blob: 3000c526f5526c06edadc598bead4cc5fd954f79 [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
Linus Torvalds1da177e2005-04-16 15:20:36 -070043
Tom Herbert7a6ae712015-12-14 11:19:47 -080044/* The interface for checksum offload between the stack and networking drivers
45 * is as follows...
46 *
47 * A. IP checksum related features
48 *
49 * Drivers advertise checksum offload capabilities in the features of a device.
Dexuan Cuidb1f00f2020-04-05 18:59:24 -070050 * From the stack's point of view these are capabilities offered by the driver.
51 * A driver typically only advertises features that it is capable of offloading
Tom Herbert7a6ae712015-12-14 11:19:47 -080052 * to its device.
53 *
54 * The checksum related features are:
55 *
56 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
57 * IP (one's complement) checksum for any combination
58 * of protocols or protocol layering. The checksum is
59 * computed and set in a packet per the CHECKSUM_PARTIAL
60 * interface (see below).
61 *
62 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
63 * TCP or UDP packets over IPv4. These are specifically
64 * unencapsulated packets of the form IPv4|TCP or
65 * IPv4|UDP where the Protocol field in the IPv4 header
Dexuan Cuidb1f00f2020-04-05 18:59:24 -070066 * is TCP or UDP. The IPv4 header may contain IP options.
Tom Herbert7a6ae712015-12-14 11:19:47 -080067 * This feature cannot be set in features for a device
68 * with NETIF_F_HW_CSUM also set. This feature is being
69 * DEPRECATED (see below).
70 *
71 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
72 * TCP or UDP packets over IPv6. These are specifically
73 * unencapsulated packets of the form IPv6|TCP or
74 * IPv4|UDP where the Next Header field in the IPv6
75 * header is either TCP or UDP. IPv6 extension headers
76 * are not supported with this feature. This feature
77 * cannot be set in features for a device with
78 * NETIF_F_HW_CSUM also set. This feature is being
79 * DEPRECATED (see below).
80 *
81 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
Dexuan Cuidb1f00f2020-04-05 18:59:24 -070082 * This flag is only used to disable the RX checksum
Tom Herbert7a6ae712015-12-14 11:19:47 -080083 * feature for a device. The stack will accept receive
84 * checksum indication in packets received on a device
85 * regardless of whether NETIF_F_RXCSUM is set.
86 *
87 * B. Checksumming of received packets by device. Indication of checksum
Dexuan Cuidb1f00f2020-04-05 18:59:24 -070088 * verification is set in skb->ip_summed. Possible values are:
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010089 *
90 * CHECKSUM_NONE:
91 *
Tom Herbert7a6ae712015-12-14 11:19:47 -080092 * Device did not checksum this packet e.g. due to lack of capabilities.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010093 * The packet contains full (though not verified) checksum in packet but
94 * not in skb->csum. Thus, skb->csum is undefined in this case.
95 *
96 * CHECKSUM_UNNECESSARY:
97 *
98 * The hardware you're dealing with doesn't calculate the full checksum
99 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
Tom Herbert77cffe22014-08-27 21:26:46 -0700100 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
101 * if their checksums are okay. skb->csum is still undefined in this case
Tom Herbert7a6ae712015-12-14 11:19:47 -0800102 * though. A driver or device must never modify the checksum field in the
103 * packet even if checksum is verified.
Tom Herbert77cffe22014-08-27 21:26:46 -0700104 *
105 * CHECKSUM_UNNECESSARY is applicable to following protocols:
106 * TCP: IPv6 and IPv4.
107 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
108 * zero UDP checksum for either IPv4 or IPv6, the networking stack
109 * may perform further validation in this case.
110 * GRE: only if the checksum is present in the header.
111 * SCTP: indicates the CRC in SCTP header has been validated.
Davide Carattib4759dc2017-05-18 15:44:43 +0200112 * FCOE: indicates the CRC in FC frame has been validated.
Tom Herbert77cffe22014-08-27 21:26:46 -0700113 *
114 * skb->csum_level indicates the number of consecutive checksums found in
115 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
116 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
117 * and a device is able to verify the checksums for UDP (possibly zero),
Dexuan Cuidb1f00f2020-04-05 18:59:24 -0700118 * GRE (checksum flag is set) and TCP, skb->csum_level would be set to
Tom Herbert77cffe22014-08-27 21:26:46 -0700119 * two. If the device were only able to verify the UDP checksum and not
Dexuan Cuidb1f00f2020-04-05 18:59:24 -0700120 * GRE, either because it doesn't support GRE checksum or because GRE
Tom Herbert77cffe22014-08-27 21:26:46 -0700121 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
122 * not considered in this case).
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100123 *
124 * CHECKSUM_COMPLETE:
125 *
126 * This is the most generic way. The device supplied checksum of the _whole_
Dexuan Cuidb1f00f2020-04-05 18:59:24 -0700127 * packet as seen by netif_rx() and fills in skb->csum. This means the
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100128 * hardware doesn't need to parse L3/L4 headers to implement this.
129 *
Davide Carattib4759dc2017-05-18 15:44:43 +0200130 * Notes:
131 * - Even if device supports only some protocols, but is able to produce
132 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
133 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100134 *
135 * CHECKSUM_PARTIAL:
136 *
Tom Herbert6edec0e2015-02-10 16:30:28 -0800137 * A checksum is set up to be offloaded to a device as described in the
138 * output description for CHECKSUM_PARTIAL. This may occur on a packet
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100139 * received directly from another Linux OS, e.g., a virtualized Linux kernel
Tom Herbert6edec0e2015-02-10 16:30:28 -0800140 * on the same host, or it may be set in the input path in GRO or remote
141 * checksum offload. For the purposes of checksum verification, the checksum
142 * referred to by skb->csum_start + skb->csum_offset and any preceding
143 * checksums in the packet are considered verified. Any checksums in the
144 * packet that are after the checksum being offloaded are not considered to
145 * be verified.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100146 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800147 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
148 * in the skb->ip_summed for a packet. Values are:
149 *
150 * CHECKSUM_PARTIAL:
151 *
152 * The driver is required to checksum the packet as seen by hard_start_xmit()
153 * from skb->csum_start up to the end, and to record/write the checksum at
154 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
155 * csum_start and csum_offset values are valid values given the length and
Dexuan Cuidb1f00f2020-04-05 18:59:24 -0700156 * offset of the packet, but it should not attempt to validate that the
157 * checksum refers to a legitimate transport layer checksum -- it is the
Tom Herbert7a6ae712015-12-14 11:19:47 -0800158 * purview of the stack to validate that csum_start and csum_offset are set
159 * correctly.
160 *
161 * When the stack requests checksum offload for a packet, the driver MUST
162 * ensure that the checksum is set correctly. A driver can either offload the
163 * checksum calculation to the device, or call skb_checksum_help (in the case
164 * that the device does not support offload for a particular checksum).
165 *
166 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
167 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
Davide Caratti43c26a12017-05-18 15:44:41 +0200168 * checksum offload capability.
169 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based
170 * on network device checksumming capabilities: if a packet does not match
171 * them, skb_checksum_help or skb_crc32c_help (depending on the value of
172 * csum_not_inet, see item D.) is called to resolve the checksum.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100173 *
174 * CHECKSUM_NONE:
175 *
176 * The skb was already checksummed by the protocol, or a checksum is not
177 * required.
178 *
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100179 * CHECKSUM_UNNECESSARY:
180 *
Dexuan Cuidb1f00f2020-04-05 18:59:24 -0700181 * This has the same meaning as CHECKSUM_NONE for checksum offload on
Tom Herbert7a6ae712015-12-14 11:19:47 -0800182 * output.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100183 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800184 * CHECKSUM_COMPLETE:
185 * Not used in checksum output. If a driver observes a packet with this value
Dexuan Cuidb1f00f2020-04-05 18:59:24 -0700186 * set in skbuff, it should treat the packet as if CHECKSUM_NONE were set.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100187 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800188 * D. Non-IP checksum (CRC) offloads
189 *
190 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
191 * offloading the SCTP CRC in a packet. To perform this offload the stack
Dexuan Cuidb1f00f2020-04-05 18:59:24 -0700192 * will set csum_start and csum_offset accordingly, set ip_summed to
Davide Carattidba00302017-05-18 15:44:40 +0200193 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
194 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
195 * A driver that supports both IP checksum offload and SCTP CRC32c offload
196 * must verify which offload is configured for a packet by testing the
197 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
198 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
Tom Herbert7a6ae712015-12-14 11:19:47 -0800199 *
200 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
201 * offloading the FCOE CRC in a packet. To perform this offload the stack
202 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
Dexuan Cuidb1f00f2020-04-05 18:59:24 -0700203 * accordingly. Note that there is no indication in the skbuff that the
204 * CHECKSUM_PARTIAL refers to an FCOE checksum, so a driver that supports
Tom Herbert7a6ae712015-12-14 11:19:47 -0800205 * both IP checksum offload and FCOE CRC offload must verify which offload
Dexuan Cuidb1f00f2020-04-05 18:59:24 -0700206 * is configured for a packet, presumably by inspecting packet headers.
Tom Herbert7a6ae712015-12-14 11:19:47 -0800207 *
208 * E. Checksumming on output with GSO.
209 *
210 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
211 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
212 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
213 * part of the GSO operation is implied. If a checksum is being offloaded
Dexuan Cuidb1f00f2020-04-05 18:59:24 -0700214 * with GSO then ip_summed is CHECKSUM_PARTIAL, and both csum_start and
215 * csum_offset are set to refer to the outermost checksum being offloaded
216 * (two offloaded checksums are possible with UDP encapsulation).
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100217 */
218
Herbert Xu60476372007-04-09 11:59:39 -0700219/* Don't change this without changing skb_csum_unnecessary! */
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100220#define CHECKSUM_NONE 0
221#define CHECKSUM_UNNECESSARY 1
222#define CHECKSUM_COMPLETE 2
223#define CHECKSUM_PARTIAL 3
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224
Tom Herbert77cffe22014-08-27 21:26:46 -0700225/* Maximum value in skb->csum_level */
226#define SKB_MAX_CSUM_LEVEL 3
227
Tobias Klauser0bec8c82014-07-22 12:06:23 +0200228#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
David S. Millerfc910a22007-03-25 20:27:59 -0700229#define SKB_WITH_OVERHEAD(X) \
Herbert Xudeea84b2007-10-21 16:27:46 -0700230 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
David S. Millerfc910a22007-03-25 20:27:59 -0700231#define SKB_MAX_ORDER(X, ORDER) \
232 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
234#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
235
Eric Dumazet87fb4b72011-10-13 07:28:54 +0000236/* return minimum truesize of one skb containing X bytes of data */
237#define SKB_TRUESIZE(X) ((X) + \
238 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
239 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
240
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241struct net_device;
David Howells716ea3a2007-04-02 20:19:53 -0700242struct scatterlist;
Jens Axboe9c55e012007-11-06 23:30:13 -0800243struct pipe_inode_info;
Herbert Xua8f820aa2014-11-07 21:22:22 +0800244struct iov_iter;
Alexander Duyckfd11a832014-12-09 19:40:49 -0800245struct napi_struct;
Petar Penkovd58e4682018-09-14 07:46:18 -0700246struct bpf_prog;
247union bpf_attr;
Florian Westphaldf5042f2018-12-18 17:15:16 +0100248struct skb_ext;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249
Pablo Neira Ayuso34666d42014-09-18 11:29:03 +0200250#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251struct nf_bridge_info {
Florian Westphal3eaf4022015-04-02 14:31:44 +0200252 enum {
253 BRNF_PROTO_UNCHANGED,
254 BRNF_PROTO_8021Q,
255 BRNF_PROTO_PPPOE
Florian Westphal7fb48c52015-05-03 22:05:28 +0200256 } orig_proto:8;
Florian Westphal72b1e5e2015-07-23 16:21:30 +0200257 u8 pkt_otherhost:1;
258 u8 in_prerouting:1;
259 u8 bridged_dnat:1;
Bernhard Thaler411ffb42015-05-30 15:28:28 +0200260 __u16 frag_max_size;
Eric Dumazetbf1ac5c2012-04-18 23:19:25 +0000261 struct net_device *physindev;
Florian Westphal63cdbc02015-09-14 17:06:27 +0200262
263 /* always valid & non-NULL from FORWARD on, for physdev match */
264 struct net_device *physoutdev;
Florian Westphal7fb48c52015-05-03 22:05:28 +0200265 union {
Florian Westphal72b1e5e2015-07-23 16:21:30 +0200266 /* prerouting: detect dnat in orig/reply direction */
Bernhard Thaler72b31f72015-05-30 15:27:40 +0200267 __be32 ipv4_daddr;
268 struct in6_addr ipv6_daddr;
Florian Westphal72b1e5e2015-07-23 16:21:30 +0200269
270 /* after prerouting + nat detected: store original source
271 * mac since neigh resolution overwrites it, only used while
272 * skb is out in neigh layer.
273 */
274 char neigh_header[8];
Bernhard Thaler72b31f72015-05-30 15:27:40 +0200275 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700276};
277#endif
278
Paul Blakey95a72332019-09-04 16:56:37 +0300279#if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
280/* Chain in tc_skb_ext will be used to share the tc chain with
281 * ovs recirc_id. It will be set to the current chain by tc
282 * and read by ovs to recirc_id.
283 */
284struct tc_skb_ext {
285 __u32 chain;
286};
287#endif
288
Linus Torvalds1da177e2005-04-16 15:20:36 -0700289struct sk_buff_head {
290 /* These two members must be first. */
291 struct sk_buff *next;
292 struct sk_buff *prev;
293
294 __u32 qlen;
295 spinlock_t lock;
296};
297
298struct sk_buff;
299
Ian Campbell9d4dde52011-12-22 23:39:14 +0000300/* To allow 64K frame to be packed as single skb without frag_list we
301 * require 64K/PAGE_SIZE pages plus 1 additional page to allow for
302 * buffers which do not start on a page boundary.
303 *
304 * Since GRO uses frags we allocate at least 16 regardless of page
305 * size.
Anton Blancharda715dea2011-03-27 14:57:26 +0000306 */
Ian Campbell9d4dde52011-12-22 23:39:14 +0000307#if (65536/PAGE_SIZE + 1) < 16
David S. Millereec00952011-03-29 23:34:08 -0700308#define MAX_SKB_FRAGS 16UL
Anton Blancharda715dea2011-03-27 14:57:26 +0000309#else
Ian Campbell9d4dde52011-12-22 23:39:14 +0000310#define MAX_SKB_FRAGS (65536/PAGE_SIZE + 1)
Anton Blancharda715dea2011-03-27 14:57:26 +0000311#endif
Hans Westgaard Ry5f74f82e2016-02-03 09:26:57 +0100312extern int sysctl_max_skb_frags;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313
Marcelo Ricardo Leitner3953c462016-06-02 15:05:40 -0300314/* Set skb_shinfo(skb)->gso_size to this in case you want skb_segment to
315 * segment using its current segmentation instead.
316 */
317#define GSO_BY_FRAGS 0xFFFF
318
Matthew Wilcox (Oracle)8842d282019-07-22 20:08:31 -0700319typedef struct bio_vec skb_frag_t;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700320
Pedro Tammela161e6132019-03-05 11:35:54 -0300321/**
Jonathan Lemon7240b602019-07-30 07:40:32 -0700322 * skb_frag_size() - Returns the size of a skb fragment
Pedro Tammela161e6132019-03-05 11:35:54 -0300323 * @frag: skb fragment
324 */
Eric Dumazet9e903e02011-10-18 21:00:24 +0000325static inline unsigned int skb_frag_size(const skb_frag_t *frag)
326{
Matthew Wilcox (Oracle)b8b576a2019-07-22 20:08:30 -0700327 return frag->bv_len;
Eric Dumazet9e903e02011-10-18 21:00:24 +0000328}
329
Pedro Tammela161e6132019-03-05 11:35:54 -0300330/**
Jonathan Lemon7240b602019-07-30 07:40:32 -0700331 * skb_frag_size_set() - Sets the size of a skb fragment
Pedro Tammela161e6132019-03-05 11:35:54 -0300332 * @frag: skb fragment
333 * @size: size of fragment
334 */
Eric Dumazet9e903e02011-10-18 21:00:24 +0000335static inline void skb_frag_size_set(skb_frag_t *frag, unsigned int size)
336{
Matthew Wilcox (Oracle)b8b576a2019-07-22 20:08:30 -0700337 frag->bv_len = size;
Eric Dumazet9e903e02011-10-18 21:00:24 +0000338}
339
Pedro Tammela161e6132019-03-05 11:35:54 -0300340/**
Jonathan Lemon7240b602019-07-30 07:40:32 -0700341 * skb_frag_size_add() - Increments the size of a skb fragment by @delta
Pedro Tammela161e6132019-03-05 11:35:54 -0300342 * @frag: skb fragment
343 * @delta: value to add
344 */
Eric Dumazet9e903e02011-10-18 21:00:24 +0000345static inline void skb_frag_size_add(skb_frag_t *frag, int delta)
346{
Matthew Wilcox (Oracle)b8b576a2019-07-22 20:08:30 -0700347 frag->bv_len += delta;
Eric Dumazet9e903e02011-10-18 21:00:24 +0000348}
349
Pedro Tammela161e6132019-03-05 11:35:54 -0300350/**
Jonathan Lemon7240b602019-07-30 07:40:32 -0700351 * skb_frag_size_sub() - Decrements the size of a skb fragment by @delta
Pedro Tammela161e6132019-03-05 11:35:54 -0300352 * @frag: skb fragment
353 * @delta: value to subtract
354 */
Eric Dumazet9e903e02011-10-18 21:00:24 +0000355static inline void skb_frag_size_sub(skb_frag_t *frag, int delta)
356{
Matthew Wilcox (Oracle)b8b576a2019-07-22 20:08:30 -0700357 frag->bv_len -= delta;
Eric Dumazet9e903e02011-10-18 21:00:24 +0000358}
359
Pedro Tammela161e6132019-03-05 11:35:54 -0300360/**
361 * skb_frag_must_loop - Test if %p is a high memory page
362 * @p: fragment's page
363 */
Willem de Bruijnc613c202017-07-31 08:15:47 -0400364static inline bool skb_frag_must_loop(struct page *p)
365{
366#if defined(CONFIG_HIGHMEM)
367 if (PageHighMem(p))
368 return true;
369#endif
370 return false;
371}
372
373/**
374 * skb_frag_foreach_page - loop over pages in a fragment
375 *
376 * @f: skb frag to operate on
Matthew Wilcox (Oracle)1dfa5bd2019-07-22 20:08:29 -0700377 * @f_off: offset from start of f->bv_page
Willem de Bruijnc613c202017-07-31 08:15:47 -0400378 * @f_len: length from f_off to loop over
379 * @p: (temp var) current page
380 * @p_off: (temp var) offset from start of current page,
381 * non-zero only on first page.
382 * @p_len: (temp var) length in current page,
383 * < PAGE_SIZE only on first and last page.
384 * @copied: (temp var) length so far, excluding current p_len.
385 *
386 * A fragment can hold a compound page, in which case per-page
387 * operations, notably kmap_atomic, must be called for each
388 * regular page.
389 */
390#define skb_frag_foreach_page(f, f_off, f_len, p, p_off, p_len, copied) \
391 for (p = skb_frag_page(f) + ((f_off) >> PAGE_SHIFT), \
392 p_off = (f_off) & (PAGE_SIZE - 1), \
393 p_len = skb_frag_must_loop(p) ? \
394 min_t(u32, f_len, PAGE_SIZE - p_off) : f_len, \
395 copied = 0; \
396 copied < f_len; \
397 copied += p_len, p++, p_off = 0, \
398 p_len = min_t(u32, f_len - copied, PAGE_SIZE)) \
399
Patrick Ohlyac45f602009-02-12 05:03:37 +0000400#define HAVE_HW_TIME_STAMP
401
402/**
Randy Dunlapd3a21be2009-03-02 03:15:58 -0800403 * struct skb_shared_hwtstamps - hardware time stamps
Patrick Ohlyac45f602009-02-12 05:03:37 +0000404 * @hwtstamp: hardware time stamp transformed into duration
405 * since arbitrary point in time
Patrick Ohlyac45f602009-02-12 05:03:37 +0000406 *
407 * Software time stamps generated by ktime_get_real() are stored in
Willem de Bruijn4d276eb2014-07-25 18:01:32 -0400408 * skb->tstamp.
Patrick Ohlyac45f602009-02-12 05:03:37 +0000409 *
410 * hwtstamps can only be compared against other hwtstamps from
411 * the same device.
412 *
413 * This structure is attached to packets as part of the
414 * &skb_shared_info. Use skb_hwtstamps() to get a pointer.
415 */
416struct skb_shared_hwtstamps {
417 ktime_t hwtstamp;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000418};
419
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000420/* Definitions for tx_flags in struct skb_shared_info */
421enum {
422 /* generate hardware time stamp */
423 SKBTX_HW_TSTAMP = 1 << 0,
424
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400425 /* generate software time stamp when queueing packet to NIC */
Oliver Hartkopp2244d072010-08-17 08:59:14 +0000426 SKBTX_SW_TSTAMP = 1 << 1,
427
428 /* device driver is going to provide hardware time stamp */
429 SKBTX_IN_PROGRESS = 1 << 2,
430
Shirley Maa6686f22011-07-06 12:22:12 +0000431 /* device driver supports TX zero-copy buffers */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000432 SKBTX_DEV_ZEROCOPY = 1 << 3,
Johannes Berg6e3e9392011-11-09 10:15:42 +0100433
434 /* generate wifi status information (where possible) */
Eric Dumazet62b1a8a2012-06-14 06:42:44 +0000435 SKBTX_WIFI_STATUS = 1 << 4,
Pravin B Shelarc9af6db2013-02-11 09:27:41 +0000436
437 /* This indicates at least one fragment might be overwritten
438 * (as in vmsplice(), sendfile() ...)
439 * If we need to compute a TX checksum, we'll need to copy
440 * all frags to avoid possible bad checksum
441 */
442 SKBTX_SHARED_FRAG = 1 << 5,
Willem de Bruijne7fd2882014-08-04 22:11:48 -0400443
444 /* generate software time stamp when entering packet scheduling */
445 SKBTX_SCHED_TSTAMP = 1 << 6,
Shirley Maa6686f22011-07-06 12:22:12 +0000446};
447
Willem de Bruijn52267792017-08-03 16:29:39 -0400448#define SKBTX_ZEROCOPY_FRAG (SKBTX_DEV_ZEROCOPY | SKBTX_SHARED_FRAG)
Willem de Bruijne1c8a602014-08-04 22:11:50 -0400449#define SKBTX_ANY_SW_TSTAMP (SKBTX_SW_TSTAMP | \
Soheil Hassas Yeganeh0a2cf202016-04-27 23:39:01 -0400450 SKBTX_SCHED_TSTAMP)
Willem de Bruijnf24b9be2014-08-04 22:11:45 -0400451#define SKBTX_ANY_TSTAMP (SKBTX_HW_TSTAMP | SKBTX_ANY_SW_TSTAMP)
452
Shirley Maa6686f22011-07-06 12:22:12 +0000453/*
454 * The callback notifies userspace to release buffers when skb DMA is done in
455 * lower device, the skb last reference should be 0 when calling this.
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000456 * The zerocopy_success argument is true if zero copy transmit occurred,
457 * false on data copy or out of memory error caused by data copy attempt.
Michael S. Tsirkinca8f4fb2012-04-09 00:24:02 +0000458 * The ctx field is used to track device context.
459 * The desc field is used to track userspace buffer index.
Shirley Maa6686f22011-07-06 12:22:12 +0000460 */
461struct ubuf_info {
Michael S. Tsirkine19d6762012-11-01 09:16:22 +0000462 void (*callback)(struct ubuf_info *, bool zerocopy_success);
Willem de Bruijn4ab6c992017-08-03 16:29:42 -0400463 union {
464 struct {
465 unsigned long desc;
466 void *ctx;
467 };
468 struct {
469 u32 id;
470 u16 len;
471 u16 zerocopy:1;
472 u32 bytelen;
473 };
474 };
Eric Dumazetc1d1b432017-08-31 16:48:22 -0700475 refcount_t refcnt;
Willem de Bruijna91dbff2017-08-03 16:29:43 -0400476
477 struct mmpin {
478 struct user_struct *user;
479 unsigned int num_pg;
480 } mmp;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000481};
482
Willem de Bruijn52267792017-08-03 16:29:39 -0400483#define skb_uarg(SKB) ((struct ubuf_info *)(skb_shinfo(SKB)->destructor_arg))
484
Sowmini Varadhan6f89dbc2018-02-15 10:49:32 -0800485int mm_account_pinned_pages(struct mmpin *mmp, size_t size);
486void mm_unaccount_pinned_pages(struct mmpin *mmp);
487
Willem de Bruijn52267792017-08-03 16:29:39 -0400488struct ubuf_info *sock_zerocopy_alloc(struct sock *sk, size_t size);
Willem de Bruijn4ab6c992017-08-03 16:29:42 -0400489struct ubuf_info *sock_zerocopy_realloc(struct sock *sk, size_t size,
490 struct ubuf_info *uarg);
Willem de Bruijn52267792017-08-03 16:29:39 -0400491
492static inline void sock_zerocopy_get(struct ubuf_info *uarg)
493{
Eric Dumazetc1d1b432017-08-31 16:48:22 -0700494 refcount_inc(&uarg->refcnt);
Willem de Bruijn52267792017-08-03 16:29:39 -0400495}
496
497void sock_zerocopy_put(struct ubuf_info *uarg);
Willem de Bruijn52900d22018-11-30 15:32:40 -0500498void sock_zerocopy_put_abort(struct ubuf_info *uarg, bool have_uref);
Willem de Bruijn52267792017-08-03 16:29:39 -0400499
500void sock_zerocopy_callback(struct ubuf_info *uarg, bool success);
501
Willem de Bruijnb5947e52018-11-30 15:32:39 -0500502int skb_zerocopy_iter_dgram(struct sk_buff *skb, struct msghdr *msg, int len);
Willem de Bruijn52267792017-08-03 16:29:39 -0400503int skb_zerocopy_iter_stream(struct sock *sk, struct sk_buff *skb,
504 struct msghdr *msg, int len,
505 struct ubuf_info *uarg);
506
Linus Torvalds1da177e2005-04-16 15:20:36 -0700507/* This data is invariant across clones and lives at
508 * the end of the header data, ie. at skb->end.
509 */
510struct skb_shared_info {
Daniel Borkmannde8f3a82017-09-25 02:25:51 +0200511 __u8 __unused;
512 __u8 meta_len;
513 __u8 nr_frags;
Ian Campbell9f42f122012-01-05 07:13:39 +0000514 __u8 tx_flags;
Herbert Xu79671682006-06-22 02:40:14 -0700515 unsigned short gso_size;
516 /* Warning: this field is not always filled in (UFO)! */
517 unsigned short gso_segs;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700518 struct sk_buff *frag_list;
Patrick Ohlyac45f602009-02-12 05:03:37 +0000519 struct skb_shared_hwtstamps hwtstamps;
Steffen Klassert7f564522017-04-08 20:36:24 +0200520 unsigned int gso_type;
Willem de Bruijn09c2d252014-08-04 22:11:47 -0400521 u32 tskey;
Eric Dumazetec7d2f22010-05-05 01:07:37 -0700522
523 /*
524 * Warning : all fields before dataref are cleared in __alloc_skb()
525 */
526 atomic_t dataref;
527
Johann Baudy69e3c752009-05-18 22:11:22 -0700528 /* Intermediate layers must ensure that destructor_arg
529 * remains valid until skb destructor */
530 void * destructor_arg;
Shirley Maa6686f22011-07-06 12:22:12 +0000531
Eric Dumazetfed66382010-07-22 19:09:08 +0000532 /* must be last field, see pskb_expand_head() */
533 skb_frag_t frags[MAX_SKB_FRAGS];
Linus Torvalds1da177e2005-04-16 15:20:36 -0700534};
535
536/* We divide dataref into two halves. The higher 16 bits hold references
537 * to the payload part of skb->data. The lower 16 bits hold references to
Patrick McHardy334a8132007-06-25 04:35:20 -0700538 * the entire skb->data. A clone of a headerless skb holds the length of
539 * the header in skb->hdr_len.
Linus Torvalds1da177e2005-04-16 15:20:36 -0700540 *
541 * All users must obey the rule that the skb->data reference count must be
542 * greater than or equal to the payload reference count.
543 *
544 * Holding a reference to the payload part means that the user does not
545 * care about modifications to the header part of skb->data.
546 */
547#define SKB_DATAREF_SHIFT 16
548#define SKB_DATAREF_MASK ((1 << SKB_DATAREF_SHIFT) - 1)
549
David S. Millerd179cd12005-08-17 14:57:30 -0700550
551enum {
Vijay Subramanianc8753d52014-10-02 10:00:43 -0700552 SKB_FCLONE_UNAVAILABLE, /* skb has no fclone (from head_cache) */
553 SKB_FCLONE_ORIG, /* orig skb (from fclone_cache) */
554 SKB_FCLONE_CLONE, /* companion fclone skb (from fclone_cache) */
David S. Millerd179cd12005-08-17 14:57:30 -0700555};
556
Herbert Xu79671682006-06-22 02:40:14 -0700557enum {
558 SKB_GSO_TCPV4 = 1 << 0,
Herbert Xu576a30e2006-06-27 13:22:38 -0700559
560 /* This indicates the skb is from an untrusted source. */
David S. Millerd9d30ad2017-07-03 07:31:57 -0700561 SKB_GSO_DODGY = 1 << 1,
Michael Chanb0da85372006-06-29 12:30:00 -0700562
563 /* This indicates the tcp segment has CWR set. */
David S. Millerd9d30ad2017-07-03 07:31:57 -0700564 SKB_GSO_TCP_ECN = 1 << 2,
Herbert Xuf83ef8c2006-06-30 13:37:03 -0700565
David S. Millerd9d30ad2017-07-03 07:31:57 -0700566 SKB_GSO_TCP_FIXEDID = 1 << 3,
Chris Leech01d5b2f2009-02-27 14:06:49 -0800567
David S. Millerd9d30ad2017-07-03 07:31:57 -0700568 SKB_GSO_TCPV6 = 1 << 4,
Pravin B Shelar68c33162013-02-14 14:02:41 +0000569
David S. Millerd9d30ad2017-07-03 07:31:57 -0700570 SKB_GSO_FCOE = 1 << 5,
Pravin B Shelar73136262013-03-07 13:21:51 +0000571
David S. Millerd9d30ad2017-07-03 07:31:57 -0700572 SKB_GSO_GRE = 1 << 6,
Simon Horman0d89d202013-05-23 21:02:52 +0000573
David S. Millerd9d30ad2017-07-03 07:31:57 -0700574 SKB_GSO_GRE_CSUM = 1 << 7,
Eric Dumazetcb32f512013-10-19 11:42:57 -0700575
David S. Millerd9d30ad2017-07-03 07:31:57 -0700576 SKB_GSO_IPXIP4 = 1 << 8,
Eric Dumazet61c1db72013-10-20 20:47:30 -0700577
David S. Millerd9d30ad2017-07-03 07:31:57 -0700578 SKB_GSO_IPXIP6 = 1 << 9,
Tom Herbert0f4f4ff2014-06-04 17:20:16 -0700579
David S. Millerd9d30ad2017-07-03 07:31:57 -0700580 SKB_GSO_UDP_TUNNEL = 1 << 10,
Tom Herbert4749c092014-06-04 17:20:23 -0700581
David S. Millerd9d30ad2017-07-03 07:31:57 -0700582 SKB_GSO_UDP_TUNNEL_CSUM = 1 << 11,
Alexander Duyckcbc53e02016-04-10 21:44:51 -0400583
David S. Millerd9d30ad2017-07-03 07:31:57 -0700584 SKB_GSO_PARTIAL = 1 << 12,
Alexander Duyck802ab552016-04-10 21:45:03 -0400585
David S. Millerd9d30ad2017-07-03 07:31:57 -0700586 SKB_GSO_TUNNEL_REMCSUM = 1 << 13,
Marcelo Ricardo Leitner90017ac2016-06-02 15:05:43 -0300587
David S. Millerd9d30ad2017-07-03 07:31:57 -0700588 SKB_GSO_SCTP = 1 << 14,
Steffen Klassertc7ef8f02017-04-14 10:05:36 +0200589
David S. Millerd9d30ad2017-07-03 07:31:57 -0700590 SKB_GSO_ESP = 1 << 15,
Willem de Bruijn0c19f8462017-11-21 10:22:25 -0500591
592 SKB_GSO_UDP = 1 << 16,
Willem de Bruijnee80d1e2018-04-26 13:42:16 -0400593
594 SKB_GSO_UDP_L4 = 1 << 17,
Steffen Klassert3b335832020-01-25 11:26:42 +0100595
596 SKB_GSO_FRAGLIST = 1 << 18,
Herbert Xu79671682006-06-22 02:40:14 -0700597};
598
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -0700599#if BITS_PER_LONG > 32
600#define NET_SKBUFF_DATA_USES_OFFSET 1
601#endif
602
603#ifdef NET_SKBUFF_DATA_USES_OFFSET
604typedef unsigned int sk_buff_data_t;
605#else
606typedef unsigned char *sk_buff_data_t;
607#endif
608
Pedro Tammela161e6132019-03-05 11:35:54 -0300609/**
Linus Torvalds1da177e2005-04-16 15:20:36 -0700610 * struct sk_buff - socket buffer
611 * @next: Next buffer in list
612 * @prev: Previous buffer in list
Eric Dumazet363ec39232014-02-26 14:02:11 -0800613 * @tstamp: Time we arrived/left
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800614 * @skb_mstamp_ns: (aka @tstamp) earliest departure time; start point
615 * for retransmit timer
Eric Dumazet56b17422014-11-03 08:19:53 -0800616 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800617 * @list: queue head
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700618 * @sk: Socket we are owned by
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800619 * @ip_defrag_offset: (aka @sk) alternate use of @sk, used in
620 * fragmentation management
Linus Torvalds1da177e2005-04-16 15:20:36 -0700621 * @dev: Device we arrived on/are leaving by
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800622 * @dev_scratch: (aka @dev) alternate use of @dev when @dev would be %NULL
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700623 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000624 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700625 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700626 * @len: Length of actual data
627 * @data_len: Data length
628 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700629 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700630 * @csum: Checksum (must include start/offset pair)
631 * @csum_start: Offset from skb->head where checksumming should start
632 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700633 * @priority: Packet queueing priority
WANG Cong60ff7462014-05-04 16:39:18 -0700634 * @ignore_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700635 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700636 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700637 * @nohdr: Payload reference only, must not modify header
638 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700639 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700640 * @ipvs_property: skbuff is owned by ipvs
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800641 * @inner_protocol_type: whether the inner protocol is
642 * ENCAP_TYPE_ETHER or ENCAP_TYPE_IPPROTO
643 * @remcsum_offload: remote checksum offload is enabled
Ido Schimmel875e8932018-12-04 08:15:10 +0000644 * @offload_fwd_mark: Packet was L2-forwarded in hardware
645 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
Willem de Bruijne7246e12017-01-07 17:06:35 -0500646 * @tc_skip_classify: do not classify packet. set by IFB device
Willem de Bruijn8dc07fd2017-01-07 17:06:37 -0500647 * @tc_at_ingress: used within tc_classify to distinguish in/egress
Pablo Neira Ayuso2c646052020-03-25 13:47:18 +0100648 * @redirected: packet was redirected by packet classifier
649 * @from_ingress: packet was redirected from the ingress path
Randy Dunlap31729362008-02-18 20:52:13 -0800650 * @peeked: this packet has been seen already, so stats have been
651 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700652 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700653 * @protocol: Packet protocol from driver
654 * @destructor: Destruct function
Eric Dumazete2080072017-10-04 12:59:58 -0700655 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
Florian Westphala9e419d2017-01-23 18:21:59 +0100656 * @_nfct: Associated connection, if any (with nfctinfo bits)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700657 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800658 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700659 * @tc_index: Traffic control index
Tom Herbert61b905d2014-03-24 15:34:47 -0700660 * @hash: the packet hash
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700661 * @queue_mapping: Queue mapping for multiqueue devices
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800662 * @head_frag: skb was allocated from page fragments,
663 * not allocated by kmalloc() or vmalloc().
Stefano Brivio8b700862018-07-11 14:39:42 +0200664 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
Florian Westphaldf5042f2018-12-18 17:15:16 +0100665 * @active_extensions: active extensions (skb_ext_id types)
Randy Dunlap553a5672008-04-20 10:51:01 -0700666 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700667 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Tom Herbert61b905d2014-03-24 15:34:47 -0700668 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
Changli Gao4ca24622011-08-19 07:26:44 -0700669 * ports.
Tom Herberta3b18dd2014-07-01 21:33:17 -0700670 * @sw_hash: indicates hash was computed in software stack
Johannes Berg6e3e9392011-11-09 10:15:42 +0100671 * @wifi_acked_valid: wifi_acked was set
672 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000673 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800674 * @encapsulation: indicates the inner headers in the skbuff are valid
675 * @encap_hdr_csum: software checksum is needed
676 * @csum_valid: checksum is already valid
Davide Carattidba00302017-05-18 15:44:40 +0200677 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800678 * @csum_complete_sw: checksum was completed by software
679 * @csum_level: indicates the number of consecutive checksums found in
680 * the packet minus one that have been verified as
681 * CHECKSUM_UNNECESSARY (max 3)
Julian Anastasov4ff06202017-02-06 23:14:12 +0200682 * @dst_pending_confirm: need to confirm neighbour
Stefano Brivioa48d1892018-07-17 11:52:57 +0200683 * @decrypted: Decrypted SKB
Pedro Tammela161e6132019-03-05 11:35:54 -0300684 * @napi_id: id of the NAPI struct this skb came from
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800685 * @sender_cpu: (aka @napi_id) source CPU in XPS
James Morris984bc162006-06-09 00:29:17 -0700686 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700687 * @mark: Generic packet mark
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800688 * @reserved_tailroom: (aka @mark) number of bytes of free space available
689 * at the tail of an sk_buff
690 * @vlan_present: VLAN tag is present
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000691 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b2008-07-14 22:49:06 -0700692 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000693 * @inner_protocol: Protocol (encapsulation)
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800694 * @inner_ipproto: (aka @inner_protocol) stores ipproto when
695 * skb->inner_protocol_type == ENCAP_TYPE_IPPROTO;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000696 * @inner_transport_header: Inner transport layer header (encapsulation)
697 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000698 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700699 * @transport_header: Transport layer header
700 * @network_header: Network layer header
701 * @mac_header: Link layer header
702 * @tail: Tail pointer
703 * @end: End pointer
704 * @head: Head of buffer
705 * @data: Data head pointer
706 * @truesize: Buffer size
707 * @users: User count - see {datagram,tcp}.c
Florian Westphaldf5042f2018-12-18 17:15:16 +0100708 * @extensions: allocated extensions, valid if active_extensions is nonzero
Linus Torvalds1da177e2005-04-16 15:20:36 -0700709 */
710
711struct sk_buff {
Eric Dumazet363ec39232014-02-26 14:02:11 -0800712 union {
Eric Dumazet56b17422014-11-03 08:19:53 -0800713 struct {
714 /* These two members must be first. */
715 struct sk_buff *next;
716 struct sk_buff *prev;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000717
Eric Dumazet56b17422014-11-03 08:19:53 -0800718 union {
Eric Dumazetbffa72c2017-09-19 05:14:24 -0700719 struct net_device *dev;
720 /* Some protocols might use this space to store information,
721 * while device pointer would be NULL.
722 * UDP receive path is one user.
723 */
724 unsigned long dev_scratch;
Eric Dumazet56b17422014-11-03 08:19:53 -0800725 };
726 };
Peter Oskolkovfa0f5272018-08-02 23:34:39 +0000727 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
David Millerd4546c22018-06-24 14:13:49 +0900728 struct list_head list;
Eric Dumazet56b17422014-11-03 08:19:53 -0800729 };
Peter Oskolkovfa0f5272018-08-02 23:34:39 +0000730
731 union {
732 struct sock *sk;
733 int ip_defrag_offset;
734 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735
Eric Dumazetc84d9492016-12-08 11:41:55 -0800736 union {
Eric Dumazetbffa72c2017-09-19 05:14:24 -0700737 ktime_t tstamp;
Eric Dumazetd3edd062018-09-21 08:51:50 -0700738 u64 skb_mstamp_ns; /* earliest departure time */
Eric Dumazetc84d9492016-12-08 11:41:55 -0800739 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700740 /*
741 * This is the control buffer. It is free to use for every
742 * layer. Please put your private variables there. If you
743 * want to keep them across layers you have to do a skb_clone()
744 * first. This is owned by whoever has the skb queued ATM.
745 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000746 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700747
Eric Dumazete2080072017-10-04 12:59:58 -0700748 union {
749 struct {
750 unsigned long _skb_refdst;
751 void (*destructor)(struct sk_buff *skb);
752 };
753 struct list_head tcp_tsorted_anchor;
754 };
755
Eric Dumazetb1937222014-09-28 22:18:47 -0700756#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +0100757 unsigned long _nfct;
Eric Dumazetb1937222014-09-28 22:18:47 -0700758#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700759 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700760 data_len;
761 __u16 mac_len,
762 hdr_len;
Eric Dumazetb1937222014-09-28 22:18:47 -0700763
764 /* Following fields are _not_ copied in __copy_skb_header()
765 * Note that queue_mapping is here mostly to fill a hole.
766 */
Eric Dumazetb1937222014-09-28 22:18:47 -0700767 __u16 queue_mapping;
Daniel Borkmann36bbef52016-09-20 00:26:13 +0200768
769/* if you move cloned around you also must adapt those constants */
770#ifdef __BIG_ENDIAN_BITFIELD
771#define CLONED_MASK (1 << 7)
772#else
773#define CLONED_MASK 1
774#endif
775#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
776
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800777 /* private: */
Daniel Borkmann36bbef52016-09-20 00:26:13 +0200778 __u8 __cloned_offset[0];
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800779 /* public: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700780 __u8 cloned:1,
Harald Welte6869c4d2005-08-09 19:24:19 -0700781 nohdr:1,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800782 fclone:2,
Herbert Xua59322b2007-12-05 01:53:40 -0800783 peeked:1,
Eric Dumazetb1937222014-09-28 22:18:47 -0700784 head_frag:1,
Stefano Brivio8b700862018-07-11 14:39:42 +0200785 pfmemalloc:1;
Florian Westphaldf5042f2018-12-18 17:15:16 +0100786#ifdef CONFIG_SKB_EXTENSIONS
787 __u8 active_extensions;
788#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700789 /* fields enclosed in headers_start/headers_end are copied
790 * using a single memcpy() in __copy_skb_header()
791 */
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700792 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700793 __u32 headers_start[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700794 /* public: */
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200795
796/* if you move pkt_type around you also must adapt those constants */
797#ifdef __BIG_ENDIAN_BITFIELD
798#define PKT_TYPE_MAX (7 << 5)
799#else
800#define PKT_TYPE_MAX 7
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100801#endif
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200802#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
803
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800804 /* private: */
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200805 __u8 __pkt_type_offset[0];
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800806 /* public: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700807 __u8 pkt_type:3;
Eric Dumazetb1937222014-09-28 22:18:47 -0700808 __u8 ignore_df:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700809 __u8 nf_trace:1;
810 __u8 ip_summed:2;
811 __u8 ooo_okay:1;
Stefano Brivio8b700862018-07-11 14:39:42 +0200812
Eric Dumazetb1937222014-09-28 22:18:47 -0700813 __u8 l4_hash:1;
814 __u8 sw_hash:1;
815 __u8 wifi_acked_valid:1;
816 __u8 wifi_acked:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700817 __u8 no_fcs:1;
818 /* Indicates the inner headers are valid in the skbuff. */
819 __u8 encapsulation:1;
820 __u8 encap_hdr_csum:1;
821 __u8 csum_valid:1;
Stefano Brivio8b700862018-07-11 14:39:42 +0200822
Michał Mirosław0c4b2d32018-11-10 19:58:36 +0100823#ifdef __BIG_ENDIAN_BITFIELD
824#define PKT_VLAN_PRESENT_BIT 7
825#else
826#define PKT_VLAN_PRESENT_BIT 0
827#endif
828#define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset)
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800829 /* private: */
Michał Mirosław0c4b2d32018-11-10 19:58:36 +0100830 __u8 __pkt_vlan_present_offset[0];
Randy Dunlapd2f273f2020-02-15 15:34:07 -0800831 /* public: */
Michał Mirosław0c4b2d32018-11-10 19:58:36 +0100832 __u8 vlan_present:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700833 __u8 csum_complete_sw:1;
834 __u8 csum_level:2;
Davide Carattidba00302017-05-18 15:44:40 +0200835 __u8 csum_not_inet:1;
Julian Anastasov4ff06202017-02-06 23:14:12 +0200836 __u8 dst_pending_confirm:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700837#ifdef CONFIG_IPV6_NDISC_NODETYPE
838 __u8 ndisc_nodetype:2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700839#endif
Stefano Brivio8b700862018-07-11 14:39:42 +0200840
Michał Mirosław0c4b2d32018-11-10 19:58:36 +0100841 __u8 ipvs_property:1;
Tom Herbert8bce6d72014-09-29 20:22:29 -0700842 __u8 inner_protocol_type:1;
Tom Herberte585f232014-11-04 09:06:54 -0800843 __u8 remcsum_offload:1;
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200844#ifdef CONFIG_NET_SWITCHDEV
845 __u8 offload_fwd_mark:1;
Ido Schimmel875e8932018-12-04 08:15:10 +0000846 __u8 offload_l3_fwd_mark:1;
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200847#endif
Willem de Bruijne7246e12017-01-07 17:06:35 -0500848#ifdef CONFIG_NET_CLS_ACT
849 __u8 tc_skip_classify:1;
Willem de Bruijn8dc07fd2017-01-07 17:06:37 -0500850 __u8 tc_at_ingress:1;
Pablo Neira Ayuso2c646052020-03-25 13:47:18 +0100851#endif
852#ifdef CONFIG_NET_REDIRECT
853 __u8 redirected:1;
854 __u8 from_ingress:1;
Willem de Bruijne7246e12017-01-07 17:06:35 -0500855#endif
Stefano Brivioa48d1892018-07-17 11:52:57 +0200856#ifdef CONFIG_TLS_DEVICE
857 __u8 decrypted:1;
858#endif
Alexander Duyck4031ae62012-01-27 06:22:53 +0000859
Linus Torvalds1da177e2005-04-16 15:20:36 -0700860#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700861 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700862#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200863
Eric Dumazetb1937222014-09-28 22:18:47 -0700864 union {
865 __wsum csum;
866 struct {
867 __u16 csum_start;
868 __u16 csum_offset;
869 };
870 };
871 __u32 priority;
872 int skb_iif;
873 __u32 hash;
874 __be16 vlan_proto;
875 __u16 vlan_tci;
Eric Dumazet2bd82482015-02-03 23:48:24 -0800876#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
877 union {
878 unsigned int napi_id;
879 unsigned int sender_cpu;
880 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700881#endif
James Morris984bc162006-06-09 00:29:17 -0700882#ifdef CONFIG_NETWORK_SECMARK
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200883 __u32 secmark;
James Morris984bc162006-06-09 00:29:17 -0700884#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -0700885
Neil Horman3b885782009-10-12 13:26:31 -0700886 union {
887 __u32 mark;
Eric Dumazet16fad692013-03-14 05:40:32 +0000888 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700889 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700890
Tom Herbert8bce6d72014-09-29 20:22:29 -0700891 union {
892 __be16 inner_protocol;
893 __u8 inner_ipproto;
894 };
895
Simon Horman1a37e412013-05-23 21:02:51 +0000896 __u16 inner_transport_header;
897 __u16 inner_network_header;
898 __u16 inner_mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700899
900 __be16 protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000901 __u16 transport_header;
902 __u16 network_header;
903 __u16 mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700904
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700905 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700906 __u32 headers_end[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700907 /* public: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700908
Linus Torvalds1da177e2005-04-16 15:20:36 -0700909 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700910 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700911 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700912 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700913 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700914 unsigned int truesize;
Reshetova, Elena63354792017-06-30 13:07:58 +0300915 refcount_t users;
Florian Westphaldf5042f2018-12-18 17:15:16 +0100916
917#ifdef CONFIG_SKB_EXTENSIONS
918 /* only useable after checking ->active_extensions != 0 */
919 struct skb_ext *extensions;
920#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700921};
922
923#ifdef __KERNEL__
924/*
925 * Handling routines are only of interest to the kernel
926 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700927
Mel Gormanc93bdd02012-07-31 16:44:19 -0700928#define SKB_ALLOC_FCLONE 0x01
929#define SKB_ALLOC_RX 0x02
Alexander Duyckfd11a832014-12-09 19:40:49 -0800930#define SKB_ALLOC_NAPI 0x04
Mel Gormanc93bdd02012-07-31 16:44:19 -0700931
Pedro Tammela161e6132019-03-05 11:35:54 -0300932/**
933 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
934 * @skb: buffer
935 */
Mel Gormanc93bdd02012-07-31 16:44:19 -0700936static inline bool skb_pfmemalloc(const struct sk_buff *skb)
937{
938 return unlikely(skb->pfmemalloc);
939}
940
Eric Dumazet7fee2262010-05-11 23:19:48 +0000941/*
942 * skb might have a dst pointer attached, refcounted or not.
943 * _skb_refdst low order bit is set if refcount was _not_ taken
944 */
945#define SKB_DST_NOREF 1UL
946#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
947
948/**
949 * skb_dst - returns skb dst_entry
950 * @skb: buffer
951 *
952 * Returns skb dst_entry, regardless of reference taken or not.
953 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000954static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
955{
Pedro Tammela161e6132019-03-05 11:35:54 -0300956 /* If refdst was not refcounted, check we still are in a
Eric Dumazet7fee2262010-05-11 23:19:48 +0000957 * rcu_read_lock section
958 */
959 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
960 !rcu_read_lock_held() &&
961 !rcu_read_lock_bh_held());
962 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000963}
964
Eric Dumazet7fee2262010-05-11 23:19:48 +0000965/**
966 * skb_dst_set - sets skb dst
967 * @skb: buffer
968 * @dst: dst entry
969 *
970 * Sets skb dst, assuming a reference was taken on dst and should
971 * be released by skb_dst_drop()
972 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000973static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
974{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000975 skb->_skb_refdst = (unsigned long)dst;
976}
977
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200978/**
979 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
980 * @skb: buffer
981 * @dst: dst entry
982 *
983 * Sets skb dst, assuming a reference was not taken on dst.
984 * If dst entry is cached, we do not take reference and dst_release
985 * will be avoided by refdst_drop. If dst entry is not cached, we take
986 * reference, so that last dst_release can destroy the dst immediately.
987 */
988static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
989{
Hannes Frederic Sowadbfc4fb2014-12-06 19:19:42 +0100990 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
991 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200992}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000993
994/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300995 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000996 * @skb: buffer
997 */
998static inline bool skb_dst_is_noref(const struct sk_buff *skb)
999{
1000 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +00001001}
1002
Pedro Tammela161e6132019-03-05 11:35:54 -03001003/**
1004 * skb_rtable - Returns the skb &rtable
1005 * @skb: buffer
1006 */
Eric Dumazet511c3f92009-06-02 05:14:27 +00001007static inline struct rtable *skb_rtable(const struct sk_buff *skb)
1008{
Eric Dumazetadf30902009-06-02 05:19:30 +00001009 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +00001010}
1011
Jamal Hadi Salim8b10cab2016-07-02 06:43:14 -04001012/* For mangling skb->pkt_type from user space side from applications
1013 * such as nft, tc, etc, we only allow a conservative subset of
1014 * possible pkt_types to be set.
1015*/
1016static inline bool skb_pkt_type_ok(u32 ptype)
1017{
1018 return ptype <= PACKET_OTHERHOST;
1019}
1020
Pedro Tammela161e6132019-03-05 11:35:54 -03001021/**
1022 * skb_napi_id - Returns the skb's NAPI id
1023 * @skb: buffer
1024 */
Miroslav Lichvar90b602f2017-05-19 17:52:37 +02001025static inline unsigned int skb_napi_id(const struct sk_buff *skb)
1026{
1027#ifdef CONFIG_NET_RX_BUSY_POLL
1028 return skb->napi_id;
1029#else
1030 return 0;
1031#endif
1032}
1033
Pedro Tammela161e6132019-03-05 11:35:54 -03001034/**
1035 * skb_unref - decrement the skb's reference count
1036 * @skb: buffer
1037 *
1038 * Returns true if we can free the skb.
1039 */
Paolo Abeni3889a8032017-06-12 11:23:41 +02001040static inline bool skb_unref(struct sk_buff *skb)
1041{
1042 if (unlikely(!skb))
1043 return false;
Reshetova, Elena63354792017-06-30 13:07:58 +03001044 if (likely(refcount_read(&skb->users) == 1))
Paolo Abeni3889a8032017-06-12 11:23:41 +02001045 smp_rmb();
Reshetova, Elena63354792017-06-30 13:07:58 +03001046 else if (likely(!refcount_dec_and_test(&skb->users)))
Paolo Abeni3889a8032017-06-12 11:23:41 +02001047 return false;
1048
1049 return true;
1050}
1051
Paolo Abeni0a463c72017-06-12 11:23:42 +02001052void skb_release_head_state(struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -07001053void kfree_skb(struct sk_buff *skb);
1054void kfree_skb_list(struct sk_buff *segs);
Willem de Bruijn64131392019-07-07 05:51:55 -04001055void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
Joe Perches7965bd42013-09-26 14:48:15 -07001056void skb_tx_error(struct sk_buff *skb);
1057void consume_skb(struct sk_buff *skb);
Paolo Abenica2c1412017-09-06 14:44:36 +02001058void __consume_stateless_skb(struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -07001059void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +00001060extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +00001061
Joe Perches7965bd42013-09-26 14:48:15 -07001062void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1063bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1064 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +00001065
Joe Perches7965bd42013-09-26 14:48:15 -07001066struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1067 int node);
Eric Dumazet2ea2f622015-04-24 16:05:01 -07001068struct sk_buff *__build_skb(void *data, unsigned int frag_size);
Joe Perches7965bd42013-09-26 14:48:15 -07001069struct sk_buff *build_skb(void *data, unsigned int frag_size);
Jesper Dangaard Brouerba0509b2019-04-12 17:07:37 +02001070struct sk_buff *build_skb_around(struct sk_buff *skb,
1071 void *data, unsigned int frag_size);
Pedro Tammela161e6132019-03-05 11:35:54 -03001072
1073/**
1074 * alloc_skb - allocate a network buffer
1075 * @size: size to allocate
1076 * @priority: allocation mask
1077 *
1078 * This function is a convenient wrapper around __alloc_skb().
1079 */
David S. Millerd179cd12005-08-17 14:57:30 -07001080static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +01001081 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -07001082{
Eric Dumazet564824b2010-10-11 19:05:25 +00001083 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -07001084}
1085
Eric Dumazet2e4e4412014-09-17 04:49:49 -07001086struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1087 unsigned long data_len,
1088 int max_page_order,
1089 int *errcode,
1090 gfp_t gfp_mask);
Jakub Kicinskida29e4b2019-06-03 15:16:58 -07001091struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
Eric Dumazet2e4e4412014-09-17 04:49:49 -07001092
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001093/* Layout of fast clones : [skb1][skb2][fclone_ref] */
1094struct sk_buff_fclones {
1095 struct sk_buff skb1;
1096
1097 struct sk_buff skb2;
1098
Reshetova, Elena26385952017-06-30 13:07:59 +03001099 refcount_t fclone_ref;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001100};
1101
1102/**
1103 * skb_fclone_busy - check if fclone is busy
Stephen Hemminger293de7d2016-10-23 09:28:29 -07001104 * @sk: socket
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001105 * @skb: buffer
1106 *
Masanari Iidabda13fe2015-12-13 16:53:02 +09001107 * Returns true if skb is a fast clone, and its clone is not freed.
Eric Dumazet39bb5e62014-10-30 10:32:34 -07001108 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1109 * so we also check that this didnt happen.
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001110 */
Eric Dumazet39bb5e62014-10-30 10:32:34 -07001111static inline bool skb_fclone_busy(const struct sock *sk,
1112 const struct sk_buff *skb)
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001113{
1114 const struct sk_buff_fclones *fclones;
1115
1116 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1117
1118 return skb->fclone == SKB_FCLONE_ORIG &&
Reshetova, Elena26385952017-06-30 13:07:59 +03001119 refcount_read(&fclones->fclone_ref) > 1 &&
Eric Dumazet39bb5e62014-10-30 10:32:34 -07001120 fclones->skb2.sk == sk;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001121}
1122
Pedro Tammela161e6132019-03-05 11:35:54 -03001123/**
1124 * alloc_skb_fclone - allocate a network buffer from fclone cache
1125 * @size: size to allocate
1126 * @priority: allocation mask
1127 *
1128 * This function is a convenient wrapper around __alloc_skb().
1129 */
David S. Millerd179cd12005-08-17 14:57:30 -07001130static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +01001131 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -07001132{
Mel Gormanc93bdd02012-07-31 16:44:19 -07001133 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -07001134}
1135
Joe Perches7965bd42013-09-26 14:48:15 -07001136struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
Toshiaki Makitab0768a82018-08-03 16:58:09 +09001137void skb_headers_offset_update(struct sk_buff *skb, int off);
Joe Perches7965bd42013-09-26 14:48:15 -07001138int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1139struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
Ilya Lesokhin08303c12018-04-30 10:16:11 +03001140void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
Joe Perches7965bd42013-09-26 14:48:15 -07001141struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
Octavian Purdilabad93e92014-06-12 01:36:26 +03001142struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1143 gfp_t gfp_mask, bool fclone);
1144static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1145 gfp_t gfp_mask)
1146{
1147 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1148}
Eric Dumazet117632e2011-12-03 21:39:53 +00001149
Joe Perches7965bd42013-09-26 14:48:15 -07001150int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1151struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1152 unsigned int headroom);
1153struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1154 int newtailroom, gfp_t priority);
Jason A. Donenfeld48a1df62017-06-04 04:16:22 +02001155int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1156 int offset, int len);
1157int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1158 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001159int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
Florian Fainellicd0a1372017-08-22 15:12:14 -07001160int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1161
1162/**
1163 * skb_pad - zero pad the tail of an skb
1164 * @skb: buffer to pad
1165 * @pad: space to pad
1166 *
1167 * Ensure that a buffer is followed by a padding area that is zero
1168 * filled. Used by network drivers which may DMA or transfer data
1169 * beyond the buffer end onto the wire.
1170 *
1171 * May return error in out of memory cases. The skb is freed on error.
1172 */
1173static inline int skb_pad(struct sk_buff *skb, int pad)
1174{
1175 return __skb_pad(skb, pad, true);
1176}
Neil Hormanead2ceb2009-03-11 09:49:55 +00001177#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001178
Hannes Frederic Sowabe12a1f2015-05-21 16:59:58 +02001179int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1180 int offset, size_t size);
1181
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08001182struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -07001183 __u32 lower_offset;
1184 __u32 upper_offset;
1185 __u32 frag_idx;
1186 __u32 stepped_offset;
1187 struct sk_buff *root_skb;
1188 struct sk_buff *cur_skb;
1189 __u8 *frag_data;
1190};
1191
Joe Perches7965bd42013-09-26 14:48:15 -07001192void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1193 unsigned int to, struct skb_seq_state *st);
1194unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1195 struct skb_seq_state *st);
1196void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -07001197
Joe Perches7965bd42013-09-26 14:48:15 -07001198unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
Bojan Prtvar059a2442015-02-22 11:46:35 +01001199 unsigned int to, struct ts_config *config);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -07001200
Tom Herbert09323cc2013-12-15 22:16:19 -08001201/*
1202 * Packet hash types specify the type of hash in skb_set_hash.
1203 *
1204 * Hash types refer to the protocol layer addresses which are used to
1205 * construct a packet's hash. The hashes are used to differentiate or identify
1206 * flows of the protocol layer for the hash type. Hash types are either
1207 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1208 *
1209 * Properties of hashes:
1210 *
1211 * 1) Two packets in different flows have different hash values
1212 * 2) Two packets in the same flow should have the same hash value
1213 *
1214 * A hash at a higher layer is considered to be more specific. A driver should
1215 * set the most specific hash possible.
1216 *
1217 * A driver cannot indicate a more specific hash than the layer at which a hash
1218 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1219 *
1220 * A driver may indicate a hash level which is less specific than the
1221 * actual layer the hash was computed on. For instance, a hash computed
1222 * at L4 may be considered an L3 hash. This should only be done if the
1223 * driver can't unambiguously determine that the HW computed the hash at
1224 * the higher layer. Note that the "should" in the second property above
1225 * permits this.
1226 */
1227enum pkt_hash_types {
1228 PKT_HASH_TYPE_NONE, /* Undefined type */
1229 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1230 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1231 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1232};
1233
Tom Herbertbcc83832015-09-01 09:24:24 -07001234static inline void skb_clear_hash(struct sk_buff *skb)
1235{
1236 skb->hash = 0;
1237 skb->sw_hash = 0;
1238 skb->l4_hash = 0;
1239}
1240
1241static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1242{
1243 if (!skb->l4_hash)
1244 skb_clear_hash(skb);
1245}
1246
1247static inline void
1248__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1249{
1250 skb->l4_hash = is_l4;
1251 skb->sw_hash = is_sw;
1252 skb->hash = hash;
1253}
1254
Tom Herbert09323cc2013-12-15 22:16:19 -08001255static inline void
1256skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1257{
Tom Herbertbcc83832015-09-01 09:24:24 -07001258 /* Used by drivers to set hash from HW */
1259 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1260}
1261
1262static inline void
1263__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1264{
1265 __skb_set_hash(skb, hash, true, is_l4);
Tom Herbert09323cc2013-12-15 22:16:19 -08001266}
1267
Tom Herberte5276932015-09-01 09:24:23 -07001268void __skb_get_hash(struct sk_buff *skb);
Florian Westphalb9177832016-10-26 18:49:46 +02001269u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
Tom Herberte5276932015-09-01 09:24:23 -07001270u32 skb_get_poff(const struct sk_buff *skb);
1271u32 __skb_get_poff(const struct sk_buff *skb, void *data,
Paolo Abeni72a338b2018-05-04 11:32:59 +02001272 const struct flow_keys_basic *keys, int hlen);
Tom Herberte5276932015-09-01 09:24:23 -07001273__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1274 void *data, int hlen_proto);
1275
1276static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1277 int thoff, u8 ip_proto)
1278{
1279 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1280}
1281
1282void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1283 const struct flow_dissector_key *key,
1284 unsigned int key_count);
1285
Willem de Bruijn2dfd1842018-09-18 16:20:18 -04001286#ifdef CONFIG_NET
Stanislav Fomichev118c8e92019-04-25 14:37:23 -07001287int skb_flow_dissector_prog_query(const union bpf_attr *attr,
1288 union bpf_attr __user *uattr);
Petar Penkovd58e4682018-09-14 07:46:18 -07001289int skb_flow_dissector_bpf_prog_attach(const union bpf_attr *attr,
1290 struct bpf_prog *prog);
1291
1292int skb_flow_dissector_bpf_prog_detach(const union bpf_attr *attr);
Willem de Bruijn2dfd1842018-09-18 16:20:18 -04001293#else
Stanislav Fomichev118c8e92019-04-25 14:37:23 -07001294static inline int skb_flow_dissector_prog_query(const union bpf_attr *attr,
1295 union bpf_attr __user *uattr)
1296{
1297 return -EOPNOTSUPP;
1298}
1299
Willem de Bruijn2dfd1842018-09-18 16:20:18 -04001300static inline int skb_flow_dissector_bpf_prog_attach(const union bpf_attr *attr,
1301 struct bpf_prog *prog)
1302{
1303 return -EOPNOTSUPP;
1304}
1305
1306static inline int skb_flow_dissector_bpf_prog_detach(const union bpf_attr *attr)
1307{
1308 return -EOPNOTSUPP;
1309}
1310#endif
Petar Penkovd58e4682018-09-14 07:46:18 -07001311
Stanislav Fomichev089b19a2019-04-22 08:55:44 -07001312struct bpf_flow_dissector;
1313bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
Stanislav Fomichev086f9562019-07-25 15:52:25 -07001314 __be16 proto, int nhoff, int hlen, unsigned int flags);
Stanislav Fomichev089b19a2019-04-22 08:55:44 -07001315
Stanislav Fomichev3cbf4ff2019-04-22 08:55:46 -07001316bool __skb_flow_dissect(const struct net *net,
1317 const struct sk_buff *skb,
Tom Herberte5276932015-09-01 09:24:23 -07001318 struct flow_dissector *flow_dissector,
1319 void *target_container,
Tom Herbertcd79a232015-09-01 09:24:27 -07001320 void *data, __be16 proto, int nhoff, int hlen,
1321 unsigned int flags);
Tom Herberte5276932015-09-01 09:24:23 -07001322
1323static inline bool skb_flow_dissect(const struct sk_buff *skb,
1324 struct flow_dissector *flow_dissector,
Tom Herbertcd79a232015-09-01 09:24:27 -07001325 void *target_container, unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001326{
Stanislav Fomichev3cbf4ff2019-04-22 08:55:46 -07001327 return __skb_flow_dissect(NULL, skb, flow_dissector,
1328 target_container, NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001329}
1330
1331static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
Tom Herbertcd79a232015-09-01 09:24:27 -07001332 struct flow_keys *flow,
1333 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001334{
1335 memset(flow, 0, sizeof(*flow));
Stanislav Fomichev3cbf4ff2019-04-22 08:55:46 -07001336 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1337 flow, NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001338}
1339
Paolo Abeni72a338b2018-05-04 11:32:59 +02001340static inline bool
Stanislav Fomichev3cbf4ff2019-04-22 08:55:46 -07001341skb_flow_dissect_flow_keys_basic(const struct net *net,
1342 const struct sk_buff *skb,
Paolo Abeni72a338b2018-05-04 11:32:59 +02001343 struct flow_keys_basic *flow, void *data,
1344 __be16 proto, int nhoff, int hlen,
1345 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001346{
1347 memset(flow, 0, sizeof(*flow));
Stanislav Fomichev3cbf4ff2019-04-22 08:55:46 -07001348 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
Tom Herbertcd79a232015-09-01 09:24:27 -07001349 data, proto, nhoff, hlen, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001350}
1351
Jiri Pirko82828b82019-06-19 09:41:02 +03001352void skb_flow_dissect_meta(const struct sk_buff *skb,
1353 struct flow_dissector *flow_dissector,
1354 void *target_container);
1355
Paul Blakey75a56752019-07-09 10:30:49 +03001356/* Gets a skb connection tracking info, ctinfo map should be a
1357 * a map of mapsize to translate enum ip_conntrack_info states
1358 * to user states.
1359 */
1360void
1361skb_flow_dissect_ct(const struct sk_buff *skb,
1362 struct flow_dissector *flow_dissector,
1363 void *target_container,
1364 u16 *ctinfo_map,
1365 size_t mapsize);
Simon Horman62b32372017-12-04 11:31:48 +01001366void
1367skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1368 struct flow_dissector *flow_dissector,
1369 void *target_container);
1370
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001371static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001372{
Tom Herberta3b18dd2014-07-01 21:33:17 -07001373 if (!skb->l4_hash && !skb->sw_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001374 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001375
Tom Herbert61b905d2014-03-24 15:34:47 -07001376 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001377}
1378
David S. Miller20a17bf2015-09-01 21:19:17 -07001379static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
Tom Herbertf70ea012015-07-31 16:52:10 -07001380{
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001381 if (!skb->l4_hash && !skb->sw_hash) {
1382 struct flow_keys keys;
Tom Herbertde4c1f82015-09-01 18:11:04 -07001383 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001384
Tom Herbertde4c1f82015-09-01 18:11:04 -07001385 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001386 }
Tom Herbertf70ea012015-07-31 16:52:10 -07001387
1388 return skb->hash;
1389}
1390
Eric Dumazet55667442019-10-22 07:57:46 -07001391__u32 skb_get_hash_perturb(const struct sk_buff *skb,
1392 const siphash_key_t *perturb);
Tom Herbert50fb7992015-05-01 11:30:12 -07001393
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001394static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1395{
Tom Herbert61b905d2014-03-24 15:34:47 -07001396 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001397}
1398
Tom Herbert3df7a742013-12-15 22:16:29 -08001399static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1400{
Tom Herbert61b905d2014-03-24 15:34:47 -07001401 to->hash = from->hash;
Tom Herberta3b18dd2014-07-01 21:33:17 -07001402 to->sw_hash = from->sw_hash;
Tom Herbert61b905d2014-03-24 15:34:47 -07001403 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -08001404};
1405
Jakub Kicinski41477662019-08-07 17:03:59 -07001406static inline void skb_copy_decrypted(struct sk_buff *to,
1407 const struct sk_buff *from)
1408{
1409#ifdef CONFIG_TLS_DEVICE
1410 to->decrypted = from->decrypted;
1411#endif
1412}
1413
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001414#ifdef NET_SKBUFF_DATA_USES_OFFSET
1415static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1416{
1417 return skb->head + skb->end;
1418}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001419
1420static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1421{
1422 return skb->end;
1423}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001424#else
1425static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1426{
1427 return skb->end;
1428}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001429
1430static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1431{
1432 return skb->end - skb->head;
1433}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001434#endif
1435
Linus Torvalds1da177e2005-04-16 15:20:36 -07001436/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001437#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001438
Patrick Ohlyac45f602009-02-12 05:03:37 +00001439static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1440{
1441 return &skb_shinfo(skb)->hwtstamps;
1442}
1443
Willem de Bruijn52267792017-08-03 16:29:39 -04001444static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1445{
1446 bool is_zcopy = skb && skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY;
1447
1448 return is_zcopy ? skb_uarg(skb) : NULL;
1449}
1450
Willem de Bruijn52900d22018-11-30 15:32:40 -05001451static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1452 bool *have_ref)
Willem de Bruijn52267792017-08-03 16:29:39 -04001453{
1454 if (skb && uarg && !skb_zcopy(skb)) {
Willem de Bruijn52900d22018-11-30 15:32:40 -05001455 if (unlikely(have_ref && *have_ref))
1456 *have_ref = false;
1457 else
1458 sock_zerocopy_get(uarg);
Willem de Bruijn52267792017-08-03 16:29:39 -04001459 skb_shinfo(skb)->destructor_arg = uarg;
1460 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1461 }
1462}
1463
Willem de Bruijn5cd8d462018-11-20 13:00:18 -05001464static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1465{
1466 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1467 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1468}
1469
1470static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1471{
1472 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1473}
1474
1475static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1476{
1477 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1478}
1479
Willem de Bruijn52267792017-08-03 16:29:39 -04001480/* Release a reference on a zerocopy structure */
1481static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy)
1482{
1483 struct ubuf_info *uarg = skb_zcopy(skb);
1484
1485 if (uarg) {
Willem de Bruijn185ce5c2019-05-15 13:29:16 -04001486 if (skb_zcopy_is_nouarg(skb)) {
1487 /* no notification callback */
1488 } else if (uarg->callback == sock_zerocopy_callback) {
Willem de Bruijn0a4a0602017-08-09 19:09:44 -04001489 uarg->zerocopy = uarg->zerocopy && zerocopy;
1490 sock_zerocopy_put(uarg);
Willem de Bruijn185ce5c2019-05-15 13:29:16 -04001491 } else {
Willem de Bruijn0a4a0602017-08-09 19:09:44 -04001492 uarg->callback(uarg, zerocopy);
1493 }
1494
Willem de Bruijn52267792017-08-03 16:29:39 -04001495 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1496 }
1497}
1498
1499/* Abort a zerocopy operation and revert zckey on error in send syscall */
1500static inline void skb_zcopy_abort(struct sk_buff *skb)
1501{
1502 struct ubuf_info *uarg = skb_zcopy(skb);
1503
1504 if (uarg) {
Willem de Bruijn52900d22018-11-30 15:32:40 -05001505 sock_zerocopy_put_abort(uarg, false);
Willem de Bruijn52267792017-08-03 16:29:39 -04001506 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1507 }
1508}
1509
David S. Millera8305bf2018-07-29 20:42:53 -07001510static inline void skb_mark_not_on_list(struct sk_buff *skb)
1511{
1512 skb->next = NULL;
1513}
1514
Jason A. Donenfelddcfea722020-01-08 16:59:02 -05001515/* Iterate through singly-linked GSO fragments of an skb. */
Jason A. Donenfeld5eee7bd2020-01-13 18:42:26 -05001516#define skb_list_walk_safe(first, skb, next_skb) \
1517 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \
1518 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL)
Jason A. Donenfelddcfea722020-01-08 16:59:02 -05001519
David S. Miller992cba72018-07-31 15:27:56 -07001520static inline void skb_list_del_init(struct sk_buff *skb)
1521{
1522 __list_del_entry(&skb->list);
1523 skb_mark_not_on_list(skb);
1524}
1525
Linus Torvalds1da177e2005-04-16 15:20:36 -07001526/**
1527 * skb_queue_empty - check if a queue is empty
1528 * @list: queue head
1529 *
1530 * Returns true if the queue is empty, false otherwise.
1531 */
1532static inline int skb_queue_empty(const struct sk_buff_head *list)
1533{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001534 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001535}
1536
1537/**
Eric Dumazetd7d16a82019-10-23 22:44:48 -07001538 * skb_queue_empty_lockless - check if a queue is empty
1539 * @list: queue head
1540 *
1541 * Returns true if the queue is empty, false otherwise.
1542 * This variant can be used in lockless contexts.
1543 */
1544static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1545{
1546 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1547}
1548
1549
1550/**
David S. Millerfc7ebb22008-09-23 00:34:07 -07001551 * skb_queue_is_last - check if skb is the last entry in the queue
1552 * @list: queue head
1553 * @skb: buffer
1554 *
1555 * Returns true if @skb is the last buffer on the list.
1556 */
1557static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1558 const struct sk_buff *skb)
1559{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001560 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -07001561}
1562
1563/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001564 * skb_queue_is_first - check if skb is the first entry in the queue
1565 * @list: queue head
1566 * @skb: buffer
1567 *
1568 * Returns true if @skb is the first buffer on the list.
1569 */
1570static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1571 const struct sk_buff *skb)
1572{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001573 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001574}
1575
1576/**
David S. Miller249c8b42008-09-23 00:44:42 -07001577 * skb_queue_next - return the next packet in the queue
1578 * @list: queue head
1579 * @skb: current buffer
1580 *
1581 * Return the next packet in @list after @skb. It is only valid to
1582 * call this if skb_queue_is_last() evaluates to false.
1583 */
1584static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1585 const struct sk_buff *skb)
1586{
1587 /* This BUG_ON may seem severe, but if we just return then we
1588 * are going to dereference garbage.
1589 */
1590 BUG_ON(skb_queue_is_last(list, skb));
1591 return skb->next;
1592}
1593
1594/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001595 * skb_queue_prev - return the prev packet in the queue
1596 * @list: queue head
1597 * @skb: current buffer
1598 *
1599 * Return the prev packet in @list before @skb. It is only valid to
1600 * call this if skb_queue_is_first() evaluates to false.
1601 */
1602static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1603 const struct sk_buff *skb)
1604{
1605 /* This BUG_ON may seem severe, but if we just return then we
1606 * are going to dereference garbage.
1607 */
1608 BUG_ON(skb_queue_is_first(list, skb));
1609 return skb->prev;
1610}
1611
1612/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001613 * skb_get - reference buffer
1614 * @skb: buffer to reference
1615 *
1616 * Makes another reference to a socket buffer and returns a pointer
1617 * to the buffer.
1618 */
1619static inline struct sk_buff *skb_get(struct sk_buff *skb)
1620{
Reshetova, Elena63354792017-06-30 13:07:58 +03001621 refcount_inc(&skb->users);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001622 return skb;
1623}
1624
1625/*
Geert Uytterhoevenf8821f92017-11-30 14:33:56 +01001626 * If users == 1, we are the only owner and can avoid redundant atomic changes.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001627 */
1628
1629/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001630 * skb_cloned - is the buffer a clone
1631 * @skb: buffer to check
1632 *
1633 * Returns true if the buffer was generated with skb_clone() and is
1634 * one of multiple shared copies of the buffer. Cloned buffers are
1635 * shared data so must not be written to under normal circumstances.
1636 */
1637static inline int skb_cloned(const struct sk_buff *skb)
1638{
1639 return skb->cloned &&
1640 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1641}
1642
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001643static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1644{
Mel Gormand0164ad2015-11-06 16:28:21 -08001645 might_sleep_if(gfpflags_allow_blocking(pri));
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001646
1647 if (skb_cloned(skb))
1648 return pskb_expand_head(skb, 0, 0, pri);
1649
1650 return 0;
1651}
1652
Linus Torvalds1da177e2005-04-16 15:20:36 -07001653/**
1654 * skb_header_cloned - is the header a clone
1655 * @skb: buffer to check
1656 *
1657 * Returns true if modifying the header part of the buffer requires
1658 * the data to be copied.
1659 */
1660static inline int skb_header_cloned(const struct sk_buff *skb)
1661{
1662 int dataref;
1663
1664 if (!skb->cloned)
1665 return 0;
1666
1667 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1668 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1669 return dataref != 1;
1670}
1671
Eric Dumazet9580bf22016-04-30 10:19:29 -07001672static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1673{
1674 might_sleep_if(gfpflags_allow_blocking(pri));
1675
1676 if (skb_header_cloned(skb))
1677 return pskb_expand_head(skb, 0, 0, pri);
1678
1679 return 0;
1680}
1681
Linus Torvalds1da177e2005-04-16 15:20:36 -07001682/**
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001683 * __skb_header_release - release reference to header
1684 * @skb: buffer to operate on
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001685 */
1686static inline void __skb_header_release(struct sk_buff *skb)
1687{
1688 skb->nohdr = 1;
1689 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1690}
1691
1692
1693/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001694 * skb_shared - is the buffer shared
1695 * @skb: buffer to check
1696 *
1697 * Returns true if more than one person has a reference to this
1698 * buffer.
1699 */
1700static inline int skb_shared(const struct sk_buff *skb)
1701{
Reshetova, Elena63354792017-06-30 13:07:58 +03001702 return refcount_read(&skb->users) != 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001703}
1704
1705/**
1706 * skb_share_check - check if buffer is shared and if so clone it
1707 * @skb: buffer to check
1708 * @pri: priority for memory allocation
1709 *
1710 * If the buffer is shared the buffer is cloned and the old copy
1711 * drops a reference. A new clone with a single reference is returned.
1712 * If the buffer is not shared the original buffer is returned. When
1713 * being called from interrupt status or with spinlocks held pri must
1714 * be GFP_ATOMIC.
1715 *
1716 * NULL is returned on a memory allocation failure.
1717 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001718static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001719{
Mel Gormand0164ad2015-11-06 16:28:21 -08001720 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001721 if (skb_shared(skb)) {
1722 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001723
1724 if (likely(nskb))
1725 consume_skb(skb);
1726 else
1727 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001728 skb = nskb;
1729 }
1730 return skb;
1731}
1732
1733/*
1734 * Copy shared buffers into a new sk_buff. We effectively do COW on
1735 * packets to handle cases where we have a local reader and forward
1736 * and a couple of other messy ones. The normal one is tcpdumping
1737 * a packet thats being forwarded.
1738 */
1739
1740/**
1741 * skb_unshare - make a copy of a shared buffer
1742 * @skb: buffer to check
1743 * @pri: priority for memory allocation
1744 *
1745 * If the socket buffer is a clone then this function creates a new
1746 * copy of the data, drops a reference count on the old copy and returns
1747 * the new copy with the reference count at 1. If the buffer is not a clone
1748 * the original buffer is returned. When called with a spinlock held or
1749 * from interrupt state @pri must be %GFP_ATOMIC
1750 *
1751 * %NULL is returned on a memory allocation failure.
1752 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001753static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001754 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001755{
Mel Gormand0164ad2015-11-06 16:28:21 -08001756 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001757 if (skb_cloned(skb)) {
1758 struct sk_buff *nskb = skb_copy(skb, pri);
Alexander Aring31eff812014-10-10 23:10:47 +02001759
1760 /* Free our shared copy */
1761 if (likely(nskb))
1762 consume_skb(skb);
1763 else
1764 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001765 skb = nskb;
1766 }
1767 return skb;
1768}
1769
1770/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001771 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001772 * @list_: list to peek at
1773 *
1774 * Peek an &sk_buff. Unlike most other operations you _MUST_
1775 * be careful with this one. A peek leaves the buffer on the
1776 * list and someone else may run off with it. You must hold
1777 * the appropriate locks or have a private queue to do this.
1778 *
1779 * Returns %NULL for an empty list or a pointer to the head element.
1780 * The reference count is not incremented and the reference is therefore
1781 * volatile. Use with caution.
1782 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001783static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001784{
Eric Dumazet18d07002012-04-30 16:31:46 +00001785 struct sk_buff *skb = list_->next;
1786
1787 if (skb == (struct sk_buff *)list_)
1788 skb = NULL;
1789 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001790}
1791
1792/**
David S. Miller8b69bd72018-08-11 18:43:38 -07001793 * __skb_peek - peek at the head of a non-empty &sk_buff_head
1794 * @list_: list to peek at
1795 *
1796 * Like skb_peek(), but the caller knows that the list is not empty.
1797 */
1798static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
1799{
1800 return list_->next;
1801}
1802
1803/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001804 * skb_peek_next - peek skb following the given one from a queue
1805 * @skb: skb to start from
1806 * @list_: list to peek at
1807 *
1808 * Returns %NULL when the end of the list is met or a pointer to the
1809 * next element. The reference count is not incremented and the
1810 * reference is therefore volatile. Use with caution.
1811 */
1812static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1813 const struct sk_buff_head *list_)
1814{
1815 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001816
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001817 if (next == (struct sk_buff *)list_)
1818 next = NULL;
1819 return next;
1820}
1821
1822/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001823 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001824 * @list_: list to peek at
1825 *
1826 * Peek an &sk_buff. Unlike most other operations you _MUST_
1827 * be careful with this one. A peek leaves the buffer on the
1828 * list and someone else may run off with it. You must hold
1829 * the appropriate locks or have a private queue to do this.
1830 *
1831 * Returns %NULL for an empty list or a pointer to the tail element.
1832 * The reference count is not incremented and the reference is therefore
1833 * volatile. Use with caution.
1834 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001835static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001836{
Eric Dumazetf8cc62c2019-11-07 18:49:43 -08001837 struct sk_buff *skb = READ_ONCE(list_->prev);
Eric Dumazet18d07002012-04-30 16:31:46 +00001838
1839 if (skb == (struct sk_buff *)list_)
1840 skb = NULL;
1841 return skb;
1842
Linus Torvalds1da177e2005-04-16 15:20:36 -07001843}
1844
1845/**
1846 * skb_queue_len - get queue length
1847 * @list_: list to measure
1848 *
1849 * Return the length of an &sk_buff queue.
1850 */
1851static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1852{
1853 return list_->qlen;
1854}
1855
David S. Miller67fed452008-09-21 22:36:24 -07001856/**
Qian Cai86b18aa2020-02-04 13:40:29 -05001857 * skb_queue_len_lockless - get queue length
1858 * @list_: list to measure
1859 *
1860 * Return the length of an &sk_buff queue.
1861 * This variant can be used in lockless contexts.
1862 */
1863static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_)
1864{
1865 return READ_ONCE(list_->qlen);
1866}
1867
1868/**
David S. Miller67fed452008-09-21 22:36:24 -07001869 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1870 * @list: queue to initialize
1871 *
1872 * This initializes only the list and queue length aspects of
1873 * an sk_buff_head object. This allows to initialize the list
1874 * aspects of an sk_buff_head without reinitializing things like
1875 * the spinlock. It can also be used for on-stack sk_buff_head
1876 * objects where the spinlock is known to not be used.
1877 */
1878static inline void __skb_queue_head_init(struct sk_buff_head *list)
1879{
1880 list->prev = list->next = (struct sk_buff *)list;
1881 list->qlen = 0;
1882}
1883
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001884/*
1885 * This function creates a split out lock class for each invocation;
1886 * this is needed for now since a whole lot of users of the skb-queue
1887 * infrastructure in drivers have different locking usage (in hardirq)
1888 * than the networking core (in softirq only). In the long run either the
1889 * network layer or drivers should need annotation to consolidate the
1890 * main types of usage into 3 classes.
1891 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001892static inline void skb_queue_head_init(struct sk_buff_head *list)
1893{
1894 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001895 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001896}
1897
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001898static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1899 struct lock_class_key *class)
1900{
1901 skb_queue_head_init(list);
1902 lockdep_set_class(&list->lock, class);
1903}
1904
Linus Torvalds1da177e2005-04-16 15:20:36 -07001905/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001906 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001907 *
1908 * The "__skb_xxxx()" functions are the non-atomic ones that
1909 * can only be called with interrupts disabled.
1910 */
Gerrit Renkerbf299272008-04-14 00:04:51 -07001911static inline void __skb_insert(struct sk_buff *newsk,
1912 struct sk_buff *prev, struct sk_buff *next,
1913 struct sk_buff_head *list)
1914{
Eric Dumazetf8cc62c2019-11-07 18:49:43 -08001915 /* See skb_queue_empty_lockless() and skb_peek_tail()
1916 * for the opposite READ_ONCE()
1917 */
Eric Dumazetd7d16a82019-10-23 22:44:48 -07001918 WRITE_ONCE(newsk->next, next);
1919 WRITE_ONCE(newsk->prev, prev);
1920 WRITE_ONCE(next->prev, newsk);
1921 WRITE_ONCE(prev->next, newsk);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001922 list->qlen++;
1923}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001924
David S. Miller67fed452008-09-21 22:36:24 -07001925static inline void __skb_queue_splice(const struct sk_buff_head *list,
1926 struct sk_buff *prev,
1927 struct sk_buff *next)
1928{
1929 struct sk_buff *first = list->next;
1930 struct sk_buff *last = list->prev;
1931
Eric Dumazetd7d16a82019-10-23 22:44:48 -07001932 WRITE_ONCE(first->prev, prev);
1933 WRITE_ONCE(prev->next, first);
David S. Miller67fed452008-09-21 22:36:24 -07001934
Eric Dumazetd7d16a82019-10-23 22:44:48 -07001935 WRITE_ONCE(last->next, next);
1936 WRITE_ONCE(next->prev, last);
David S. Miller67fed452008-09-21 22:36:24 -07001937}
1938
1939/**
1940 * skb_queue_splice - join two skb lists, this is designed for stacks
1941 * @list: the new list to add
1942 * @head: the place to add it in the first list
1943 */
1944static inline void skb_queue_splice(const struct sk_buff_head *list,
1945 struct sk_buff_head *head)
1946{
1947 if (!skb_queue_empty(list)) {
1948 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001949 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001950 }
1951}
1952
1953/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001954 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001955 * @list: the new list to add
1956 * @head: the place to add it in the first list
1957 *
1958 * The list at @list is reinitialised
1959 */
1960static inline void skb_queue_splice_init(struct sk_buff_head *list,
1961 struct sk_buff_head *head)
1962{
1963 if (!skb_queue_empty(list)) {
1964 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001965 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001966 __skb_queue_head_init(list);
1967 }
1968}
1969
1970/**
1971 * skb_queue_splice_tail - join two skb lists, each list being a queue
1972 * @list: the new list to add
1973 * @head: the place to add it in the first list
1974 */
1975static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1976 struct sk_buff_head *head)
1977{
1978 if (!skb_queue_empty(list)) {
1979 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001980 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001981 }
1982}
1983
1984/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001985 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001986 * @list: the new list to add
1987 * @head: the place to add it in the first list
1988 *
1989 * Each of the lists is a queue.
1990 * The list at @list is reinitialised
1991 */
1992static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1993 struct sk_buff_head *head)
1994{
1995 if (!skb_queue_empty(list)) {
1996 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001997 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001998 __skb_queue_head_init(list);
1999 }
2000}
2001
Linus Torvalds1da177e2005-04-16 15:20:36 -07002002/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08002003 * __skb_queue_after - queue a buffer at the list head
2004 * @list: list to use
2005 * @prev: place after this buffer
2006 * @newsk: buffer to queue
2007 *
2008 * Queue a buffer int the middle of a list. This function takes no locks
2009 * and you must therefore hold required locks before calling it.
2010 *
2011 * A buffer cannot be placed on two lists at the same time.
2012 */
2013static inline void __skb_queue_after(struct sk_buff_head *list,
2014 struct sk_buff *prev,
2015 struct sk_buff *newsk)
2016{
Gerrit Renkerbf299272008-04-14 00:04:51 -07002017 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08002018}
2019
Joe Perches7965bd42013-09-26 14:48:15 -07002020void skb_append(struct sk_buff *old, struct sk_buff *newsk,
2021 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07002022
Gerrit Renkerf5572852008-04-14 00:05:28 -07002023static inline void __skb_queue_before(struct sk_buff_head *list,
2024 struct sk_buff *next,
2025 struct sk_buff *newsk)
2026{
2027 __skb_insert(newsk, next->prev, next, list);
2028}
2029
Stephen Hemminger300ce172005-10-30 13:47:34 -08002030/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002031 * __skb_queue_head - queue a buffer at the list head
2032 * @list: list to use
2033 * @newsk: buffer to queue
2034 *
2035 * Queue a buffer at the start of a list. This function takes no locks
2036 * and you must therefore hold required locks before calling it.
2037 *
2038 * A buffer cannot be placed on two lists at the same time.
2039 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002040static inline void __skb_queue_head(struct sk_buff_head *list,
2041 struct sk_buff *newsk)
2042{
Stephen Hemminger300ce172005-10-30 13:47:34 -08002043 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002044}
Brian Norris4ea7b0c2019-02-11 13:02:25 -08002045void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002046
2047/**
2048 * __skb_queue_tail - queue a buffer at the list tail
2049 * @list: list to use
2050 * @newsk: buffer to queue
2051 *
2052 * Queue a buffer at the end of a list. This function takes no locks
2053 * and you must therefore hold required locks before calling it.
2054 *
2055 * A buffer cannot be placed on two lists at the same time.
2056 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002057static inline void __skb_queue_tail(struct sk_buff_head *list,
2058 struct sk_buff *newsk)
2059{
Gerrit Renkerf5572852008-04-14 00:05:28 -07002060 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002061}
Brian Norris4ea7b0c2019-02-11 13:02:25 -08002062void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002063
Linus Torvalds1da177e2005-04-16 15:20:36 -07002064/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07002065 * remove sk_buff from list. _Must_ be called atomically, and with
2066 * the list known..
2067 */
Joe Perches7965bd42013-09-26 14:48:15 -07002068void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002069static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2070{
2071 struct sk_buff *next, *prev;
2072
Qian Cai86b18aa2020-02-04 13:40:29 -05002073 WRITE_ONCE(list->qlen, list->qlen - 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002074 next = skb->next;
2075 prev = skb->prev;
2076 skb->next = skb->prev = NULL;
Eric Dumazetd7d16a82019-10-23 22:44:48 -07002077 WRITE_ONCE(next->prev, prev);
2078 WRITE_ONCE(prev->next, next);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002079}
2080
Gerrit Renkerf525c062008-04-14 00:04:12 -07002081/**
2082 * __skb_dequeue - remove from the head of the queue
2083 * @list: list to dequeue from
2084 *
2085 * Remove the head of the list. This function does not take any locks
2086 * so must be used with appropriate locks held only. The head item is
2087 * returned or %NULL if the list is empty.
2088 */
Gerrit Renkerf525c062008-04-14 00:04:12 -07002089static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2090{
2091 struct sk_buff *skb = skb_peek(list);
2092 if (skb)
2093 __skb_unlink(skb, list);
2094 return skb;
2095}
Brian Norris4ea7b0c2019-02-11 13:02:25 -08002096struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002097
2098/**
2099 * __skb_dequeue_tail - remove from the tail of the queue
2100 * @list: list to dequeue from
2101 *
2102 * Remove the tail of the list. This function does not take any locks
2103 * so must be used with appropriate locks held only. The tail item is
2104 * returned or %NULL if the list is empty.
2105 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002106static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2107{
2108 struct sk_buff *skb = skb_peek_tail(list);
2109 if (skb)
2110 __skb_unlink(skb, list);
2111 return skb;
2112}
Brian Norris4ea7b0c2019-02-11 13:02:25 -08002113struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002114
2115
David S. Millerbdcc0922012-03-07 20:53:36 -05002116static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002117{
2118 return skb->data_len;
2119}
2120
2121static inline unsigned int skb_headlen(const struct sk_buff *skb)
2122{
2123 return skb->len - skb->data_len;
2124}
2125
Willem de Bruijn3ece7822017-08-03 16:29:38 -04002126static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002127{
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03002128 unsigned int i, len = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002129
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03002130 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00002131 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Willem de Bruijn3ece7822017-08-03 16:29:38 -04002132 return len;
2133}
2134
2135static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2136{
2137 return skb_headlen(skb) + __skb_pagelen(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002138}
2139
Ian Campbell131ea662011-08-19 06:25:00 +00002140/**
2141 * __skb_fill_page_desc - initialise a paged fragment in an skb
2142 * @skb: buffer containing fragment to be initialised
2143 * @i: paged fragment index to initialise
2144 * @page: the page to use for this fragment
2145 * @off: the offset to the data with @page
2146 * @size: the length of the data
2147 *
2148 * Initialises the @i'th fragment of @skb to point to &size bytes at
2149 * offset @off within @page.
2150 *
2151 * Does not take any additional reference on the fragment.
2152 */
2153static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2154 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002155{
2156 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2157
Mel Gormanc48a11c2012-07-31 16:44:23 -07002158 /*
Michal Hocko2f064f32015-08-21 14:11:51 -07002159 * Propagate page pfmemalloc to the skb if we can. The problem is
2160 * that not all callers have unique ownership of the page but rely
2161 * on page_is_pfmemalloc doing the right thing(tm).
Mel Gormanc48a11c2012-07-31 16:44:23 -07002162 */
Matthew Wilcox (Oracle)1dfa5bd2019-07-22 20:08:29 -07002163 frag->bv_page = page;
Jonathan Lemon65c84f12019-07-30 07:40:34 -07002164 frag->bv_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00002165 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00002166
2167 page = compound_head(page);
Michal Hocko2f064f32015-08-21 14:11:51 -07002168 if (page_is_pfmemalloc(page))
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00002169 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00002170}
2171
2172/**
2173 * skb_fill_page_desc - initialise a paged fragment in an skb
2174 * @skb: buffer containing fragment to be initialised
2175 * @i: paged fragment index to initialise
2176 * @page: the page to use for this fragment
2177 * @off: the offset to the data with @page
2178 * @size: the length of the data
2179 *
2180 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01002181 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00002182 * addition updates @skb such that @i is the last fragment.
2183 *
2184 * Does not take any additional reference on the fragment.
2185 */
2186static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2187 struct page *page, int off, int size)
2188{
2189 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002190 skb_shinfo(skb)->nr_frags = i + 1;
2191}
2192
Joe Perches7965bd42013-09-26 14:48:15 -07002193void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2194 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07002195
Jason Wangf8e617e2013-11-01 14:07:47 +08002196void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2197 unsigned int truesize);
2198
Linus Torvalds1da177e2005-04-16 15:20:36 -07002199#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2200
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002201#ifdef NET_SKBUFF_DATA_USES_OFFSET
2202static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2203{
2204 return skb->head + skb->tail;
2205}
2206
2207static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2208{
2209 skb->tail = skb->data - skb->head;
2210}
2211
2212static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2213{
2214 skb_reset_tail_pointer(skb);
2215 skb->tail += offset;
2216}
Simon Horman7cc46192013-05-28 20:34:29 +00002217
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002218#else /* NET_SKBUFF_DATA_USES_OFFSET */
2219static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2220{
2221 return skb->tail;
2222}
2223
2224static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2225{
2226 skb->tail = skb->data;
2227}
2228
2229static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2230{
2231 skb->tail = skb->data + offset;
2232}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07002233
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002234#endif /* NET_SKBUFF_DATA_USES_OFFSET */
2235
Linus Torvalds1da177e2005-04-16 15:20:36 -07002236/*
2237 * Add data to an sk_buff
2238 */
Johannes Berg4df864c2017-06-16 14:29:21 +02002239void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2240void *skb_put(struct sk_buff *skb, unsigned int len);
2241static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002242{
Johannes Berg4df864c2017-06-16 14:29:21 +02002243 void *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002244 SKB_LINEAR_ASSERT(skb);
2245 skb->tail += len;
2246 skb->len += len;
2247 return tmp;
2248}
2249
yuan linyude77b962017-06-18 22:48:17 +08002250static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2251{
2252 void *tmp = __skb_put(skb, len);
2253
2254 memset(tmp, 0, len);
2255 return tmp;
2256}
2257
2258static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2259 unsigned int len)
2260{
2261 void *tmp = __skb_put(skb, len);
2262
2263 memcpy(tmp, data, len);
2264 return tmp;
2265}
2266
2267static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2268{
2269 *(u8 *)__skb_put(skb, 1) = val;
2270}
2271
Johannes Berg83ad3572017-06-14 22:17:20 +02002272static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
Johannes Berge45a79d2017-05-24 09:07:47 +02002273{
Johannes Berg83ad3572017-06-14 22:17:20 +02002274 void *tmp = skb_put(skb, len);
Johannes Berge45a79d2017-05-24 09:07:47 +02002275
2276 memset(tmp, 0, len);
2277
2278 return tmp;
2279}
2280
Johannes Berg59ae1d12017-06-16 14:29:20 +02002281static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2282 unsigned int len)
2283{
2284 void *tmp = skb_put(skb, len);
2285
2286 memcpy(tmp, data, len);
2287
2288 return tmp;
2289}
2290
Johannes Berg634fef62017-06-16 14:29:24 +02002291static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2292{
2293 *(u8 *)skb_put(skb, 1) = val;
2294}
2295
Johannes Bergd58ff352017-06-16 14:29:23 +02002296void *skb_push(struct sk_buff *skb, unsigned int len);
2297static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002298{
2299 skb->data -= len;
2300 skb->len += len;
2301 return skb->data;
2302}
2303
Johannes Bergaf728682017-06-16 14:29:22 +02002304void *skb_pull(struct sk_buff *skb, unsigned int len);
2305static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002306{
2307 skb->len -= len;
2308 BUG_ON(skb->len < skb->data_len);
2309 return skb->data += len;
2310}
2311
Johannes Bergaf728682017-06-16 14:29:22 +02002312static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
David S. Miller47d29642010-05-02 02:21:44 -07002313{
2314 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2315}
2316
Johannes Bergaf728682017-06-16 14:29:22 +02002317void *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002318
Johannes Bergaf728682017-06-16 14:29:22 +02002319static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002320{
2321 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07002322 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002323 return NULL;
2324 skb->len -= len;
2325 return skb->data += len;
2326}
2327
Johannes Bergaf728682017-06-16 14:29:22 +02002328static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002329{
2330 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2331}
2332
Heiner Kallweitb9df4fd2019-10-06 18:19:54 +02002333static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002334{
2335 if (likely(len <= skb_headlen(skb)))
Heiner Kallweitb9df4fd2019-10-06 18:19:54 +02002336 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002337 if (unlikely(len > skb->len))
Heiner Kallweitb9df4fd2019-10-06 18:19:54 +02002338 return false;
Gerrit Renker987c4022008-08-11 18:17:17 -07002339 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002340}
2341
Eric Dumazetc8c8b122016-12-07 09:19:33 -08002342void skb_condense(struct sk_buff *skb);
2343
Linus Torvalds1da177e2005-04-16 15:20:36 -07002344/**
2345 * skb_headroom - bytes at buffer head
2346 * @skb: buffer to check
2347 *
2348 * Return the number of bytes of free space at the head of an &sk_buff.
2349 */
Chuck Leverc2636b42007-10-23 21:07:32 -07002350static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002351{
2352 return skb->data - skb->head;
2353}
2354
2355/**
2356 * skb_tailroom - bytes at buffer end
2357 * @skb: buffer to check
2358 *
2359 * Return the number of bytes of free space at the tail of an sk_buff
2360 */
2361static inline int skb_tailroom(const struct sk_buff *skb)
2362{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07002363 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002364}
2365
2366/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00002367 * skb_availroom - bytes at buffer end
2368 * @skb: buffer to check
2369 *
2370 * Return the number of bytes of free space at the tail of an sk_buff
2371 * allocated by sk_stream_alloc()
2372 */
2373static inline int skb_availroom(const struct sk_buff *skb)
2374{
Eric Dumazet16fad692013-03-14 05:40:32 +00002375 if (skb_is_nonlinear(skb))
2376 return 0;
2377
2378 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00002379}
2380
2381/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002382 * skb_reserve - adjust headroom
2383 * @skb: buffer to alter
2384 * @len: bytes to move
2385 *
2386 * Increase the headroom of an empty &sk_buff by reducing the tail
2387 * room. This is only allowed for an empty buffer.
2388 */
David S. Miller8243126c2006-01-17 02:54:21 -08002389static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002390{
2391 skb->data += len;
2392 skb->tail += len;
2393}
2394
Benjamin Poirier1837b2e2016-02-29 15:03:33 -08002395/**
2396 * skb_tailroom_reserve - adjust reserved_tailroom
2397 * @skb: buffer to alter
2398 * @mtu: maximum amount of headlen permitted
2399 * @needed_tailroom: minimum amount of reserved_tailroom
2400 *
2401 * Set reserved_tailroom so that headlen can be as large as possible but
2402 * not larger than mtu and tailroom cannot be smaller than
2403 * needed_tailroom.
2404 * The required headroom should already have been reserved before using
2405 * this function.
2406 */
2407static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2408 unsigned int needed_tailroom)
2409{
2410 SKB_LINEAR_ASSERT(skb);
2411 if (mtu < skb_tailroom(skb) - needed_tailroom)
2412 /* use at most mtu */
2413 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2414 else
2415 /* use up to all available space */
2416 skb->reserved_tailroom = needed_tailroom;
2417}
2418
Tom Herbert8bce6d72014-09-29 20:22:29 -07002419#define ENCAP_TYPE_ETHER 0
2420#define ENCAP_TYPE_IPPROTO 1
2421
2422static inline void skb_set_inner_protocol(struct sk_buff *skb,
2423 __be16 protocol)
2424{
2425 skb->inner_protocol = protocol;
2426 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2427}
2428
2429static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2430 __u8 ipproto)
2431{
2432 skb->inner_ipproto = ipproto;
2433 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2434}
2435
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002436static inline void skb_reset_inner_headers(struct sk_buff *skb)
2437{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002438 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002439 skb->inner_network_header = skb->network_header;
2440 skb->inner_transport_header = skb->transport_header;
2441}
2442
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00002443static inline void skb_reset_mac_len(struct sk_buff *skb)
2444{
2445 skb->mac_len = skb->network_header - skb->mac_header;
2446}
2447
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002448static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2449 *skb)
2450{
2451 return skb->head + skb->inner_transport_header;
2452}
2453
Tom Herbert55dc5a92015-12-14 11:19:40 -08002454static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2455{
2456 return skb_inner_transport_header(skb) - skb->data;
2457}
2458
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002459static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2460{
2461 skb->inner_transport_header = skb->data - skb->head;
2462}
2463
2464static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2465 const int offset)
2466{
2467 skb_reset_inner_transport_header(skb);
2468 skb->inner_transport_header += offset;
2469}
2470
2471static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2472{
2473 return skb->head + skb->inner_network_header;
2474}
2475
2476static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2477{
2478 skb->inner_network_header = skb->data - skb->head;
2479}
2480
2481static inline void skb_set_inner_network_header(struct sk_buff *skb,
2482 const int offset)
2483{
2484 skb_reset_inner_network_header(skb);
2485 skb->inner_network_header += offset;
2486}
2487
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002488static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2489{
2490 return skb->head + skb->inner_mac_header;
2491}
2492
2493static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2494{
2495 skb->inner_mac_header = skb->data - skb->head;
2496}
2497
2498static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2499 const int offset)
2500{
2501 skb_reset_inner_mac_header(skb);
2502 skb->inner_mac_header += offset;
2503}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002504static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2505{
Cong Wang35d04612013-05-29 15:16:05 +08002506 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002507}
2508
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002509static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2510{
2511 return skb->head + skb->transport_header;
2512}
2513
2514static inline void skb_reset_transport_header(struct sk_buff *skb)
2515{
2516 skb->transport_header = skb->data - skb->head;
2517}
2518
2519static inline void skb_set_transport_header(struct sk_buff *skb,
2520 const int offset)
2521{
2522 skb_reset_transport_header(skb);
2523 skb->transport_header += offset;
2524}
2525
2526static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2527{
2528 return skb->head + skb->network_header;
2529}
2530
2531static inline void skb_reset_network_header(struct sk_buff *skb)
2532{
2533 skb->network_header = skb->data - skb->head;
2534}
2535
2536static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2537{
2538 skb_reset_network_header(skb);
2539 skb->network_header += offset;
2540}
2541
2542static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2543{
2544 return skb->head + skb->mac_header;
2545}
2546
Amir Vadaiea6da4f2017-02-07 09:56:06 +02002547static inline int skb_mac_offset(const struct sk_buff *skb)
2548{
2549 return skb_mac_header(skb) - skb->data;
2550}
2551
Daniel Borkmann0daf4342017-07-02 02:13:25 +02002552static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2553{
2554 return skb->network_header - skb->mac_header;
2555}
2556
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002557static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2558{
Cong Wang35d04612013-05-29 15:16:05 +08002559 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002560}
2561
2562static inline void skb_reset_mac_header(struct sk_buff *skb)
2563{
2564 skb->mac_header = skb->data - skb->head;
2565}
2566
2567static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2568{
2569 skb_reset_mac_header(skb);
2570 skb->mac_header += offset;
2571}
2572
Timo Teräs0e3da5b2013-12-16 11:02:09 +02002573static inline void skb_pop_mac_header(struct sk_buff *skb)
2574{
2575 skb->mac_header = skb->network_header;
2576}
2577
Maxim Mikityanskiyd2aa1252019-02-21 12:39:57 +00002578static inline void skb_probe_transport_header(struct sk_buff *skb)
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002579{
Paolo Abeni72a338b2018-05-04 11:32:59 +02002580 struct flow_keys_basic keys;
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002581
2582 if (skb_transport_header_was_set(skb))
2583 return;
Paolo Abeni72a338b2018-05-04 11:32:59 +02002584
Stanislav Fomichev3cbf4ff2019-04-22 08:55:46 -07002585 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2586 NULL, 0, 0, 0, 0))
Tom Herbert42aecaa2015-06-04 09:16:39 -07002587 skb_set_transport_header(skb, keys.control.thoff);
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002588}
2589
Eric Dumazet03606892012-02-23 10:55:02 +00002590static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2591{
2592 if (skb_mac_header_was_set(skb)) {
2593 const unsigned char *old_mac = skb_mac_header(skb);
2594
2595 skb_set_mac_header(skb, -skb->mac_len);
2596 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2597 }
2598}
2599
Michał Mirosław04fb4512010-12-14 15:24:08 +00002600static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2601{
2602 return skb->csum_start - skb_headroom(skb);
2603}
2604
Alexander Duyck08b64fc2016-02-05 15:27:49 -08002605static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2606{
2607 return skb->head + skb->csum_start;
2608}
2609
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002610static inline int skb_transport_offset(const struct sk_buff *skb)
2611{
2612 return skb_transport_header(skb) - skb->data;
2613}
2614
2615static inline u32 skb_network_header_len(const struct sk_buff *skb)
2616{
2617 return skb->transport_header - skb->network_header;
2618}
2619
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002620static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2621{
2622 return skb->inner_transport_header - skb->inner_network_header;
2623}
2624
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002625static inline int skb_network_offset(const struct sk_buff *skb)
2626{
2627 return skb_network_header(skb) - skb->data;
2628}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03002629
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002630static inline int skb_inner_network_offset(const struct sk_buff *skb)
2631{
2632 return skb_inner_network_header(skb) - skb->data;
2633}
2634
Changli Gaof9599ce2010-08-04 04:43:44 +00002635static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2636{
2637 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2638}
2639
Linus Torvalds1da177e2005-04-16 15:20:36 -07002640/*
2641 * CPUs often take a performance hit when accessing unaligned memory
2642 * locations. The actual performance hit varies, it can be small if the
2643 * hardware handles it or large if we have to take an exception and fix it
2644 * in software.
2645 *
2646 * Since an ethernet header is 14 bytes network drivers often end up with
2647 * the IP header at an unaligned offset. The IP header can be aligned by
2648 * shifting the start of the packet by 2 bytes. Drivers should do this
2649 * with:
2650 *
Tobias Klauser8660c122009-07-13 22:48:16 +00002651 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002652 *
2653 * The downside to this alignment of the IP header is that the DMA is now
2654 * unaligned. On some architectures the cost of an unaligned DMA is high
2655 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00002656 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002657 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2658 * to be overridden.
2659 */
2660#ifndef NET_IP_ALIGN
2661#define NET_IP_ALIGN 2
2662#endif
2663
Anton Blanchard025be812006-03-31 02:27:06 -08002664/*
2665 * The networking layer reserves some headroom in skb data (via
2666 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2667 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08002668 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08002669 *
2670 * Unfortunately this headroom changes the DMA alignment of the resulting
2671 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2672 * on some architectures. An architecture can override this value,
2673 * perhaps setting it to a cacheline in size (since that will maintain
2674 * cacheline alignment of the DMA). It must be a power of 2.
2675 *
David S. Millerd6301d32009-02-08 19:24:13 -08002676 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08002677 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07002678 *
2679 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2680 * to reduce average number of cache lines per packet.
2681 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07002682 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08002683 */
2684#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07002685#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08002686#endif
2687
Joe Perches7965bd42013-09-26 14:48:15 -07002688int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002689
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002690static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002691{
Yangtao Li5e1abdc2018-11-06 10:45:36 -05002692 if (WARN_ON(skb_is_nonlinear(skb)))
Herbert Xu3cc0e872006-06-09 16:13:38 -07002693 return;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002694 skb->len = len;
2695 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002696}
2697
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002698static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2699{
2700 __skb_set_length(skb, len);
2701}
2702
Joe Perches7965bd42013-09-26 14:48:15 -07002703void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002704
2705static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2706{
Herbert Xu3cc0e872006-06-09 16:13:38 -07002707 if (skb->data_len)
2708 return ___pskb_trim(skb, len);
2709 __skb_trim(skb, len);
2710 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002711}
2712
2713static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2714{
2715 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2716}
2717
2718/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07002719 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2720 * @skb: buffer to alter
2721 * @len: new length
2722 *
2723 * This is identical to pskb_trim except that the caller knows that
2724 * the skb is not cloned so we should never get an error due to out-
2725 * of-memory.
2726 */
2727static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2728{
2729 int err = pskb_trim(skb, len);
2730 BUG_ON(err);
2731}
2732
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002733static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2734{
2735 unsigned int diff = len - skb->len;
2736
2737 if (skb_tailroom(skb) < diff) {
2738 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2739 GFP_ATOMIC);
2740 if (ret)
2741 return ret;
2742 }
2743 __skb_set_length(skb, len);
2744 return 0;
2745}
2746
Herbert Xue9fa4f72006-08-13 20:12:58 -07002747/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002748 * skb_orphan - orphan a buffer
2749 * @skb: buffer to orphan
2750 *
2751 * If a buffer currently has an owner then we call the owner's
2752 * destructor function and make the @skb unowned. The buffer continues
2753 * to exist but is no longer charged to its former owner.
2754 */
2755static inline void skb_orphan(struct sk_buff *skb)
2756{
Eric Dumazetc34a7612013-07-30 16:11:15 -07002757 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002758 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002759 skb->destructor = NULL;
2760 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07002761 } else {
2762 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002763 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002764}
2765
2766/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002767 * skb_orphan_frags - orphan the frags contained in a buffer
2768 * @skb: buffer to orphan frags from
2769 * @gfp_mask: allocation mask for replacement pages
2770 *
2771 * For each frag in the SKB which needs a destructor (i.e. has an
2772 * owner) create a copy of that frag and release the original
2773 * page by calling the destructor.
2774 */
2775static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2776{
Willem de Bruijn1f8b9772017-08-03 16:29:41 -04002777 if (likely(!skb_zcopy(skb)))
2778 return 0;
Willem de Bruijn185ce5c2019-05-15 13:29:16 -04002779 if (!skb_zcopy_is_nouarg(skb) &&
2780 skb_uarg(skb)->callback == sock_zerocopy_callback)
Willem de Bruijn1f8b9772017-08-03 16:29:41 -04002781 return 0;
2782 return skb_copy_ubufs(skb, gfp_mask);
2783}
2784
2785/* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2786static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2787{
2788 if (likely(!skb_zcopy(skb)))
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002789 return 0;
2790 return skb_copy_ubufs(skb, gfp_mask);
2791}
2792
2793/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002794 * __skb_queue_purge - empty a list
2795 * @list: list to empty
2796 *
2797 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2798 * the list and one reference dropped. This function does not take the
2799 * list lock and the caller must hold the relevant locks to use it.
2800 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002801static inline void __skb_queue_purge(struct sk_buff_head *list)
2802{
2803 struct sk_buff *skb;
2804 while ((skb = __skb_dequeue(list)) != NULL)
2805 kfree_skb(skb);
2806}
Brian Norris4ea7b0c2019-02-11 13:02:25 -08002807void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002808
Peter Oskolkov385114d2018-08-02 23:34:38 +00002809unsigned int skb_rbtree_purge(struct rb_root *root);
Yaogong Wang9f5afea2016-09-07 14:49:28 -07002810
Joe Perches7965bd42013-09-26 14:48:15 -07002811void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002812
Joe Perches7965bd42013-09-26 14:48:15 -07002813struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2814 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07002815
2816/**
2817 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2818 * @dev: network device to receive on
2819 * @length: length to allocate
2820 *
2821 * Allocate a new &sk_buff and assign it a usage count of one. The
2822 * buffer has unspecified headroom built in. Users should allocate
2823 * the headroom they think they need without accounting for the
2824 * built in space. The built in space is used for optimisations.
2825 *
2826 * %NULL is returned if there is no free memory. Although this function
2827 * allocates memory it can be called from an interrupt.
2828 */
2829static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002830 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002831{
2832 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2833}
2834
Eric Dumazet6f532612012-05-18 05:12:12 +00002835/* legacy helper around __netdev_alloc_skb() */
2836static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2837 gfp_t gfp_mask)
2838{
2839 return __netdev_alloc_skb(NULL, length, gfp_mask);
2840}
2841
2842/* legacy helper around netdev_alloc_skb() */
2843static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2844{
2845 return netdev_alloc_skb(NULL, length);
2846}
2847
2848
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002849static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2850 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002851{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002852 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002853
2854 if (NET_IP_ALIGN && skb)
2855 skb_reserve(skb, NET_IP_ALIGN);
2856 return skb;
2857}
2858
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002859static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2860 unsigned int length)
2861{
2862 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2863}
2864
Alexander Duyck181edb22015-05-06 21:12:03 -07002865static inline void skb_free_frag(void *addr)
2866{
Alexander Duyck8c2dd3e2017-01-10 16:58:06 -08002867 page_frag_free(addr);
Alexander Duyck181edb22015-05-06 21:12:03 -07002868}
2869
Alexander Duyckffde7322014-12-09 19:40:42 -08002870void *napi_alloc_frag(unsigned int fragsz);
Alexander Duyckfd11a832014-12-09 19:40:49 -08002871struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2872 unsigned int length, gfp_t gfp_mask);
2873static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2874 unsigned int length)
2875{
2876 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2877}
Jesper Dangaard Brouer795bb1c2016-02-08 13:14:59 +01002878void napi_consume_skb(struct sk_buff *skb, int budget);
2879
2880void __kfree_skb_flush(void);
Jesper Dangaard Brouer15fad712016-02-08 13:15:04 +01002881void __kfree_skb_defer(struct sk_buff *skb);
Alexander Duyckffde7322014-12-09 19:40:42 -08002882
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002883/**
Alexander Duyck71dfda52014-11-11 09:26:34 -08002884 * __dev_alloc_pages - allocate page for network Rx
2885 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2886 * @order: size of the allocation
2887 *
2888 * Allocate a new page.
2889 *
2890 * %NULL is returned if there is no free memory.
2891*/
2892static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2893 unsigned int order)
2894{
2895 /* This piece of code contains several assumptions.
2896 * 1. This is for device Rx, therefor a cold page is preferred.
2897 * 2. The expectation is the user wants a compound page.
2898 * 3. If requesting a order 0 page it will not be compound
2899 * due to the check to see if order has a value in prep_new_page
2900 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2901 * code in gfp_to_alloc_flags that should be enforcing this.
2902 */
Mel Gorman453f85d2017-11-15 17:38:03 -08002903 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
Alexander Duyck71dfda52014-11-11 09:26:34 -08002904
2905 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2906}
2907
2908static inline struct page *dev_alloc_pages(unsigned int order)
2909{
Neil Horman95829b32016-05-19 11:30:54 -04002910 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002911}
2912
2913/**
2914 * __dev_alloc_page - allocate a page for network Rx
2915 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2916 *
2917 * Allocate a new page.
2918 *
2919 * %NULL is returned if there is no free memory.
2920 */
2921static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2922{
2923 return __dev_alloc_pages(gfp_mask, 0);
2924}
2925
2926static inline struct page *dev_alloc_page(void)
2927{
Neil Horman95829b32016-05-19 11:30:54 -04002928 return dev_alloc_pages(0);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002929}
2930
2931/**
Mel Gorman06140022012-07-31 16:44:24 -07002932 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2933 * @page: The page that was allocated from skb_alloc_page
2934 * @skb: The skb that may need pfmemalloc set
2935 */
2936static inline void skb_propagate_pfmemalloc(struct page *page,
2937 struct sk_buff *skb)
2938{
Michal Hocko2f064f32015-08-21 14:11:51 -07002939 if (page_is_pfmemalloc(page))
Mel Gorman06140022012-07-31 16:44:24 -07002940 skb->pfmemalloc = true;
2941}
2942
Eric Dumazet564824b2010-10-11 19:05:25 +00002943/**
Jonathan Lemon7240b602019-07-30 07:40:32 -07002944 * skb_frag_off() - Returns the offset of a skb fragment
2945 * @frag: the paged fragment
2946 */
2947static inline unsigned int skb_frag_off(const skb_frag_t *frag)
2948{
Jonathan Lemon65c84f12019-07-30 07:40:34 -07002949 return frag->bv_offset;
Jonathan Lemon7240b602019-07-30 07:40:32 -07002950}
2951
2952/**
2953 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
2954 * @frag: skb fragment
2955 * @delta: value to add
2956 */
2957static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
2958{
Jonathan Lemon65c84f12019-07-30 07:40:34 -07002959 frag->bv_offset += delta;
Jonathan Lemon7240b602019-07-30 07:40:32 -07002960}
2961
2962/**
2963 * skb_frag_off_set() - Sets the offset of a skb fragment
2964 * @frag: skb fragment
2965 * @offset: offset of fragment
2966 */
2967static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
2968{
Jonathan Lemon65c84f12019-07-30 07:40:34 -07002969 frag->bv_offset = offset;
Jonathan Lemon7240b602019-07-30 07:40:32 -07002970}
2971
2972/**
2973 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
2974 * @fragto: skb fragment where offset is set
2975 * @fragfrom: skb fragment offset is copied from
2976 */
2977static inline void skb_frag_off_copy(skb_frag_t *fragto,
2978 const skb_frag_t *fragfrom)
2979{
Jonathan Lemon65c84f12019-07-30 07:40:34 -07002980 fragto->bv_offset = fragfrom->bv_offset;
Jonathan Lemon7240b602019-07-30 07:40:32 -07002981}
2982
2983/**
Masanari Iidae2278672014-02-18 22:54:36 +09002984 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002985 * @frag: the paged fragment
2986 *
2987 * Returns the &struct page associated with @frag.
2988 */
2989static inline struct page *skb_frag_page(const skb_frag_t *frag)
2990{
Matthew Wilcox (Oracle)1dfa5bd2019-07-22 20:08:29 -07002991 return frag->bv_page;
Ian Campbell131ea662011-08-19 06:25:00 +00002992}
2993
2994/**
2995 * __skb_frag_ref - take an addition reference on a paged fragment.
2996 * @frag: the paged fragment
2997 *
2998 * Takes an additional reference on the paged fragment @frag.
2999 */
3000static inline void __skb_frag_ref(skb_frag_t *frag)
3001{
3002 get_page(skb_frag_page(frag));
3003}
3004
3005/**
3006 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
3007 * @skb: the buffer
3008 * @f: the fragment offset.
3009 *
3010 * Takes an additional reference on the @f'th paged fragment of @skb.
3011 */
3012static inline void skb_frag_ref(struct sk_buff *skb, int f)
3013{
3014 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
3015}
3016
3017/**
3018 * __skb_frag_unref - release a reference on a paged fragment.
3019 * @frag: the paged fragment
3020 *
3021 * Releases a reference on the paged fragment @frag.
3022 */
3023static inline void __skb_frag_unref(skb_frag_t *frag)
3024{
3025 put_page(skb_frag_page(frag));
3026}
3027
3028/**
3029 * skb_frag_unref - release a reference on a paged fragment of an skb.
3030 * @skb: the buffer
3031 * @f: the fragment offset
3032 *
3033 * Releases a reference on the @f'th paged fragment of @skb.
3034 */
3035static inline void skb_frag_unref(struct sk_buff *skb, int f)
3036{
3037 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
3038}
3039
3040/**
3041 * skb_frag_address - gets the address of the data contained in a paged fragment
3042 * @frag: the paged fragment buffer
3043 *
3044 * Returns the address of the data within @frag. The page must already
3045 * be mapped.
3046 */
3047static inline void *skb_frag_address(const skb_frag_t *frag)
3048{
Jonathan Lemon7240b602019-07-30 07:40:32 -07003049 return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
Ian Campbell131ea662011-08-19 06:25:00 +00003050}
3051
3052/**
3053 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3054 * @frag: the paged fragment buffer
3055 *
3056 * Returns the address of the data within @frag. Checks that the page
3057 * is mapped and returns %NULL otherwise.
3058 */
3059static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3060{
3061 void *ptr = page_address(skb_frag_page(frag));
3062 if (unlikely(!ptr))
3063 return NULL;
3064
Jonathan Lemon7240b602019-07-30 07:40:32 -07003065 return ptr + skb_frag_off(frag);
3066}
3067
3068/**
3069 * skb_frag_page_copy() - sets the page in a fragment from another fragment
3070 * @fragto: skb fragment where page is set
3071 * @fragfrom: skb fragment page is copied from
3072 */
3073static inline void skb_frag_page_copy(skb_frag_t *fragto,
3074 const skb_frag_t *fragfrom)
3075{
3076 fragto->bv_page = fragfrom->bv_page;
Ian Campbell131ea662011-08-19 06:25:00 +00003077}
3078
3079/**
3080 * __skb_frag_set_page - sets the page contained in a paged fragment
3081 * @frag: the paged fragment
3082 * @page: the page to set
3083 *
3084 * Sets the fragment @frag to contain @page.
3085 */
3086static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3087{
Matthew Wilcox (Oracle)1dfa5bd2019-07-22 20:08:29 -07003088 frag->bv_page = page;
Ian Campbell131ea662011-08-19 06:25:00 +00003089}
3090
3091/**
3092 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3093 * @skb: the buffer
3094 * @f: the fragment offset
3095 * @page: the page to set
3096 *
3097 * Sets the @f'th fragment of @skb to contain @page.
3098 */
3099static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3100 struct page *page)
3101{
3102 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3103}
3104
Eric Dumazet400dfd32013-10-17 16:27:07 -07003105bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3106
Ian Campbell131ea662011-08-19 06:25:00 +00003107/**
3108 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00003109 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00003110 * @frag: the paged fragment to map
3111 * @offset: the offset within the fragment (starting at the
3112 * fragment's own offset)
3113 * @size: the number of bytes to map
Mauro Carvalho Chehab771b00a2017-05-12 09:19:29 -03003114 * @dir: the direction of the mapping (``PCI_DMA_*``)
Ian Campbell131ea662011-08-19 06:25:00 +00003115 *
3116 * Maps the page associated with @frag to @device.
3117 */
3118static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3119 const skb_frag_t *frag,
3120 size_t offset, size_t size,
3121 enum dma_data_direction dir)
3122{
3123 return dma_map_page(dev, skb_frag_page(frag),
Jonathan Lemon7240b602019-07-30 07:40:32 -07003124 skb_frag_off(frag) + offset, size, dir);
Ian Campbell131ea662011-08-19 06:25:00 +00003125}
3126
Eric Dumazet117632e2011-12-03 21:39:53 +00003127static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3128 gfp_t gfp_mask)
3129{
3130 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3131}
3132
Octavian Purdilabad93e92014-06-12 01:36:26 +03003133
3134static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3135 gfp_t gfp_mask)
3136{
3137 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3138}
3139
3140
Ian Campbell131ea662011-08-19 06:25:00 +00003141/**
Patrick McHardy334a8132007-06-25 04:35:20 -07003142 * skb_clone_writable - is the header of a clone writable
3143 * @skb: buffer to check
3144 * @len: length up to which to write
3145 *
3146 * Returns true if modifying the header part of the cloned buffer
3147 * does not requires the data to be copied.
3148 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003149static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07003150{
3151 return !skb_header_cloned(skb) &&
3152 skb_headroom(skb) + len <= skb->hdr_len;
3153}
3154
Daniel Borkmann36976492016-02-19 23:05:25 +01003155static inline int skb_try_make_writable(struct sk_buff *skb,
3156 unsigned int write_len)
3157{
3158 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3159 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3160}
3161
Herbert Xud9cc2042007-09-16 16:21:16 -07003162static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3163 int cloned)
3164{
3165 int delta = 0;
3166
Herbert Xud9cc2042007-09-16 16:21:16 -07003167 if (headroom > skb_headroom(skb))
3168 delta = headroom - skb_headroom(skb);
3169
3170 if (delta || cloned)
3171 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3172 GFP_ATOMIC);
3173 return 0;
3174}
3175
Patrick McHardy334a8132007-06-25 04:35:20 -07003176/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07003177 * skb_cow - copy header of skb when it is required
3178 * @skb: buffer to cow
3179 * @headroom: needed headroom
3180 *
3181 * If the skb passed lacks sufficient headroom or its data part
3182 * is shared, data is reallocated. If reallocation fails, an error
3183 * is returned and original skb is not changed.
3184 *
3185 * The result is skb with writable area skb->head...skb->tail
3186 * and at least @headroom of space at head.
3187 */
3188static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3189{
Herbert Xud9cc2042007-09-16 16:21:16 -07003190 return __skb_cow(skb, headroom, skb_cloned(skb));
3191}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003192
Herbert Xud9cc2042007-09-16 16:21:16 -07003193/**
3194 * skb_cow_head - skb_cow but only making the head writable
3195 * @skb: buffer to cow
3196 * @headroom: needed headroom
3197 *
3198 * This function is identical to skb_cow except that we replace the
3199 * skb_cloned check by skb_header_cloned. It should be used when
3200 * you only need to push on some header and do not need to modify
3201 * the data.
3202 */
3203static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3204{
3205 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07003206}
3207
3208/**
3209 * skb_padto - pad an skbuff up to a minimal size
3210 * @skb: buffer to pad
3211 * @len: minimal length
3212 *
3213 * Pads up a buffer to ensure the trailing bytes exist and are
3214 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07003215 * is untouched. Otherwise it is extended. Returns zero on
3216 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003217 */
Herbert Xu5b057c62006-06-23 02:06:41 -07003218static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003219{
3220 unsigned int size = skb->len;
3221 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07003222 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07003223 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003224}
3225
Alexander Duyck9c0c1122014-12-03 08:17:33 -08003226/**
Brian Norris4ea7b0c2019-02-11 13:02:25 -08003227 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
Alexander Duyck9c0c1122014-12-03 08:17:33 -08003228 * @skb: buffer to pad
3229 * @len: minimal length
Florian Fainellicd0a1372017-08-22 15:12:14 -07003230 * @free_on_error: free buffer on error
3231 *
3232 * Pads up a buffer to ensure the trailing bytes exist and are
3233 * blanked. If the buffer already contains sufficient data it
3234 * is untouched. Otherwise it is extended. Returns zero on
3235 * success. The skb is freed on error if @free_on_error is true.
3236 */
3237static inline int __skb_put_padto(struct sk_buff *skb, unsigned int len,
3238 bool free_on_error)
3239{
3240 unsigned int size = skb->len;
3241
3242 if (unlikely(size < len)) {
3243 len -= size;
3244 if (__skb_pad(skb, len, free_on_error))
3245 return -ENOMEM;
3246 __skb_put(skb, len);
3247 }
3248 return 0;
3249}
3250
3251/**
3252 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3253 * @skb: buffer to pad
3254 * @len: minimal length
Alexander Duyck9c0c1122014-12-03 08:17:33 -08003255 *
3256 * Pads up a buffer to ensure the trailing bytes exist and are
3257 * blanked. If the buffer already contains sufficient data it
3258 * is untouched. Otherwise it is extended. Returns zero on
3259 * success. The skb is freed on error.
3260 */
3261static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
3262{
Florian Fainellicd0a1372017-08-22 15:12:14 -07003263 return __skb_put_padto(skb, len, true);
Alexander Duyck9c0c1122014-12-03 08:17:33 -08003264}
3265
Linus Torvalds1da177e2005-04-16 15:20:36 -07003266static inline int skb_add_data(struct sk_buff *skb,
Al Viroaf2b0402014-11-27 21:44:24 -05003267 struct iov_iter *from, int copy)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003268{
3269 const int off = skb->len;
3270
3271 if (skb->ip_summed == CHECKSUM_NONE) {
Al Viroaf2b0402014-11-27 21:44:24 -05003272 __wsum csum = 0;
Al Viro15e6cb42016-11-01 22:42:45 -04003273 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3274 &csum, from)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003275 skb->csum = csum_block_add(skb->csum, csum, off);
3276 return 0;
3277 }
Al Viro15e6cb42016-11-01 22:42:45 -04003278 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003279 return 0;
3280
3281 __skb_trim(skb, off);
3282 return -EFAULT;
3283}
3284
Eric Dumazet38ba0a62012-04-23 17:48:27 +00003285static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3286 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003287{
Willem de Bruijn1f8b9772017-08-03 16:29:41 -04003288 if (skb_zcopy(skb))
3289 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003290 if (i) {
Matthew Wilcox (Oracle)d8e18a52019-07-22 20:08:26 -07003291 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07003292
Ian Campbellea2ab692011-08-22 23:44:58 +00003293 return page == skb_frag_page(frag) &&
Jonathan Lemon7240b602019-07-30 07:40:32 -07003294 off == skb_frag_off(frag) + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003295 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00003296 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003297}
3298
Herbert Xu364c6ba2006-06-09 16:10:40 -07003299static inline int __skb_linearize(struct sk_buff *skb)
3300{
3301 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3302}
3303
Linus Torvalds1da177e2005-04-16 15:20:36 -07003304/**
3305 * skb_linearize - convert paged skb to linear one
3306 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07003307 *
3308 * If there is no free memory -ENOMEM is returned, otherwise zero
3309 * is returned and the old skb data released.
3310 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07003311static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003312{
Herbert Xu364c6ba2006-06-09 16:10:40 -07003313 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3314}
3315
3316/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00003317 * skb_has_shared_frag - can any frag be overwritten
3318 * @skb: buffer to test
3319 *
3320 * Return true if the skb has at least one frag that might be modified
3321 * by an external entity (as in vmsplice()/sendfile())
3322 */
3323static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3324{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00003325 return skb_is_nonlinear(skb) &&
3326 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00003327}
3328
3329/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07003330 * skb_linearize_cow - make sure skb is linear and writable
3331 * @skb: buffer to process
3332 *
3333 * If there is no free memory -ENOMEM is returned, otherwise zero
3334 * is returned and the old skb data released.
3335 */
3336static inline int skb_linearize_cow(struct sk_buff *skb)
3337{
3338 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3339 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003340}
3341
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003342static __always_inline void
3343__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3344 unsigned int off)
3345{
3346 if (skb->ip_summed == CHECKSUM_COMPLETE)
3347 skb->csum = csum_block_sub(skb->csum,
3348 csum_partial(start, len, 0), off);
3349 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3350 skb_checksum_start_offset(skb) < 0)
3351 skb->ip_summed = CHECKSUM_NONE;
3352}
3353
Linus Torvalds1da177e2005-04-16 15:20:36 -07003354/**
3355 * skb_postpull_rcsum - update checksum for received skb after pull
3356 * @skb: buffer to update
3357 * @start: start of data before pull
3358 * @len: length of data pulled
3359 *
3360 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07003361 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3362 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003363 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003364static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08003365 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003366{
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003367 __skb_postpull_rcsum(skb, start, len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003368}
3369
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003370static __always_inline void
3371__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3372 unsigned int off)
3373{
3374 if (skb->ip_summed == CHECKSUM_COMPLETE)
3375 skb->csum = csum_block_add(skb->csum,
3376 csum_partial(start, len, 0), off);
3377}
Herbert Xucbb042f2006-03-20 22:43:56 -08003378
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003379/**
3380 * skb_postpush_rcsum - update checksum for received skb after push
3381 * @skb: buffer to update
3382 * @start: start of data after push
3383 * @len: length of data pushed
3384 *
3385 * After doing a push on a received packet, you need to call this to
3386 * update the CHECKSUM_COMPLETE checksum.
3387 */
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01003388static inline void skb_postpush_rcsum(struct sk_buff *skb,
3389 const void *start, unsigned int len)
3390{
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003391 __skb_postpush_rcsum(skb, start, len, 0);
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01003392}
3393
Johannes Bergaf728682017-06-16 14:29:22 +02003394void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003395
David S. Miller7ce5a272013-12-02 17:26:05 -05003396/**
WANG Cong82a31b92016-06-30 10:15:22 -07003397 * skb_push_rcsum - push skb and update receive checksum
3398 * @skb: buffer to update
3399 * @len: length of data pulled
3400 *
3401 * This function performs an skb_push on the packet and updates
3402 * the CHECKSUM_COMPLETE checksum. It should be used on
3403 * receive path processing instead of skb_push unless you know
3404 * that the checksum difference is zero (e.g., a valid IP header)
3405 * or you are setting ip_summed to CHECKSUM_NONE.
3406 */
Johannes Bergd58ff352017-06-16 14:29:23 +02003407static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
WANG Cong82a31b92016-06-30 10:15:22 -07003408{
3409 skb_push(skb, len);
3410 skb_postpush_rcsum(skb, skb->data, len);
3411 return skb->data;
3412}
3413
Eric Dumazet88078d92018-04-18 11:43:15 -07003414int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
WANG Cong82a31b92016-06-30 10:15:22 -07003415/**
David S. Miller7ce5a272013-12-02 17:26:05 -05003416 * pskb_trim_rcsum - trim received skb and update checksum
3417 * @skb: buffer to trim
3418 * @len: new length
3419 *
3420 * This is exactly the same as pskb_trim except that it ensures the
3421 * checksum of received packets are still valid after the operation.
Ross Lagerwall6c57f042019-01-17 15:34:38 +00003422 * It can change skb pointers.
David S. Miller7ce5a272013-12-02 17:26:05 -05003423 */
3424
3425static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3426{
3427 if (likely(len >= skb->len))
3428 return 0;
Eric Dumazet88078d92018-04-18 11:43:15 -07003429 return pskb_trim_rcsum_slow(skb, len);
David S. Miller7ce5a272013-12-02 17:26:05 -05003430}
3431
Daniel Borkmann5293efe2016-08-18 01:00:39 +02003432static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3433{
3434 if (skb->ip_summed == CHECKSUM_COMPLETE)
3435 skb->ip_summed = CHECKSUM_NONE;
3436 __skb_trim(skb, len);
3437 return 0;
3438}
3439
3440static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3441{
3442 if (skb->ip_summed == CHECKSUM_COMPLETE)
3443 skb->ip_summed = CHECKSUM_NONE;
3444 return __skb_grow(skb, len);
3445}
3446
Eric Dumazet18a4c0e2017-10-05 22:21:21 -07003447#define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3448#define skb_rb_first(root) rb_to_skb(rb_first(root))
3449#define skb_rb_last(root) rb_to_skb(rb_last(root))
3450#define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3451#define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3452
Linus Torvalds1da177e2005-04-16 15:20:36 -07003453#define skb_queue_walk(queue, skb) \
3454 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003455 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07003456 skb = skb->next)
3457
James Chapman46f89142007-04-30 00:07:31 -07003458#define skb_queue_walk_safe(queue, skb, tmp) \
3459 for (skb = (queue)->next, tmp = skb->next; \
3460 skb != (struct sk_buff *)(queue); \
3461 skb = tmp, tmp = skb->next)
3462
David S. Miller1164f522008-09-23 00:49:44 -07003463#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003464 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07003465 skb = skb->next)
3466
Eric Dumazet18a4c0e2017-10-05 22:21:21 -07003467#define skb_rbtree_walk(skb, root) \
3468 for (skb = skb_rb_first(root); skb != NULL; \
3469 skb = skb_rb_next(skb))
3470
3471#define skb_rbtree_walk_from(skb) \
3472 for (; skb != NULL; \
3473 skb = skb_rb_next(skb))
3474
3475#define skb_rbtree_walk_from_safe(skb, tmp) \
3476 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3477 skb = tmp)
3478
David S. Miller1164f522008-09-23 00:49:44 -07003479#define skb_queue_walk_from_safe(queue, skb, tmp) \
3480 for (tmp = skb->next; \
3481 skb != (struct sk_buff *)(queue); \
3482 skb = tmp, tmp = skb->next)
3483
Stephen Hemminger300ce172005-10-30 13:47:34 -08003484#define skb_queue_reverse_walk(queue, skb) \
3485 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003486 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08003487 skb = skb->prev)
3488
David S. Miller686a2952011-01-20 22:47:32 -08003489#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3490 for (skb = (queue)->prev, tmp = skb->prev; \
3491 skb != (struct sk_buff *)(queue); \
3492 skb = tmp, tmp = skb->prev)
3493
3494#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3495 for (tmp = skb->prev; \
3496 skb != (struct sk_buff *)(queue); \
3497 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003498
David S. Miller21dc3302010-08-23 00:13:46 -07003499static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07003500{
3501 return skb_shinfo(skb)->frag_list != NULL;
3502}
3503
3504static inline void skb_frag_list_init(struct sk_buff *skb)
3505{
3506 skb_shinfo(skb)->frag_list = NULL;
3507}
3508
David S. Milleree039872009-06-09 00:17:13 -07003509#define skb_walk_frags(skb, iter) \
3510 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3511
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003512
Sabrina Dubrocab50b0582019-11-25 14:48:57 +01003513int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue,
3514 int *err, long *timeo_p,
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003515 const struct sk_buff *skb);
Paolo Abeni65101ae2017-05-16 11:20:13 +02003516struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3517 struct sk_buff_head *queue,
3518 unsigned int flags,
Paolo Abenifd69c392019-04-08 10:15:59 +02003519 int *off, int *err,
Paolo Abeni65101ae2017-05-16 11:20:13 +02003520 struct sk_buff **last);
Sabrina Dubrocab50b0582019-11-25 14:48:57 +01003521struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
3522 struct sk_buff_head *queue,
Paolo Abenie427cad2020-02-28 14:45:22 +01003523 unsigned int flags, int *off, int *err,
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003524 struct sk_buff **last);
Sabrina Dubrocab50b0582019-11-25 14:48:57 +01003525struct sk_buff *__skb_recv_datagram(struct sock *sk,
3526 struct sk_buff_head *sk_queue,
Paolo Abenie427cad2020-02-28 14:45:22 +01003527 unsigned int flags, int *off, int *err);
Joe Perches7965bd42013-09-26 14:48:15 -07003528struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3529 int *err);
Linus Torvaldsa11e1d42018-06-28 09:43:44 -07003530__poll_t datagram_poll(struct file *file, struct socket *sock,
3531 struct poll_table_struct *wait);
Al Viroc0371da2014-11-24 10:42:55 -05003532int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3533 struct iov_iter *to, int size);
David S. Miller51f3d022014-11-05 16:46:40 -05003534static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3535 struct msghdr *msg, int size)
3536{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003537 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
David S. Miller51f3d022014-11-05 16:46:40 -05003538}
Al Viroe5a4b0b2014-11-24 18:17:55 -05003539int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3540 struct msghdr *msg);
Sagi Grimberg65d69e22018-12-03 17:52:10 -08003541int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3542 struct iov_iter *to, int len,
3543 struct ahash_request *hash);
Al Viro3a654f92014-06-19 14:15:22 -04003544int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3545 struct iov_iter *from, int len);
Al Viro3a654f92014-06-19 14:15:22 -04003546int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
Joe Perches7965bd42013-09-26 14:48:15 -07003547void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
samanthakumar627d2d62016-04-05 12:41:16 -04003548void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3549static inline void skb_free_datagram_locked(struct sock *sk,
3550 struct sk_buff *skb)
3551{
3552 __skb_free_datagram_locked(sk, skb, 0);
3553}
Joe Perches7965bd42013-09-26 14:48:15 -07003554int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07003555int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3556int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3557__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3558 int len, __wsum csum);
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +02003559int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
Joe Perches7965bd42013-09-26 14:48:15 -07003560 struct pipe_inode_info *pipe, unsigned int len,
Al Viro25869262016-09-17 21:02:10 -04003561 unsigned int flags);
Tom Herbert20bf50d2017-07-28 16:22:42 -07003562int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3563 int len);
Joe Perches7965bd42013-09-26 14:48:15 -07003564void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01003565unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00003566int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3567 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07003568void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3569int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3570void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Daniel Axtens779b7932018-03-01 17:13:37 +11003571bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
Daniel Axtens2b16f042018-01-31 14:15:33 +11003572bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
Joe Perches7965bd42013-09-26 14:48:15 -07003573struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Steffen Klassert3a1296a2020-01-25 11:26:44 +01003574struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features,
3575 unsigned int offset);
Vlad Yasevich0d5501c2014-08-08 14:42:13 -04003576struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
Jiri Pirkoe2195122014-11-19 14:05:01 +01003577int skb_ensure_writable(struct sk_buff *skb, int write_len);
Shmulik Ladkanibfca4c52016-09-19 19:11:09 +03003578int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
Jiri Pirko93515d52014-11-19 14:05:02 +01003579int skb_vlan_pop(struct sk_buff *skb);
3580int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
Davide Carattifa4e0f82019-10-12 13:55:07 +02003581int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
Martin Varghesed04ac222019-12-05 05:57:22 +05303582 int mac_len, bool ethernet);
Martin Varghese040b5cf2019-12-02 10:49:51 +05303583int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
3584 bool ethernet);
John Hurleyd27cf5c2019-07-07 15:01:56 +01003585int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
John Hurley2a2ea502019-07-07 15:01:57 +01003586int skb_mpls_dec_ttl(struct sk_buff *skb);
Sowmini Varadhan6fa01cc2016-04-22 18:36:35 -07003587struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3588 gfp_t gfp);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003589
Al Viro6ce8e9c2014-04-06 21:25:44 -04003590static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3591{
Al Viro3073f072017-02-17 23:13:25 -05003592 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
Al Viro6ce8e9c2014-04-06 21:25:44 -04003593}
3594
Al Viro7eab8d92014-04-06 21:51:23 -04003595static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3596{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003597 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro7eab8d92014-04-06 21:51:23 -04003598}
3599
Daniel Borkmann2817a332013-10-30 11:50:51 +01003600struct skb_checksum_ops {
3601 __wsum (*update)(const void *mem, int len, __wsum wsum);
3602 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3603};
3604
Davide Caratti96178132017-05-18 15:44:37 +02003605extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3606
Daniel Borkmann2817a332013-10-30 11:50:51 +01003607__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3608 __wsum csum, const struct skb_checksum_ops *ops);
3609__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3610 __wsum csum);
3611
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003612static inline void * __must_check
3613__skb_header_pointer(const struct sk_buff *skb, int offset,
3614 int len, void *data, int hlen, void *buffer)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003615{
Patrick McHardy55820ee2005-07-05 14:08:10 -07003616 if (hlen - offset >= len)
David S. Miller690e36e2014-08-23 12:13:41 -07003617 return data + offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003618
David S. Miller690e36e2014-08-23 12:13:41 -07003619 if (!skb ||
3620 skb_copy_bits(skb, offset, buffer, len) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003621 return NULL;
3622
3623 return buffer;
3624}
3625
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003626static inline void * __must_check
3627skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
David S. Miller690e36e2014-08-23 12:13:41 -07003628{
3629 return __skb_header_pointer(skb, offset, len, skb->data,
3630 skb_headlen(skb), buffer);
3631}
3632
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01003633/**
3634 * skb_needs_linearize - check if we need to linearize a given skb
3635 * depending on the given device features.
3636 * @skb: socket buffer to check
3637 * @features: net device features
3638 *
3639 * Returns true if either:
3640 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3641 * 2. skb is fragmented and the device does not support SG.
3642 */
3643static inline bool skb_needs_linearize(struct sk_buff *skb,
3644 netdev_features_t features)
3645{
3646 return skb_is_nonlinear(skb) &&
3647 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3648 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3649}
3650
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03003651static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3652 void *to,
3653 const unsigned int len)
3654{
3655 memcpy(to, skb->data, len);
3656}
3657
3658static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3659 const int offset, void *to,
3660 const unsigned int len)
3661{
3662 memcpy(to, skb->data + offset, len);
3663}
3664
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03003665static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3666 const void *from,
3667 const unsigned int len)
3668{
3669 memcpy(skb->data, from, len);
3670}
3671
3672static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3673 const int offset,
3674 const void *from,
3675 const unsigned int len)
3676{
3677 memcpy(skb->data + offset, from, len);
3678}
3679
Joe Perches7965bd42013-09-26 14:48:15 -07003680void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003681
Patrick Ohlyac45f602009-02-12 05:03:37 +00003682static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3683{
3684 return skb->tstamp;
3685}
3686
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003687/**
3688 * skb_get_timestamp - get timestamp from a skb
3689 * @skb: skb to get stamp from
Deepa Dinamani13c6ee22019-02-02 07:34:48 -08003690 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003691 *
3692 * Timestamps are stored in the skb as offsets to a base timestamp.
3693 * This function converts the offset back to a struct timeval and stores
3694 * it in stamp.
3695 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00003696static inline void skb_get_timestamp(const struct sk_buff *skb,
Deepa Dinamani13c6ee22019-02-02 07:34:48 -08003697 struct __kernel_old_timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003698{
Deepa Dinamani13c6ee22019-02-02 07:34:48 -08003699 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003700}
3701
Deepa Dinamani887feae2019-02-02 07:34:50 -08003702static inline void skb_get_new_timestamp(const struct sk_buff *skb,
3703 struct __kernel_sock_timeval *stamp)
3704{
3705 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3706
3707 stamp->tv_sec = ts.tv_sec;
3708 stamp->tv_usec = ts.tv_nsec / 1000;
3709}
3710
Patrick Ohlyac45f602009-02-12 05:03:37 +00003711static inline void skb_get_timestampns(const struct sk_buff *skb,
Arnd Bergmanndf1b4ba2019-10-25 22:04:46 +02003712 struct __kernel_old_timespec *stamp)
Patrick Ohlyac45f602009-02-12 05:03:37 +00003713{
Arnd Bergmanndf1b4ba2019-10-25 22:04:46 +02003714 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3715
3716 stamp->tv_sec = ts.tv_sec;
3717 stamp->tv_nsec = ts.tv_nsec;
Patrick Ohlyac45f602009-02-12 05:03:37 +00003718}
3719
Deepa Dinamani887feae2019-02-02 07:34:50 -08003720static inline void skb_get_new_timestampns(const struct sk_buff *skb,
3721 struct __kernel_timespec *stamp)
3722{
3723 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3724
3725 stamp->tv_sec = ts.tv_sec;
3726 stamp->tv_nsec = ts.tv_nsec;
3727}
3728
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003729static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003730{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003731 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003732}
3733
Stephen Hemminger164891a2007-04-23 22:26:16 -07003734static inline ktime_t net_timedelta(ktime_t t)
3735{
3736 return ktime_sub(ktime_get_real(), t);
3737}
3738
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003739static inline ktime_t net_invalid_timestamp(void)
3740{
Thomas Gleixner8b0e1952016-12-25 12:30:41 +01003741 return 0;
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003742}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003743
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003744static inline u8 skb_metadata_len(const struct sk_buff *skb)
3745{
3746 return skb_shinfo(skb)->meta_len;
3747}
3748
3749static inline void *skb_metadata_end(const struct sk_buff *skb)
3750{
3751 return skb_mac_header(skb);
3752}
3753
3754static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3755 const struct sk_buff *skb_b,
3756 u8 meta_len)
3757{
3758 const void *a = skb_metadata_end(skb_a);
3759 const void *b = skb_metadata_end(skb_b);
3760 /* Using more efficient varaiant than plain call to memcmp(). */
3761#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3762 u64 diffs = 0;
3763
3764 switch (meta_len) {
3765#define __it(x, op) (x -= sizeof(u##op))
3766#define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3767 case 32: diffs |= __it_diff(a, b, 64);
Gustavo A. R. Silva82385b02018-10-17 15:01:37 +02003768 /* fall through */
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003769 case 24: diffs |= __it_diff(a, b, 64);
Gustavo A. R. Silva82385b02018-10-17 15:01:37 +02003770 /* fall through */
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003771 case 16: diffs |= __it_diff(a, b, 64);
Gustavo A. R. Silva82385b02018-10-17 15:01:37 +02003772 /* fall through */
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003773 case 8: diffs |= __it_diff(a, b, 64);
3774 break;
3775 case 28: diffs |= __it_diff(a, b, 64);
Gustavo A. R. Silva82385b02018-10-17 15:01:37 +02003776 /* fall through */
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003777 case 20: diffs |= __it_diff(a, b, 64);
Gustavo A. R. Silva82385b02018-10-17 15:01:37 +02003778 /* fall through */
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003779 case 12: diffs |= __it_diff(a, b, 64);
Gustavo A. R. Silva82385b02018-10-17 15:01:37 +02003780 /* fall through */
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003781 case 4: diffs |= __it_diff(a, b, 32);
3782 break;
3783 }
3784 return diffs;
3785#else
3786 return memcmp(a - meta_len, b - meta_len, meta_len);
3787#endif
3788}
3789
3790static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3791 const struct sk_buff *skb_b)
3792{
3793 u8 len_a = skb_metadata_len(skb_a);
3794 u8 len_b = skb_metadata_len(skb_b);
3795
3796 if (!(len_a | len_b))
3797 return false;
3798
3799 return len_a != len_b ?
3800 true : __skb_metadata_differs(skb_a, skb_b, len_a);
3801}
3802
3803static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3804{
3805 skb_shinfo(skb)->meta_len = meta_len;
3806}
3807
3808static inline void skb_metadata_clear(struct sk_buff *skb)
3809{
3810 skb_metadata_set(skb, 0);
3811}
3812
Alexander Duyck62bccb82014-09-04 13:31:35 -04003813struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3814
Richard Cochranc1f19b52010-07-17 08:49:36 +00003815#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3816
Joe Perches7965bd42013-09-26 14:48:15 -07003817void skb_clone_tx_timestamp(struct sk_buff *skb);
3818bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00003819
3820#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3821
3822static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3823{
3824}
3825
3826static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3827{
3828 return false;
3829}
3830
3831#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3832
3833/**
3834 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3835 *
Richard Cochranda92b192011-10-21 00:49:15 +00003836 * PHY drivers may accept clones of transmitted packets for
3837 * timestamping via their phy_driver.txtstamp method. These drivers
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003838 * must call this function to return the skb back to the stack with a
3839 * timestamp.
Richard Cochranda92b192011-10-21 00:49:15 +00003840 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00003841 * @skb: clone of the the original outgoing packet
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003842 * @hwtstamps: hardware time stamps
Richard Cochranc1f19b52010-07-17 08:49:36 +00003843 *
3844 */
3845void skb_complete_tx_timestamp(struct sk_buff *skb,
3846 struct skb_shared_hwtstamps *hwtstamps);
3847
Willem de Bruijne7fd2882014-08-04 22:11:48 -04003848void __skb_tstamp_tx(struct sk_buff *orig_skb,
3849 struct skb_shared_hwtstamps *hwtstamps,
3850 struct sock *sk, int tstype);
3851
Patrick Ohlyac45f602009-02-12 05:03:37 +00003852/**
3853 * skb_tstamp_tx - queue clone of skb with send time stamps
3854 * @orig_skb: the original outgoing packet
3855 * @hwtstamps: hardware time stamps, may be NULL if not available
3856 *
3857 * If the skb has a socket associated, then this function clones the
3858 * skb (thus sharing the actual data and optional structures), stores
3859 * the optional hardware time stamping information (if non NULL) or
3860 * generates a software time stamp (otherwise), then queues the clone
3861 * to the error queue of the socket. Errors are silently ignored.
3862 */
Joe Perches7965bd42013-09-26 14:48:15 -07003863void skb_tstamp_tx(struct sk_buff *orig_skb,
3864 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00003865
Richard Cochran4507a712010-07-17 08:48:28 +00003866/**
3867 * skb_tx_timestamp() - Driver hook for transmit timestamping
3868 *
3869 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00003870 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00003871 *
David S. Miller73409f32013-12-27 13:04:33 -05003872 * Specifically, one should make absolutely sure that this function is
3873 * called before TX completion of this packet can trigger. Otherwise
3874 * the packet could potentially already be freed.
3875 *
Richard Cochran4507a712010-07-17 08:48:28 +00003876 * @skb: A socket buffer.
3877 */
3878static inline void skb_tx_timestamp(struct sk_buff *skb)
3879{
Richard Cochranc1f19b52010-07-17 08:49:36 +00003880 skb_clone_tx_timestamp(skb);
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +02003881 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3882 skb_tstamp_tx(skb, NULL);
Richard Cochran4507a712010-07-17 08:48:28 +00003883}
3884
Johannes Berg6e3e9392011-11-09 10:15:42 +01003885/**
3886 * skb_complete_wifi_ack - deliver skb with wifi status
3887 *
3888 * @skb: the original outgoing packet
3889 * @acked: ack status
3890 *
3891 */
3892void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3893
Joe Perches7965bd42013-09-26 14:48:15 -07003894__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3895__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003896
Herbert Xu60476372007-04-09 11:59:39 -07003897static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3898{
Tom Herbert6edec0e2015-02-10 16:30:28 -08003899 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3900 skb->csum_valid ||
3901 (skb->ip_summed == CHECKSUM_PARTIAL &&
3902 skb_checksum_start_offset(skb) >= 0));
Herbert Xu60476372007-04-09 11:59:39 -07003903}
3904
Herbert Xufb286bb2005-11-10 13:01:24 -08003905/**
3906 * skb_checksum_complete - Calculate checksum of an entire packet
3907 * @skb: packet to process
3908 *
3909 * This function calculates the checksum over the entire packet plus
3910 * the value of skb->csum. The latter can be used to supply the
3911 * checksum of a pseudo header as used by TCP/UDP. It returns the
3912 * checksum.
3913 *
3914 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3915 * this function can be used to verify that checksum on received
3916 * packets. In that case the function should return zero if the
3917 * checksum is correct. In particular, this function will return zero
3918 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3919 * hardware has already verified the correctness of the checksum.
3920 */
Al Viro4381ca32007-07-15 21:00:11 +01003921static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08003922{
Herbert Xu60476372007-04-09 11:59:39 -07003923 return skb_csum_unnecessary(skb) ?
3924 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003925}
3926
Tom Herbert77cffe22014-08-27 21:26:46 -07003927static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3928{
3929 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3930 if (skb->csum_level == 0)
3931 skb->ip_summed = CHECKSUM_NONE;
3932 else
3933 skb->csum_level--;
3934 }
3935}
3936
3937static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3938{
3939 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3940 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3941 skb->csum_level++;
3942 } else if (skb->ip_summed == CHECKSUM_NONE) {
3943 skb->ip_summed = CHECKSUM_UNNECESSARY;
3944 skb->csum_level = 0;
3945 }
3946}
3947
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003948/* Check if we need to perform checksum complete validation.
3949 *
3950 * Returns true if checksum complete is needed, false otherwise
3951 * (either checksum is unnecessary or zero checksum is allowed).
3952 */
3953static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3954 bool zero_okay,
3955 __sum16 check)
3956{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003957 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3958 skb->csum_valid = 1;
Tom Herbert77cffe22014-08-27 21:26:46 -07003959 __skb_decr_checksum_unnecessary(skb);
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003960 return false;
3961 }
3962
3963 return true;
3964}
3965
David S. Millerda279882018-02-16 15:52:42 -05003966/* For small packets <= CHECKSUM_BREAK perform checksum complete directly
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003967 * in checksum_init.
3968 */
3969#define CHECKSUM_BREAK 76
3970
Tom Herbert4e18b9a2015-04-20 14:10:04 -07003971/* Unset checksum-complete
3972 *
3973 * Unset checksum complete can be done when packet is being modified
3974 * (uncompressed for instance) and checksum-complete value is
3975 * invalidated.
3976 */
3977static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3978{
3979 if (skb->ip_summed == CHECKSUM_COMPLETE)
3980 skb->ip_summed = CHECKSUM_NONE;
3981}
3982
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003983/* Validate (init) checksum based on checksum complete.
3984 *
3985 * Return values:
3986 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3987 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3988 * checksum is stored in skb->csum for use in __skb_checksum_complete
3989 * non-zero: value of invalid checksum
3990 *
3991 */
3992static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3993 bool complete,
3994 __wsum psum)
3995{
3996 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3997 if (!csum_fold(csum_add(psum, skb->csum))) {
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003998 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003999 return 0;
4000 }
4001 }
4002
4003 skb->csum = psum;
4004
Tom Herbert5d0c2b92014-06-10 18:54:13 -07004005 if (complete || skb->len <= CHECKSUM_BREAK) {
4006 __sum16 csum;
4007
4008 csum = __skb_checksum_complete(skb);
4009 skb->csum_valid = !csum;
4010 return csum;
4011 }
Tom Herbert76ba0aa2014-05-02 16:29:18 -07004012
4013 return 0;
4014}
4015
4016static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
4017{
4018 return 0;
4019}
4020
4021/* Perform checksum validate (init). Note that this is a macro since we only
4022 * want to calculate the pseudo header which is an input function if necessary.
4023 * First we try to validate without any computation (checksum unnecessary) and
4024 * then calculate based on checksum complete calling the function to compute
4025 * pseudo header.
4026 *
4027 * Return values:
4028 * 0: checksum is validated or try to in skb_checksum_complete
4029 * non-zero: value of invalid checksum
4030 */
4031#define __skb_checksum_validate(skb, proto, complete, \
4032 zero_okay, check, compute_pseudo) \
4033({ \
4034 __sum16 __ret = 0; \
Tom Herbert5d0c2b92014-06-10 18:54:13 -07004035 skb->csum_valid = 0; \
Tom Herbert76ba0aa2014-05-02 16:29:18 -07004036 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
4037 __ret = __skb_checksum_validate_complete(skb, \
4038 complete, compute_pseudo(skb, proto)); \
4039 __ret; \
4040})
4041
4042#define skb_checksum_init(skb, proto, compute_pseudo) \
4043 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
4044
4045#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
4046 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4047
4048#define skb_checksum_validate(skb, proto, compute_pseudo) \
4049 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4050
4051#define skb_checksum_validate_zero_check(skb, proto, check, \
4052 compute_pseudo) \
Sabrina Dubroca096a4cf2015-02-06 18:54:19 +01004053 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
Tom Herbert76ba0aa2014-05-02 16:29:18 -07004054
4055#define skb_checksum_simple_validate(skb) \
4056 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4057
Tom Herbertd96535a2014-08-31 15:12:42 -07004058static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4059{
Davide Caratti219f1d792017-05-18 15:44:39 +02004060 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
Tom Herbertd96535a2014-08-31 15:12:42 -07004061}
4062
Li RongQinge4aa33a2019-07-04 17:03:26 +08004063static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
Tom Herbertd96535a2014-08-31 15:12:42 -07004064{
4065 skb->csum = ~pseudo;
4066 skb->ip_summed = CHECKSUM_COMPLETE;
4067}
4068
Li RongQinge4aa33a2019-07-04 17:03:26 +08004069#define skb_checksum_try_convert(skb, proto, compute_pseudo) \
Tom Herbertd96535a2014-08-31 15:12:42 -07004070do { \
4071 if (__skb_checksum_convert_check(skb)) \
Li RongQinge4aa33a2019-07-04 17:03:26 +08004072 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
Tom Herbertd96535a2014-08-31 15:12:42 -07004073} while (0)
4074
Tom Herbert15e23962015-02-10 16:30:31 -08004075static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4076 u16 start, u16 offset)
4077{
4078 skb->ip_summed = CHECKSUM_PARTIAL;
4079 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4080 skb->csum_offset = offset - start;
4081}
4082
Tom Herbertdcdc8992015-02-02 16:07:34 -08004083/* Update skbuf and packet to reflect the remote checksum offload operation.
4084 * When called, ptr indicates the starting point for skb->csum when
4085 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4086 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4087 */
4088static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
Tom Herbert15e23962015-02-10 16:30:31 -08004089 int start, int offset, bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08004090{
4091 __wsum delta;
4092
Tom Herbert15e23962015-02-10 16:30:31 -08004093 if (!nopartial) {
4094 skb_remcsum_adjust_partial(skb, ptr, start, offset);
4095 return;
4096 }
4097
Tom Herbertdcdc8992015-02-02 16:07:34 -08004098 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4099 __skb_checksum_complete(skb);
4100 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4101 }
4102
4103 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4104
4105 /* Adjust skb->csum since we changed the packet */
4106 skb->csum = csum_add(skb->csum, delta);
4107}
4108
Florian Westphalcb9c6832017-01-23 18:21:56 +01004109static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4110{
4111#if IS_ENABLED(CONFIG_NF_CONNTRACK)
Jeremy Sowden261db6c2019-09-13 09:13:14 +01004112 return (void *)(skb->_nfct & NFCT_PTRMASK);
Florian Westphalcb9c6832017-01-23 18:21:56 +01004113#else
4114 return NULL;
4115#endif
4116}
4117
Jeremy Sowden261db6c2019-09-13 09:13:14 +01004118static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07004119{
Jeremy Sowden261db6c2019-09-13 09:13:14 +01004120#if IS_ENABLED(CONFIG_NF_CONNTRACK)
4121 return skb->_nfct;
4122#else
4123 return 0UL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01004124#endif
Jeremy Sowden261db6c2019-09-13 09:13:14 +01004125}
4126
4127static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
4128{
4129#if IS_ENABLED(CONFIG_NF_CONNTRACK)
4130 skb->_nfct = nfct;
4131#endif
4132}
Florian Westphaldf5042f2018-12-18 17:15:16 +01004133
4134#ifdef CONFIG_SKB_EXTENSIONS
4135enum skb_ext_id {
4136#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4137 SKB_EXT_BRIDGE_NF,
4138#endif
Florian Westphal41650792018-12-18 17:15:27 +01004139#ifdef CONFIG_XFRM
4140 SKB_EXT_SEC_PATH,
4141#endif
Paul Blakey95a72332019-09-04 16:56:37 +03004142#if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4143 TC_SKB_EXT,
4144#endif
Mat Martineau3ee17bc2020-01-09 07:59:19 -08004145#if IS_ENABLED(CONFIG_MPTCP)
4146 SKB_EXT_MPTCP,
4147#endif
Florian Westphaldf5042f2018-12-18 17:15:16 +01004148 SKB_EXT_NUM, /* must be last */
4149};
4150
4151/**
4152 * struct skb_ext - sk_buff extensions
4153 * @refcnt: 1 on allocation, deallocated on 0
4154 * @offset: offset to add to @data to obtain extension address
4155 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4156 * @data: start of extension data, variable sized
4157 *
4158 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4159 * to use 'u8' types while allowing up to 2kb worth of extension data.
4160 */
4161struct skb_ext {
4162 refcount_t refcnt;
4163 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4164 u8 chunks; /* same */
Gustavo A. R. Silva5c91aa12020-03-23 19:22:24 -05004165 char data[] __aligned(8);
Florian Westphaldf5042f2018-12-18 17:15:16 +01004166};
4167
Paolo Abeni8b69a802020-01-09 07:59:24 -08004168struct skb_ext *__skb_ext_alloc(void);
4169void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id,
4170 struct skb_ext *ext);
Florian Westphaldf5042f2018-12-18 17:15:16 +01004171void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4172void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4173void __skb_ext_put(struct skb_ext *ext);
4174
4175static inline void skb_ext_put(struct sk_buff *skb)
4176{
4177 if (skb->active_extensions)
4178 __skb_ext_put(skb->extensions);
4179}
4180
Florian Westphaldf5042f2018-12-18 17:15:16 +01004181static inline void __skb_ext_copy(struct sk_buff *dst,
4182 const struct sk_buff *src)
4183{
4184 dst->active_extensions = src->active_extensions;
4185
4186 if (src->active_extensions) {
4187 struct skb_ext *ext = src->extensions;
4188
4189 refcount_inc(&ext->refcnt);
4190 dst->extensions = ext;
4191 }
4192}
4193
4194static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4195{
4196 skb_ext_put(dst);
4197 __skb_ext_copy(dst, src);
4198}
4199
4200static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4201{
4202 return !!ext->offset[i];
4203}
4204
4205static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4206{
4207 return skb->active_extensions & (1 << id);
4208}
4209
4210static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4211{
4212 if (skb_ext_exist(skb, id))
4213 __skb_ext_del(skb, id);
4214}
4215
4216static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4217{
4218 if (skb_ext_exist(skb, id)) {
4219 struct skb_ext *ext = skb->extensions;
4220
4221 return (void *)ext + (ext->offset[id] << 3);
4222 }
4223
4224 return NULL;
4225}
Florian Westphal174e2382019-09-26 20:37:05 +02004226
4227static inline void skb_ext_reset(struct sk_buff *skb)
4228{
4229 if (unlikely(skb->active_extensions)) {
4230 __skb_ext_put(skb->extensions);
4231 skb->active_extensions = 0;
4232 }
4233}
Florian Westphal677bf082019-11-21 06:56:23 +01004234
4235static inline bool skb_has_extensions(struct sk_buff *skb)
4236{
4237 return unlikely(skb->active_extensions);
4238}
Florian Westphaldf5042f2018-12-18 17:15:16 +01004239#else
4240static inline void skb_ext_put(struct sk_buff *skb) {}
Florian Westphal174e2382019-09-26 20:37:05 +02004241static inline void skb_ext_reset(struct sk_buff *skb) {}
Florian Westphaldf5042f2018-12-18 17:15:16 +01004242static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
4243static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {}
4244static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
Florian Westphal677bf082019-11-21 06:56:23 +01004245static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
Florian Westphaldf5042f2018-12-18 17:15:16 +01004246#endif /* CONFIG_SKB_EXTENSIONS */
4247
Florian Westphal895b5c92019-09-29 20:54:03 +02004248static inline void nf_reset_ct(struct sk_buff *skb)
Patrick McHardya193a4a2006-03-20 19:23:05 -08004249{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07004250#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01004251 nf_conntrack_put(skb_nfct(skb));
4252 skb->_nfct = 0;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01004253#endif
Patrick McHardya193a4a2006-03-20 19:23:05 -08004254}
4255
Patrick McHardy124dff02013-04-05 20:42:05 +02004256static inline void nf_reset_trace(struct sk_buff *skb)
4257{
Florian Westphal478b3602014-02-15 23:48:45 +01004258#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00004259 skb->nf_trace = 0;
4260#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07004261}
4262
Ye Yin2b5ec1a2017-10-26 16:57:05 +08004263static inline void ipvs_reset(struct sk_buff *skb)
4264{
4265#if IS_ENABLED(CONFIG_IP_VS)
4266 skb->ipvs_property = 0;
4267#endif
4268}
4269
Florian Westphalde8bda12018-12-18 17:15:17 +01004270/* Note: This doesn't put any conntrack info in dst. */
Eric Dumazetb1937222014-09-28 22:18:47 -07004271static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4272 bool copy)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07004273{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07004274#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01004275 dst->_nfct = src->_nfct;
4276 nf_conntrack_get(skb_nfct(src));
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01004277#endif
Florian Westphal478b3602014-02-15 23:48:45 +01004278#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Eric Dumazetb1937222014-09-28 22:18:47 -07004279 if (copy)
4280 dst->nf_trace = src->nf_trace;
Florian Westphal478b3602014-02-15 23:48:45 +01004281#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07004282}
4283
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07004284static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4285{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07004286#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01004287 nf_conntrack_put(skb_nfct(dst));
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01004288#endif
Eric Dumazetb1937222014-09-28 22:18:47 -07004289 __nf_copy(dst, src, true);
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07004290}
4291
James Morris984bc162006-06-09 00:29:17 -07004292#ifdef CONFIG_NETWORK_SECMARK
4293static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4294{
4295 to->secmark = from->secmark;
4296}
4297
4298static inline void skb_init_secmark(struct sk_buff *skb)
4299{
4300 skb->secmark = 0;
4301}
4302#else
4303static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4304{ }
4305
4306static inline void skb_init_secmark(struct sk_buff *skb)
4307{ }
4308#endif
4309
Florian Westphal7af8f4c2018-12-18 17:15:19 +01004310static inline int secpath_exists(const struct sk_buff *skb)
4311{
4312#ifdef CONFIG_XFRM
Florian Westphal41650792018-12-18 17:15:27 +01004313 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
Florian Westphal7af8f4c2018-12-18 17:15:19 +01004314#else
4315 return 0;
4316#endif
4317}
4318
Eric W. Biederman574f7192014-04-01 12:20:24 -07004319static inline bool skb_irq_freeable(const struct sk_buff *skb)
4320{
4321 return !skb->destructor &&
Florian Westphal7af8f4c2018-12-18 17:15:19 +01004322 !secpath_exists(skb) &&
Florian Westphalcb9c6832017-01-23 18:21:56 +01004323 !skb_nfct(skb) &&
Eric W. Biederman574f7192014-04-01 12:20:24 -07004324 !skb->_skb_refdst &&
4325 !skb_has_frag_list(skb);
4326}
4327
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07004328static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4329{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07004330 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07004331}
4332
Stephen Hemminger92477442009-03-21 13:39:26 -07004333static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07004334{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07004335 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07004336}
4337
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07004338static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4339{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07004340 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07004341}
4342
David S. Millerd5a9e242009-01-27 16:22:11 -08004343static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4344{
4345 skb->queue_mapping = rx_queue + 1;
4346}
4347
Stephen Hemminger92477442009-03-21 13:39:26 -07004348static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08004349{
4350 return skb->queue_mapping - 1;
4351}
4352
Stephen Hemminger92477442009-03-21 13:39:26 -07004353static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08004354{
Eric Dumazeta02cec22010-09-22 20:43:57 +00004355 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08004356}
4357
Julian Anastasov4ff06202017-02-06 23:14:12 +02004358static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4359{
4360 skb->dst_pending_confirm = val;
4361}
4362
4363static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4364{
4365 return skb->dst_pending_confirm != 0;
4366}
4367
Florian Westphal2294be0f2018-12-18 17:15:20 +01004368static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04004369{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07004370#ifdef CONFIG_XFRM
Florian Westphal41650792018-12-18 17:15:27 +01004371 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07004372#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07004373 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07004374#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04004375}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07004376
Pravin B Shelar68c33162013-02-14 14:02:41 +00004377/* Keeps track of mac header offset relative to skb->head.
4378 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4379 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07004380 * tunnel skb it points to outer mac header.
4381 * Keeps track of level of encapsulation of network headers.
4382 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00004383struct skb_gso_cb {
Alexander Duyck802ab552016-04-10 21:45:03 -04004384 union {
4385 int mac_offset;
4386 int data_offset;
4387 };
Eric Dumazet3347c962013-10-19 11:42:56 -07004388 int encap_level;
Alexander Duyck76443452016-02-05 15:27:37 -08004389 __wsum csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07004390 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00004391};
Cambda Zhua08e7fd2020-03-26 15:33:14 +08004392#define SKB_GSO_CB_OFFSET 32
4393#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_GSO_CB_OFFSET))
Pravin B Shelar68c33162013-02-14 14:02:41 +00004394
4395static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4396{
4397 return (skb_mac_header(inner_skb) - inner_skb->head) -
4398 SKB_GSO_CB(inner_skb)->mac_offset;
4399}
4400
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00004401static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4402{
4403 int new_headroom, headroom;
4404 int ret;
4405
4406 headroom = skb_headroom(skb);
4407 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4408 if (ret)
4409 return ret;
4410
4411 new_headroom = skb_headroom(skb);
4412 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4413 return 0;
4414}
4415
Alexander Duyck08b64fc2016-02-05 15:27:49 -08004416static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4417{
4418 /* Do not update partial checksums if remote checksum is enabled. */
4419 if (skb->remcsum_offload)
4420 return;
4421
4422 SKB_GSO_CB(skb)->csum = res;
4423 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4424}
4425
Tom Herbert7e2b10c2014-06-04 17:20:02 -07004426/* Compute the checksum for a gso segment. First compute the checksum value
4427 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4428 * then add in skb->csum (checksum from csum_start to end of packet).
4429 * skb->csum and csum_start are then updated to reflect the checksum of the
4430 * resultant packet starting from the transport header-- the resultant checksum
4431 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4432 * header.
4433 */
4434static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4435{
Alexander Duyck76443452016-02-05 15:27:37 -08004436 unsigned char *csum_start = skb_transport_header(skb);
4437 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4438 __wsum partial = SKB_GSO_CB(skb)->csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07004439
Alexander Duyck76443452016-02-05 15:27:37 -08004440 SKB_GSO_CB(skb)->csum = res;
4441 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07004442
Alexander Duyck76443452016-02-05 15:27:37 -08004443 return csum_fold(csum_partial(csum_start, plen, partial));
Tom Herbert7e2b10c2014-06-04 17:20:02 -07004444}
4445
David S. Millerbdcc0922012-03-07 20:53:36 -05004446static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07004447{
4448 return skb_shinfo(skb)->gso_size;
4449}
4450
Eric Dumazet36a8f392013-09-29 01:21:32 -07004451/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05004452static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02004453{
4454 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4455}
4456
Daniel Axtensd02f51c2018-03-03 03:03:46 +01004457/* Note: Should be called only if skb_is_gso(skb) is true */
4458static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4459{
4460 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4461}
4462
Willem de Bruijn4c3024d2019-03-06 14:35:15 -05004463/* Note: Should be called only if skb_is_gso(skb) is true */
Willem de Bruijnb90efd22019-02-07 14:54:16 -05004464static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4465{
Willem de Bruijn4c3024d2019-03-06 14:35:15 -05004466 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
Willem de Bruijnb90efd22019-02-07 14:54:16 -05004467}
4468
Daniel Borkmann5293efe2016-08-18 01:00:39 +02004469static inline void skb_gso_reset(struct sk_buff *skb)
4470{
4471 skb_shinfo(skb)->gso_size = 0;
4472 skb_shinfo(skb)->gso_segs = 0;
4473 skb_shinfo(skb)->gso_type = 0;
4474}
4475
Daniel Axtensd02f51c2018-03-03 03:03:46 +01004476static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4477 u16 increment)
4478{
4479 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4480 return;
4481 shinfo->gso_size += increment;
4482}
4483
4484static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4485 u16 decrement)
4486{
4487 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4488 return;
4489 shinfo->gso_size -= decrement;
4490}
4491
Joe Perches7965bd42013-09-26 14:48:15 -07004492void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07004493
4494static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4495{
4496 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4497 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04004498 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4499
Alexander Duyckb78462e2010-06-02 12:24:37 +00004500 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4501 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07004502 __skb_warn_lro_forwarding(skb);
4503 return true;
4504 }
4505 return false;
4506}
4507
Herbert Xu35fc92a2007-03-26 23:22:20 -07004508static inline void skb_forward_csum(struct sk_buff *skb)
4509{
4510 /* Unfortunately we don't support this one. Any brave souls? */
4511 if (skb->ip_summed == CHECKSUM_COMPLETE)
4512 skb->ip_summed = CHECKSUM_NONE;
4513}
4514
Eric Dumazetbc8acf22010-09-02 13:07:41 -07004515/**
4516 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4517 * @skb: skb to check
4518 *
4519 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4520 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4521 * use this helper, to document places where we make this assertion.
4522 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04004523static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07004524{
4525#ifdef DEBUG
4526 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4527#endif
4528}
4529
Rusty Russellf35d9d82008-02-04 23:49:54 -05004530bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00004531
Paul Durranted1f50c2014-01-09 10:02:46 +00004532int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
Linus Lüssing9afd85c2015-05-02 14:01:07 +02004533struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4534 unsigned int transport_len,
4535 __sum16(*skb_chkf)(struct sk_buff *skb));
Paul Durranted1f50c2014-01-09 10:02:46 +00004536
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00004537/**
4538 * skb_head_is_locked - Determine if the skb->head is locked down
4539 * @skb: skb to check
4540 *
4541 * The head on skbs build around a head frag can be removed if they are
4542 * not cloned. This function returns true if the skb head is locked down
4543 * due to either being allocated via kmalloc, or by being a clone with
4544 * multiple references to the head.
4545 */
4546static inline bool skb_head_is_locked(const struct sk_buff *skb)
4547{
4548 return !skb->head_frag || skb_cloned(skb);
4549}
Florian Westphalfe6cc552014-02-13 23:09:12 +01004550
Edward Cree179bc672016-02-11 20:48:04 +00004551/* Local Checksum Offload.
4552 * Compute outer checksum based on the assumption that the
4553 * inner checksum will be offloaded later.
Otto Sabartd0dcde62019-01-06 00:29:15 +01004554 * See Documentation/networking/checksum-offloads.rst for
Edward Creee8ae7b02016-02-11 21:03:37 +00004555 * explanation of how this works.
Edward Cree179bc672016-02-11 20:48:04 +00004556 * Fill in outer checksum adjustment (e.g. with sum of outer
4557 * pseudo-header) before calling.
4558 * Also ensure that inner checksum is in linear data area.
4559 */
4560static inline __wsum lco_csum(struct sk_buff *skb)
4561{
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004562 unsigned char *csum_start = skb_checksum_start(skb);
4563 unsigned char *l4_hdr = skb_transport_header(skb);
4564 __wsum partial;
Edward Cree179bc672016-02-11 20:48:04 +00004565
4566 /* Start with complement of inner checksum adjustment */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004567 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4568 skb->csum_offset));
4569
Edward Cree179bc672016-02-11 20:48:04 +00004570 /* Add in checksum of our headers (incl. outer checksum
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004571 * adjustment filled in by caller) and return result.
Edward Cree179bc672016-02-11 20:48:04 +00004572 */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004573 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
Edward Cree179bc672016-02-11 20:48:04 +00004574}
4575
Pablo Neira Ayuso2c646052020-03-25 13:47:18 +01004576static inline bool skb_is_redirected(const struct sk_buff *skb)
4577{
4578#ifdef CONFIG_NET_REDIRECT
4579 return skb->redirected;
4580#else
4581 return false;
4582#endif
4583}
4584
4585static inline void skb_set_redirected(struct sk_buff *skb, bool from_ingress)
4586{
4587#ifdef CONFIG_NET_REDIRECT
4588 skb->redirected = 1;
4589 skb->from_ingress = from_ingress;
4590 if (skb->from_ingress)
4591 skb->tstamp = 0;
4592#endif
4593}
4594
4595static inline void skb_reset_redirect(struct sk_buff *skb)
4596{
4597#ifdef CONFIG_NET_REDIRECT
4598 skb->redirected = 0;
4599#endif
4600}
4601
Linus Torvalds1da177e2005-04-16 15:20:36 -07004602#endif /* __KERNEL__ */
4603#endif /* _LINUX_SKBUFF_H */