer is considered to be more specific. A driver should
1233 * set the most specific hash possible.
1234 *
1235 * A driver cannot indicate a more specific hash than the layer at which a hash
1236 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1237 *
1238 * A driver may indicate a hash level which is less specific than the
1239 * actual layer the hash was computed on. For instance, a hash computed
1240 * at L4 may be considered an L3 hash. This should only be done if the
1241 * driver can't unambiguously determine that the HW computed the hash at
1242 * the higher layer. Note that the "should" in the second property above
1243 * permits this.
1244 */
1245enum pkt_hash_types {
1246 PKT_HASH_TYPE_NONE, /* Undefined type */
1247 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1248 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1249 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1250};
1251
Tom Herbertbcc83832015-09-01 09:24:24 -07001252static inline void skb_clear_hash(struct sk_buff *skb)
1253{
1254 skb->hash = 0;
1255 skb->sw_hash = 0;
1256 skb->l4_hash = 0;
1257}
1258
1259static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1260{
1261 if (!skb->l4_hash)
1262 skb_clear_hash(skb);
1263}
1264
1265static inline void
1266__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1267{
1268 skb->l4_hash = is_l4;
1269 skb->sw_hash = is_sw;
1270 skb->hash = hash;
1271}
1272
Tom Herbert09323cc2013-12-15 22:16:19 -08001273static inline void
1274skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1275{
Tom Herbertbcc83832015-09-01 09:24:24 -07001276 /* Used by drivers to set hash from HW */
1277 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1278}
1279
1280static inline void
1281__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1282{
1283 __skb_set_hash(skb, hash, true, is_l4);
Tom Herbert09323cc2013-12-15 22:16:19 -08001284}
1285
Tom Herberte5276932015-09-01 09:24:23 -07001286void __skb_get_hash(struct sk_buff *skb);
Florian Westphalb9177832016-10-26 18:49:46 +02001287u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
Tom Herberte5276932015-09-01 09:24:23 -07001288u32 skb_get_poff(const struct sk_buff *skb);
1289u32 __skb_get_poff(const struct sk_buff *skb, void *data,
Paolo Abeni72a338b2018-05-04 11:32:59 +02001290 const struct flow_keys_basic *keys, int hlen);
Tom Herberte5276932015-09-01 09:24:23 -07001291__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1292 void *data, int hlen_proto);
1293
1294static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1295 int thoff, u8 ip_proto)
1296{
1297 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1298}
1299
1300void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1301 const struct flow_dissector_key *key,
1302 unsigned int key_count);
1303
Stanislav Fomichev089b19a2019-04-22 08:55:44 -07001304struct bpf_flow_dissector;
1305bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
Stanislav Fomichev086f9562019-07-25 15:52:25 -07001306 __be16 proto, int nhoff, int hlen, unsigned int flags);
Stanislav Fomichev089b19a2019-04-22 08:55:44 -07001307
Stanislav Fomichev3cbf4ff2019-04-22 08:55:46 -07001308bool __skb_flow_dissect(const struct net *net,
1309 const struct sk_buff *skb,
Tom Herberte5276932015-09-01 09:24:23 -07001310 struct flow_dissector *flow_dissector,
1311 void *target_container,
Tom Herbertcd79a232015-09-01 09:24:27 -07001312 void *data, __be16 proto, int nhoff, int hlen,
1313 unsigned int flags);
Tom Herberte5276932015-09-01 09:24:23 -07001314
1315static inline bool skb_flow_dissect(const struct sk_buff *skb,
1316 struct flow_dissector *flow_dissector,
Tom Herbertcd79a232015-09-01 09:24:27 -07001317 void *target_container, unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001318{
Stanislav Fomichev3cbf4ff2019-04-22 08:55:46 -07001319 return __skb_flow_dissect(NULL, skb, flow_dissector,
1320 target_container, NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001321}
1322
1323static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
Tom Herbertcd79a232015-09-01 09:24:27 -07001324 struct flow_keys *flow,
1325 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001326{
1327 memset(flow, 0, sizeof(*flow));
Stanislav Fomichev3cbf4ff2019-04-22 08:55:46 -07001328 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1329 flow, NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001330}
1331
Paolo Abeni72a338b2018-05-04 11:32:59 +02001332static inline bool
Stanislav Fomichev3cbf4ff2019-04-22 08:55:46 -07001333skb_flow_dissect_flow_keys_basic(const struct net *net,
1334 const struct sk_buff *skb,
Paolo Abeni72a338b2018-05-04 11:32:59 +02001335 struct flow_keys_basic *flow, void *data,
1336 __be16 proto, int nhoff, int hlen,
1337 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001338{
1339 memset(flow, 0, sizeof(*flow));
Stanislav Fomichev3cbf4ff2019-04-22 08:55:46 -07001340 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
Tom Herbertcd79a232015-09-01 09:24:27 -07001341 data, proto, nhoff, hlen, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001342}
1343
Jiri Pirko82828b82019-06-19 09:41:02 +03001344void skb_flow_dissect_meta(const struct sk_buff *skb,
1345 struct flow_dissector *flow_dissector,
1346 void *target_container);
1347
Paul Blakey75a56752019-07-09 10:30:49 +03001348/* Gets a skb connection tracking info, ctinfo map should be a
Randy Dunlap2ff17112020-07-15 09:42:39 -07001349 * map of mapsize to translate enum ip_conntrack_info states
Paul Blakey75a56752019-07-09 10:30:49 +03001350 * to user states.
1351 */
1352void
1353skb_flow_dissect_ct(const struct sk_buff *skb,
1354 struct flow_dissector *flow_dissector,
1355 void *target_container,
1356 u16 *ctinfo_map,
1357 size_t mapsize);
Simon Horman62b32372017-12-04 11:31:48 +01001358void
1359skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1360 struct flow_dissector *flow_dissector,
1361 void *target_container);
1362
Ariel Levkovich0cb09af2020-07-23 01:03:00 +03001363void skb_flow_dissect_hash(const struct sk_buff *skb,
1364 struct flow_dissector *flow_dissector,
1365 void *target_container);
1366
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001367static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001368{
Tom Herberta3b18dd2014-07-01 21:33:17 -07001369 if (!skb->l4_hash && !skb->sw_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001370 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001371
Tom Herbert61b905d2014-03-24 15:34:47 -07001372 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001373}
1374
David S. Miller20a17bf2015-09-01 21:19:17 -07001375static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
Tom Herbertf70ea012015-07-31 16:52:10 -07001376{
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001377 if (!skb->l4_hash && !skb->sw_hash) {
1378 struct flow_keys keys;
Tom Herbertde4c1f82015-09-01 18:11:04 -07001379 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001380
Tom Herbertde4c1f82015-09-01 18:11:04 -07001381 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001382 }
Tom Herbertf70ea012015-07-31 16:52:10 -07001383
1384 return skb->hash;
1385}
1386
Eric Dumazet55667442019-10-22 07:57:46 -07001387__u32 skb_get_hash_perturb(const struct sk_buff *skb,
1388 const siphash_key_t *perturb);
Tom Herbert50fb7992015-05-01 11:30:12 -07001389
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001390static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1391{
Tom Herbert61b905d2014-03-24 15:34:47 -07001392 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001393}
1394
Tom Herbert3df7a742013-12-15 22:16:29 -08001395static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1396{
Tom Herbert61b905d2014-03-24 15:34:47 -07001397 to->hash = from->hash;
Tom Herberta3b18dd2014-07-01 21:33:17 -07001398 to->sw_hash = from->sw_hash;
Tom Herbert61b905d2014-03-24 15:34:47 -07001399 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -08001400};
1401
Jakub Kicinski41477662019-08-07 17:03:59 -07001402static inline void skb_copy_decrypted(struct sk_buff *to,
1403 const struct sk_buff *from)
1404{
1405#ifdef CONFIG_TLS_DEVICE
1406 to->decrypted = from->decrypted;
1407#endif
1408}
1409
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001410#ifdef NET_SKBUFF_DATA_USES_OFFSET
1411static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1412{
1413 return skb->head + skb->end;
1414}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001415
1416static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1417{
1418 return skb->end;
1419}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001420#else
1421static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1422{
1423 return skb->end;
1424}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001425
1426static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1427{
1428 return skb->end - skb->head;
1429}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001430#endif
1431
Linus Torvalds1da177e2005-04-16 15:20:36 -07001432/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001433#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001434
Patrick Ohlyac45f602009-02-12 05:03:37 +00001435static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1436{
1437 return &skb_shinfo(skb)->hwtstamps;
1438}
1439
Willem de Bruijn52267792017-08-03 16:29:39 -04001440static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1441{
1442 bool is_zcopy = skb && skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY;
1443
1444 return is_zcopy ? skb_uarg(skb) : NULL;
1445}
1446
Willem de Bruijn52900d22018-11-30 15:32:40 -05001447static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1448 bool *have_ref)
Willem de Bruijn52267792017-08-03 16:29:39 -04001449{
1450 if (skb && uarg && !skb_zcopy(skb)) {
Willem de Bruijn52900d22018-11-30 15:32:40 -05001451 if (unlikely(have_ref && *have_ref))
1452 *have_ref = false;
1453 else
1454 sock_zerocopy_get(uarg);
Willem de Bruijn52267792017-08-03 16:29:39 -04001455 skb_shinfo(skb)->destructor_arg = uarg;
1456 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1457 }
1458}
1459
Willem de Bruijn5cd8d462018-11-20 13:00:18 -05001460static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1461{
1462 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1463 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1464}
1465
1466static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1467{
1468 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1469}
1470
1471static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1472{
1473 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1474}
1475
Willem de Bruijn52267792017-08-03 16:29:39 -04001476/* Release a reference on a zerocopy structure */
1477static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy)
1478{
1479 struct ubuf_info *uarg = skb_zcopy(skb);
1480
1481 if (uarg) {
Willem de Bruijn185ce5c2019-05-15 13:29:16 -04001482 if (skb_zcopy_is_nouarg(skb)) {
1483 /* no notification callback */
1484 } else if (uarg->callback == sock_zerocopy_callback) {
Willem de Bruijn0a4a0602017-08-09 19:09:44 -04001485 uarg->zerocopy = uarg->zerocopy && zerocopy;
1486 sock_zerocopy_put(uarg);
Willem de Bruijn185ce5c2019-05-15 13:29:16 -04001487 } else {
Willem de Bruijn0a4a0602017-08-09 19:09:44 -04001488 uarg->callback(uarg, zerocopy);
1489 }
1490
Willem de Bruijn52267792017-08-03 16:29:39 -04001491 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1492 }
1493}
1494
1495/* Abort a zerocopy operation and revert zckey on error in send syscall */
1496static inline void skb_zcopy_abort(struct sk_buff *skb)
1497{
1498 struct ubuf_info *uarg = skb_zcopy(skb);
1499
1500 if (uarg) {
Willem de Bruijn52900d22018-11-30 15:32:40 -05001501 sock_zerocopy_put_abort(uarg, false);
Willem de Bruijn52267792017-08-03 16:29:39 -04001502 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1503 }
1504}
1505
David S. Millera8305bf2018-07-29 20:42:53 -07001506static inline void skb_mark_not_on_list(struct sk_buff *skb)
1507{
1508 skb->next = NULL;
1509}
1510
Jason A. Donenfelddcfea722020-01-08 16:59:02 -05001511/* Iterate through singly-linked GSO fragments of an skb. */
Jason A. Donenfeld5eee7bd2020-01-13 18:42:26 -05001512#define skb_list_walk_safe(first, skb, next_skb) \
1513 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \
1514 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL)
Jason A. Donenfelddcfea722020-01-08 16:59:02 -05001515
David S. Miller992cba72018-07-31 15:27:56 -07001516static inline void skb_list_del_init(struct sk_buff *skb)
1517{
1518 __list_del_entry(&skb->list);
1519 skb_mark_not_on_list(skb);
1520}
1521
Linus Torvalds1da177e2005-04-16 15:20:36 -07001522/**
1523 * skb_queue_empty - check if a queue is empty
1524 * @list: queue head
1525 *
1526 * Returns true if the queue is empty, false otherwise.
1527 */
1528static inline int skb_queue_empty(const struct sk_buff_head *list)
1529{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001530 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001531}
1532
1533/**
Eric Dumazetd7d16a82019-10-23 22:44:48 -07001534 * skb_queue_empty_lockless - check if a queue is empty
1535 * @list: queue head
1536 *
1537 * Returns true if the queue is empty, false otherwise.
1538 * This variant can be used in lockless contexts.
1539 */
1540static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1541{
1542 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1543}
1544
1545
1546/**
David S. Millerfc7ebb22008-09-23 00:34:07 -07001547 * skb_queue_is_last - check if skb is the last entry in the queue
1548 * @list: queue head
1549 * @skb: buffer
1550 *
1551 * Returns true if @skb is the last buffer on the list.
1552 */
1553static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1554 const struct sk_buff *skb)
1555{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001556 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -07001557}
1558
1559/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001560 * skb_queue_is_first - check if skb is the first entry in the queue
1561 * @list: queue head
1562 * @skb: buffer
1563 *
1564 * Returns true if @skb is the first buffer on the list.
1565 */
1566static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1567 const struct sk_buff *skb)
1568{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001569 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001570}
1571
1572/**
David S. Miller249c8b42008-09-23 00:44:42 -07001573 * skb_queue_next - return the next packet in the queue
1574 * @list: queue head
1575 * @skb: current buffer
1576 *
1577 * Return the next packet in @list after @skb. It is only valid to
1578 * call this if skb_queue_is_last() evaluates to false.
1579 */
1580static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1581 const struct sk_buff *skb)
1582{
1583 /* This BUG_ON may seem severe, but if we just return then we
1584 * are going to dereference garbage.
1585 */
1586 BUG_ON(skb_queue_is_last(list, skb));
1587 return skb->next;
1588}
1589
1590/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001591 * skb_queue_prev - return the prev packet in the queue
1592 * @list: queue head
1593 * @skb: current buffer
1594 *
1595 * Return the prev packet in @list before @skb. It is only valid to
1596 * call this if skb_queue_is_first() evaluates to false.
1597 */
1598static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1599 const struct sk_buff *skb)
1600{
1601 /* This BUG_ON may seem severe, but if we just return then we
1602 * are going to dereference garbage.
1603 */
1604 BUG_ON(skb_queue_is_first(list, skb));
1605 return skb->prev;
1606}
1607
1608/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001609 * skb_get - reference buffer
1610 * @skb: buffer to reference
1611 *
1612 * Makes another reference to a socket buffer and returns a pointer
1613 * to the buffer.
1614 */
1615static inline struct sk_buff *skb_get(struct sk_buff *skb)
1616{
Reshetova, Elena63354792017-06-30 13:07:58 +03001617 refcount_inc(&skb->users);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001618 return skb;
1619}
1620
1621/*
Geert Uytterhoevenf8821f92017-11-30 14:33:56 +01001622 * If users == 1, we are the only owner and can avoid redundant atomic changes.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001623 */
1624
1625/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001626 * skb_cloned - is the buffer a clone
1627 * @skb: buffer to check
1628 *
1629 * Returns true if the buffer was generated with skb_clone() and is
1630 * one of multiple shared copies of the buffer. Cloned buffers are
1631 * shared data so must not be written to under normal circumstances.
1632 */
1633static inline int skb_cloned(const struct sk_buff *skb)
1634{
1635 return skb->cloned &&
1636 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1637}
1638
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001639static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1640{
Mel Gormand0164ad2015-11-06 16:28:21 -08001641 might_sleep_if(gfpflags_allow_blocking(pri));
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001642
1643 if (skb_cloned(skb))
1644 return pskb_expand_head(skb, 0, 0, pri);
1645
1646 return 0;
1647}
1648
Linus Torvalds1da177e2005-04-16 15:20:36 -07001649/**
1650 * skb_header_cloned - is the header a clone
1651 * @skb: buffer to check
1652 *
1653 * Returns true if modifying the header part of the buffer requires
1654 * the data to be copied.
1655 */
1656static inline int skb_header_cloned(const struct sk_buff *skb)
1657{
1658 int dataref;
1659
1660 if (!skb->cloned)
1661 return 0;
1662
1663 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1664 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1665 return dataref != 1;
1666}
1667
Eric Dumazet9580bf22016-04-30 10:19:29 -07001668static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1669{
1670 might_sleep_if(gfpflags_allow_blocking(pri));
1671
1672 if (skb_header_cloned(skb))
1673 return pskb_expand_head(skb, 0, 0, pri);
1674
1675 return 0;
1676}
1677
Linus Torvalds1da177e2005-04-16 15:20:36 -07001678/**
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001679 * __skb_header_release - release reference to header
1680 * @skb: buffer to operate on
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001681 */
1682static inline void __skb_header_release(struct sk_buff *skb)
1683{
1684 skb->nohdr = 1;
1685 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1686}
1687
1688
1689/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001690 * skb_shared - is the buffer shared
1691 * @skb: buffer to check
1692 *
1693 * Returns true if more than one person has a reference to this
1694 * buffer.
1695 */
1696static inline int skb_shared(const struct sk_buff *skb)
1697{
Reshetova, Elena63354792017-06-30 13:07:58 +03001698 return refcount_read(&skb->users) != 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001699}
1700
1701/**
1702 * skb_share_check - check if buffer is shared and if so clone it
1703 * @skb: buffer to check
1704 * @pri: priority for memory allocation
1705 *
1706 * If the buffer is shared the buffer is cloned and the old copy
1707 * drops a reference. A new clone with a single reference is returned.
1708 * If the buffer is not shared the original buffer is returned. When
1709 * being called from interrupt status or with spinlocks held pri must
1710 * be GFP_ATOMIC.
1711 *
1712 * NULL is returned on a memory allocation failure.
1713 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001714static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001715{
Mel Gormand0164ad2015-11-06 16:28:21 -08001716 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001717 if (skb_shared(skb)) {
1718 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001719
1720 if (likely(nskb))
1721 consume_skb(skb);
1722 else
1723 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001724 skb = nskb;
1725 }
1726 return skb;
1727}
1728
1729/*
1730 * Copy shared buffers into a new sk_buff. We effectively do COW on
1731 * packets to handle cases where we have a local reader and forward
1732 * and a couple of other messy ones. The normal one is tcpdumping
1733 * a packet thats being forwarded.
1734 */
1735
1736/**
1737 * skb_unshare - make a copy of a shared buffer
1738 * @skb: buffer to check
1739 * @pri: priority for memory allocation
1740 *
1741 * If the socket buffer is a clone then this function creates a new
1742 * copy of the data, drops a reference count on the old copy and returns
1743 * the new copy with the reference count at 1. If the buffer is not a clone
1744 * the original buffer is returned. When called with a spinlock held or
1745 * from interrupt state @pri must be %GFP_ATOMIC
1746 *
1747 * %NULL is returned on a memory allocation failure.
1748 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001749static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001750 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001751{
Mel Gormand0164ad2015-11-06 16:28:21 -08001752 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001753 if (skb_cloned(skb)) {
1754 struct sk_buff *nskb = skb_copy(skb, pri);
Alexander Aring31eff812014-10-10 23:10:47 +02001755
1756 /* Free our shared copy */
1757 if (likely(nskb))
1758 consume_skb(skb);
1759 else
1760 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001761 skb = nskb;
1762 }
1763 return skb;
1764}
1765
1766/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001767 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001768 * @list_: list to peek at
1769 *
1770 * Peek an &sk_buff. Unlike most other operations you _MUST_
1771 * be careful with this one. A peek leaves the buffer on the
1772 * list and someone else may run off with it. You must hold
1773 * the appropriate locks or have a private queue to do this.
1774 *
1775 * Returns %NULL for an empty list or a pointer to the head element.
1776 * The reference count is not incremented and the reference is therefore
1777 * volatile. Use with caution.
1778 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001779static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001780{
Eric Dumazet18d07002012-04-30 16:31:46 +00001781 struct sk_buff *skb = list_->next;
1782
1783 if (skb == (struct sk_buff *)list_)
1784 skb = NULL;
1785 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001786}
1787
1788/**
David S. Miller8b69bd72018-08-11 18:43:38 -07001789 * __skb_peek - peek at the head of a non-empty &sk_buff_head
1790 * @list_: list to peek at
1791 *
1792 * Like skb_peek(), but the caller knows that the list is not empty.
1793 */
1794static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
1795{
1796 return list_->next;
1797}
1798
1799/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001800 * skb_peek_next - peek skb following the given one from a queue
1801 * @skb: skb to start from
1802 * @list_: list to peek at
1803 *
1804 * Returns %NULL when the end of the list is met or a pointer to the
1805 * next element. The reference count is not incremented and the
1806 * reference is therefore volatile. Use with caution.
1807 */
1808static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1809 const struct sk_buff_head *list_)
1810{
1811 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001812
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001813 if (next == (struct sk_buff *)list_)
1814 next = NULL;
1815 return next;
1816}
1817
1818/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001819 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001820 * @list_: list to peek at
1821 *
1822 * Peek an &sk_buff. Unlike most other operations you _MUST_
1823 * be careful with this one. A peek leaves the buffer on the
1824 * list and someone else may run off with it. You must hold
1825 * the appropriate locks or have a private queue to do this.
1826 *
1827 * Returns %NULL for an empty list or a pointer to the tail element.
1828 * The reference count is not incremented and the reference is therefore
1829 * volatile. Use with caution.
1830 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001831static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001832{
Eric Dumazetf8cc62c2019-11-07 18:49:43 -08001833 struct sk_buff *skb = READ_ONCE(list_->prev);
Eric Dumazet18d07002012-04-30 16:31:46 +00001834
1835 if (skb == (struct sk_buff *)list_)
1836 skb = NULL;
1837 return skb;
1838
Linus Torvalds1da177e2005-04-16 15:20:36 -07001839}
1840
1841/**
1842 * skb_queue_len - get queue length
1843 * @list_: list to measure
1844 *
1845 * Return the length of an &sk_buff queue.
1846 */
1847static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1848{
1849 return list_->qlen;
1850}
1851
David S. Miller67fed452008-09-21 22:36:24 -07001852/**
Qian Cai86b18aa2020-02-04 13:40:29 -05001853 * skb_queue_len_lockless - get queue length
1854 * @list_: list to measure
1855 *
1856 * Return the length of an &sk_buff queue.
1857 * This variant can be used in lockless contexts.
1858 */
1859static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_)
1860{
1861 return READ_ONCE(list_->qlen);
1862}
1863
1864/**
David S. Miller67fed452008-09-21 22:36:24 -07001865 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1866 * @list: queue to initialize
1867 *
1868 * This initializes only the list and queue length aspects of
1869 * an sk_buff_head object. This allows to initialize the list
1870 * aspects of an sk_buff_head without reinitializing things like
1871 * the spinlock. It can also be used for on-stack sk_buff_head
1872 * objects where the spinlock is known to not be used.
1873 */
1874static inline void __skb_queue_head_init(struct sk_buff_head *list)
1875{
1876 list->prev = list->next = (struct sk_buff *)list;
1877 list->qlen = 0;
1878}
1879
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001880/*
1881 * This function creates a split out lock class for each invocation;
1882 * this is needed for now since a whole lot of users of the skb-queue
1883 * infrastructure in drivers have different locking usage (in hardirq)
1884 * than the networking core (in softirq only). In the long run either the
1885 * network layer or drivers should need annotation to consolidate the
1886 * main types of usage into 3 classes.
1887 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001888static inline void skb_queue_head_init(struct sk_buff_head *list)
1889{
1890 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001891 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001892}
1893
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001894static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1895 struct lock_class_key *class)
1896{
1897 skb_queue_head_init(list);
1898 lockdep_set_class(&list->lock, class);
1899}
1900
Linus Torvalds1da177e2005-04-16 15:20:36 -07001901/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001902 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001903 *
1904 * The "__skb_xxxx()" functions are the non-atomic ones that
1905 * can only be called with interrupts disabled.
1906 */
Gerrit Renkerbf299272008-04-14 00:04:51 -07001907static inline void __skb_insert(struct sk_buff *newsk,
1908 struct sk_buff *prev, struct sk_buff *next,
1909 struct sk_buff_head *list)
1910{
Eric Dumazetf8cc62c2019-11-07 18:49:43 -08001911 /* See skb_queue_empty_lockless() and skb_peek_tail()
1912 * for the opposite READ_ONCE()
1913 */
Eric Dumazetd7d16a82019-10-23 22:44:48 -07001914 WRITE_ONCE(newsk->next, next);
1915 WRITE_ONCE(newsk->prev, prev);
1916 WRITE_ONCE(next->prev, newsk);
1917 WRITE_ONCE(prev->next, newsk);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001918 list->qlen++;
1919}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001920
David S. Miller67fed452008-09-21 22:36:24 -07001921static inline void __skb_queue_splice(const struct sk_buff_head *list,
1922 struct sk_buff *prev,
1923 struct sk_buff *next)
1924{
1925 struct sk_buff *first = list->next;
1926 struct sk_buff *last = list->prev;
1927
Eric Dumazetd7d16a82019-10-23 22:44:48 -07001928 WRITE_ONCE(first->prev, prev);
1929 WRITE_ONCE(prev->next, first);
David S. Miller67fed452008-09-21 22:36:24 -07001930
Eric Dumazetd7d16a82019-10-23 22:44:48 -07001931 WRITE_ONCE(last->next, next);
1932 WRITE_ONCE(next->prev, last);
David S. Miller67fed452008-09-21 22:36:24 -07001933}
1934
1935/**
1936 * skb_queue_splice - join two skb lists, this is designed for stacks
1937 * @list: the new list to add
1938 * @head: the place to add it in the first list
1939 */
1940static inline void skb_queue_splice(const struct sk_buff_head *list,
1941 struct sk_buff_head *head)
1942{
1943 if (!skb_queue_empty(list)) {
1944 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001945 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001946 }
1947}
1948
1949/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001950 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001951 * @list: the new list to add
1952 * @head: the place to add it in the first list
1953 *
1954 * The list at @list is reinitialised
1955 */
1956static inline void skb_queue_splice_init(struct sk_buff_head *list,
1957 struct sk_buff_head *head)
1958{
1959 if (!skb_queue_empty(list)) {
1960 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001961 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001962 __skb_queue_head_init(list);
1963 }
1964}
1965
1966/**
1967 * skb_queue_splice_tail - join two skb lists, each list being a queue
1968 * @list: the new list to add
1969 * @head: the place to add it in the first list
1970 */
1971static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1972 struct sk_buff_head *head)
1973{
1974 if (!skb_queue_empty(list)) {
1975 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001976 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001977 }
1978}
1979
1980/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001981 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001982 * @list: the new list to add
1983 * @head: the place to add it in the first list
1984 *
1985 * Each of the lists is a queue.
1986 * The list at @list is reinitialised
1987 */
1988static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1989 struct sk_buff_head *head)
1990{
1991 if (!skb_queue_empty(list)) {
1992 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001993 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001994 __skb_queue_head_init(list);
1995 }
1996}
1997
Linus Torvalds1da177e2005-04-16 15:20:36 -07001998/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001999 * __skb_queue_after - queue a buffer at the list head
2000 * @list: list to use
2001 * @prev: place after this buffer
2002 * @newsk: buffer to queue
2003 *
2004 * Queue a buffer int the middle of a list. This function takes no locks
2005 * and you must therefore hold required locks before calling it.
2006 *
2007 * A buffer cannot be placed on two lists at the same time.
2008 */
2009static inline void __skb_queue_after(struct sk_buff_head *list,
2010 struct sk_buff *prev,
2011 struct sk_buff *newsk)
2012{
Gerrit Renkerbf299272008-04-14 00:04:51 -07002013 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08002014}
2015
Joe Perches7965bd42013-09-26 14:48:15 -07002016void skb_append(struct sk_buff *old, struct sk_buff *newsk,
2017 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07002018
Gerrit Renkerf5572852008-04-14 00:05:28 -07002019static inline void __skb_queue_before(struct sk_buff_head *list,
2020 struct sk_buff *next,
2021 struct sk_buff *newsk)
2022{
2023 __skb_insert(newsk, next->prev, next, list);
2024}
2025
Stephen Hemminger300ce172005-10-30 13:47:34 -08002026/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002027 * __skb_queue_head - queue a buffer at the list head
2028 * @list: list to use
2029 * @newsk: buffer to queue
2030 *
2031 * Queue a buffer at the start of a list. This function takes no locks
2032 * and you must therefore hold required locks before calling it.
2033 *
2034 * A buffer cannot be placed on two lists at the same time.
2035 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002036static inline void __skb_queue_head(struct sk_buff_head *list,
2037 struct sk_buff *newsk)
2038{
Stephen Hemminger300ce172005-10-30 13:47:34 -08002039 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002040}
Brian Norris4ea7b0c2019-02-11 13:02:25 -08002041void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002042
2043/**
2044 * __skb_queue_tail - queue a buffer at the list tail
2045 * @list: list to use
2046 * @newsk: buffer to queue
2047 *
2048 * Queue a buffer at the end of a list. This function takes no locks
2049 * and you must therefore hold required locks before calling it.
2050 *
2051 * A buffer cannot be placed on two lists at the same time.
2052 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002053static inline void __skb_queue_tail(struct sk_buff_head *list,
2054 struct sk_buff *newsk)
2055{
Gerrit Renkerf5572852008-04-14 00:05:28 -07002056 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002057}
Brian Norris4ea7b0c2019-02-11 13:02:25 -08002058void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002059
Linus Torvalds1da177e2005-04-16 15:20:36 -07002060/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07002061 * remove sk_buff from list. _Must_ be called atomically, and with
2062 * the list known..
2063 */
Joe Perches7965bd42013-09-26 14:48:15 -07002064void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002065static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2066{
2067 struct sk_buff *next, *prev;
2068
Qian Cai86b18aa2020-02-04 13:40:29 -05002069 WRITE_ONCE(list->qlen, list->qlen - 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002070 next = skb->next;
2071 prev = skb->prev;
2072 skb->next = skb->prev = NULL;
Eric Dumazetd7d16a82019-10-23 22:44:48 -07002073 WRITE_ONCE(next->prev, prev);
2074 WRITE_ONCE(prev->next, next);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002075}
2076
Gerrit Renkerf525c062008-04-14 00:04:12 -07002077/**
2078 * __skb_dequeue - remove from the head of the queue
2079 * @list: list to dequeue from
2080 *
2081 * Remove the head of the list. This function does not take any locks
2082 * so must be used with appropriate locks held only. The head item is
2083 * returned or %NULL if the list is empty.
2084 */
Gerrit Renkerf525c062008-04-14 00:04:12 -07002085static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2086{
2087 struct sk_buff *skb = skb_peek(list);
2088 if (skb)
2089 __skb_unlink(skb, list);
2090 return skb;
2091}
Brian Norris4ea7b0c2019-02-11 13:02:25 -08002092struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002093
2094/**
2095 * __skb_dequeue_tail - remove from the tail of the queue
2096 * @list: list to dequeue from
2097 *
2098 * Remove the tail of the list. This function does not take any locks
2099 * so must be used with appropriate locks held only. The tail item is
2100 * returned or %NULL if the list is empty.
2101 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002102static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2103{
2104 struct sk_buff *skb = skb_peek_tail(list);
2105 if (skb)
2106 __skb_unlink(skb, list);
2107 return skb;
2108}
Brian Norris4ea7b0c2019-02-11 13:02:25 -08002109struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002110
2111
David S. Millerbdcc0922012-03-07 20:53:36 -05002112static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002113{
2114 return skb->data_len;
2115}
2116
2117static inline unsigned int skb_headlen(const struct sk_buff *skb)
2118{
2119 return skb->len - skb->data_len;
2120}
2121
Willem de Bruijn3ece7822017-08-03 16:29:38 -04002122static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002123{
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03002124 unsigned int i, len = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002125
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03002126 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00002127 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Willem de Bruijn3ece7822017-08-03 16:29:38 -04002128 return len;
2129}
2130
2131static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2132{
2133 return skb_headlen(skb) + __skb_pagelen(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002134}
2135
Ian Campbell131ea662011-08-19 06:25:00 +00002136/**
2137 * __skb_fill_page_desc - initialise a paged fragment in an skb
2138 * @skb: buffer containing fragment to be initialised
2139 * @i: paged fragment index to initialise
2140 * @page: the page to use for this fragment
2141 * @off: the offset to the data with @page
2142 * @size: the length of the data
2143 *
2144 * Initialises the @i'th fragment of @skb to point to &size bytes at
2145 * offset @off within @page.
2146 *
2147 * Does not take any additional reference on the fragment.
2148 */
2149static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2150 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002151{
2152 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2153
Mel Gormanc48a11c2012-07-31 16:44:23 -07002154 /*
Michal Hocko2f064f32015-08-21 14:11:51 -07002155 * Propagate page pfmemalloc to the skb if we can. The problem is
2156 * that not all callers have unique ownership of the page but rely
2157 * on page_is_pfmemalloc doing the right thing(tm).
Mel Gormanc48a11c2012-07-31 16:44:23 -07002158 */
Matthew Wilcox (Oracle)1dfa5bd2019-07-22 20:08:29 -07002159 frag->bv_page = page;
Jonathan Lemon65c84f12019-07-30 07:40:34 -07002160 frag->bv_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00002161 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00002162
2163 page = compound_head(page);
Michal Hocko2f064f32015-08-21 14:11:51 -07002164 if (page_is_pfmemalloc(page))
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00002165 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00002166}
2167
2168/**
2169 * skb_fill_page_desc - initialise a paged fragment in an skb
2170 * @skb: buffer containing fragment to be initialised
2171 * @i: paged fragment index to initialise
2172 * @page: the page to use for this fragment
2173 * @off: the offset to the data with @page
2174 * @size: the length of the data
2175 *
2176 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01002177 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00002178 * addition updates @skb such that @i is the last fragment.
2179 *
2180 * Does not take any additional reference on the fragment.
2181 */
2182static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2183 struct page *page, int off, int size)
2184{
2185 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002186 skb_shinfo(skb)->nr_frags = i + 1;
2187}
2188
Joe Perches7965bd42013-09-26 14:48:15 -07002189void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2190 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07002191
Jason Wangf8e617e2013-11-01 14:07:47 +08002192void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2193 unsigned int truesize);
2194
Linus Torvalds1da177e2005-04-16 15:20:36 -07002195#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2196
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002197#ifdef NET_SKBUFF_DATA_USES_OFFSET
2198static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2199{
2200 return skb->head + skb->tail;
2201}
2202
2203static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2204{
2205 skb->tail = skb->data - skb->head;
2206}
2207
2208static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2209{
2210 skb_reset_tail_pointer(skb);
2211 skb->tail += offset;
2212}
Simon Horman7cc46192013-05-28 20:34:29 +00002213
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002214#else /* NET_SKBUFF_DATA_USES_OFFSET */
2215static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2216{
2217 return skb->tail;
2218}
2219
2220static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2221{
2222 skb->tail = skb->data;
2223}
2224
2225static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2226{
2227 skb->tail = skb->data + offset;
2228}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07002229
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002230#endif /* NET_SKBUFF_DATA_USES_OFFSET */
2231
Linus Torvalds1da177e2005-04-16 15:20:36 -07002232/*
2233 * Add data to an sk_buff
2234 */
Johannes Berg4df864c2017-06-16 14:29:21 +02002235void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2236void *skb_put(struct sk_buff *skb, unsigned int len);
2237static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002238{
Johannes Berg4df864c2017-06-16 14:29:21 +02002239 void *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002240 SKB_LINEAR_ASSERT(skb);
2241 skb->tail += len;
2242 skb->len += len;
2243 return tmp;
2244}
2245
yuan linyude77b962017-06-18 22:48:17 +08002246static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2247{
2248 void *tmp = __skb_put(skb, len);
2249
2250 memset(tmp, 0, len);
2251 return tmp;
2252}
2253
2254static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2255 unsigned int len)
2256{
2257 void *tmp = __skb_put(skb, len);
2258
2259 memcpy(tmp, data, len);
2260 return tmp;
2261}
2262
2263static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2264{
2265 *(u8 *)__skb_put(skb, 1) = val;
2266}
2267
Johannes Berg83ad3572017-06-14 22:17:20 +02002268static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
Johannes Berge45a79d2017-05-24 09:07:47 +02002269{
Johannes Berg83ad3572017-06-14 22:17:20 +02002270 void *tmp = skb_put(skb, len);
Johannes Berge45a79d2017-05-24 09:07:47 +02002271
2272 memset(tmp, 0, len);
2273
2274 return tmp;
2275}
2276
Johannes Berg59ae1d12017-06-16 14:29:20 +02002277static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2278 unsigned int len)
2279{
2280 void *tmp = skb_put(skb, len);
2281
2282 memcpy(tmp, data, len);
2283
2284 return tmp;
2285}
2286
Johannes Berg634fef62017-06-16 14:29:24 +02002287static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2288{
2289 *(u8 *)skb_put(skb, 1) = val;
2290}
2291
Johannes Bergd58ff352017-06-16 14:29:23 +02002292void *skb_push(struct sk_buff *skb, unsigned int len);
2293static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002294{
2295 skb->data -= len;
2296 skb->len += len;
2297 return skb->data;
2298}
2299
Johannes Bergaf728682017-06-16 14:29:22 +02002300void *skb_pull(struct sk_buff *skb, unsigned int len);
2301static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002302{
2303 skb->len -= len;
2304 BUG_ON(skb->len < skb->data_len);
2305 return skb->data += len;
2306}
2307
Johannes Bergaf728682017-06-16 14:29:22 +02002308static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
David S. Miller47d29642010-05-02 02:21:44 -07002309{
2310 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2311}
2312
Johannes Bergaf728682017-06-16 14:29:22 +02002313void *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002314
Johannes Bergaf728682017-06-16 14:29:22 +02002315static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002316{
2317 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07002318 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002319 return NULL;
2320 skb->len -= len;
2321 return skb->data += len;
2322}
2323
Johannes Bergaf728682017-06-16 14:29:22 +02002324static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002325{
2326 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2327}
2328
Heiner Kallweitb9df4fd2019-10-06 18:19:54 +02002329static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002330{
2331 if (likely(len <= skb_headlen(skb)))
Heiner Kallweitb9df4fd2019-10-06 18:19:54 +02002332 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002333 if (unlikely(len > skb->len))
Heiner Kallweitb9df4fd2019-10-06 18:19:54 +02002334 return false;
Gerrit Renker987c4022008-08-11 18:17:17 -07002335 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002336}
2337
Eric Dumazetc8c8b122016-12-07 09:19:33 -08002338void skb_condense(struct sk_buff *skb);
2339
Linus Torvalds1da177e2005-04-16 15:20:36 -07002340/**
2341 * skb_headroom - bytes at buffer head
2342 * @skb: buffer to check
2343 *
2344 * Return the number of bytes of free space at the head of an &sk_buff.
2345 */
Chuck Leverc2636b42007-10-23 21:07:32 -07002346static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002347{
2348 return skb->data - skb->head;
2349}
2350
2351/**
2352 * skb_tailroom - bytes at buffer end
2353 * @skb: buffer to check
2354 *
2355 * Return the number of bytes of free space at the tail of an sk_buff
2356 */
2357static inline int skb_tailroom(const struct sk_buff *skb)
2358{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07002359 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002360}
2361
2362/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00002363 * skb_availroom - bytes at buffer end
2364 * @skb: buffer to check
2365 *
2366 * Return the number of bytes of free space at the tail of an sk_buff
2367 * allocated by sk_stream_alloc()
2368 */
2369static inline int skb_availroom(const struct sk_buff *skb)
2370{
Eric Dumazet16fad692013-03-14 05:40:32 +00002371 if (skb_is_nonlinear(skb))
2372 return 0;
2373
2374 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00002375}
2376
2377/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002378 * skb_reserve - adjust headroom
2379 * @skb: buffer to alter
2380 * @len: bytes to move
2381 *
2382 * Increase the headroom of an empty &sk_buff by reducing the tail
2383 * room. This is only allowed for an empty buffer.
2384 */
David S. Miller8243126c2006-01-17 02:54:21 -08002385static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002386{
2387 skb->data += len;
2388 skb->tail += len;
2389}
2390
Benjamin Poirier1837b2e2016-02-29 15:03:33 -08002391/**
2392 * skb_tailroom_reserve - adjust reserved_tailroom
2393 * @skb: buffer to alter
2394 * @mtu: maximum amount of headlen permitted
2395 * @needed_tailroom: minimum amount of reserved_tailroom
2396 *
2397 * Set reserved_tailroom so that headlen can be as large as possible but
2398 * not larger than mtu and tailroom cannot be smaller than
2399 * needed_tailroom.
2400 * The required headroom should already have been reserved before using
2401 * this function.
2402 */
2403static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2404 unsigned int needed_tailroom)
2405{
2406 SKB_LINEAR_ASSERT(skb);
2407 if (mtu < skb_tailroom(skb) - needed_tailroom)
2408 /* use at most mtu */
2409 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2410 else
2411 /* use up to all available space */
2412 skb->reserved_tailroom = needed_tailroom;
2413}
2414
Tom Herbert8bce6d72014-09-29 20:22:29 -07002415#define ENCAP_TYPE_ETHER 0
2416#define ENCAP_TYPE_IPPROTO 1
2417
2418static inline void skb_set_inner_protocol(struct sk_buff *skb,
2419 __be16 protocol)
2420{
2421 skb->inner_protocol = protocol;
2422 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2423}
2424
2425static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2426 __u8 ipproto)
2427{
2428 skb->inner_ipproto = ipproto;
2429 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2430}
2431
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002432static inline void skb_reset_inner_headers(struct sk_buff *skb)
2433{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002434 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002435 skb->inner_network_header = skb->network_header;
2436 skb->inner_transport_header = skb->transport_header;
2437}
2438
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00002439static inline void skb_reset_mac_len(struct sk_buff *skb)
2440{
2441 skb->mac_len = skb->network_header - skb->mac_header;
2442}
2443
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002444static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2445 *skb)
2446{
2447 return skb->head + skb->inner_transport_header;
2448}
2449
Tom Herbert55dc5a92015-12-14 11:19:40 -08002450static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2451{
2452 return skb_inner_transport_header(skb) - skb->data;
2453}
2454
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002455static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2456{
2457 skb->inner_transport_header = skb->data - skb->head;
2458}
2459
2460static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2461 const int offset)
2462{
2463 skb_reset_inner_transport_header(skb);
2464 skb->inner_transport_header += offset;
2465}
2466
2467static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2468{
2469 return skb->head + skb->inner_network_header;
2470}
2471
2472static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2473{
2474 skb->inner_network_header = skb->data - skb->head;
2475}
2476
2477static inline void skb_set_inner_network_header(struct sk_buff *skb,
2478 const int offset)
2479{
2480 skb_reset_inner_network_header(skb);
2481 skb->inner_network_header += offset;
2482}
2483
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002484static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2485{
2486 return skb->head + skb->inner_mac_header;
2487}
2488
2489static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2490{
2491 skb->inner_mac_header = skb->data - skb->head;
2492}
2493
2494static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2495 const int offset)
2496{
2497 skb_reset_inner_mac_header(skb);
2498 skb->inner_mac_header += offset;
2499}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002500static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2501{
Cong Wang35d04612013-05-29 15:16:05 +08002502 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002503}
2504
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002505static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2506{
2507 return skb->head + skb->transport_header;
2508}
2509
2510static inline void skb_reset_transport_header(struct sk_buff *skb)
2511{
2512 skb->transport_header = skb->data - skb->head;
2513}
2514
2515static inline void skb_set_transport_header(struct sk_buff *skb,
2516 const int offset)
2517{
2518 skb_reset_transport_header(skb);
2519 skb->transport_header += offset;
2520}
2521
2522static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2523{
2524 return skb->head + skb->network_header;
2525}
2526
2527static inline void skb_reset_network_header(struct sk_buff *skb)
2528{
2529 skb->network_header = skb->data - skb->head;
2530}
2531
2532static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2533{
2534 skb_reset_network_header(skb);
2535 skb->network_header += offset;
2536}
2537
2538static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2539{
2540 return skb->head + skb->mac_header;
2541}
2542
Amir Vadaiea6da4f2017-02-07 09:56:06 +02002543static inline int skb_mac_offset(const struct sk_buff *skb)
2544{
2545 return skb_mac_header(skb) - skb->data;
2546}
2547
Daniel Borkmann0daf4342017-07-02 02:13:25 +02002548static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2549{
2550 return skb->network_header - skb->mac_header;
2551}
2552
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002553static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2554{
Cong Wang35d04612013-05-29 15:16:05 +08002555 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002556}
2557
Daniel Borkmannb4ab3142020-09-30 17:18:17 +02002558static inline void skb_unset_mac_header(struct sk_buff *skb)
2559{
2560 skb->mac_header = (typeof(skb->mac_header))~0U;
2561}
2562
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002563static inline void skb_reset_mac_header(struct sk_buff *skb)
2564{
2565 skb->mac_header = skb->data - skb->head;
2566}
2567
2568static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2569{
2570 skb_reset_mac_header(skb);
2571 skb->mac_header += offset;
2572}
2573
Timo Teräs0e3da5b2013-12-16 11:02:09 +02002574static inline void skb_pop_mac_header(struct sk_buff *skb)
2575{
2576 skb->mac_header = skb->network_header;
2577}
2578
Maxim Mikityanskiyd2aa1252019-02-21 12:39:57 +00002579static inline void skb_probe_transport_header(struct sk_buff *skb)
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002580{
Paolo Abeni72a338b2018-05-04 11:32:59 +02002581 struct flow_keys_basic keys;
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002582
2583 if (skb_transport_header_was_set(skb))
2584 return;
Paolo Abeni72a338b2018-05-04 11:32:59 +02002585
Stanislav Fomichev3cbf4ff2019-04-22 08:55:46 -07002586 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2587 NULL, 0, 0, 0, 0))
Tom Herbert42aecaa2015-06-04 09:16:39 -07002588 skb_set_transport_header(skb, keys.control.thoff);
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002589}
2590
Eric Dumazet03606892012-02-23 10:55:02 +00002591static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2592{
2593 if (skb_mac_header_was_set(skb)) {
2594 const unsigned char *old_mac = skb_mac_header(skb);
2595
2596 skb_set_mac_header(skb, -skb->mac_len);
2597 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2598 }
2599}
2600
Michał Mirosław04fb4512010-12-14 15:24:08 +00002601static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2602{
2603 return skb->csum_start - skb_headroom(skb);
2604}
2605
Alexander Duyck08b64fc2016-02-05 15:27:49 -08002606static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2607{
2608 return skb->head + skb->csum_start;
2609}
2610
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002611static inline int skb_transport_offset(const struct sk_buff *skb)
2612{
2613 return skb_transport_header(skb) - skb->data;
2614}
2615
2616static inline u32 skb_network_header_len(const struct sk_buff *skb)
2617{
2618 return skb->transport_header - skb->network_header;
2619}
2620
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002621static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2622{
2623 return skb->inner_transport_header - skb->inner_network_header;
2624}
2625
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002626static inline int skb_network_offset(const struct sk_buff *skb)
2627{
2628 return skb_network_header(skb) - skb->data;
2629}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03002630
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002631static inline int skb_inner_network_offset(const struct sk_buff *skb)
2632{
2633 return skb_inner_network_header(skb) - skb->data;
2634}
2635
Changli Gaof9599ce2010-08-04 04:43:44 +00002636static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2637{
2638 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2639}
2640
Linus Torvalds1da177e2005-04-16 15:20:36 -07002641/*
2642 * CPUs often take a performance hit when accessing unaligned memory
2643 * locations. The actual performance hit varies, it can be small if the
2644 * hardware handles it or large if we have to take an exception and fix it
2645 * in software.
2646 *
2647 * Since an ethernet header is 14 bytes network drivers often end up with
2648 * the IP header at an unaligned offset. The IP header can be aligned by
2649 * shifting the start of the packet by 2 bytes. Drivers should do this
2650 * with:
2651 *
Tobias Klauser8660c122009-07-13 22:48:16 +00002652 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002653 *
2654 * The downside to this alignment of the IP header is that the DMA is now
2655 * unaligned. On some architectures the cost of an unaligned DMA is high
2656 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00002657 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002658 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2659 * to be overridden.
2660 */
2661#ifndef NET_IP_ALIGN
2662#define NET_IP_ALIGN 2
2663#endif
2664
Anton Blanchard025be812006-03-31 02:27:06 -08002665/*
2666 * The networking layer reserves some headroom in skb data (via
2667 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2668 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08002669 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08002670 *
2671 * Unfortunately this headroom changes the DMA alignment of the resulting
2672 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2673 * on some architectures. An architecture can override this value,
2674 * perhaps setting it to a cacheline in size (since that will maintain
2675 * cacheline alignment of the DMA). It must be a power of 2.
2676 *
David S. Millerd6301d32009-02-08 19:24:13 -08002677 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08002678 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07002679 *
2680 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2681 * to reduce average number of cache lines per packet.
Miaohe Lin645f0892020-08-27 07:27:49 -04002682 * get_rps_cpu() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07002683 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08002684 */
2685#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07002686#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08002687#endif
2688
Joe Perches7965bd42013-09-26 14:48:15 -07002689int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002690
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002691static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002692{
Yangtao Li5e1abdc2018-11-06 10:45:36 -05002693 if (WARN_ON(skb_is_nonlinear(skb)))
Herbert Xu3cc0e872006-06-09 16:13:38 -07002694 return;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002695 skb->len = len;
2696 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002697}
2698
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002699static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2700{
2701 __skb_set_length(skb, len);
2702}
2703
Joe Perches7965bd42013-09-26 14:48:15 -07002704void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002705
2706static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2707{
Herbert Xu3cc0e872006-06-09 16:13:38 -07002708 if (skb->data_len)
2709 return ___pskb_trim(skb, len);
2710 __skb_trim(skb, len);
2711 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002712}
2713
2714static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2715{
2716 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2717}
2718
2719/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07002720 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2721 * @skb: buffer to alter
2722 * @len: new length
2723 *
2724 * This is identical to pskb_trim except that the caller knows that
2725 * the skb is not cloned so we should never get an error due to out-
2726 * of-memory.
2727 */
2728static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2729{
2730 int err = pskb_trim(skb, len);
2731 BUG_ON(err);
2732}
2733
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002734static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2735{
2736 unsigned int diff = len - skb->len;
2737
2738 if (skb_tailroom(skb) < diff) {
2739 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2740 GFP_ATOMIC);
2741 if (ret)
2742 return ret;
2743 }
2744 __skb_set_length(skb, len);
2745 return 0;
2746}
2747
Herbert Xue9fa4f72006-08-13 20:12:58 -07002748/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002749 * skb_orphan - orphan a buffer
2750 * @skb: buffer to orphan
2751 *
2752 * If a buffer currently has an owner then we call the owner's
2753 * destructor function and make the @skb unowned. The buffer continues
2754 * to exist but is no longer charged to its former owner.
2755 */
2756static inline void skb_orphan(struct sk_buff *skb)
2757{
Eric Dumazetc34a7612013-07-30 16:11:15 -07002758 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002759 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002760 skb->destructor = NULL;
2761 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07002762 } else {
2763 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002764 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002765}
2766
2767/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002768 * skb_orphan_frags - orphan the frags contained in a buffer
2769 * @skb: buffer to orphan frags from
2770 * @gfp_mask: allocation mask for replacement pages
2771 *
2772 * For each frag in the SKB which needs a destructor (i.e. has an
2773 * owner) create a copy of that frag and release the original
2774 * page by calling the destructor.
2775 */
2776static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2777{
Willem de Bruijn1f8b9772017-08-03 16:29:41 -04002778 if (likely(!skb_zcopy(skb)))
2779 return 0;
Willem de Bruijn185ce5c2019-05-15 13:29:16 -04002780 if (!skb_zcopy_is_nouarg(skb) &&
2781 skb_uarg(skb)->callback == sock_zerocopy_callback)
Willem de Bruijn1f8b9772017-08-03 16:29:41 -04002782 return 0;
2783 return skb_copy_ubufs(skb, gfp_mask);
2784}
2785
2786/* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2787static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2788{
2789 if (likely(!skb_zcopy(skb)))
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002790 return 0;
2791 return skb_copy_ubufs(skb, gfp_mask);
2792}
2793
2794/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002795 * __skb_queue_purge - empty a list
2796 * @list: list to empty
2797 *
2798 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2799 * the list and one reference dropped. This function does not take the
2800 * list lock and the caller must hold the relevant locks to use it.
2801 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002802static inline void __skb_queue_purge(struct sk_buff_head *list)
2803{
2804 struct sk_buff *skb;
2805 while ((skb = __skb_dequeue(list)) != NULL)
2806 kfree_skb(skb);
2807}
Brian Norris4ea7b0c2019-02-11 13:02:25 -08002808void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002809
Peter Oskolkov385114d2018-08-02 23:34:38 +00002810unsigned int skb_rbtree_purge(struct rb_root *root);
Yaogong Wang9f5afea2016-09-07 14:49:28 -07002811
Joe Perches7965bd42013-09-26 14:48:15 -07002812void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002813
Joe Perches7965bd42013-09-26 14:48:15 -07002814struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2815 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07002816
2817/**
2818 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2819 * @dev: network device to receive on
2820 * @length: length to allocate
2821 *
2822 * Allocate a new &sk_buff and assign it a usage count of one. The
2823 * buffer has unspecified headroom built in. Users should allocate
2824 * the headroom they think they need without accounting for the
2825 * built in space. The built in space is used for optimisations.
2826 *
2827 * %NULL is returned if there is no free memory. Although this function
2828 * allocates memory it can be called from an interrupt.
2829 */
2830static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002831 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002832{
2833 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2834}
2835
Eric Dumazet6f532612012-05-18 05:12:12 +00002836/* legacy helper around __netdev_alloc_skb() */
2837static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2838 gfp_t gfp_mask)
2839{
2840 return __netdev_alloc_skb(NULL, length, gfp_mask);
2841}
2842
2843/* legacy helper around netdev_alloc_skb() */
2844static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2845{
2846 return netdev_alloc_skb(NULL, length);
2847}
2848
2849
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002850static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2851 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002852{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002853 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002854
2855 if (NET_IP_ALIGN && skb)
2856 skb_reserve(skb, NET_IP_ALIGN);
2857 return skb;
2858}
2859
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002860static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2861 unsigned int length)
2862{
2863 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2864}
2865
Alexander Duyck181edb22015-05-06 21:12:03 -07002866static inline void skb_free_frag(void *addr)
2867{
Alexander Duyck8c2dd3e2017-01-10 16:58:06 -08002868 page_frag_free(addr);
Alexander Duyck181edb22015-05-06 21:12:03 -07002869}
2870
Alexander Duyckffde7322014-12-09 19:40:42 -08002871void *napi_alloc_frag(unsigned int fragsz);
Alexander Duyckfd11a832014-12-09 19:40:49 -08002872struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2873 unsigned int length, gfp_t gfp_mask);
2874static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2875 unsigned int length)
2876{
2877 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2878}
Jesper Dangaard Brouer795bb1c2016-02-08 13:14:59 +01002879void napi_consume_skb(struct sk_buff *skb, int budget);
2880
2881void __kfree_skb_flush(void);
Jesper Dangaard Brouer15fad712016-02-08 13:15:04 +01002882void __kfree_skb_defer(struct sk_buff *skb);
Alexander Duyckffde7322014-12-09 19:40:42 -08002883
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002884/**
Alexander Duyck71dfda52014-11-11 09:26:34 -08002885 * __dev_alloc_pages - allocate page for network Rx
2886 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2887 * @order: size of the allocation
2888 *
2889 * Allocate a new page.
2890 *
2891 * %NULL is returned if there is no free memory.
2892*/
2893static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2894 unsigned int order)
2895{
2896 /* This piece of code contains several assumptions.
2897 * 1. This is for device Rx, therefor a cold page is preferred.
2898 * 2. The expectation is the user wants a compound page.
2899 * 3. If requesting a order 0 page it will not be compound
2900 * due to the check to see if order has a value in prep_new_page
2901 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2902 * code in gfp_to_alloc_flags that should be enforcing this.
2903 */
Mel Gorman453f85d2017-11-15 17:38:03 -08002904 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
Alexander Duyck71dfda52014-11-11 09:26:34 -08002905
2906 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2907}
2908
2909static inline struct page *dev_alloc_pages(unsigned int order)
2910{
Neil Horman95829b32016-05-19 11:30:54 -04002911 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002912}
2913
2914/**
2915 * __dev_alloc_page - allocate a page for network Rx
2916 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2917 *
2918 * Allocate a new page.
2919 *
2920 * %NULL is returned if there is no free memory.
2921 */
2922static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2923{
2924 return __dev_alloc_pages(gfp_mask, 0);
2925}
2926
2927static inline struct page *dev_alloc_page(void)
2928{
Neil Horman95829b32016-05-19 11:30:54 -04002929 return dev_alloc_pages(0);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002930}
2931
2932/**
Mel Gorman06140022012-07-31 16:44:24 -07002933 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2934 * @page: The page that was allocated from skb_alloc_page
2935 * @skb: The skb that may need pfmemalloc set
2936 */
2937static inline void skb_propagate_pfmemalloc(struct page *page,
2938 struct sk_buff *skb)
2939{
Michal Hocko2f064f32015-08-21 14:11:51 -07002940 if (page_is_pfmemalloc(page))
Mel Gorman06140022012-07-31 16:44:24 -07002941 skb->pfmemalloc = true;
2942}
2943
Eric Dumazet564824b2010-10-11 19:05:25 +00002944/**
Jonathan Lemon7240b602019-07-30 07:40:32 -07002945 * skb_frag_off() - Returns the offset of a skb fragment
2946 * @frag: the paged fragment
2947 */
2948static inline unsigned int skb_frag_off(const skb_frag_t *frag)
2949{
Jonathan Lemon65c84f12019-07-30 07:40:34 -07002950 return frag->bv_offset;
Jonathan Lemon7240b602019-07-30 07:40:32 -07002951}
2952
2953/**
2954 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
2955 * @frag: skb fragment
2956 * @delta: value to add
2957 */
2958static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
2959{
Jonathan Lemon65c84f12019-07-30 07:40:34 -07002960 frag->bv_offset += delta;
Jonathan Lemon7240b602019-07-30 07:40:32 -07002961}
2962
2963/**
2964 * skb_frag_off_set() - Sets the offset of a skb fragment
2965 * @frag: skb fragment
2966 * @offset: offset of fragment
2967 */
2968static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
2969{
Jonathan Lemon65c84f12019-07-30 07:40:34 -07002970 frag->bv_offset = offset;
Jonathan Lemon7240b602019-07-30 07:40:32 -07002971}
2972
2973/**
2974 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
2975 * @fragto: skb fragment where offset is set
2976 * @fragfrom: skb fragment offset is copied from
2977 */
2978static inline void skb_frag_off_copy(skb_frag_t *fragto,
2979 const skb_frag_t *fragfrom)
2980{
Jonathan Lemon65c84f12019-07-30 07:40:34 -07002981 fragto->bv_offset = fragfrom->bv_offset;
Jonathan Lemon7240b602019-07-30 07:40:32 -07002982}
2983
2984/**
Masanari Iidae2278672014-02-18 22:54:36 +09002985 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002986 * @frag: the paged fragment
2987 *
2988 * Returns the &struct page associated with @frag.
2989 */
2990static inline struct page *skb_frag_page(const skb_frag_t *frag)
2991{
Matthew Wilcox (Oracle)1dfa5bd2019-07-22 20:08:29 -07002992 return frag->bv_page;
Ian Campbell131ea662011-08-19 06:25:00 +00002993}
2994
2995/**
2996 * __skb_frag_ref - take an addition reference on a paged fragment.
2997 * @frag: the paged fragment
2998 *
2999 * Takes an additional reference on the paged fragment @frag.
3000 */
3001static inline void __skb_frag_ref(skb_frag_t *frag)
3002{
3003 get_page(skb_frag_page(frag));
3004}
3005
3006/**
3007 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
3008 * @skb: the buffer
3009 * @f: the fragment offset.
3010 *
3011 * Takes an additional reference on the @f'th paged fragment of @skb.
3012 */
3013static inline void skb_frag_ref(struct sk_buff *skb, int f)
3014{
3015 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
3016}
3017
3018/**
3019 * __skb_frag_unref - release a reference on a paged fragment.
3020 * @frag: the paged fragment
3021 *
3022 * Releases a reference on the paged fragment @frag.
3023 */
3024static inline void __skb_frag_unref(skb_frag_t *frag)
3025{
3026 put_page(skb_frag_page(frag));
3027}
3028
3029/**
3030 * skb_frag_unref - release a reference on a paged fragment of an skb.
3031 * @skb: the buffer
3032 * @f: the fragment offset
3033 *
3034 * Releases a reference on the @f'th paged fragment of @skb.
3035 */
3036static inline void skb_frag_unref(struct sk_buff *skb, int f)
3037{
3038 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
3039}
3040
3041/**
3042 * skb_frag_address - gets the address of the data contained in a paged fragment
3043 * @frag: the paged fragment buffer
3044 *
3045 * Returns the address of the data within @frag. The page must already
3046 * be mapped.
3047 */
3048static inline void *skb_frag_address(const skb_frag_t *frag)
3049{
Jonathan Lemon7240b602019-07-30 07:40:32 -07003050 return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
Ian Campbell131ea662011-08-19 06:25:00 +00003051}
3052
3053/**
3054 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3055 * @frag: the paged fragment buffer
3056 *
3057 * Returns the address of the data within @frag. Checks that the page
3058 * is mapped and returns %NULL otherwise.
3059 */
3060static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3061{
3062 void *ptr = page_address(skb_frag_page(frag));
3063 if (unlikely(!ptr))
3064 return NULL;
3065
Jonathan Lemon7240b602019-07-30 07:40:32 -07003066 return ptr + skb_frag_off(frag);
3067}
3068
3069/**
3070 * skb_frag_page_copy() - sets the page in a fragment from another fragment
3071 * @fragto: skb fragment where page is set
3072 * @fragfrom: skb fragment page is copied from
3073 */
3074static inline void skb_frag_page_copy(skb_frag_t *fragto,
3075 const skb_frag_t *fragfrom)
3076{
3077 fragto->bv_page = fragfrom->bv_page;
Ian Campbell131ea662011-08-19 06:25:00 +00003078}
3079
3080/**
3081 * __skb_frag_set_page - sets the page contained in a paged fragment
3082 * @frag: the paged fragment
3083 * @page: the page to set
3084 *
3085 * Sets the fragment @frag to contain @page.
3086 */
3087static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3088{
Matthew Wilcox (Oracle)1dfa5bd2019-07-22 20:08:29 -07003089 frag->bv_page = page;
Ian Campbell131ea662011-08-19 06:25:00 +00003090}
3091
3092/**
3093 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3094 * @skb: the buffer
3095 * @f: the fragment offset
3096 * @page: the page to set
3097 *
3098 * Sets the @f'th fragment of @skb to contain @page.
3099 */
3100static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3101 struct page *page)
3102{
3103 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3104}
3105
Eric Dumazet400dfd32013-10-17 16:27:07 -07003106bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3107
Ian Campbell131ea662011-08-19 06:25:00 +00003108/**
3109 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00003110 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00003111 * @frag: the paged fragment to map
3112 * @offset: the offset within the fragment (starting at the
3113 * fragment's own offset)
3114 * @size: the number of bytes to map
Mauro Carvalho Chehab771b00a2017-05-12 09:19:29 -03003115 * @dir: the direction of the mapping (``PCI_DMA_*``)
Ian Campbell131ea662011-08-19 06:25:00 +00003116 *
3117 * Maps the page associated with @frag to @device.
3118 */
3119static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3120 const skb_frag_t *frag,
3121 size_t offset, size_t size,
3122 enum dma_data_direction dir)
3123{
3124 return dma_map_page(dev, skb_frag_page(frag),
Jonathan Lemon7240b602019-07-30 07:40:32 -07003125 skb_frag_off(frag) + offset, size, dir);
Ian Campbell131ea662011-08-19 06:25:00 +00003126}
3127
Eric Dumazet117632e2011-12-03 21:39:53 +00003128static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3129 gfp_t gfp_mask)
3130{
3131 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3132}
3133
Octavian Purdilabad93e92014-06-12 01:36:26 +03003134
3135static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3136 gfp_t gfp_mask)
3137{
3138 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3139}
3140
3141
Ian Campbell131ea662011-08-19 06:25:00 +00003142/**
Patrick McHardy334a8132007-06-25 04:35:20 -07003143 * skb_clone_writable - is the header of a clone writable
3144 * @skb: buffer to check
3145 * @len: length up to which to write
3146 *
3147 * Returns true if modifying the header part of the cloned buffer
3148 * does not requires the data to be copied.
3149 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003150static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07003151{
3152 return !skb_header_cloned(skb) &&
3153 skb_headroom(skb) + len <= skb->hdr_len;
3154}
3155
Daniel Borkmann36976492016-02-19 23:05:25 +01003156static inline int skb_try_make_writable(struct sk_buff *skb,
3157 unsigned int write_len)
3158{
3159 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3160 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3161}
3162
Herbert Xud9cc2042007-09-16 16:21:16 -07003163static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3164 int cloned)
3165{
3166 int delta = 0;
3167
Herbert Xud9cc2042007-09-16 16:21:16 -07003168 if (headroom > skb_headroom(skb))
3169 delta = headroom - skb_headroom(skb);
3170
3171 if (delta || cloned)
3172 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3173 GFP_ATOMIC);
3174 return 0;
3175}
3176
Patrick McHardy334a8132007-06-25 04:35:20 -07003177/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07003178 * skb_cow - copy header of skb when it is required
3179 * @skb: buffer to cow
3180 * @headroom: needed headroom
3181 *
3182 * If the skb passed lacks sufficient headroom or its data part
3183 * is shared, data is reallocated. If reallocation fails, an error
3184 * is returned and original skb is not changed.
3185 *
3186 * The result is skb with writable area skb->head...skb->tail
3187 * and at least @headroom of space at head.
3188 */
3189static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3190{
Herbert Xud9cc2042007-09-16 16:21:16 -07003191 return __skb_cow(skb, headroom, skb_cloned(skb));
3192}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003193
Herbert Xud9cc2042007-09-16 16:21:16 -07003194/**
3195 * skb_cow_head - skb_cow but only making the head writable
3196 * @skb: buffer to cow
3197 * @headroom: needed headroom
3198 *
3199 * This function is identical to skb_cow except that we replace the
3200 * skb_cloned check by skb_header_cloned. It should be used when
3201 * you only need to push on some header and do not need to modify
3202 * the data.
3203 */
3204static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3205{
3206 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07003207}
3208
3209/**
3210 * skb_padto - pad an skbuff up to a minimal size
3211 * @skb: buffer to pad
3212 * @len: minimal length
3213 *
3214 * Pads up a buffer to ensure the trailing bytes exist and are
3215 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07003216 * is untouched. Otherwise it is extended. Returns zero on
3217 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003218 */
Herbert Xu5b057c62006-06-23 02:06:41 -07003219static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003220{
3221 unsigned int size = skb->len;
3222 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07003223 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07003224 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003225}
3226
Alexander Duyck9c0c1122014-12-03 08:17:33 -08003227/**
Brian Norris4ea7b0c2019-02-11 13:02:25 -08003228 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
Alexander Duyck9c0c1122014-12-03 08:17:33 -08003229 * @skb: buffer to pad
3230 * @len: minimal length
Florian Fainellicd0a1372017-08-22 15:12:14 -07003231 * @free_on_error: free buffer on error
3232 *
3233 * Pads up a buffer to ensure the trailing bytes exist and are
3234 * blanked. If the buffer already contains sufficient data it
3235 * is untouched. Otherwise it is extended. Returns zero on
3236 * success. The skb is freed on error if @free_on_error is true.
3237 */
Eric Dumazet4a009cb2020-09-09 01:27:40 -07003238static inline int __must_check __skb_put_padto(struct sk_buff *skb,
3239 unsigned int len,
3240 bool free_on_error)
Florian Fainellicd0a1372017-08-22 15:12:14 -07003241{
3242 unsigned int size = skb->len;
3243
3244 if (unlikely(size < len)) {
3245 len -= size;
3246 if (__skb_pad(skb, len, free_on_error))
3247 return -ENOMEM;
3248 __skb_put(skb, len);
3249 }
3250 return 0;
3251}
3252
3253/**
3254 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3255 * @skb: buffer to pad
3256 * @len: minimal length
Alexander Duyck9c0c1122014-12-03 08:17:33 -08003257 *
3258 * Pads up a buffer to ensure the trailing bytes exist and are
3259 * blanked. If the buffer already contains sufficient data it
3260 * is untouched. Otherwise it is extended. Returns zero on
3261 * success. The skb is freed on error.
3262 */
Eric Dumazet4a009cb2020-09-09 01:27:40 -07003263static inline int __must_check skb_put_padto(struct sk_buff *skb, unsigned int len)
Alexander Duyck9c0c1122014-12-03 08:17:33 -08003264{
Florian Fainellicd0a1372017-08-22 15:12:14 -07003265 return __skb_put_padto(skb, len, true);
Alexander Duyck9c0c1122014-12-03 08:17:33 -08003266}
3267
Linus Torvalds1da177e2005-04-16 15:20:36 -07003268static inline int skb_add_data(struct sk_buff *skb,
Al Viroaf2b0402014-11-27 21:44:24 -05003269 struct iov_iter *from, int copy)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003270{
3271 const int off = skb->len;
3272
3273 if (skb->ip_summed == CHECKSUM_NONE) {
Al Viroaf2b0402014-11-27 21:44:24 -05003274 __wsum csum = 0;
Al Viro15e6cb42016-11-01 22:42:45 -04003275 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3276 &csum, from)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003277 skb->csum = csum_block_add(skb->csum, csum, off);
3278 return 0;
3279 }
Al Viro15e6cb42016-11-01 22:42:45 -04003280 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003281 return 0;
3282
3283 __skb_trim(skb, off);
3284 return -EFAULT;
3285}
3286
Eric Dumazet38ba0a62012-04-23 17:48:27 +00003287static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3288 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003289{
Willem de Bruijn1f8b9772017-08-03 16:29:41 -04003290 if (skb_zcopy(skb))
3291 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003292 if (i) {
Matthew Wilcox (Oracle)d8e18a52019-07-22 20:08:26 -07003293 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07003294
Ian Campbellea2ab692011-08-22 23:44:58 +00003295 return page == skb_frag_page(frag) &&
Jonathan Lemon7240b602019-07-30 07:40:32 -07003296 off == skb_frag_off(frag) + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003297 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00003298 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003299}
3300
Herbert Xu364c6ba2006-06-09 16:10:40 -07003301static inline int __skb_linearize(struct sk_buff *skb)
3302{
3303 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3304}
3305
Linus Torvalds1da177e2005-04-16 15:20:36 -07003306/**
3307 * skb_linearize - convert paged skb to linear one
3308 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07003309 *
3310 * If there is no free memory -ENOMEM is returned, otherwise zero
3311 * is returned and the old skb data released.
3312 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07003313static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003314{
Herbert Xu364c6ba2006-06-09 16:10:40 -07003315 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3316}
3317
3318/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00003319 * skb_has_shared_frag - can any frag be overwritten
3320 * @skb: buffer to test
3321 *
3322 * Return true if the skb has at least one frag that might be modified
3323 * by an external entity (as in vmsplice()/sendfile())
3324 */
3325static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3326{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00003327 return skb_is_nonlinear(skb) &&
3328 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00003329}
3330
3331/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07003332 * skb_linearize_cow - make sure skb is linear and writable
3333 * @skb: buffer to process
3334 *
3335 * If there is no free memory -ENOMEM is returned, otherwise zero
3336 * is returned and the old skb data released.
3337 */
3338static inline int skb_linearize_cow(struct sk_buff *skb)
3339{
3340 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3341 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003342}
3343
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003344static __always_inline void
3345__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3346 unsigned int off)
3347{
3348 if (skb->ip_summed == CHECKSUM_COMPLETE)
3349 skb->csum = csum_block_sub(skb->csum,
3350 csum_partial(start, len, 0), off);
3351 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3352 skb_checksum_start_offset(skb) < 0)
3353 skb->ip_summed = CHECKSUM_NONE;
3354}
3355
Linus Torvalds1da177e2005-04-16 15:20:36 -07003356/**
3357 * skb_postpull_rcsum - update checksum for received skb after pull
3358 * @skb: buffer to update
3359 * @start: start of data before pull
3360 * @len: length of data pulled
3361 *
3362 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07003363 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3364 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003365 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003366static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08003367 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003368{
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003369 __skb_postpull_rcsum(skb, start, len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003370}
3371
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003372static __always_inline void
3373__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3374 unsigned int off)
3375{
3376 if (skb->ip_summed == CHECKSUM_COMPLETE)
3377 skb->csum = csum_block_add(skb->csum,
3378 csum_partial(start, len, 0), off);
3379}
Herbert Xucbb042f2006-03-20 22:43:56 -08003380
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003381/**
3382 * skb_postpush_rcsum - update checksum for received skb after push
3383 * @skb: buffer to update
3384 * @start: start of data after push
3385 * @len: length of data pushed
3386 *
3387 * After doing a push on a received packet, you need to call this to
3388 * update the CHECKSUM_COMPLETE checksum.
3389 */
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01003390static inline void skb_postpush_rcsum(struct sk_buff *skb,
3391 const void *start, unsigned int len)
3392{
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003393 __skb_postpush_rcsum(skb, start, len, 0);
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01003394}
3395
Johannes Bergaf728682017-06-16 14:29:22 +02003396void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003397
David S. Miller7ce5a272013-12-02 17:26:05 -05003398/**
WANG Cong82a31b92016-06-30 10:15:22 -07003399 * skb_push_rcsum - push skb and update receive checksum
3400 * @skb: buffer to update
3401 * @len: length of data pulled
3402 *
3403 * This function performs an skb_push on the packet and updates
3404 * the CHECKSUM_COMPLETE checksum. It should be used on
3405 * receive path processing instead of skb_push unless you know
3406 * that the checksum difference is zero (e.g., a valid IP header)
3407 * or you are setting ip_summed to CHECKSUM_NONE.
3408 */
Johannes Bergd58ff352017-06-16 14:29:23 +02003409static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
WANG Cong82a31b92016-06-30 10:15:22 -07003410{
3411 skb_push(skb, len);
3412 skb_postpush_rcsum(skb, skb->data, len);
3413 return skb->data;
3414}
3415
Eric Dumazet88078d92018-04-18 11:43:15 -07003416int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
WANG Cong82a31b92016-06-30 10:15:22 -07003417/**
David S. Miller7ce5a272013-12-02 17:26:05 -05003418 * pskb_trim_rcsum - trim received skb and update checksum
3419 * @skb: buffer to trim
3420 * @len: new length
3421 *
3422 * This is exactly the same as pskb_trim except that it ensures the
3423 * checksum of received packets are still valid after the operation.
Ross Lagerwall6c57f042019-01-17 15:34:38 +00003424 * It can change skb pointers.
David S. Miller7ce5a272013-12-02 17:26:05 -05003425 */
3426
3427static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3428{
3429 if (likely(len >= skb->len))
3430 return 0;
Eric Dumazet88078d92018-04-18 11:43:15 -07003431 return pskb_trim_rcsum_slow(skb, len);
David S. Miller7ce5a272013-12-02 17:26:05 -05003432}
3433
Daniel Borkmann5293efe2016-08-18 01:00:39 +02003434static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3435{
3436 if (skb->ip_summed == CHECKSUM_COMPLETE)
3437 skb->ip_summed = CHECKSUM_NONE;
3438 __skb_trim(skb, len);
3439 return 0;
3440}
3441
3442static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3443{
3444 if (skb->ip_summed == CHECKSUM_COMPLETE)
3445 skb->ip_summed = CHECKSUM_NONE;
3446 return __skb_grow(skb, len);
3447}
3448
Eric Dumazet18a4c0e2017-10-05 22:21:21 -07003449#define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3450#define skb_rb_first(root) rb_to_skb(rb_first(root))
3451#define skb_rb_last(root) rb_to_skb(rb_last(root))
3452#define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3453#define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3454
Linus Torvalds1da177e2005-04-16 15:20:36 -07003455#define skb_queue_walk(queue, skb) \
3456 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003457 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07003458 skb = skb->next)
3459
James Chapman46f89142007-04-30 00:07:31 -07003460#define skb_queue_walk_safe(queue, skb, tmp) \
3461 for (skb = (queue)->next, tmp = skb->next; \
3462 skb != (struct sk_buff *)(queue); \
3463 skb = tmp, tmp = skb->next)
3464
David S. Miller1164f522008-09-23 00:49:44 -07003465#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003466 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07003467 skb = skb->next)
3468
Eric Dumazet18a4c0e2017-10-05 22:21:21 -07003469#define skb_rbtree_walk(skb, root) \
3470 for (skb = skb_rb_first(root); skb != NULL; \
3471 skb = skb_rb_next(skb))
3472
3473#define skb_rbtree_walk_from(skb) \
3474 for (; skb != NULL; \
3475 skb = skb_rb_next(skb))
3476
3477#define skb_rbtree_walk_from_safe(skb, tmp) \
3478 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3479 skb = tmp)
3480
David S. Miller1164f522008-09-23 00:49:44 -07003481#define skb_queue_walk_from_safe(queue, skb, tmp) \
3482 for (tmp = skb->next; \
3483 skb != (struct sk_buff *)(queue); \
3484 skb = tmp, tmp = skb->next)
3485
Stephen Hemminger300ce172005-10-30 13:47:34 -08003486#define skb_queue_reverse_walk(queue, skb) \
3487 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003488 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08003489 skb = skb->prev)
3490
David S. Miller686a2952011-01-20 22:47:32 -08003491#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3492 for (skb = (queue)->prev, tmp = skb->prev; \
3493 skb != (struct sk_buff *)(queue); \
3494 skb = tmp, tmp = skb->prev)
3495
3496#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3497 for (tmp = skb->prev; \
3498 skb != (struct sk_buff *)(queue); \
3499 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003500
David S. Miller21dc3302010-08-23 00:13:46 -07003501static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07003502{
3503 return skb_shinfo(skb)->frag_list != NULL;
3504}
3505
3506static inline void skb_frag_list_init(struct sk_buff *skb)
3507{
3508 skb_shinfo(skb)->frag_list = NULL;
3509}
3510
David S. Milleree039872009-06-09 00:17:13 -07003511#define skb_walk_frags(skb, iter) \
3512 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3513
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003514
Sabrina Dubrocab50b0582019-11-25 14:48:57 +01003515int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue,
3516 int *err, long *timeo_p,
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003517 const struct sk_buff *skb);
Paolo Abeni65101ae2017-05-16 11:20:13 +02003518struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3519 struct sk_buff_head *queue,
3520 unsigned int flags,
Paolo Abenifd69c392019-04-08 10:15:59 +02003521 int *off, int *err,
Paolo Abeni65101ae2017-05-16 11:20:13 +02003522 struct sk_buff **last);
Sabrina Dubrocab50b0582019-11-25 14:48:57 +01003523struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
3524 struct sk_buff_head *queue,
Paolo Abenie427cad2020-02-28 14:45:22 +01003525 unsigned int flags, int *off, int *err,
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003526 struct sk_buff **last);
Sabrina Dubrocab50b0582019-11-25 14:48:57 +01003527struct sk_buff *__skb_recv_datagram(struct sock *sk,
3528 struct sk_buff_head *sk_queue,
Paolo Abenie427cad2020-02-28 14:45:22 +01003529 unsigned int flags, int *off, int *err);
Joe Perches7965bd42013-09-26 14:48:15 -07003530struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3531 int *err);
Linus Torvaldsa11e1d42018-06-28 09:43:44 -07003532__poll_t datagram_poll(struct file *file, struct socket *sock,
3533 struct poll_table_struct *wait);
Al Viroc0371da2014-11-24 10:42:55 -05003534int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3535 struct iov_iter *to, int size);
David S. Miller51f3d022014-11-05 16:46:40 -05003536static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3537 struct msghdr *msg, int size)
3538{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003539 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
David S. Miller51f3d022014-11-05 16:46:40 -05003540}
Al Viroe5a4b0b2014-11-24 18:17:55 -05003541int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3542 struct msghdr *msg);
Sagi Grimberg65d69e22018-12-03 17:52:10 -08003543int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3544 struct iov_iter *to, int len,
3545 struct ahash_request *hash);
Al Viro3a654f92014-06-19 14:15:22 -04003546int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3547 struct iov_iter *from, int len);
Al Viro3a654f92014-06-19 14:15:22 -04003548int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
Joe Perches7965bd42013-09-26 14:48:15 -07003549void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
samanthakumar627d2d62016-04-05 12:41:16 -04003550void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3551static inline void skb_free_datagram_locked(struct sock *sk,
3552 struct sk_buff *skb)
3553{
3554 __skb_free_datagram_locked(sk, skb, 0);
3555}
Joe Perches7965bd42013-09-26 14:48:15 -07003556int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07003557int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3558int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3559__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
Al Viro8d5930d2020-07-10 20:07:10 -04003560 int len);
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +02003561int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
Joe Perches7965bd42013-09-26 14:48:15 -07003562 struct pipe_inode_info *pipe, unsigned int len,
Al Viro25869262016-09-17 21:02:10 -04003563 unsigned int flags);
Tom Herbert20bf50d2017-07-28 16:22:42 -07003564int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3565 int len);
Joe Perches7965bd42013-09-26 14:48:15 -07003566void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01003567unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00003568int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3569 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07003570void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3571int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3572void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Daniel Axtens779b7932018-03-01 17:13:37 +11003573bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
Daniel Axtens2b16f042018-01-31 14:15:33 +11003574bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
Joe Perches7965bd42013-09-26 14:48:15 -07003575struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Steffen Klassert3a1296a2020-01-25 11:26:44 +01003576struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features,
3577 unsigned int offset);
Vlad Yasevich0d5501c2014-08-08 14:42:13 -04003578struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
Jiri Pirkoe2195122014-11-19 14:05:01 +01003579int skb_ensure_writable(struct sk_buff *skb, int write_len);
Shmulik Ladkanibfca4c52016-09-19 19:11:09 +03003580int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
Jiri Pirko93515d52014-11-19 14:05:02 +01003581int skb_vlan_pop(struct sk_buff *skb);
3582int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
Guillaume Nault19fbcb32020-10-03 00:44:28 +02003583int skb_eth_pop(struct sk_buff *skb);
3584int skb_eth_push(struct sk_buff *skb, const unsigned char *dst,
3585 const unsigned char *src);
Davide Carattifa4e0f82019-10-12 13:55:07 +02003586int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
Martin Varghesed04ac222019-12-05 05:57:22 +05303587 int mac_len, bool ethernet);
Martin Varghese040b5cf2019-12-02 10:49:51 +05303588int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
3589 bool ethernet);
John Hurleyd27cf5c2019-07-07 15:01:56 +01003590int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
John Hurley2a2ea502019-07-07 15:01:57 +01003591int skb_mpls_dec_ttl(struct sk_buff *skb);
Sowmini Varadhan6fa01cc2016-04-22 18:36:35 -07003592struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3593 gfp_t gfp);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003594
Al Viro6ce8e9c2014-04-06 21:25:44 -04003595static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3596{
Al Viro3073f072017-02-17 23:13:25 -05003597 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
Al Viro6ce8e9c2014-04-06 21:25:44 -04003598}
3599
Al Viro7eab8d92014-04-06 21:51:23 -04003600static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3601{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003602 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro7eab8d92014-04-06 21:51:23 -04003603}
3604
Daniel Borkmann2817a332013-10-30 11:50:51 +01003605struct skb_checksum_ops {
3606 __wsum (*update)(const void *mem, int len, __wsum wsum);
3607 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3608};
3609
Davide Caratti96178132017-05-18 15:44:37 +02003610extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3611
Daniel Borkmann2817a332013-10-30 11:50:51 +01003612__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3613 __wsum csum, const struct skb_checksum_ops *ops);
3614__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3615 __wsum csum);
3616
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003617static inline void * __must_check
3618__skb_header_pointer(const struct sk_buff *skb, int offset,
3619 int len, void *data, int hlen, void *buffer)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003620{
Patrick McHardy55820ee2005-07-05 14:08:10 -07003621 if (hlen - offset >= len)
David S. Miller690e36e2014-08-23 12:13:41 -07003622 return data + offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003623
David S. Miller690e36e2014-08-23 12:13:41 -07003624 if (!skb ||
3625 skb_copy_bits(skb, offset, buffer, len) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003626 return NULL;
3627
3628 return buffer;
3629}
3630
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003631static inline void * __must_check
3632skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
David S. Miller690e36e2014-08-23 12:13:41 -07003633{
3634 return __skb_header_pointer(skb, offset, len, skb->data,
3635 skb_headlen(skb), buffer);
3636}
3637
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01003638/**
3639 * skb_needs_linearize - check if we need to linearize a given skb
3640 * depending on the given device features.
3641 * @skb: socket buffer to check
3642 * @features: net device features
3643 *
3644 * Returns true if either:
3645 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3646 * 2. skb is fragmented and the device does not support SG.
3647 */
3648static inline bool skb_needs_linearize(struct sk_buff *skb,
3649 netdev_features_t features)
3650{
3651 return skb_is_nonlinear(skb) &&
3652 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3653 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3654}
3655
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03003656static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3657 void *to,
3658 const unsigned int len)
3659{
3660 memcpy(to, skb->data, len);
3661}
3662
3663static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3664 const int offset, void *to,
3665 const unsigned int len)
3666{
3667 memcpy(to, skb->data + offset, len);
3668}
3669
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03003670static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3671 const void *from,
3672 const unsigned int len)
3673{
3674 memcpy(skb->data, from, len);
3675}
3676
3677static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3678 const int offset,
3679 const void *from,
3680 const unsigned int len)
3681{
3682 memcpy(skb->data + offset, from, len);
3683}
3684
Joe Perches7965bd42013-09-26 14:48:15 -07003685void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003686
Patrick Ohlyac45f602009-02-12 05:03:37 +00003687static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3688{
3689 return skb->tstamp;
3690}
3691
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003692/**
3693 * skb_get_timestamp - get timestamp from a skb
3694 * @skb: skb to get stamp from
Deepa Dinamani13c6ee22019-02-02 07:34:48 -08003695 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003696 *
3697 * Timestamps are stored in the skb as offsets to a base timestamp.
3698 * This function converts the offset back to a struct timeval and stores
3699 * it in stamp.
3700 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00003701static inline void skb_get_timestamp(const struct sk_buff *skb,
Deepa Dinamani13c6ee22019-02-02 07:34:48 -08003702 struct __kernel_old_timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003703{
Deepa Dinamani13c6ee22019-02-02 07:34:48 -08003704 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003705}
3706
Deepa Dinamani887feae2019-02-02 07:34:50 -08003707static inline void skb_get_new_timestamp(const struct sk_buff *skb,
3708 struct __kernel_sock_timeval *stamp)
3709{
3710 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3711
3712 stamp->tv_sec = ts.tv_sec;
3713 stamp->tv_usec = ts.tv_nsec / 1000;
3714}
3715
Patrick Ohlyac45f602009-02-12 05:03:37 +00003716static inline void skb_get_timestampns(const struct sk_buff *skb,
Arnd Bergmanndf1b4ba2019-10-25 22:04:46 +02003717 struct __kernel_old_timespec *stamp)
Patrick Ohlyac45f602009-02-12 05:03:37 +00003718{
Arnd Bergmanndf1b4ba2019-10-25 22:04:46 +02003719 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3720
3721 stamp->tv_sec = ts.tv_sec;
3722 stamp->tv_nsec = ts.tv_nsec;
Patrick Ohlyac45f602009-02-12 05:03:37 +00003723}
3724
Deepa Dinamani887feae2019-02-02 07:34:50 -08003725static inline void skb_get_new_timestampns(const struct sk_buff *skb,
3726 struct __kernel_timespec *stamp)
3727{
3728 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3729
3730 stamp->tv_sec = ts.tv_sec;
3731 stamp->tv_nsec = ts.tv_nsec;
3732}
3733
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003734static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003735{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003736 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003737}
3738
Stephen Hemminger164891a2007-04-23 22:26:16 -07003739static inline ktime_t net_timedelta(ktime_t t)
3740{
3741 return ktime_sub(ktime_get_real(), t);
3742}
3743
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003744static inline ktime_t net_invalid_timestamp(void)
3745{
Thomas Gleixner8b0e1952016-12-25 12:30:41 +01003746 return 0;
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003747}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003748
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003749static inline u8 skb_metadata_len(const struct sk_buff *skb)
3750{
3751 return skb_shinfo(skb)->meta_len;
3752}
3753
3754static inline void *skb_metadata_end(const struct sk_buff *skb)
3755{
3756 return skb_mac_header(skb);
3757}
3758
3759static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3760 const struct sk_buff *skb_b,
3761 u8 meta_len)
3762{
3763 const void *a = skb_metadata_end(skb_a);
3764 const void *b = skb_metadata_end(skb_b);
3765 /* Using more efficient varaiant than plain call to memcmp(). */
3766#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3767 u64 diffs = 0;
3768
3769 switch (meta_len) {
3770#define __it(x, op) (x -= sizeof(u##op))
3771#define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3772 case 32: diffs |= __it_diff(a, b, 64);
Gustavo A. R. Silvadf561f662020-08-23 17:36:59 -05003773 fallthrough;
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003774 case 24: diffs |= __it_diff(a, b, 64);
Gustavo A. R. Silvadf561f662020-08-23 17:36:59 -05003775 fallthrough;
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003776 case 16: diffs |= __it_diff(a, b, 64);
Gustavo A. R. Silvadf561f662020-08-23 17:36:59 -05003777 fallthrough;
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003778 case 8: diffs |= __it_diff(a, b, 64);
3779 break;
3780 case 28: diffs |= __it_diff(a, b, 64);
Gustavo A. R. Silvadf561f662020-08-23 17:36:59 -05003781 fallthrough;
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003782 case 20: diffs |= __it_diff(a, b, 64);
Gustavo A. R. Silvadf561f662020-08-23 17:36:59 -05003783 fallthrough;
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003784 case 12: diffs |= __it_diff(a, b, 64);
Gustavo A. R. Silvadf561f662020-08-23 17:36:59 -05003785 fallthrough;
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003786 case 4: diffs |= __it_diff(a, b, 32);
3787 break;
3788 }
3789 return diffs;
3790#else
3791 return memcmp(a - meta_len, b - meta_len, meta_len);
3792#endif
3793}
3794
3795static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3796 const struct sk_buff *skb_b)
3797{
3798 u8 len_a = skb_metadata_len(skb_a);
3799 u8 len_b = skb_metadata_len(skb_b);
3800
3801 if (!(len_a | len_b))
3802 return false;
3803
3804 return len_a != len_b ?
3805 true : __skb_metadata_differs(skb_a, skb_b, len_a);
3806}
3807
3808static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3809{
3810 skb_shinfo(skb)->meta_len = meta_len;
3811}
3812
3813static inline void skb_metadata_clear(struct sk_buff *skb)
3814{
3815 skb_metadata_set(skb, 0);
3816}
3817
Alexander Duyck62bccb82014-09-04 13:31:35 -04003818struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3819
Richard Cochranc1f19b52010-07-17 08:49:36 +00003820#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3821
Joe Perches7965bd42013-09-26 14:48:15 -07003822void skb_clone_tx_timestamp(struct sk_buff *skb);
3823bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00003824
3825#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3826
3827static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3828{
3829}
3830
3831static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3832{
3833 return false;
3834}
3835
3836#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3837
3838/**
3839 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3840 *
Richard Cochranda92b192011-10-21 00:49:15 +00003841 * PHY drivers may accept clones of transmitted packets for
3842 * timestamping via their phy_driver.txtstamp method. These drivers
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003843 * must call this function to return the skb back to the stack with a
3844 * timestamp.
Richard Cochranda92b192011-10-21 00:49:15 +00003845 *
Randy Dunlap2ff17112020-07-15 09:42:39 -07003846 * @skb: clone of the original outgoing packet
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003847 * @hwtstamps: hardware time stamps
Richard Cochranc1f19b52010-07-17 08:49:36 +00003848 *
3849 */
3850void skb_complete_tx_timestamp(struct sk_buff *skb,
3851 struct skb_shared_hwtstamps *hwtstamps);
3852
Willem de Bruijne7fd2882014-08-04 22:11:48 -04003853void __skb_tstamp_tx(struct sk_buff *orig_skb,
3854 struct skb_shared_hwtstamps *hwtstamps,
3855 struct sock *sk, int tstype);
3856
Patrick Ohlyac45f602009-02-12 05:03:37 +00003857/**
3858 * skb_tstamp_tx - queue clone of skb with send time stamps
3859 * @orig_skb: the original outgoing packet
3860 * @hwtstamps: hardware time stamps, may be NULL if not available
3861 *
3862 * If the skb has a socket associated, then this function clones the
3863 * skb (thus sharing the actual data and optional structures), stores
3864 * the optional hardware time stamping information (if non NULL) or
3865 * generates a software time stamp (otherwise), then queues the clone
3866 * to the error queue of the socket. Errors are silently ignored.
3867 */
Joe Perches7965bd42013-09-26 14:48:15 -07003868void skb_tstamp_tx(struct sk_buff *orig_skb,
3869 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00003870
Richard Cochran4507a712010-07-17 08:48:28 +00003871/**
3872 * skb_tx_timestamp() - Driver hook for transmit timestamping
3873 *
3874 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00003875 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00003876 *
David S. Miller73409f32013-12-27 13:04:33 -05003877 * Specifically, one should make absolutely sure that this function is
3878 * called before TX completion of this packet can trigger. Otherwise
3879 * the packet could potentially already be freed.
3880 *
Richard Cochran4507a712010-07-17 08:48:28 +00003881 * @skb: A socket buffer.
3882 */
3883static inline void skb_tx_timestamp(struct sk_buff *skb)
3884{
Richard Cochranc1f19b52010-07-17 08:49:36 +00003885 skb_clone_tx_timestamp(skb);
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +02003886 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3887 skb_tstamp_tx(skb, NULL);
Richard Cochran4507a712010-07-17 08:48:28 +00003888}
3889
Johannes Berg6e3e9392011-11-09 10:15:42 +01003890/**
3891 * skb_complete_wifi_ack - deliver skb with wifi status
3892 *
3893 * @skb: the original outgoing packet
3894 * @acked: ack status
3895 *
3896 */
3897void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3898
Joe Perches7965bd42013-09-26 14:48:15 -07003899__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3900__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003901
Herbert Xu60476372007-04-09 11:59:39 -07003902static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3903{
Tom Herbert6edec0e2015-02-10 16:30:28 -08003904 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3905 skb->csum_valid ||
3906 (skb->ip_summed == CHECKSUM_PARTIAL &&
3907 skb_checksum_start_offset(skb) >= 0));
Herbert Xu60476372007-04-09 11:59:39 -07003908}
3909
Herbert Xufb286bb2005-11-10 13:01:24 -08003910/**
3911 * skb_checksum_complete - Calculate checksum of an entire packet
3912 * @skb: packet to process
3913 *
3914 * This function calculates the checksum over the entire packet plus
3915 * the value of skb->csum. The latter can be used to supply the
3916 * checksum of a pseudo header as used by TCP/UDP. It returns the
3917 * checksum.
3918 *
3919 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3920 * this function can be used to verify that checksum on received
3921 * packets. In that case the function should return zero if the
3922 * checksum is correct. In particular, this function will return zero
3923 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3924 * hardware has already verified the correctness of the checksum.
3925 */
Al Viro4381ca32007-07-15 21:00:11 +01003926static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08003927{
Herbert Xu60476372007-04-09 11:59:39 -07003928 return skb_csum_unnecessary(skb) ?
3929 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003930}
3931
Tom Herbert77cffe22014-08-27 21:26:46 -07003932static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3933{
3934 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3935 if (skb->csum_level == 0)
3936 skb->ip_summed = CHECKSUM_NONE;
3937 else
3938 skb->csum_level--;
3939 }
3940}
3941
3942static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3943{
3944 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3945 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3946 skb->csum_level++;
3947 } else if (skb->ip_summed == CHECKSUM_NONE) {
3948 skb->ip_summed = CHECKSUM_UNNECESSARY;
3949 skb->csum_level = 0;
3950 }
3951}
3952
Daniel Borkmann836e66c2020-06-02 16:58:32 +02003953static inline void __skb_reset_checksum_unnecessary(struct sk_buff *skb)
3954{
3955 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3956 skb->ip_summed = CHECKSUM_NONE;
3957 skb->csum_level = 0;
3958 }
3959}
3960
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003961/* Check if we need to perform checksum complete validation.
3962 *
3963 * Returns true if checksum complete is needed, false otherwise
3964 * (either checksum is unnecessary or zero checksum is allowed).
3965 */
3966static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3967 bool zero_okay,
3968 __sum16 check)
3969{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003970 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3971 skb->csum_valid = 1;
Tom Herbert77cffe22014-08-27 21:26:46 -07003972 __skb_decr_checksum_unnecessary(skb);
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003973 return false;
3974 }
3975
3976 return true;
3977}
3978
David S. Millerda279882018-02-16 15:52:42 -05003979/* For small packets <= CHECKSUM_BREAK perform checksum complete directly
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003980 * in checksum_init.
3981 */
3982#define CHECKSUM_BREAK 76
3983
Tom Herbert4e18b9a2015-04-20 14:10:04 -07003984/* Unset checksum-complete
3985 *
3986 * Unset checksum complete can be done when packet is being modified
3987 * (uncompressed for instance) and checksum-complete value is
3988 * invalidated.
3989 */
3990static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3991{
3992 if (skb->ip_summed == CHECKSUM_COMPLETE)
3993 skb->ip_summed = CHECKSUM_NONE;
3994}
3995
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003996/* Validate (init) checksum based on checksum complete.
3997 *
3998 * Return values:
3999 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
4000 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
4001 * checksum is stored in skb->csum for use in __skb_checksum_complete
4002 * non-zero: value of invalid checksum
4003 *
4004 */
4005static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
4006 bool complete,
4007 __wsum psum)
4008{
4009 if (skb->ip_summed == CHECKSUM_COMPLETE) {
4010 if (!csum_fold(csum_add(psum, skb->csum))) {
Tom Herbert5d0c2b92014-06-10 18:54:13 -07004011 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07004012 return 0;
4013 }
4014 }
4015
4016 skb->csum = psum;
4017
Tom Herbert5d0c2b92014-06-10 18:54:13 -07004018 if (complete || skb->len <= CHECKSUM_BREAK) {
4019 __sum16 csum;
4020
4021 csum = __skb_checksum_complete(skb);
4022 skb->csum_valid = !csum;
4023 return csum;
4024 }
Tom Herbert76ba0aa2014-05-02 16:29:18 -07004025
4026 return 0;
4027}
4028
4029static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
4030{
4031 return 0;
4032}
4033
4034/* Perform checksum validate (init). Note that this is a macro since we only
4035 * want to calculate the pseudo header which is an input function if necessary.
4036 * First we try to validate without any computation (checksum unnecessary) and
4037 * then calculate based on checksum complete calling the function to compute
4038 * pseudo header.
4039 *
4040 * Return values:
4041 * 0: checksum is validated or try to in skb_checksum_complete
4042 * non-zero: value of invalid checksum
4043 */
4044#define __skb_checksum_validate(skb, proto, complete, \
4045 zero_okay, check, compute_pseudo) \
4046({ \
4047 __sum16 __ret = 0; \
Tom Herbert5d0c2b92014-06-10 18:54:13 -07004048 skb->csum_valid = 0; \
Tom Herbert76ba0aa2014-05-02 16:29:18 -07004049 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
4050 __ret = __skb_checksum_validate_complete(skb, \
4051 complete, compute_pseudo(skb, proto)); \
4052 __ret; \
4053})
4054
4055#define skb_checksum_init(skb, proto, compute_pseudo) \
4056 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
4057
4058#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
4059 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4060
4061#define skb_checksum_validate(skb, proto, compute_pseudo) \
4062 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4063
4064#define skb_checksum_validate_zero_check(skb, proto, check, \
4065 compute_pseudo) \
Sabrina Dubroca096a4cf2015-02-06 18:54:19 +01004066 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
Tom Herbert76ba0aa2014-05-02 16:29:18 -07004067
4068#define skb_checksum_simple_validate(skb) \
4069 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4070
Tom Herbertd96535a2014-08-31 15:12:42 -07004071static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4072{
Davide Caratti219f1d792017-05-18 15:44:39 +02004073 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
Tom Herbertd96535a2014-08-31 15:12:42 -07004074}
4075
Li RongQinge4aa33a2019-07-04 17:03:26 +08004076static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
Tom Herbertd96535a2014-08-31 15:12:42 -07004077{
4078 skb->csum = ~pseudo;
4079 skb->ip_summed = CHECKSUM_COMPLETE;
4080}
4081
Li RongQinge4aa33a2019-07-04 17:03:26 +08004082#define skb_checksum_try_convert(skb, proto, compute_pseudo) \
Tom Herbertd96535a2014-08-31 15:12:42 -07004083do { \
4084 if (__skb_checksum_convert_check(skb)) \
Li RongQinge4aa33a2019-07-04 17:03:26 +08004085 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
Tom Herbertd96535a2014-08-31 15:12:42 -07004086} while (0)
4087
Tom Herbert15e23962015-02-10 16:30:31 -08004088static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4089 u16 start, u16 offset)
4090{
4091 skb->ip_summed = CHECKSUM_PARTIAL;
4092 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4093 skb->csum_offset = offset - start;
4094}
4095
Tom Herbertdcdc8992015-02-02 16:07:34 -08004096/* Update skbuf and packet to reflect the remote checksum offload operation.
4097 * When called, ptr indicates the starting point for skb->csum when
4098 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4099 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4100 */
4101static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
Tom Herbert15e23962015-02-10 16:30:31 -08004102 int start, int offset, bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08004103{
4104 __wsum delta;
4105
Tom Herbert15e23962015-02-10 16:30:31 -08004106 if (!nopartial) {
4107 skb_remcsum_adjust_partial(skb, ptr, start, offset);
4108 return;
4109 }
4110
Tom Herbertdcdc8992015-02-02 16:07:34 -08004111 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4112 __skb_checksum_complete(skb);
4113 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4114 }
4115
4116 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4117
4118 /* Adjust skb->csum since we changed the packet */
4119 skb->csum = csum_add(skb->csum, delta);
4120}
4121
Florian Westphalcb9c6832017-01-23 18:21:56 +01004122static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4123{
4124#if IS_ENABLED(CONFIG_NF_CONNTRACK)
Jeremy Sowden261db6c2019-09-13 09:13:14 +01004125 return (void *)(skb->_nfct & NFCT_PTRMASK);
Florian Westphalcb9c6832017-01-23 18:21:56 +01004126#else
4127 return NULL;
4128#endif
4129}
4130
Jeremy Sowden261db6c2019-09-13 09:13:14 +01004131static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07004132{
Jeremy Sowden261db6c2019-09-13 09:13:14 +01004133#if IS_ENABLED(CONFIG_NF_CONNTRACK)
4134 return skb->_nfct;
4135#else
4136 return 0UL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01004137#endif
Jeremy Sowden261db6c2019-09-13 09:13:14 +01004138}
4139
4140static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
4141{
4142#if IS_ENABLED(CONFIG_NF_CONNTRACK)
4143 skb->_nfct = nfct;
4144#endif
4145}
Florian Westphaldf5042f2018-12-18 17:15:16 +01004146
4147#ifdef CONFIG_SKB_EXTENSIONS
4148enum skb_ext_id {
4149#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4150 SKB_EXT_BRIDGE_NF,
4151#endif
Florian Westphal41650792018-12-18 17:15:27 +01004152#ifdef CONFIG_XFRM
4153 SKB_EXT_SEC_PATH,
4154#endif
Paul Blakey95a72332019-09-04 16:56:37 +03004155#if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4156 TC_SKB_EXT,
4157#endif
Mat Martineau3ee17bc2020-01-09 07:59:19 -08004158#if IS_ENABLED(CONFIG_MPTCP)
4159 SKB_EXT_MPTCP,
4160#endif
Florian Westphaldf5042f2018-12-18 17:15:16 +01004161 SKB_EXT_NUM, /* must be last */
4162};
4163
4164/**
4165 * struct skb_ext - sk_buff extensions
4166 * @refcnt: 1 on allocation, deallocated on 0
4167 * @offset: offset to add to @data to obtain extension address
4168 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4169 * @data: start of extension data, variable sized
4170 *
4171 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4172 * to use 'u8' types while allowing up to 2kb worth of extension data.
4173 */
4174struct skb_ext {
4175 refcount_t refcnt;
4176 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4177 u8 chunks; /* same */
Gustavo A. R. Silva5c91aa12020-03-23 19:22:24 -05004178 char data[] __aligned(8);
Florian Westphaldf5042f2018-12-18 17:15:16 +01004179};
4180
Florian Westphal4930f482020-05-16 10:46:23 +02004181struct skb_ext *__skb_ext_alloc(gfp_t flags);
Paolo Abeni8b69a802020-01-09 07:59:24 -08004182void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id,
4183 struct skb_ext *ext);
Florian Westphaldf5042f2018-12-18 17:15:16 +01004184void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4185void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4186void __skb_ext_put(struct skb_ext *ext);
4187
4188static inline void skb_ext_put(struct sk_buff *skb)
4189{
4190 if (skb->active_extensions)
4191 __skb_ext_put(skb->extensions);
4192}
4193
Florian Westphaldf5042f2018-12-18 17:15:16 +01004194static inline void __skb_ext_copy(struct sk_buff *dst,
4195 const struct sk_buff *src)
4196{
4197 dst->active_extensions = src->active_extensions;
4198
4199 if (src->active_extensions) {
4200 struct skb_ext *ext = src->extensions;
4201
4202 refcount_inc(&ext->refcnt);
4203 dst->extensions = ext;
4204 }
4205}
4206
4207static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4208{
4209 skb_ext_put(dst);
4210 __skb_ext_copy(dst, src);
4211}
4212
4213static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4214{
4215 return !!ext->offset[i];
4216}
4217
4218static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4219{
4220 return skb->active_extensions & (1 << id);
4221}
4222
4223static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4224{
4225 if (skb_ext_exist(skb, id))
4226 __skb_ext_del(skb, id);
4227}
4228
4229static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4230{
4231 if (skb_ext_exist(skb, id)) {
4232 struct skb_ext *ext = skb->extensions;
4233
4234 return (void *)ext + (ext->offset[id] << 3);
4235 }
4236
4237 return NULL;
4238}
Florian Westphal174e2382019-09-26 20:37:05 +02004239
4240static inline void skb_ext_reset(struct sk_buff *skb)
4241{
4242 if (unlikely(skb->active_extensions)) {
4243 __skb_ext_put(skb->extensions);
4244 skb->active_extensions = 0;
4245 }
4246}
Florian Westphal677bf082019-11-21 06:56:23 +01004247
4248static inline bool skb_has_extensions(struct sk_buff *skb)
4249{
4250 return unlikely(skb->active_extensions);
4251}
Florian Westphaldf5042f2018-12-18 17:15:16 +01004252#else
4253static inline void skb_ext_put(struct sk_buff *skb) {}
Florian Westphal174e2382019-09-26 20:37:05 +02004254static inline void skb_ext_reset(struct sk_buff *skb) {}
Florian Westphaldf5042f2018-12-18 17:15:16 +01004255static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
4256static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {}
4257static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
Florian Westphal677bf082019-11-21 06:56:23 +01004258static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
Florian Westphaldf5042f2018-12-18 17:15:16 +01004259#endif /* CONFIG_SKB_EXTENSIONS */
4260
Florian Westphal895b5c92019-09-29 20:54:03 +02004261static inline void nf_reset_ct(struct sk_buff *skb)
Patrick McHardya193a4a2006-03-20 19:23:05 -08004262{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07004263#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01004264 nf_conntrack_put(skb_nfct(skb));
4265 skb->_nfct = 0;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01004266#endif
Patrick McHardya193a4a2006-03-20 19:23:05 -08004267}
4268
Patrick McHardy124dff02013-04-05 20:42:05 +02004269static inline void nf_reset_trace(struct sk_buff *skb)
4270{
Florian Westphal478b3602014-02-15 23:48:45 +01004271#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00004272 skb->nf_trace = 0;
4273#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07004274}
4275
Ye Yin2b5ec1a2017-10-26 16:57:05 +08004276static inline void ipvs_reset(struct sk_buff *skb)
4277{
4278#if IS_ENABLED(CONFIG_IP_VS)
4279 skb->ipvs_property = 0;
4280#endif
4281}
4282
Florian Westphalde8bda12018-12-18 17:15:17 +01004283/* Note: This doesn't put any conntrack info in dst. */
Eric Dumazetb1937222014-09-28 22:18:47 -07004284static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4285 bool copy)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07004286{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07004287#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01004288 dst->_nfct = src->_nfct;
4289 nf_conntrack_get(skb_nfct(src));
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01004290#endif
Florian Westphal478b3602014-02-15 23:48:45 +01004291#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Eric Dumazetb1937222014-09-28 22:18:47 -07004292 if (copy)
4293 dst->nf_trace = src->nf_trace;
Florian Westphal478b3602014-02-15 23:48:45 +01004294#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07004295}
4296
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07004297static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4298{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07004299#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01004300 nf_conntrack_put(skb_nfct(dst));
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01004301#endif
Eric Dumazetb1937222014-09-28 22:18:47 -07004302 __nf_copy(dst, src, true);
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07004303}
4304
James Morris984bc162006-06-09 00:29:17 -07004305#ifdef CONFIG_NETWORK_SECMARK
4306static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4307{
4308 to->secmark = from->secmark;
4309}
4310
4311static inline void skb_init_secmark(struct sk_buff *skb)
4312{
4313 skb->secmark = 0;
4314}
4315#else
4316static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4317{ }
4318
4319static inline void skb_init_secmark(struct sk_buff *skb)
4320{ }
4321#endif
4322
Florian Westphal7af8f4c2018-12-18 17:15:19 +01004323static inline int secpath_exists(const struct sk_buff *skb)
4324{
4325#ifdef CONFIG_XFRM
Florian Westphal41650792018-12-18 17:15:27 +01004326 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
Florian Westphal7af8f4c2018-12-18 17:15:19 +01004327#else
4328 return 0;
4329#endif
4330}
4331
Eric W. Biederman574f7192014-04-01 12:20:24 -07004332static inline bool skb_irq_freeable(const struct sk_buff *skb)
4333{
4334 return !skb->destructor &&
Florian Westphal7af8f4c2018-12-18 17:15:19 +01004335 !secpath_exists(skb) &&
Florian Westphalcb9c6832017-01-23 18:21:56 +01004336 !skb_nfct(skb) &&
Eric W. Biederman574f7192014-04-01 12:20:24 -07004337 !skb->_skb_refdst &&
4338 !skb_has_frag_list(skb);
4339}
4340
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07004341static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4342{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07004343 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07004344}
4345
Stephen Hemminger92477442009-03-21 13:39:26 -07004346static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07004347{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07004348 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07004349}
4350
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07004351static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4352{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07004353 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07004354}
4355
David S. Millerd5a9e242009-01-27 16:22:11 -08004356static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4357{
4358 skb->queue_mapping = rx_queue + 1;
4359}
4360
Stephen Hemminger92477442009-03-21 13:39:26 -07004361static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08004362{
4363 return skb->queue_mapping - 1;
4364}
4365
Stephen Hemminger92477442009-03-21 13:39:26 -07004366static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08004367{
Eric Dumazeta02cec22010-09-22 20:43:57 +00004368 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08004369}
4370
Julian Anastasov4ff06202017-02-06 23:14:12 +02004371static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4372{
4373 skb->dst_pending_confirm = val;
4374}
4375
4376static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4377{
4378 return skb->dst_pending_confirm != 0;
4379}
4380
Florian Westphal2294be0f2018-12-18 17:15:20 +01004381static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04004382{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07004383#ifdef CONFIG_XFRM
Florian Westphal41650792018-12-18 17:15:27 +01004384 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07004385#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07004386 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07004387#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04004388}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07004389
Pravin B Shelar68c33162013-02-14 14:02:41 +00004390/* Keeps track of mac header offset relative to skb->head.
4391 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4392 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07004393 * tunnel skb it points to outer mac header.
4394 * Keeps track of level of encapsulation of network headers.
4395 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00004396struct skb_gso_cb {
Alexander Duyck802ab552016-04-10 21:45:03 -04004397 union {
4398 int mac_offset;
4399 int data_offset;
4400 };
Eric Dumazet3347c962013-10-19 11:42:56 -07004401 int encap_level;
Alexander Duyck76443452016-02-05 15:27:37 -08004402 __wsum csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07004403 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00004404};
Cambda Zhua08e7fd2020-03-26 15:33:14 +08004405#define SKB_GSO_CB_OFFSET 32
4406#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET))
Pravin B Shelar68c33162013-02-14 14:02:41 +00004407
4408static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4409{
4410 return (skb_mac_header(inner_skb) - inner_skb->head) -
4411 SKB_GSO_CB(inner_skb)->mac_offset;
4412}
4413
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00004414static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4415{
4416 int new_headroom, headroom;
4417 int ret;
4418
4419 headroom = skb_headroom(skb);
4420 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4421 if (ret)
4422 return ret;
4423
4424 new_headroom = skb_headroom(skb);
4425 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4426 return 0;
4427}
4428
Alexander Duyck08b64fc2016-02-05 15:27:49 -08004429static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4430{
4431 /* Do not update partial checksums if remote checksum is enabled. */
4432 if (skb->remcsum_offload)
4433 return;
4434
4435 SKB_GSO_CB(skb)->csum = res;
4436 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4437}
4438
Tom Herbert7e2b10c2014-06-04 17:20:02 -07004439/* Compute the checksum for a gso segment. First compute the checksum value
4440 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4441 * then add in skb->csum (checksum from csum_start to end of packet).
4442 * skb->csum and csum_start are then updated to reflect the checksum of the
4443 * resultant packet starting from the transport header-- the resultant checksum
4444 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4445 * header.
4446 */
4447static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4448{
Alexander Duyck76443452016-02-05 15:27:37 -08004449 unsigned char *csum_start = skb_transport_header(skb);
4450 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4451 __wsum partial = SKB_GSO_CB(skb)->csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07004452
Alexander Duyck76443452016-02-05 15:27:37 -08004453 SKB_GSO_CB(skb)->csum = res;
4454 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07004455
Alexander Duyck76443452016-02-05 15:27:37 -08004456 return csum_fold(csum_partial(csum_start, plen, partial));
Tom Herbert7e2b10c2014-06-04 17:20:02 -07004457}
4458
David S. Millerbdcc0922012-03-07 20:53:36 -05004459static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07004460{
4461 return skb_shinfo(skb)->gso_size;
4462}
4463
Eric Dumazet36a8f392013-09-29 01:21:32 -07004464/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05004465static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02004466{
4467 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4468}
4469
Daniel Axtensd02f51c2018-03-03 03:03:46 +01004470/* Note: Should be called only if skb_is_gso(skb) is true */
4471static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4472{
4473 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4474}
4475
Willem de Bruijn4c3024d2019-03-06 14:35:15 -05004476/* Note: Should be called only if skb_is_gso(skb) is true */
Willem de Bruijnb90efd22019-02-07 14:54:16 -05004477static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4478{
Willem de Bruijn4c3024d2019-03-06 14:35:15 -05004479 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
Willem de Bruijnb90efd22019-02-07 14:54:16 -05004480}
4481
Daniel Borkmann5293efe2016-08-18 01:00:39 +02004482static inline void skb_gso_reset(struct sk_buff *skb)
4483{
4484 skb_shinfo(skb)->gso_size = 0;
4485 skb_shinfo(skb)->gso_segs = 0;
4486 skb_shinfo(skb)->gso_type = 0;
4487}
4488
Daniel Axtensd02f51c2018-03-03 03:03:46 +01004489static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4490 u16 increment)
4491{
4492 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4493 return;
4494 shinfo->gso_size += increment;
4495}
4496
4497static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4498 u16 decrement)
4499{
4500 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4501 return;
4502 shinfo->gso_size -= decrement;
4503}
4504
Joe Perches7965bd42013-09-26 14:48:15 -07004505void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07004506
4507static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4508{
4509 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4510 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04004511 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4512
Alexander Duyckb78462e2010-06-02 12:24:37 +00004513 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4514 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07004515 __skb_warn_lro_forwarding(skb);
4516 return true;
4517 }
4518 return false;
4519}
4520
Herbert Xu35fc92a2007-03-26 23:22:20 -07004521static inline void skb_forward_csum(struct sk_buff *skb)
4522{
4523 /* Unfortunately we don't support this one. Any brave souls? */
4524 if (skb->ip_summed == CHECKSUM_COMPLETE)
4525 skb->ip_summed = CHECKSUM_NONE;
4526}
4527
Eric Dumazetbc8acf22010-09-02 13:07:41 -07004528/**
4529 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4530 * @skb: skb to check
4531 *
4532 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4533 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4534 * use this helper, to document places where we make this assertion.
4535 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04004536static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07004537{
4538#ifdef DEBUG
4539 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4540#endif
4541}
4542
Rusty Russellf35d9d82008-02-04 23:49:54 -05004543bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00004544
Paul Durranted1f50c2014-01-09 10:02:46 +00004545int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
Linus Lüssing9afd85c2015-05-02 14:01:07 +02004546struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4547 unsigned int transport_len,
4548 __sum16(*skb_chkf)(struct sk_buff *skb));
Paul Durranted1f50c2014-01-09 10:02:46 +00004549
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00004550/**
4551 * skb_head_is_locked - Determine if the skb->head is locked down
4552 * @skb: skb to check
4553 *
4554 * The head on skbs build around a head frag can be removed if they are
4555 * not cloned. This function returns true if the skb head is locked down
4556 * due to either being allocated via kmalloc, or by being a clone with
4557 * multiple references to the head.
4558 */
4559static inline bool skb_head_is_locked(const struct sk_buff *skb)
4560{
4561 return !skb->head_frag || skb_cloned(skb);
4562}
Florian Westphalfe6cc552014-02-13 23:09:12 +01004563
Edward Cree179bc672016-02-11 20:48:04 +00004564/* Local Checksum Offload.
4565 * Compute outer checksum based on the assumption that the
4566 * inner checksum will be offloaded later.
Otto Sabartd0dcde62019-01-06 00:29:15 +01004567 * See Documentation/networking/checksum-offloads.rst for
Edward Creee8ae7b02016-02-11 21:03:37 +00004568 * explanation of how this works.
Edward Cree179bc672016-02-11 20:48:04 +00004569 * Fill in outer checksum adjustment (e.g. with sum of outer
4570 * pseudo-header) before calling.
4571 * Also ensure that inner checksum is in linear data area.
4572 */
4573static inline __wsum lco_csum(struct sk_buff *skb)
4574{
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004575 unsigned char *csum_start = skb_checksum_start(skb);
4576 unsigned char *l4_hdr = skb_transport_header(skb);
4577 __wsum partial;
Edward Cree179bc672016-02-11 20:48:04 +00004578
4579 /* Start with complement of inner checksum adjustment */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004580 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4581 skb->csum_offset));
4582
Edward Cree179bc672016-02-11 20:48:04 +00004583 /* Add in checksum of our headers (incl. outer checksum
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004584 * adjustment filled in by caller) and return result.
Edward Cree179bc672016-02-11 20:48:04 +00004585 */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004586 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
Edward Cree179bc672016-02-11 20:48:04 +00004587}
4588
Pablo Neira Ayuso2c646052020-03-25 13:47:18 +01004589static inline bool skb_is_redirected(const struct sk_buff *skb)
4590{
4591#ifdef CONFIG_NET_REDIRECT
4592 return skb->redirected;
4593#else
4594 return false;
4595#endif
4596}
4597
4598static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress)
4599{
4600#ifdef CONFIG_NET_REDIRECT
4601 skb->redirected = 1;
4602 skb->from_ingress = from_ingress;
4603 if (skb->from_ingress)
4604 skb->tstamp = 0;
4605#endif
4606}
4607
4608static inline void skb_reset_redirect(struct sk_buff *skb)
4609{
4610#ifdef CONFIG_NET_REDIRECT
4611 skb->redirected = 0;
4612#endif
4613}
4614
Linus Torvalds1da177e2005-04-16 15:20:36 -07004615#endif /* __KERNEL__ */
4616#endif /* _LINUX_SKBUFF_H */