blob: ca8806b693882e9ae933d70d0bb6c4cbd3ca81bd [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.
50 * From the stack's point of view these are capabilities offered by the driver,
51 * a driver typically only advertises features that it is capable of offloading
52 * to its device.
53 *
54 * The checksum related features are:
55 *
56 * NETIF_F_HW_CSUM - The driver (or its device) is able to compute one
57 * IP (one's complement) checksum for any combination
58 * of protocols or protocol layering. The checksum is
59 * computed and set in a packet per the CHECKSUM_PARTIAL
60 * interface (see below).
61 *
62 * NETIF_F_IP_CSUM - Driver (device) is only able to checksum plain
63 * TCP or UDP packets over IPv4. These are specifically
64 * unencapsulated packets of the form IPv4|TCP or
65 * IPv4|UDP where the Protocol field in the IPv4 header
66 * is TCP or UDP. The IPv4 header may contain IP options
67 * This feature cannot be set in features for a device
68 * with NETIF_F_HW_CSUM also set. This feature is being
69 * DEPRECATED (see below).
70 *
71 * NETIF_F_IPV6_CSUM - Driver (device) is only able to checksum plain
72 * TCP or UDP packets over IPv6. These are specifically
73 * unencapsulated packets of the form IPv6|TCP or
74 * IPv4|UDP where the Next Header field in the IPv6
75 * header is either TCP or UDP. IPv6 extension headers
76 * are not supported with this feature. This feature
77 * cannot be set in features for a device with
78 * NETIF_F_HW_CSUM also set. This feature is being
79 * DEPRECATED (see below).
80 *
81 * NETIF_F_RXCSUM - Driver (device) performs receive checksum offload.
82 * This flag is used only used to disable the RX checksum
83 * feature for a device. The stack will accept receive
84 * checksum indication in packets received on a device
85 * regardless of whether NETIF_F_RXCSUM is set.
86 *
87 * B. Checksumming of received packets by device. Indication of checksum
88 * verification is in set skb->ip_summed. Possible values are:
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010089 *
90 * CHECKSUM_NONE:
91 *
Tom Herbert7a6ae712015-12-14 11:19:47 -080092 * Device did not checksum this packet e.g. due to lack of capabilities.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +010093 * The packet contains full (though not verified) checksum in packet but
94 * not in skb->csum. Thus, skb->csum is undefined in this case.
95 *
96 * CHECKSUM_UNNECESSARY:
97 *
98 * The hardware you're dealing with doesn't calculate the full checksum
99 * (as in CHECKSUM_COMPLETE), but it does parse headers and verify checksums
Tom Herbert77cffe22014-08-27 21:26:46 -0700100 * for specific protocols. For such packets it will set CHECKSUM_UNNECESSARY
101 * if their checksums are okay. skb->csum is still undefined in this case
Tom Herbert7a6ae712015-12-14 11:19:47 -0800102 * though. A driver or device must never modify the checksum field in the
103 * packet even if checksum is verified.
Tom Herbert77cffe22014-08-27 21:26:46 -0700104 *
105 * CHECKSUM_UNNECESSARY is applicable to following protocols:
106 * TCP: IPv6 and IPv4.
107 * UDP: IPv4 and IPv6. A device may apply CHECKSUM_UNNECESSARY to a
108 * zero UDP checksum for either IPv4 or IPv6, the networking stack
109 * may perform further validation in this case.
110 * GRE: only if the checksum is present in the header.
111 * SCTP: indicates the CRC in SCTP header has been validated.
Davide Carattib4759dc2017-05-18 15:44:43 +0200112 * FCOE: indicates the CRC in FC frame has been validated.
Tom Herbert77cffe22014-08-27 21:26:46 -0700113 *
114 * skb->csum_level indicates the number of consecutive checksums found in
115 * the packet minus one that have been verified as CHECKSUM_UNNECESSARY.
116 * For instance if a device receives an IPv6->UDP->GRE->IPv4->TCP packet
117 * and a device is able to verify the checksums for UDP (possibly zero),
118 * GRE (checksum flag is set), and TCP-- skb->csum_level would be set to
119 * two. If the device were only able to verify the UDP checksum and not
120 * GRE, either because it doesn't support GRE checksum of because GRE
121 * checksum is bad, skb->csum_level would be set to zero (TCP checksum is
122 * not considered in this case).
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100123 *
124 * CHECKSUM_COMPLETE:
125 *
126 * This is the most generic way. The device supplied checksum of the _whole_
127 * packet as seen by netif_rx() and fills out in skb->csum. Meaning, the
128 * hardware doesn't need to parse L3/L4 headers to implement this.
129 *
Davide Carattib4759dc2017-05-18 15:44:43 +0200130 * Notes:
131 * - Even if device supports only some protocols, but is able to produce
132 * skb->csum, it MUST use CHECKSUM_COMPLETE, not CHECKSUM_UNNECESSARY.
133 * - CHECKSUM_COMPLETE is not applicable to SCTP and FCoE protocols.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100134 *
135 * CHECKSUM_PARTIAL:
136 *
Tom Herbert6edec0e2015-02-10 16:30:28 -0800137 * A checksum is set up to be offloaded to a device as described in the
138 * output description for CHECKSUM_PARTIAL. This may occur on a packet
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100139 * received directly from another Linux OS, e.g., a virtualized Linux kernel
Tom Herbert6edec0e2015-02-10 16:30:28 -0800140 * on the same host, or it may be set in the input path in GRO or remote
141 * checksum offload. For the purposes of checksum verification, the checksum
142 * referred to by skb->csum_start + skb->csum_offset and any preceding
143 * checksums in the packet are considered verified. Any checksums in the
144 * packet that are after the checksum being offloaded are not considered to
145 * be verified.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100146 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800147 * C. Checksumming on transmit for non-GSO. The stack requests checksum offload
148 * in the skb->ip_summed for a packet. Values are:
149 *
150 * CHECKSUM_PARTIAL:
151 *
152 * The driver is required to checksum the packet as seen by hard_start_xmit()
153 * from skb->csum_start up to the end, and to record/write the checksum at
154 * offset skb->csum_start + skb->csum_offset. A driver may verify that the
155 * csum_start and csum_offset values are valid values given the length and
156 * offset of the packet, however they should not attempt to validate that the
157 * checksum refers to a legitimate transport layer checksum-- it is the
158 * purview of the stack to validate that csum_start and csum_offset are set
159 * correctly.
160 *
161 * When the stack requests checksum offload for a packet, the driver MUST
162 * ensure that the checksum is set correctly. A driver can either offload the
163 * checksum calculation to the device, or call skb_checksum_help (in the case
164 * that the device does not support offload for a particular checksum).
165 *
166 * NETIF_F_IP_CSUM and NETIF_F_IPV6_CSUM are being deprecated in favor of
167 * NETIF_F_HW_CSUM. New devices should use NETIF_F_HW_CSUM to indicate
Davide Caratti43c26a12017-05-18 15:44:41 +0200168 * checksum offload capability.
169 * skb_csum_hwoffload_help() can be called to resolve CHECKSUM_PARTIAL based
170 * on network device checksumming capabilities: if a packet does not match
171 * them, skb_checksum_help or skb_crc32c_help (depending on the value of
172 * csum_not_inet, see item D.) is called to resolve the checksum.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100173 *
174 * CHECKSUM_NONE:
175 *
176 * The skb was already checksummed by the protocol, or a checksum is not
177 * required.
178 *
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100179 * CHECKSUM_UNNECESSARY:
180 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800181 * This has the same meaning on as CHECKSUM_NONE for checksum offload on
182 * output.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100183 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800184 * CHECKSUM_COMPLETE:
185 * Not used in checksum output. If a driver observes a packet with this value
186 * set in skbuff, if should treat as CHECKSUM_NONE being set.
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100187 *
Tom Herbert7a6ae712015-12-14 11:19:47 -0800188 * D. Non-IP checksum (CRC) offloads
189 *
190 * NETIF_F_SCTP_CRC - This feature indicates that a device is capable of
191 * offloading the SCTP CRC in a packet. To perform this offload the stack
Davide Carattidba00302017-05-18 15:44:40 +0200192 * will set set csum_start and csum_offset accordingly, set ip_summed to
193 * CHECKSUM_PARTIAL and set csum_not_inet to 1, to provide an indication in
194 * the skbuff that the CHECKSUM_PARTIAL refers to CRC32c.
195 * A driver that supports both IP checksum offload and SCTP CRC32c offload
196 * must verify which offload is configured for a packet by testing the
197 * value of skb->csum_not_inet; skb_crc32c_csum_help is provided to resolve
198 * CHECKSUM_PARTIAL on skbs where csum_not_inet is set to 1.
Tom Herbert7a6ae712015-12-14 11:19:47 -0800199 *
200 * NETIF_F_FCOE_CRC - This feature indicates that a device is capable of
201 * offloading the FCOE CRC in a packet. To perform this offload the stack
202 * will set ip_summed to CHECKSUM_PARTIAL and set csum_start and csum_offset
203 * accordingly. Note the there is no indication in the skbuff that the
204 * CHECKSUM_PARTIAL refers to an FCOE checksum, a driver that supports
205 * both IP checksum offload and FCOE CRC offload must verify which offload
206 * is configured for a packet presumably by inspecting packet headers.
207 *
208 * E. Checksumming on output with GSO.
209 *
210 * In the case of a GSO packet (skb_is_gso(skb) is true), checksum offload
211 * is implied by the SKB_GSO_* flags in gso_type. Most obviously, if the
212 * gso_type is SKB_GSO_TCPV4 or SKB_GSO_TCPV6, TCP checksum offload as
213 * part of the GSO operation is implied. If a checksum is being offloaded
214 * with GSO then ip_summed is CHECKSUM_PARTIAL, csum_start and csum_offset
215 * are set to refer to the outermost checksum being offload (two offloaded
216 * checksums are possible with UDP encapsulation).
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100217 */
218
Herbert Xu60476372007-04-09 11:59:39 -0700219/* Don't change this without changing skb_csum_unnecessary! */
Daniel Borkmann78ea85f2013-12-16 23:27:09 +0100220#define CHECKSUM_NONE 0
221#define CHECKSUM_UNNECESSARY 1
222#define CHECKSUM_COMPLETE 2
223#define CHECKSUM_PARTIAL 3
Linus Torvalds1da177e2005-04-16 15:20:36 -0700224
Tom Herbert77cffe22014-08-27 21:26:46 -0700225/* Maximum value in skb->csum_level */
226#define SKB_MAX_CSUM_LEVEL 3
227
Tobias Klauser0bec8c82014-07-22 12:06:23 +0200228#define SKB_DATA_ALIGN(X) ALIGN(X, SMP_CACHE_BYTES)
David S. Millerfc910a22007-03-25 20:27:59 -0700229#define SKB_WITH_OVERHEAD(X) \
Herbert Xudeea84b2007-10-21 16:27:46 -0700230 ((X) - SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
David S. Millerfc910a22007-03-25 20:27:59 -0700231#define SKB_MAX_ORDER(X, ORDER) \
232 SKB_WITH_OVERHEAD((PAGE_SIZE << (ORDER)) - (X))
Linus Torvalds1da177e2005-04-16 15:20:36 -0700233#define SKB_MAX_HEAD(X) (SKB_MAX_ORDER((X), 0))
234#define SKB_MAX_ALLOC (SKB_MAX_ORDER(0, 2))
235
Eric Dumazet87fb4b72011-10-13 07:28:54 +0000236/* return minimum truesize of one skb containing X bytes of data */
237#define SKB_TRUESIZE(X) ((X) + \
238 SKB_DATA_ALIGN(sizeof(struct sk_buff)) + \
239 SKB_DATA_ALIGN(sizeof(struct skb_shared_info)))
240
Linus Torvalds1da177e2005-04-16 15:20:36 -0700241struct net_device;
David Howells716ea3a2007-04-02 20:19:53 -0700242struct scatterlist;
Jens Axboe9c55e012007-11-06 23:30:13 -0800243struct pipe_inode_info;
Herbert Xua8f820aa2014-11-07 21:22:22 +0800244struct iov_iter;
Alexander Duyckfd11a832014-12-09 19:40:49 -0800245struct napi_struct;
Petar Penkovd58e4682018-09-14 07:46:18 -0700246struct bpf_prog;
247union bpf_attr;
Florian Westphaldf5042f2018-12-18 17:15:16 +0100248struct skb_ext;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700249
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
Eric Dumazet56b17422014-11-03 08:19:53 -0800614 * @rbnode: RB tree node, alternative to next/prev for netem/tcp
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700615 * @sk: Socket we are owned by
Linus Torvalds1da177e2005-04-16 15:20:36 -0700616 * @dev: Device we arrived on/are leaving by
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700617 * @cb: Control buffer. Free for use by every layer. Put private vars here
Eric Dumazet7fee2262010-05-11 23:19:48 +0000618 * @_skb_refdst: destination entry (with norefcount bit)
Martin Waitz67be2dd2005-05-01 08:59:26 -0700619 * @sp: the security path, used for xfrm
Linus Torvalds1da177e2005-04-16 15:20:36 -0700620 * @len: Length of actual data
621 * @data_len: Data length
622 * @mac_len: Length of link layer header
Patrick McHardy334a8132007-06-25 04:35:20 -0700623 * @hdr_len: writable header length of cloned skb
Herbert Xu663ead32007-04-09 11:59:07 -0700624 * @csum: Checksum (must include start/offset pair)
625 * @csum_start: Offset from skb->head where checksumming should start
626 * @csum_offset: Offset from csum_start where checksum should be stored
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700627 * @priority: Packet queueing priority
WANG Cong60ff7462014-05-04 16:39:18 -0700628 * @ignore_df: allow local fragmentation
Linus Torvalds1da177e2005-04-16 15:20:36 -0700629 * @cloned: Head may be cloned (check refcnt to be sure)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700630 * @ip_summed: Driver fed us an IP checksum
Linus Torvalds1da177e2005-04-16 15:20:36 -0700631 * @nohdr: Payload reference only, must not modify header
632 * @pkt_type: Packet class
Randy Dunlapc83c2482005-10-18 22:07:41 -0700633 * @fclone: skbuff clone status
Randy Dunlapc83c2482005-10-18 22:07:41 -0700634 * @ipvs_property: skbuff is owned by ipvs
Ido Schimmel875e8932018-12-04 08:15:10 +0000635 * @offload_fwd_mark: Packet was L2-forwarded in hardware
636 * @offload_l3_fwd_mark: Packet was L3-forwarded in hardware
Willem de Bruijne7246e12017-01-07 17:06:35 -0500637 * @tc_skip_classify: do not classify packet. set by IFB device
Willem de Bruijn8dc07fd2017-01-07 17:06:37 -0500638 * @tc_at_ingress: used within tc_classify to distinguish in/egress
Willem de Bruijnbc31c902017-01-07 17:06:38 -0500639 * @tc_redirected: packet was redirected by a tc action
640 * @tc_from_ingress: if tc_redirected, tc_at_ingress at time of redirect
Randy Dunlap31729362008-02-18 20:52:13 -0800641 * @peeked: this packet has been seen already, so stats have been
642 * done for it, don't do them again
Jozsef Kadlecsikba9dda32007-07-07 22:21:23 -0700643 * @nf_trace: netfilter packet trace flag
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700644 * @protocol: Packet protocol from driver
645 * @destructor: Destruct function
Eric Dumazete2080072017-10-04 12:59:58 -0700646 * @tcp_tsorted_anchor: list structure for TCP (tp->tsorted_sent_queue)
Florian Westphala9e419d2017-01-23 18:21:59 +0100647 * @_nfct: Associated connection, if any (with nfctinfo bits)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700648 * @nf_bridge: Saved data about a bridged frame - see br_netfilter.c
Eric Dumazet8964be42009-11-20 15:35:04 -0800649 * @skb_iif: ifindex of device we arrived on
Linus Torvalds1da177e2005-04-16 15:20:36 -0700650 * @tc_index: Traffic control index
Tom Herbert61b905d2014-03-24 15:34:47 -0700651 * @hash: the packet hash
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700652 * @queue_mapping: Queue mapping for multiqueue devices
Stefano Brivio8b700862018-07-11 14:39:42 +0200653 * @pfmemalloc: skbuff was allocated from PFMEMALLOC reserves
Florian Westphaldf5042f2018-12-18 17:15:16 +0100654 * @active_extensions: active extensions (skb_ext_id types)
Randy Dunlap553a5672008-04-20 10:51:01 -0700655 * @ndisc_nodetype: router type (from link layer)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700656 * @ooo_okay: allow the mapping of a socket to a queue to be changed
Tom Herbert61b905d2014-03-24 15:34:47 -0700657 * @l4_hash: indicate hash is a canonical 4-tuple hash over transport
Changli Gao4ca24622011-08-19 07:26:44 -0700658 * ports.
Tom Herberta3b18dd2014-07-01 21:33:17 -0700659 * @sw_hash: indicates hash was computed in software stack
Johannes Berg6e3e9392011-11-09 10:15:42 +0100660 * @wifi_acked_valid: wifi_acked was set
661 * @wifi_acked: whether frame was acked on wifi or not
Ben Greear3bdc0eb2012-02-11 15:39:30 +0000662 * @no_fcs: Request NIC to treat last 4 bytes as Ethernet FCS
Davide Carattidba00302017-05-18 15:44:40 +0200663 * @csum_not_inet: use CRC32c to resolve CHECKSUM_PARTIAL
Julian Anastasov4ff06202017-02-06 23:14:12 +0200664 * @dst_pending_confirm: need to confirm neighbour
Stefano Brivioa48d1892018-07-17 11:52:57 +0200665 * @decrypted: Decrypted SKB
Pedro Tammela161e6132019-03-05 11:35:54 -0300666 * @napi_id: id of the NAPI struct this skb came from
James Morris984bc162006-06-09 00:29:17 -0700667 * @secmark: security marking
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700668 * @mark: Generic packet mark
Patrick McHardy86a9bad2013-04-19 02:04:30 +0000669 * @vlan_proto: vlan encapsulation protocol
Patrick McHardy6aa895b2008-07-14 22:49:06 -0700670 * @vlan_tci: vlan tag control information
Simon Horman0d89d202013-05-23 21:02:52 +0000671 * @inner_protocol: Protocol (encapsulation)
Joseph Gasparakis6a674e92012-12-07 14:14:14 +0000672 * @inner_transport_header: Inner transport layer header (encapsulation)
673 * @inner_network_header: Network layer header (encapsulation)
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +0000674 * @inner_mac_header: Link layer header (encapsulation)
Daniel Balutad84e0bd2011-07-10 07:04:04 -0700675 * @transport_header: Transport layer header
676 * @network_header: Network layer header
677 * @mac_header: Link layer header
678 * @tail: Tail pointer
679 * @end: End pointer
680 * @head: Head of buffer
681 * @data: Data head pointer
682 * @truesize: Buffer size
683 * @users: User count - see {datagram,tcp}.c
Florian Westphaldf5042f2018-12-18 17:15:16 +0100684 * @extensions: allocated extensions, valid if active_extensions is nonzero
Linus Torvalds1da177e2005-04-16 15:20:36 -0700685 */
686
687struct sk_buff {
Eric Dumazet363ec39232014-02-26 14:02:11 -0800688 union {
Eric Dumazet56b17422014-11-03 08:19:53 -0800689 struct {
690 /* These two members must be first. */
691 struct sk_buff *next;
692 struct sk_buff *prev;
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000693
Eric Dumazet56b17422014-11-03 08:19:53 -0800694 union {
Eric Dumazetbffa72c2017-09-19 05:14:24 -0700695 struct net_device *dev;
696 /* Some protocols might use this space to store information,
697 * while device pointer would be NULL.
698 * UDP receive path is one user.
699 */
700 unsigned long dev_scratch;
Eric Dumazet56b17422014-11-03 08:19:53 -0800701 };
702 };
Peter Oskolkovfa0f5272018-08-02 23:34:39 +0000703 struct rb_node rbnode; /* used in netem, ip4 defrag, and tcp stack */
David Millerd4546c22018-06-24 14:13:49 +0900704 struct list_head list;
Eric Dumazet56b17422014-11-03 08:19:53 -0800705 };
Peter Oskolkovfa0f5272018-08-02 23:34:39 +0000706
707 union {
708 struct sock *sk;
709 int ip_defrag_offset;
710 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700711
Eric Dumazetc84d9492016-12-08 11:41:55 -0800712 union {
Eric Dumazetbffa72c2017-09-19 05:14:24 -0700713 ktime_t tstamp;
Eric Dumazetd3edd062018-09-21 08:51:50 -0700714 u64 skb_mstamp_ns; /* earliest departure time */
Eric Dumazetc84d9492016-12-08 11:41:55 -0800715 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700716 /*
717 * This is the control buffer. It is free to use for every
718 * layer. Please put your private variables there. If you
719 * want to keep them across layers you have to do a skb_clone()
720 * first. This is owned by whoever has the skb queued ATM.
721 */
Felix Fietkauda3f5cf2010-02-23 11:45:51 +0000722 char cb[48] __aligned(8);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700723
Eric Dumazete2080072017-10-04 12:59:58 -0700724 union {
725 struct {
726 unsigned long _skb_refdst;
727 void (*destructor)(struct sk_buff *skb);
728 };
729 struct list_head tcp_tsorted_anchor;
730 };
731
Eric Dumazetb1937222014-09-28 22:18:47 -0700732#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +0100733 unsigned long _nfct;
Eric Dumazetb1937222014-09-28 22:18:47 -0700734#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700735 unsigned int len,
Patrick McHardy334a8132007-06-25 04:35:20 -0700736 data_len;
737 __u16 mac_len,
738 hdr_len;
Eric Dumazetb1937222014-09-28 22:18:47 -0700739
740 /* Following fields are _not_ copied in __copy_skb_header()
741 * Note that queue_mapping is here mostly to fill a hole.
742 */
Eric Dumazetb1937222014-09-28 22:18:47 -0700743 __u16 queue_mapping;
Daniel Borkmann36bbef52016-09-20 00:26:13 +0200744
745/* if you move cloned around you also must adapt those constants */
746#ifdef __BIG_ENDIAN_BITFIELD
747#define CLONED_MASK (1 << 7)
748#else
749#define CLONED_MASK 1
750#endif
751#define CLONED_OFFSET() offsetof(struct sk_buff, __cloned_offset)
752
753 __u8 __cloned_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700754 __u8 cloned:1,
Harald Welte6869c4d2005-08-09 19:24:19 -0700755 nohdr:1,
Patrick McHardyb84f4cc2005-11-20 21:19:21 -0800756 fclone:2,
Herbert Xua59322b2007-12-05 01:53:40 -0800757 peeked:1,
Eric Dumazetb1937222014-09-28 22:18:47 -0700758 head_frag:1,
Stefano Brivio8b700862018-07-11 14:39:42 +0200759 pfmemalloc:1;
Florian Westphaldf5042f2018-12-18 17:15:16 +0100760#ifdef CONFIG_SKB_EXTENSIONS
761 __u8 active_extensions;
762#endif
Eric Dumazetb1937222014-09-28 22:18:47 -0700763 /* fields enclosed in headers_start/headers_end are copied
764 * using a single memcpy() in __copy_skb_header()
765 */
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700766 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700767 __u32 headers_start[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700768 /* public: */
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200769
770/* if you move pkt_type around you also must adapt those constants */
771#ifdef __BIG_ENDIAN_BITFIELD
772#define PKT_TYPE_MAX (7 << 5)
773#else
774#define PKT_TYPE_MAX 7
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +0100775#endif
Hannes Frederic Sowa233577a2014-09-12 14:04:43 +0200776#define PKT_TYPE_OFFSET() offsetof(struct sk_buff, __pkt_type_offset)
777
778 __u8 __pkt_type_offset[0];
Eric Dumazetb1937222014-09-28 22:18:47 -0700779 __u8 pkt_type:3;
Eric Dumazetb1937222014-09-28 22:18:47 -0700780 __u8 ignore_df:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700781 __u8 nf_trace:1;
782 __u8 ip_summed:2;
783 __u8 ooo_okay:1;
Stefano Brivio8b700862018-07-11 14:39:42 +0200784
Eric Dumazetb1937222014-09-28 22:18:47 -0700785 __u8 l4_hash:1;
786 __u8 sw_hash:1;
787 __u8 wifi_acked_valid:1;
788 __u8 wifi_acked:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700789 __u8 no_fcs:1;
790 /* Indicates the inner headers are valid in the skbuff. */
791 __u8 encapsulation:1;
792 __u8 encap_hdr_csum:1;
793 __u8 csum_valid:1;
Stefano Brivio8b700862018-07-11 14:39:42 +0200794
Michał Mirosław0c4b2d32018-11-10 19:58:36 +0100795#ifdef __BIG_ENDIAN_BITFIELD
796#define PKT_VLAN_PRESENT_BIT 7
797#else
798#define PKT_VLAN_PRESENT_BIT 0
799#endif
800#define PKT_VLAN_PRESENT_OFFSET() offsetof(struct sk_buff, __pkt_vlan_present_offset)
801 __u8 __pkt_vlan_present_offset[0];
802 __u8 vlan_present:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700803 __u8 csum_complete_sw:1;
804 __u8 csum_level:2;
Davide Carattidba00302017-05-18 15:44:40 +0200805 __u8 csum_not_inet:1;
Julian Anastasov4ff06202017-02-06 23:14:12 +0200806 __u8 dst_pending_confirm:1;
Eric Dumazetb1937222014-09-28 22:18:47 -0700807#ifdef CONFIG_IPV6_NDISC_NODETYPE
808 __u8 ndisc_nodetype:2;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700809#endif
Stefano Brivio8b700862018-07-11 14:39:42 +0200810
Michał Mirosław0c4b2d32018-11-10 19:58:36 +0100811 __u8 ipvs_property:1;
Tom Herbert8bce6d72014-09-29 20:22:29 -0700812 __u8 inner_protocol_type:1;
Tom Herberte585f232014-11-04 09:06:54 -0800813 __u8 remcsum_offload:1;
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200814#ifdef CONFIG_NET_SWITCHDEV
815 __u8 offload_fwd_mark:1;
Ido Schimmel875e8932018-12-04 08:15:10 +0000816 __u8 offload_l3_fwd_mark:1;
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200817#endif
Willem de Bruijne7246e12017-01-07 17:06:35 -0500818#ifdef CONFIG_NET_CLS_ACT
819 __u8 tc_skip_classify:1;
Willem de Bruijn8dc07fd2017-01-07 17:06:37 -0500820 __u8 tc_at_ingress:1;
Willem de Bruijnbc31c902017-01-07 17:06:38 -0500821 __u8 tc_redirected:1;
822 __u8 tc_from_ingress:1;
Willem de Bruijne7246e12017-01-07 17:06:35 -0500823#endif
Stefano Brivioa48d1892018-07-17 11:52:57 +0200824#ifdef CONFIG_TLS_DEVICE
825 __u8 decrypted:1;
826#endif
Alexander Duyck4031ae62012-01-27 06:22:53 +0000827
Linus Torvalds1da177e2005-04-16 15:20:36 -0700828#ifdef CONFIG_NET_SCHED
Patrick McHardyb6b99eb2005-08-09 19:33:51 -0700829 __u16 tc_index; /* traffic control index */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700830#endif
Vegard Nossumfe55f6d2008-08-30 12:16:35 +0200831
Eric Dumazetb1937222014-09-28 22:18:47 -0700832 union {
833 __wsum csum;
834 struct {
835 __u16 csum_start;
836 __u16 csum_offset;
837 };
838 };
839 __u32 priority;
840 int skb_iif;
841 __u32 hash;
842 __be16 vlan_proto;
843 __u16 vlan_tci;
Eric Dumazet2bd82482015-02-03 23:48:24 -0800844#if defined(CONFIG_NET_RX_BUSY_POLL) || defined(CONFIG_XPS)
845 union {
846 unsigned int napi_id;
847 unsigned int sender_cpu;
848 };
Chris Leech97fc2f02006-05-23 17:55:33 -0700849#endif
James Morris984bc162006-06-09 00:29:17 -0700850#ifdef CONFIG_NETWORK_SECMARK
Ido Schimmel6bc506b2016-08-25 18:42:37 +0200851 __u32 secmark;
James Morris984bc162006-06-09 00:29:17 -0700852#endif
Scott Feldman0c4f6912015-07-18 18:24:48 -0700853
Neil Horman3b885782009-10-12 13:26:31 -0700854 union {
855 __u32 mark;
Eric Dumazet16fad692013-03-14 05:40:32 +0000856 __u32 reserved_tailroom;
Neil Horman3b885782009-10-12 13:26:31 -0700857 };
Linus Torvalds1da177e2005-04-16 15:20:36 -0700858
Tom Herbert8bce6d72014-09-29 20:22:29 -0700859 union {
860 __be16 inner_protocol;
861 __u8 inner_ipproto;
862 };
863
Simon Horman1a37e412013-05-23 21:02:51 +0000864 __u16 inner_transport_header;
865 __u16 inner_network_header;
866 __u16 inner_mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700867
868 __be16 protocol;
Simon Horman1a37e412013-05-23 21:02:51 +0000869 __u16 transport_header;
870 __u16 network_header;
871 __u16 mac_header;
Eric Dumazetb1937222014-09-28 22:18:47 -0700872
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700873 /* private: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700874 __u32 headers_end[0];
Randy Dunlapebcf34f2014-10-26 19:14:06 -0700875 /* public: */
Eric Dumazetb1937222014-09-28 22:18:47 -0700876
Linus Torvalds1da177e2005-04-16 15:20:36 -0700877 /* These elements must be at the end, see alloc_skb() for details. */
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700878 sk_buff_data_t tail;
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700879 sk_buff_data_t end;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700880 unsigned char *head,
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -0700881 *data;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -0700882 unsigned int truesize;
Reshetova, Elena63354792017-06-30 13:07:58 +0300883 refcount_t users;
Florian Westphaldf5042f2018-12-18 17:15:16 +0100884
885#ifdef CONFIG_SKB_EXTENSIONS
886 /* only useable after checking ->active_extensions != 0 */
887 struct skb_ext *extensions;
888#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -0700889};
890
891#ifdef __KERNEL__
892/*
893 * Handling routines are only of interest to the kernel
894 */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700895
Mel Gormanc93bdd02012-07-31 16:44:19 -0700896#define SKB_ALLOC_FCLONE 0x01
897#define SKB_ALLOC_RX 0x02
Alexander Duyckfd11a832014-12-09 19:40:49 -0800898#define SKB_ALLOC_NAPI 0x04
Mel Gormanc93bdd02012-07-31 16:44:19 -0700899
Pedro Tammela161e6132019-03-05 11:35:54 -0300900/**
901 * skb_pfmemalloc - Test if the skb was allocated from PFMEMALLOC reserves
902 * @skb: buffer
903 */
Mel Gormanc93bdd02012-07-31 16:44:19 -0700904static inline bool skb_pfmemalloc(const struct sk_buff *skb)
905{
906 return unlikely(skb->pfmemalloc);
907}
908
Eric Dumazet7fee2262010-05-11 23:19:48 +0000909/*
910 * skb might have a dst pointer attached, refcounted or not.
911 * _skb_refdst low order bit is set if refcount was _not_ taken
912 */
913#define SKB_DST_NOREF 1UL
914#define SKB_DST_PTRMASK ~(SKB_DST_NOREF)
915
916/**
917 * skb_dst - returns skb dst_entry
918 * @skb: buffer
919 *
920 * Returns skb dst_entry, regardless of reference taken or not.
921 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000922static inline struct dst_entry *skb_dst(const struct sk_buff *skb)
923{
Pedro Tammela161e6132019-03-05 11:35:54 -0300924 /* If refdst was not refcounted, check we still are in a
Eric Dumazet7fee2262010-05-11 23:19:48 +0000925 * rcu_read_lock section
926 */
927 WARN_ON((skb->_skb_refdst & SKB_DST_NOREF) &&
928 !rcu_read_lock_held() &&
929 !rcu_read_lock_bh_held());
930 return (struct dst_entry *)(skb->_skb_refdst & SKB_DST_PTRMASK);
Eric Dumazetadf30902009-06-02 05:19:30 +0000931}
932
Eric Dumazet7fee2262010-05-11 23:19:48 +0000933/**
934 * skb_dst_set - sets skb dst
935 * @skb: buffer
936 * @dst: dst entry
937 *
938 * Sets skb dst, assuming a reference was taken on dst and should
939 * be released by skb_dst_drop()
940 */
Eric Dumazetadf30902009-06-02 05:19:30 +0000941static inline void skb_dst_set(struct sk_buff *skb, struct dst_entry *dst)
942{
Eric Dumazet7fee2262010-05-11 23:19:48 +0000943 skb->_skb_refdst = (unsigned long)dst;
944}
945
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200946/**
947 * skb_dst_set_noref - sets skb dst, hopefully, without taking reference
948 * @skb: buffer
949 * @dst: dst entry
950 *
951 * Sets skb dst, assuming a reference was not taken on dst.
952 * If dst entry is cached, we do not take reference and dst_release
953 * will be avoided by refdst_drop. If dst entry is not cached, we take
954 * reference, so that last dst_release can destroy the dst immediately.
955 */
956static inline void skb_dst_set_noref(struct sk_buff *skb, struct dst_entry *dst)
957{
Hannes Frederic Sowadbfc4fb2014-12-06 19:19:42 +0100958 WARN_ON(!rcu_read_lock_held() && !rcu_read_lock_bh_held());
959 skb->_skb_refdst = (unsigned long)dst | SKB_DST_NOREF;
Julian Anastasov932bc4d2013-03-21 11:57:58 +0200960}
Eric Dumazet7fee2262010-05-11 23:19:48 +0000961
962/**
Lucas De Marchi25985ed2011-03-30 22:57:33 -0300963 * skb_dst_is_noref - Test if skb dst isn't refcounted
Eric Dumazet7fee2262010-05-11 23:19:48 +0000964 * @skb: buffer
965 */
966static inline bool skb_dst_is_noref(const struct sk_buff *skb)
967{
968 return (skb->_skb_refdst & SKB_DST_NOREF) && skb_dst(skb);
Eric Dumazetadf30902009-06-02 05:19:30 +0000969}
970
Pedro Tammela161e6132019-03-05 11:35:54 -0300971/**
972 * skb_rtable - Returns the skb &rtable
973 * @skb: buffer
974 */
Eric Dumazet511c3f92009-06-02 05:14:27 +0000975static inline struct rtable *skb_rtable(const struct sk_buff *skb)
976{
Eric Dumazetadf30902009-06-02 05:19:30 +0000977 return (struct rtable *)skb_dst(skb);
Eric Dumazet511c3f92009-06-02 05:14:27 +0000978}
979
Jamal Hadi Salim8b10cab2016-07-02 06:43:14 -0400980/* For mangling skb->pkt_type from user space side from applications
981 * such as nft, tc, etc, we only allow a conservative subset of
982 * possible pkt_types to be set.
983*/
984static inline bool skb_pkt_type_ok(u32 ptype)
985{
986 return ptype <= PACKET_OTHERHOST;
987}
988
Pedro Tammela161e6132019-03-05 11:35:54 -0300989/**
990 * skb_napi_id - Returns the skb's NAPI id
991 * @skb: buffer
992 */
Miroslav Lichvar90b602f2017-05-19 17:52:37 +0200993static inline unsigned int skb_napi_id(const struct sk_buff *skb)
994{
995#ifdef CONFIG_NET_RX_BUSY_POLL
996 return skb->napi_id;
997#else
998 return 0;
999#endif
1000}
1001
Pedro Tammela161e6132019-03-05 11:35:54 -03001002/**
1003 * skb_unref - decrement the skb's reference count
1004 * @skb: buffer
1005 *
1006 * Returns true if we can free the skb.
1007 */
Paolo Abeni3889a8032017-06-12 11:23:41 +02001008static inline bool skb_unref(struct sk_buff *skb)
1009{
1010 if (unlikely(!skb))
1011 return false;
Reshetova, Elena63354792017-06-30 13:07:58 +03001012 if (likely(refcount_read(&skb->users) == 1))
Paolo Abeni3889a8032017-06-12 11:23:41 +02001013 smp_rmb();
Reshetova, Elena63354792017-06-30 13:07:58 +03001014 else if (likely(!refcount_dec_and_test(&skb->users)))
Paolo Abeni3889a8032017-06-12 11:23:41 +02001015 return false;
1016
1017 return true;
1018}
1019
Paolo Abeni0a463c72017-06-12 11:23:42 +02001020void skb_release_head_state(struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -07001021void kfree_skb(struct sk_buff *skb);
1022void kfree_skb_list(struct sk_buff *segs);
Willem de Bruijn64131392019-07-07 05:51:55 -04001023void skb_dump(const char *level, const struct sk_buff *skb, bool full_pkt);
Joe Perches7965bd42013-09-26 14:48:15 -07001024void skb_tx_error(struct sk_buff *skb);
1025void consume_skb(struct sk_buff *skb);
Paolo Abenica2c1412017-09-06 14:44:36 +02001026void __consume_stateless_skb(struct sk_buff *skb);
Joe Perches7965bd42013-09-26 14:48:15 -07001027void __kfree_skb(struct sk_buff *skb);
Eric Dumazetd7e88832012-04-30 08:10:34 +00001028extern struct kmem_cache *skbuff_head_cache;
Eric Dumazetbad43ca2012-05-19 03:02:02 +00001029
Joe Perches7965bd42013-09-26 14:48:15 -07001030void kfree_skb_partial(struct sk_buff *skb, bool head_stolen);
1031bool skb_try_coalesce(struct sk_buff *to, struct sk_buff *from,
1032 bool *fragstolen, int *delta_truesize);
Eric Dumazetbad43ca2012-05-19 03:02:02 +00001033
Joe Perches7965bd42013-09-26 14:48:15 -07001034struct sk_buff *__alloc_skb(unsigned int size, gfp_t priority, int flags,
1035 int node);
Eric Dumazet2ea2f622015-04-24 16:05:01 -07001036struct sk_buff *__build_skb(void *data, unsigned int frag_size);
Joe Perches7965bd42013-09-26 14:48:15 -07001037struct sk_buff *build_skb(void *data, unsigned int frag_size);
Jesper Dangaard Brouerba0509b2019-04-12 17:07:37 +02001038struct sk_buff *build_skb_around(struct sk_buff *skb,
1039 void *data, unsigned int frag_size);
Pedro Tammela161e6132019-03-05 11:35:54 -03001040
1041/**
1042 * alloc_skb - allocate a network buffer
1043 * @size: size to allocate
1044 * @priority: allocation mask
1045 *
1046 * This function is a convenient wrapper around __alloc_skb().
1047 */
David S. Millerd179cd12005-08-17 14:57:30 -07001048static inline struct sk_buff *alloc_skb(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +01001049 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -07001050{
Eric Dumazet564824b2010-10-11 19:05:25 +00001051 return __alloc_skb(size, priority, 0, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -07001052}
1053
Eric Dumazet2e4e4412014-09-17 04:49:49 -07001054struct sk_buff *alloc_skb_with_frags(unsigned long header_len,
1055 unsigned long data_len,
1056 int max_page_order,
1057 int *errcode,
1058 gfp_t gfp_mask);
Jakub Kicinskida29e4b2019-06-03 15:16:58 -07001059struct sk_buff *alloc_skb_for_msg(struct sk_buff *first);
Eric Dumazet2e4e4412014-09-17 04:49:49 -07001060
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001061/* Layout of fast clones : [skb1][skb2][fclone_ref] */
1062struct sk_buff_fclones {
1063 struct sk_buff skb1;
1064
1065 struct sk_buff skb2;
1066
Reshetova, Elena26385952017-06-30 13:07:59 +03001067 refcount_t fclone_ref;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001068};
1069
1070/**
1071 * skb_fclone_busy - check if fclone is busy
Stephen Hemminger293de7d2016-10-23 09:28:29 -07001072 * @sk: socket
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001073 * @skb: buffer
1074 *
Masanari Iidabda13fe2015-12-13 16:53:02 +09001075 * Returns true if skb is a fast clone, and its clone is not freed.
Eric Dumazet39bb5e62014-10-30 10:32:34 -07001076 * Some drivers call skb_orphan() in their ndo_start_xmit(),
1077 * so we also check that this didnt happen.
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001078 */
Eric Dumazet39bb5e62014-10-30 10:32:34 -07001079static inline bool skb_fclone_busy(const struct sock *sk,
1080 const struct sk_buff *skb)
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001081{
1082 const struct sk_buff_fclones *fclones;
1083
1084 fclones = container_of(skb, struct sk_buff_fclones, skb1);
1085
1086 return skb->fclone == SKB_FCLONE_ORIG &&
Reshetova, Elena26385952017-06-30 13:07:59 +03001087 refcount_read(&fclones->fclone_ref) > 1 &&
Eric Dumazet39bb5e62014-10-30 10:32:34 -07001088 fclones->skb2.sk == sk;
Eric Dumazetd0bf4a92014-09-29 13:29:15 -07001089}
1090
Pedro Tammela161e6132019-03-05 11:35:54 -03001091/**
1092 * alloc_skb_fclone - allocate a network buffer from fclone cache
1093 * @size: size to allocate
1094 * @priority: allocation mask
1095 *
1096 * This function is a convenient wrapper around __alloc_skb().
1097 */
David S. Millerd179cd12005-08-17 14:57:30 -07001098static inline struct sk_buff *alloc_skb_fclone(unsigned int size,
Al Virodd0fc662005-10-07 07:46:04 +01001099 gfp_t priority)
David S. Millerd179cd12005-08-17 14:57:30 -07001100{
Mel Gormanc93bdd02012-07-31 16:44:19 -07001101 return __alloc_skb(size, priority, SKB_ALLOC_FCLONE, NUMA_NO_NODE);
David S. Millerd179cd12005-08-17 14:57:30 -07001102}
1103
Joe Perches7965bd42013-09-26 14:48:15 -07001104struct sk_buff *skb_morph(struct sk_buff *dst, struct sk_buff *src);
Toshiaki Makitab0768a82018-08-03 16:58:09 +09001105void skb_headers_offset_update(struct sk_buff *skb, int off);
Joe Perches7965bd42013-09-26 14:48:15 -07001106int skb_copy_ubufs(struct sk_buff *skb, gfp_t gfp_mask);
1107struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t priority);
Ilya Lesokhin08303c12018-04-30 10:16:11 +03001108void skb_copy_header(struct sk_buff *new, const struct sk_buff *old);
Joe Perches7965bd42013-09-26 14:48:15 -07001109struct sk_buff *skb_copy(const struct sk_buff *skb, gfp_t priority);
Octavian Purdilabad93e92014-06-12 01:36:26 +03001110struct sk_buff *__pskb_copy_fclone(struct sk_buff *skb, int headroom,
1111 gfp_t gfp_mask, bool fclone);
1112static inline struct sk_buff *__pskb_copy(struct sk_buff *skb, int headroom,
1113 gfp_t gfp_mask)
1114{
1115 return __pskb_copy_fclone(skb, headroom, gfp_mask, false);
1116}
Eric Dumazet117632e2011-12-03 21:39:53 +00001117
Joe Perches7965bd42013-09-26 14:48:15 -07001118int pskb_expand_head(struct sk_buff *skb, int nhead, int ntail, gfp_t gfp_mask);
1119struct sk_buff *skb_realloc_headroom(struct sk_buff *skb,
1120 unsigned int headroom);
1121struct sk_buff *skb_copy_expand(const struct sk_buff *skb, int newheadroom,
1122 int newtailroom, gfp_t priority);
Jason A. Donenfeld48a1df62017-06-04 04:16:22 +02001123int __must_check skb_to_sgvec_nomark(struct sk_buff *skb, struct scatterlist *sg,
1124 int offset, int len);
1125int __must_check skb_to_sgvec(struct sk_buff *skb, struct scatterlist *sg,
1126 int offset, int len);
Joe Perches7965bd42013-09-26 14:48:15 -07001127int skb_cow_data(struct sk_buff *skb, int tailbits, struct sk_buff **trailer);
Florian Fainellicd0a1372017-08-22 15:12:14 -07001128int __skb_pad(struct sk_buff *skb, int pad, bool free_on_error);
1129
1130/**
1131 * skb_pad - zero pad the tail of an skb
1132 * @skb: buffer to pad
1133 * @pad: space to pad
1134 *
1135 * Ensure that a buffer is followed by a padding area that is zero
1136 * filled. Used by network drivers which may DMA or transfer data
1137 * beyond the buffer end onto the wire.
1138 *
1139 * May return error in out of memory cases. The skb is freed on error.
1140 */
1141static inline int skb_pad(struct sk_buff *skb, int pad)
1142{
1143 return __skb_pad(skb, pad, true);
1144}
Neil Hormanead2ceb2009-03-11 09:49:55 +00001145#define dev_kfree_skb(a) consume_skb(a)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001146
Hannes Frederic Sowabe12a1f2015-05-21 16:59:58 +02001147int skb_append_pagefrags(struct sk_buff *skb, struct page *page,
1148 int offset, size_t size);
1149
Eric Dumazetd94d9fe2009-11-04 09:50:58 -08001150struct skb_seq_state {
Thomas Graf677e90e2005-06-23 20:59:51 -07001151 __u32 lower_offset;
1152 __u32 upper_offset;
1153 __u32 frag_idx;
1154 __u32 stepped_offset;
1155 struct sk_buff *root_skb;
1156 struct sk_buff *cur_skb;
1157 __u8 *frag_data;
1158};
1159
Joe Perches7965bd42013-09-26 14:48:15 -07001160void skb_prepare_seq_read(struct sk_buff *skb, unsigned int from,
1161 unsigned int to, struct skb_seq_state *st);
1162unsigned int skb_seq_read(unsigned int consumed, const u8 **data,
1163 struct skb_seq_state *st);
1164void skb_abort_seq_read(struct skb_seq_state *st);
Thomas Graf677e90e2005-06-23 20:59:51 -07001165
Joe Perches7965bd42013-09-26 14:48:15 -07001166unsigned int skb_find_text(struct sk_buff *skb, unsigned int from,
Bojan Prtvar059a2442015-02-22 11:46:35 +01001167 unsigned int to, struct ts_config *config);
Thomas Graf3fc7e8a2005-06-23 21:00:17 -07001168
Tom Herbert09323cc2013-12-15 22:16:19 -08001169/*
1170 * Packet hash types specify the type of hash in skb_set_hash.
1171 *
1172 * Hash types refer to the protocol layer addresses which are used to
1173 * construct a packet's hash. The hashes are used to differentiate or identify
1174 * flows of the protocol layer for the hash type. Hash types are either
1175 * layer-2 (L2), layer-3 (L3), or layer-4 (L4).
1176 *
1177 * Properties of hashes:
1178 *
1179 * 1) Two packets in different flows have different hash values
1180 * 2) Two packets in the same flow should have the same hash value
1181 *
1182 * A hash at a higher layer is considered to be more specific. A driver should
1183 * set the most specific hash possible.
1184 *
1185 * A driver cannot indicate a more specific hash than the layer at which a hash
1186 * was computed. For instance an L3 hash cannot be set as an L4 hash.
1187 *
1188 * A driver may indicate a hash level which is less specific than the
1189 * actual layer the hash was computed on. For instance, a hash computed
1190 * at L4 may be considered an L3 hash. This should only be done if the
1191 * driver can't unambiguously determine that the HW computed the hash at
1192 * the higher layer. Note that the "should" in the second property above
1193 * permits this.
1194 */
1195enum pkt_hash_types {
1196 PKT_HASH_TYPE_NONE, /* Undefined type */
1197 PKT_HASH_TYPE_L2, /* Input: src_MAC, dest_MAC */
1198 PKT_HASH_TYPE_L3, /* Input: src_IP, dst_IP */
1199 PKT_HASH_TYPE_L4, /* Input: src_IP, dst_IP, src_port, dst_port */
1200};
1201
Tom Herbertbcc83832015-09-01 09:24:24 -07001202static inline void skb_clear_hash(struct sk_buff *skb)
1203{
1204 skb->hash = 0;
1205 skb->sw_hash = 0;
1206 skb->l4_hash = 0;
1207}
1208
1209static inline void skb_clear_hash_if_not_l4(struct sk_buff *skb)
1210{
1211 if (!skb->l4_hash)
1212 skb_clear_hash(skb);
1213}
1214
1215static inline void
1216__skb_set_hash(struct sk_buff *skb, __u32 hash, bool is_sw, bool is_l4)
1217{
1218 skb->l4_hash = is_l4;
1219 skb->sw_hash = is_sw;
1220 skb->hash = hash;
1221}
1222
Tom Herbert09323cc2013-12-15 22:16:19 -08001223static inline void
1224skb_set_hash(struct sk_buff *skb, __u32 hash, enum pkt_hash_types type)
1225{
Tom Herbertbcc83832015-09-01 09:24:24 -07001226 /* Used by drivers to set hash from HW */
1227 __skb_set_hash(skb, hash, false, type == PKT_HASH_TYPE_L4);
1228}
1229
1230static inline void
1231__skb_set_sw_hash(struct sk_buff *skb, __u32 hash, bool is_l4)
1232{
1233 __skb_set_hash(skb, hash, true, is_l4);
Tom Herbert09323cc2013-12-15 22:16:19 -08001234}
1235
Tom Herberte5276932015-09-01 09:24:23 -07001236void __skb_get_hash(struct sk_buff *skb);
Florian Westphalb9177832016-10-26 18:49:46 +02001237u32 __skb_get_hash_symmetric(const struct sk_buff *skb);
Tom Herberte5276932015-09-01 09:24:23 -07001238u32 skb_get_poff(const struct sk_buff *skb);
1239u32 __skb_get_poff(const struct sk_buff *skb, void *data,
Paolo Abeni72a338b2018-05-04 11:32:59 +02001240 const struct flow_keys_basic *keys, int hlen);
Tom Herberte5276932015-09-01 09:24:23 -07001241__be32 __skb_flow_get_ports(const struct sk_buff *skb, int thoff, u8 ip_proto,
1242 void *data, int hlen_proto);
1243
1244static inline __be32 skb_flow_get_ports(const struct sk_buff *skb,
1245 int thoff, u8 ip_proto)
1246{
1247 return __skb_flow_get_ports(skb, thoff, ip_proto, NULL, 0);
1248}
1249
1250void skb_flow_dissector_init(struct flow_dissector *flow_dissector,
1251 const struct flow_dissector_key *key,
1252 unsigned int key_count);
1253
Willem de Bruijn2dfd1842018-09-18 16:20:18 -04001254#ifdef CONFIG_NET
Stanislav Fomichev118c8e92019-04-25 14:37:23 -07001255int skb_flow_dissector_prog_query(const union bpf_attr *attr,
1256 union bpf_attr __user *uattr);
Petar Penkovd58e4682018-09-14 07:46:18 -07001257int skb_flow_dissector_bpf_prog_attach(const union bpf_attr *attr,
1258 struct bpf_prog *prog);
1259
1260int skb_flow_dissector_bpf_prog_detach(const union bpf_attr *attr);
Willem de Bruijn2dfd1842018-09-18 16:20:18 -04001261#else
Stanislav Fomichev118c8e92019-04-25 14:37:23 -07001262static inline int skb_flow_dissector_prog_query(const union bpf_attr *attr,
1263 union bpf_attr __user *uattr)
1264{
1265 return -EOPNOTSUPP;
1266}
1267
Willem de Bruijn2dfd1842018-09-18 16:20:18 -04001268static inline int skb_flow_dissector_bpf_prog_attach(const union bpf_attr *attr,
1269 struct bpf_prog *prog)
1270{
1271 return -EOPNOTSUPP;
1272}
1273
1274static inline int skb_flow_dissector_bpf_prog_detach(const union bpf_attr *attr)
1275{
1276 return -EOPNOTSUPP;
1277}
1278#endif
Petar Penkovd58e4682018-09-14 07:46:18 -07001279
Stanislav Fomichev089b19a2019-04-22 08:55:44 -07001280struct bpf_flow_dissector;
1281bool bpf_flow_dissect(struct bpf_prog *prog, struct bpf_flow_dissector *ctx,
Stanislav Fomichev086f9562019-07-25 15:52:25 -07001282 __be16 proto, int nhoff, int hlen, unsigned int flags);
Stanislav Fomichev089b19a2019-04-22 08:55:44 -07001283
Stanislav Fomichev3cbf4ff2019-04-22 08:55:46 -07001284bool __skb_flow_dissect(const struct net *net,
1285 const struct sk_buff *skb,
Tom Herberte5276932015-09-01 09:24:23 -07001286 struct flow_dissector *flow_dissector,
1287 void *target_container,
Tom Herbertcd79a232015-09-01 09:24:27 -07001288 void *data, __be16 proto, int nhoff, int hlen,
1289 unsigned int flags);
Tom Herberte5276932015-09-01 09:24:23 -07001290
1291static inline bool skb_flow_dissect(const struct sk_buff *skb,
1292 struct flow_dissector *flow_dissector,
Tom Herbertcd79a232015-09-01 09:24:27 -07001293 void *target_container, unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001294{
Stanislav Fomichev3cbf4ff2019-04-22 08:55:46 -07001295 return __skb_flow_dissect(NULL, skb, flow_dissector,
1296 target_container, NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001297}
1298
1299static inline bool skb_flow_dissect_flow_keys(const struct sk_buff *skb,
Tom Herbertcd79a232015-09-01 09:24:27 -07001300 struct flow_keys *flow,
1301 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001302{
1303 memset(flow, 0, sizeof(*flow));
Stanislav Fomichev3cbf4ff2019-04-22 08:55:46 -07001304 return __skb_flow_dissect(NULL, skb, &flow_keys_dissector,
1305 flow, NULL, 0, 0, 0, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001306}
1307
Paolo Abeni72a338b2018-05-04 11:32:59 +02001308static inline bool
Stanislav Fomichev3cbf4ff2019-04-22 08:55:46 -07001309skb_flow_dissect_flow_keys_basic(const struct net *net,
1310 const struct sk_buff *skb,
Paolo Abeni72a338b2018-05-04 11:32:59 +02001311 struct flow_keys_basic *flow, void *data,
1312 __be16 proto, int nhoff, int hlen,
1313 unsigned int flags)
Tom Herberte5276932015-09-01 09:24:23 -07001314{
1315 memset(flow, 0, sizeof(*flow));
Stanislav Fomichev3cbf4ff2019-04-22 08:55:46 -07001316 return __skb_flow_dissect(net, skb, &flow_keys_basic_dissector, flow,
Tom Herbertcd79a232015-09-01 09:24:27 -07001317 data, proto, nhoff, hlen, flags);
Tom Herberte5276932015-09-01 09:24:23 -07001318}
1319
Jiri Pirko82828b82019-06-19 09:41:02 +03001320void skb_flow_dissect_meta(const struct sk_buff *skb,
1321 struct flow_dissector *flow_dissector,
1322 void *target_container);
1323
Paul Blakey75a56752019-07-09 10:30:49 +03001324/* Gets a skb connection tracking info, ctinfo map should be a
1325 * a map of mapsize to translate enum ip_conntrack_info states
1326 * to user states.
1327 */
1328void
1329skb_flow_dissect_ct(const struct sk_buff *skb,
1330 struct flow_dissector *flow_dissector,
1331 void *target_container,
1332 u16 *ctinfo_map,
1333 size_t mapsize);
Simon Horman62b32372017-12-04 11:31:48 +01001334void
1335skb_flow_dissect_tunnel_info(const struct sk_buff *skb,
1336 struct flow_dissector *flow_dissector,
1337 void *target_container);
1338
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001339static inline __u32 skb_get_hash(struct sk_buff *skb)
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001340{
Tom Herberta3b18dd2014-07-01 21:33:17 -07001341 if (!skb->l4_hash && !skb->sw_hash)
Tom Herbert3958afa1b2013-12-15 22:12:06 -08001342 __skb_get_hash(skb);
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001343
Tom Herbert61b905d2014-03-24 15:34:47 -07001344 return skb->hash;
Krishna Kumarbfb564e2010-08-04 06:15:52 +00001345}
1346
David S. Miller20a17bf2015-09-01 21:19:17 -07001347static inline __u32 skb_get_hash_flowi6(struct sk_buff *skb, const struct flowi6 *fl6)
Tom Herbertf70ea012015-07-31 16:52:10 -07001348{
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001349 if (!skb->l4_hash && !skb->sw_hash) {
1350 struct flow_keys keys;
Tom Herbertde4c1f82015-09-01 18:11:04 -07001351 __u32 hash = __get_hash_from_flowi6(fl6, &keys);
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001352
Tom Herbertde4c1f82015-09-01 18:11:04 -07001353 __skb_set_sw_hash(skb, hash, flow_keys_have_l4(&keys));
Tom Herbertc6cc1ca2015-09-01 09:24:25 -07001354 }
Tom Herbertf70ea012015-07-31 16:52:10 -07001355
1356 return skb->hash;
1357}
1358
Eric Dumazet55667442019-10-22 07:57:46 -07001359__u32 skb_get_hash_perturb(const struct sk_buff *skb,
1360 const siphash_key_t *perturb);
Tom Herbert50fb7992015-05-01 11:30:12 -07001361
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001362static inline __u32 skb_get_hash_raw(const struct sk_buff *skb)
1363{
Tom Herbert61b905d2014-03-24 15:34:47 -07001364 return skb->hash;
Tom Herbert57bdf7f42014-01-15 08:57:54 -08001365}
1366
Tom Herbert3df7a742013-12-15 22:16:29 -08001367static inline void skb_copy_hash(struct sk_buff *to, const struct sk_buff *from)
1368{
Tom Herbert61b905d2014-03-24 15:34:47 -07001369 to->hash = from->hash;
Tom Herberta3b18dd2014-07-01 21:33:17 -07001370 to->sw_hash = from->sw_hash;
Tom Herbert61b905d2014-03-24 15:34:47 -07001371 to->l4_hash = from->l4_hash;
Tom Herbert3df7a742013-12-15 22:16:29 -08001372};
1373
Jakub Kicinski41477662019-08-07 17:03:59 -07001374static inline void skb_copy_decrypted(struct sk_buff *to,
1375 const struct sk_buff *from)
1376{
1377#ifdef CONFIG_TLS_DEVICE
1378 to->decrypted = from->decrypted;
1379#endif
1380}
1381
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001382#ifdef NET_SKBUFF_DATA_USES_OFFSET
1383static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1384{
1385 return skb->head + skb->end;
1386}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001387
1388static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1389{
1390 return skb->end;
1391}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001392#else
1393static inline unsigned char *skb_end_pointer(const struct sk_buff *skb)
1394{
1395 return skb->end;
1396}
Alexander Duyckec47ea82012-05-04 14:26:56 +00001397
1398static inline unsigned int skb_end_offset(const struct sk_buff *skb)
1399{
1400 return skb->end - skb->head;
1401}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001402#endif
1403
Linus Torvalds1da177e2005-04-16 15:20:36 -07001404/* Internal */
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07001405#define skb_shinfo(SKB) ((struct skb_shared_info *)(skb_end_pointer(SKB)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07001406
Patrick Ohlyac45f602009-02-12 05:03:37 +00001407static inline struct skb_shared_hwtstamps *skb_hwtstamps(struct sk_buff *skb)
1408{
1409 return &skb_shinfo(skb)->hwtstamps;
1410}
1411
Willem de Bruijn52267792017-08-03 16:29:39 -04001412static inline struct ubuf_info *skb_zcopy(struct sk_buff *skb)
1413{
1414 bool is_zcopy = skb && skb_shinfo(skb)->tx_flags & SKBTX_DEV_ZEROCOPY;
1415
1416 return is_zcopy ? skb_uarg(skb) : NULL;
1417}
1418
Willem de Bruijn52900d22018-11-30 15:32:40 -05001419static inline void skb_zcopy_set(struct sk_buff *skb, struct ubuf_info *uarg,
1420 bool *have_ref)
Willem de Bruijn52267792017-08-03 16:29:39 -04001421{
1422 if (skb && uarg && !skb_zcopy(skb)) {
Willem de Bruijn52900d22018-11-30 15:32:40 -05001423 if (unlikely(have_ref && *have_ref))
1424 *have_ref = false;
1425 else
1426 sock_zerocopy_get(uarg);
Willem de Bruijn52267792017-08-03 16:29:39 -04001427 skb_shinfo(skb)->destructor_arg = uarg;
1428 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1429 }
1430}
1431
Willem de Bruijn5cd8d462018-11-20 13:00:18 -05001432static inline void skb_zcopy_set_nouarg(struct sk_buff *skb, void *val)
1433{
1434 skb_shinfo(skb)->destructor_arg = (void *)((uintptr_t) val | 0x1UL);
1435 skb_shinfo(skb)->tx_flags |= SKBTX_ZEROCOPY_FRAG;
1436}
1437
1438static inline bool skb_zcopy_is_nouarg(struct sk_buff *skb)
1439{
1440 return (uintptr_t) skb_shinfo(skb)->destructor_arg & 0x1UL;
1441}
1442
1443static inline void *skb_zcopy_get_nouarg(struct sk_buff *skb)
1444{
1445 return (void *)((uintptr_t) skb_shinfo(skb)->destructor_arg & ~0x1UL);
1446}
1447
Willem de Bruijn52267792017-08-03 16:29:39 -04001448/* Release a reference on a zerocopy structure */
1449static inline void skb_zcopy_clear(struct sk_buff *skb, bool zerocopy)
1450{
1451 struct ubuf_info *uarg = skb_zcopy(skb);
1452
1453 if (uarg) {
Willem de Bruijn185ce5c2019-05-15 13:29:16 -04001454 if (skb_zcopy_is_nouarg(skb)) {
1455 /* no notification callback */
1456 } else if (uarg->callback == sock_zerocopy_callback) {
Willem de Bruijn0a4a0602017-08-09 19:09:44 -04001457 uarg->zerocopy = uarg->zerocopy && zerocopy;
1458 sock_zerocopy_put(uarg);
Willem de Bruijn185ce5c2019-05-15 13:29:16 -04001459 } else {
Willem de Bruijn0a4a0602017-08-09 19:09:44 -04001460 uarg->callback(uarg, zerocopy);
1461 }
1462
Willem de Bruijn52267792017-08-03 16:29:39 -04001463 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1464 }
1465}
1466
1467/* Abort a zerocopy operation and revert zckey on error in send syscall */
1468static inline void skb_zcopy_abort(struct sk_buff *skb)
1469{
1470 struct ubuf_info *uarg = skb_zcopy(skb);
1471
1472 if (uarg) {
Willem de Bruijn52900d22018-11-30 15:32:40 -05001473 sock_zerocopy_put_abort(uarg, false);
Willem de Bruijn52267792017-08-03 16:29:39 -04001474 skb_shinfo(skb)->tx_flags &= ~SKBTX_ZEROCOPY_FRAG;
1475 }
1476}
1477
David S. Millera8305bf2018-07-29 20:42:53 -07001478static inline void skb_mark_not_on_list(struct sk_buff *skb)
1479{
1480 skb->next = NULL;
1481}
1482
Jason A. Donenfelddcfea722020-01-08 16:59:02 -05001483/* Iterate through singly-linked GSO fragments of an skb. */
Jason A. Donenfeld5eee7bd2020-01-13 18:42:26 -05001484#define skb_list_walk_safe(first, skb, next_skb) \
1485 for ((skb) = (first), (next_skb) = (skb) ? (skb)->next : NULL; (skb); \
1486 (skb) = (next_skb), (next_skb) = (skb) ? (skb)->next : NULL)
Jason A. Donenfelddcfea722020-01-08 16:59:02 -05001487
David S. Miller992cba72018-07-31 15:27:56 -07001488static inline void skb_list_del_init(struct sk_buff *skb)
1489{
1490 __list_del_entry(&skb->list);
1491 skb_mark_not_on_list(skb);
1492}
1493
Linus Torvalds1da177e2005-04-16 15:20:36 -07001494/**
1495 * skb_queue_empty - check if a queue is empty
1496 * @list: queue head
1497 *
1498 * Returns true if the queue is empty, false otherwise.
1499 */
1500static inline int skb_queue_empty(const struct sk_buff_head *list)
1501{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001502 return list->next == (const struct sk_buff *) list;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001503}
1504
1505/**
Eric Dumazetd7d16a82019-10-23 22:44:48 -07001506 * skb_queue_empty_lockless - check if a queue is empty
1507 * @list: queue head
1508 *
1509 * Returns true if the queue is empty, false otherwise.
1510 * This variant can be used in lockless contexts.
1511 */
1512static inline bool skb_queue_empty_lockless(const struct sk_buff_head *list)
1513{
1514 return READ_ONCE(list->next) == (const struct sk_buff *) list;
1515}
1516
1517
1518/**
David S. Millerfc7ebb22008-09-23 00:34:07 -07001519 * skb_queue_is_last - check if skb is the last entry in the queue
1520 * @list: queue head
1521 * @skb: buffer
1522 *
1523 * Returns true if @skb is the last buffer on the list.
1524 */
1525static inline bool skb_queue_is_last(const struct sk_buff_head *list,
1526 const struct sk_buff *skb)
1527{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001528 return skb->next == (const struct sk_buff *) list;
David S. Millerfc7ebb22008-09-23 00:34:07 -07001529}
1530
1531/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001532 * skb_queue_is_first - check if skb is the first entry in the queue
1533 * @list: queue head
1534 * @skb: buffer
1535 *
1536 * Returns true if @skb is the first buffer on the list.
1537 */
1538static inline bool skb_queue_is_first(const struct sk_buff_head *list,
1539 const struct sk_buff *skb)
1540{
Daniel Borkmannfd44b932014-01-07 23:23:44 +01001541 return skb->prev == (const struct sk_buff *) list;
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001542}
1543
1544/**
David S. Miller249c8b42008-09-23 00:44:42 -07001545 * skb_queue_next - return the next packet in the queue
1546 * @list: queue head
1547 * @skb: current buffer
1548 *
1549 * Return the next packet in @list after @skb. It is only valid to
1550 * call this if skb_queue_is_last() evaluates to false.
1551 */
1552static inline struct sk_buff *skb_queue_next(const struct sk_buff_head *list,
1553 const struct sk_buff *skb)
1554{
1555 /* This BUG_ON may seem severe, but if we just return then we
1556 * are going to dereference garbage.
1557 */
1558 BUG_ON(skb_queue_is_last(list, skb));
1559 return skb->next;
1560}
1561
1562/**
Ilpo Järvinen832d11c2008-11-24 21:20:15 -08001563 * skb_queue_prev - return the prev packet in the queue
1564 * @list: queue head
1565 * @skb: current buffer
1566 *
1567 * Return the prev packet in @list before @skb. It is only valid to
1568 * call this if skb_queue_is_first() evaluates to false.
1569 */
1570static inline struct sk_buff *skb_queue_prev(const struct sk_buff_head *list,
1571 const struct sk_buff *skb)
1572{
1573 /* This BUG_ON may seem severe, but if we just return then we
1574 * are going to dereference garbage.
1575 */
1576 BUG_ON(skb_queue_is_first(list, skb));
1577 return skb->prev;
1578}
1579
1580/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001581 * skb_get - reference buffer
1582 * @skb: buffer to reference
1583 *
1584 * Makes another reference to a socket buffer and returns a pointer
1585 * to the buffer.
1586 */
1587static inline struct sk_buff *skb_get(struct sk_buff *skb)
1588{
Reshetova, Elena63354792017-06-30 13:07:58 +03001589 refcount_inc(&skb->users);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001590 return skb;
1591}
1592
1593/*
Geert Uytterhoevenf8821f92017-11-30 14:33:56 +01001594 * If users == 1, we are the only owner and can avoid redundant atomic changes.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001595 */
1596
1597/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001598 * skb_cloned - is the buffer a clone
1599 * @skb: buffer to check
1600 *
1601 * Returns true if the buffer was generated with skb_clone() and is
1602 * one of multiple shared copies of the buffer. Cloned buffers are
1603 * shared data so must not be written to under normal circumstances.
1604 */
1605static inline int skb_cloned(const struct sk_buff *skb)
1606{
1607 return skb->cloned &&
1608 (atomic_read(&skb_shinfo(skb)->dataref) & SKB_DATAREF_MASK) != 1;
1609}
1610
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001611static inline int skb_unclone(struct sk_buff *skb, gfp_t pri)
1612{
Mel Gormand0164ad2015-11-06 16:28:21 -08001613 might_sleep_if(gfpflags_allow_blocking(pri));
Pravin B Shelar14bbd6a2013-02-14 09:44:49 +00001614
1615 if (skb_cloned(skb))
1616 return pskb_expand_head(skb, 0, 0, pri);
1617
1618 return 0;
1619}
1620
Linus Torvalds1da177e2005-04-16 15:20:36 -07001621/**
1622 * skb_header_cloned - is the header a clone
1623 * @skb: buffer to check
1624 *
1625 * Returns true if modifying the header part of the buffer requires
1626 * the data to be copied.
1627 */
1628static inline int skb_header_cloned(const struct sk_buff *skb)
1629{
1630 int dataref;
1631
1632 if (!skb->cloned)
1633 return 0;
1634
1635 dataref = atomic_read(&skb_shinfo(skb)->dataref);
1636 dataref = (dataref & SKB_DATAREF_MASK) - (dataref >> SKB_DATAREF_SHIFT);
1637 return dataref != 1;
1638}
1639
Eric Dumazet9580bf22016-04-30 10:19:29 -07001640static inline int skb_header_unclone(struct sk_buff *skb, gfp_t pri)
1641{
1642 might_sleep_if(gfpflags_allow_blocking(pri));
1643
1644 if (skb_header_cloned(skb))
1645 return pskb_expand_head(skb, 0, 0, pri);
1646
1647 return 0;
1648}
1649
Linus Torvalds1da177e2005-04-16 15:20:36 -07001650/**
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001651 * __skb_header_release - release reference to header
1652 * @skb: buffer to operate on
Eric Dumazetf4a775d2014-09-22 16:29:32 -07001653 */
1654static inline void __skb_header_release(struct sk_buff *skb)
1655{
1656 skb->nohdr = 1;
1657 atomic_set(&skb_shinfo(skb)->dataref, 1 + (1 << SKB_DATAREF_SHIFT));
1658}
1659
1660
1661/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001662 * skb_shared - is the buffer shared
1663 * @skb: buffer to check
1664 *
1665 * Returns true if more than one person has a reference to this
1666 * buffer.
1667 */
1668static inline int skb_shared(const struct sk_buff *skb)
1669{
Reshetova, Elena63354792017-06-30 13:07:58 +03001670 return refcount_read(&skb->users) != 1;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001671}
1672
1673/**
1674 * skb_share_check - check if buffer is shared and if so clone it
1675 * @skb: buffer to check
1676 * @pri: priority for memory allocation
1677 *
1678 * If the buffer is shared the buffer is cloned and the old copy
1679 * drops a reference. A new clone with a single reference is returned.
1680 * If the buffer is not shared the original buffer is returned. When
1681 * being called from interrupt status or with spinlocks held pri must
1682 * be GFP_ATOMIC.
1683 *
1684 * NULL is returned on a memory allocation failure.
1685 */
Eric Dumazet47061bc2012-08-03 20:54:15 +00001686static inline struct sk_buff *skb_share_check(struct sk_buff *skb, gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001687{
Mel Gormand0164ad2015-11-06 16:28:21 -08001688 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001689 if (skb_shared(skb)) {
1690 struct sk_buff *nskb = skb_clone(skb, pri);
Eric Dumazet47061bc2012-08-03 20:54:15 +00001691
1692 if (likely(nskb))
1693 consume_skb(skb);
1694 else
1695 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001696 skb = nskb;
1697 }
1698 return skb;
1699}
1700
1701/*
1702 * Copy shared buffers into a new sk_buff. We effectively do COW on
1703 * packets to handle cases where we have a local reader and forward
1704 * and a couple of other messy ones. The normal one is tcpdumping
1705 * a packet thats being forwarded.
1706 */
1707
1708/**
1709 * skb_unshare - make a copy of a shared buffer
1710 * @skb: buffer to check
1711 * @pri: priority for memory allocation
1712 *
1713 * If the socket buffer is a clone then this function creates a new
1714 * copy of the data, drops a reference count on the old copy and returns
1715 * the new copy with the reference count at 1. If the buffer is not a clone
1716 * the original buffer is returned. When called with a spinlock held or
1717 * from interrupt state @pri must be %GFP_ATOMIC
1718 *
1719 * %NULL is returned on a memory allocation failure.
1720 */
Victor Fuscoe2bf5212005-07-18 13:36:38 -07001721static inline struct sk_buff *skb_unshare(struct sk_buff *skb,
Al Virodd0fc662005-10-07 07:46:04 +01001722 gfp_t pri)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001723{
Mel Gormand0164ad2015-11-06 16:28:21 -08001724 might_sleep_if(gfpflags_allow_blocking(pri));
Linus Torvalds1da177e2005-04-16 15:20:36 -07001725 if (skb_cloned(skb)) {
1726 struct sk_buff *nskb = skb_copy(skb, pri);
Alexander Aring31eff812014-10-10 23:10:47 +02001727
1728 /* Free our shared copy */
1729 if (likely(nskb))
1730 consume_skb(skb);
1731 else
1732 kfree_skb(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001733 skb = nskb;
1734 }
1735 return skb;
1736}
1737
1738/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001739 * skb_peek - peek at the head of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001740 * @list_: list to peek at
1741 *
1742 * Peek an &sk_buff. Unlike most other operations you _MUST_
1743 * be careful with this one. A peek leaves the buffer on the
1744 * list and someone else may run off with it. You must hold
1745 * the appropriate locks or have a private queue to do this.
1746 *
1747 * Returns %NULL for an empty list or a pointer to the head element.
1748 * The reference count is not incremented and the reference is therefore
1749 * volatile. Use with caution.
1750 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001751static inline struct sk_buff *skb_peek(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001752{
Eric Dumazet18d07002012-04-30 16:31:46 +00001753 struct sk_buff *skb = list_->next;
1754
1755 if (skb == (struct sk_buff *)list_)
1756 skb = NULL;
1757 return skb;
Linus Torvalds1da177e2005-04-16 15:20:36 -07001758}
1759
1760/**
David S. Miller8b69bd72018-08-11 18:43:38 -07001761 * __skb_peek - peek at the head of a non-empty &sk_buff_head
1762 * @list_: list to peek at
1763 *
1764 * Like skb_peek(), but the caller knows that the list is not empty.
1765 */
1766static inline struct sk_buff *__skb_peek(const struct sk_buff_head *list_)
1767{
1768 return list_->next;
1769}
1770
1771/**
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001772 * skb_peek_next - peek skb following the given one from a queue
1773 * @skb: skb to start from
1774 * @list_: list to peek at
1775 *
1776 * Returns %NULL when the end of the list is met or a pointer to the
1777 * next element. The reference count is not incremented and the
1778 * reference is therefore volatile. Use with caution.
1779 */
1780static inline struct sk_buff *skb_peek_next(struct sk_buff *skb,
1781 const struct sk_buff_head *list_)
1782{
1783 struct sk_buff *next = skb->next;
Eric Dumazet18d07002012-04-30 16:31:46 +00001784
Pavel Emelyanovda5ef6e2012-02-21 07:31:18 +00001785 if (next == (struct sk_buff *)list_)
1786 next = NULL;
1787 return next;
1788}
1789
1790/**
Ben Hutchings1a5778a2010-02-14 22:35:47 -08001791 * skb_peek_tail - peek at the tail of an &sk_buff_head
Linus Torvalds1da177e2005-04-16 15:20:36 -07001792 * @list_: list to peek at
1793 *
1794 * Peek an &sk_buff. Unlike most other operations you _MUST_
1795 * be careful with this one. A peek leaves the buffer on the
1796 * list and someone else may run off with it. You must hold
1797 * the appropriate locks or have a private queue to do this.
1798 *
1799 * Returns %NULL for an empty list or a pointer to the tail element.
1800 * The reference count is not incremented and the reference is therefore
1801 * volatile. Use with caution.
1802 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04001803static inline struct sk_buff *skb_peek_tail(const struct sk_buff_head *list_)
Linus Torvalds1da177e2005-04-16 15:20:36 -07001804{
Eric Dumazetf8cc62c2019-11-07 18:49:43 -08001805 struct sk_buff *skb = READ_ONCE(list_->prev);
Eric Dumazet18d07002012-04-30 16:31:46 +00001806
1807 if (skb == (struct sk_buff *)list_)
1808 skb = NULL;
1809 return skb;
1810
Linus Torvalds1da177e2005-04-16 15:20:36 -07001811}
1812
1813/**
1814 * skb_queue_len - get queue length
1815 * @list_: list to measure
1816 *
1817 * Return the length of an &sk_buff queue.
1818 */
1819static inline __u32 skb_queue_len(const struct sk_buff_head *list_)
1820{
1821 return list_->qlen;
1822}
1823
David S. Miller67fed452008-09-21 22:36:24 -07001824/**
Qian Cai86b18aa2020-02-04 13:40:29 -05001825 * skb_queue_len_lockless - get queue length
1826 * @list_: list to measure
1827 *
1828 * Return the length of an &sk_buff queue.
1829 * This variant can be used in lockless contexts.
1830 */
1831static inline __u32 skb_queue_len_lockless(const struct sk_buff_head *list_)
1832{
1833 return READ_ONCE(list_->qlen);
1834}
1835
1836/**
David S. Miller67fed452008-09-21 22:36:24 -07001837 * __skb_queue_head_init - initialize non-spinlock portions of sk_buff_head
1838 * @list: queue to initialize
1839 *
1840 * This initializes only the list and queue length aspects of
1841 * an sk_buff_head object. This allows to initialize the list
1842 * aspects of an sk_buff_head without reinitializing things like
1843 * the spinlock. It can also be used for on-stack sk_buff_head
1844 * objects where the spinlock is known to not be used.
1845 */
1846static inline void __skb_queue_head_init(struct sk_buff_head *list)
1847{
1848 list->prev = list->next = (struct sk_buff *)list;
1849 list->qlen = 0;
1850}
1851
Arjan van de Ven76f10ad2006-08-02 14:06:55 -07001852/*
1853 * This function creates a split out lock class for each invocation;
1854 * this is needed for now since a whole lot of users of the skb-queue
1855 * infrastructure in drivers have different locking usage (in hardirq)
1856 * than the networking core (in softirq only). In the long run either the
1857 * network layer or drivers should need annotation to consolidate the
1858 * main types of usage into 3 classes.
1859 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07001860static inline void skb_queue_head_init(struct sk_buff_head *list)
1861{
1862 spin_lock_init(&list->lock);
David S. Miller67fed452008-09-21 22:36:24 -07001863 __skb_queue_head_init(list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07001864}
1865
Pavel Emelianovc2ecba72007-04-17 12:45:31 -07001866static inline void skb_queue_head_init_class(struct sk_buff_head *list,
1867 struct lock_class_key *class)
1868{
1869 skb_queue_head_init(list);
1870 lockdep_set_class(&list->lock, class);
1871}
1872
Linus Torvalds1da177e2005-04-16 15:20:36 -07001873/*
Gerrit Renkerbf299272008-04-14 00:04:51 -07001874 * Insert an sk_buff on a list.
Linus Torvalds1da177e2005-04-16 15:20:36 -07001875 *
1876 * The "__skb_xxxx()" functions are the non-atomic ones that
1877 * can only be called with interrupts disabled.
1878 */
Gerrit Renkerbf299272008-04-14 00:04:51 -07001879static inline void __skb_insert(struct sk_buff *newsk,
1880 struct sk_buff *prev, struct sk_buff *next,
1881 struct sk_buff_head *list)
1882{
Eric Dumazetf8cc62c2019-11-07 18:49:43 -08001883 /* See skb_queue_empty_lockless() and skb_peek_tail()
1884 * for the opposite READ_ONCE()
1885 */
Eric Dumazetd7d16a82019-10-23 22:44:48 -07001886 WRITE_ONCE(newsk->next, next);
1887 WRITE_ONCE(newsk->prev, prev);
1888 WRITE_ONCE(next->prev, newsk);
1889 WRITE_ONCE(prev->next, newsk);
Gerrit Renkerbf299272008-04-14 00:04:51 -07001890 list->qlen++;
1891}
Linus Torvalds1da177e2005-04-16 15:20:36 -07001892
David S. Miller67fed452008-09-21 22:36:24 -07001893static inline void __skb_queue_splice(const struct sk_buff_head *list,
1894 struct sk_buff *prev,
1895 struct sk_buff *next)
1896{
1897 struct sk_buff *first = list->next;
1898 struct sk_buff *last = list->prev;
1899
Eric Dumazetd7d16a82019-10-23 22:44:48 -07001900 WRITE_ONCE(first->prev, prev);
1901 WRITE_ONCE(prev->next, first);
David S. Miller67fed452008-09-21 22:36:24 -07001902
Eric Dumazetd7d16a82019-10-23 22:44:48 -07001903 WRITE_ONCE(last->next, next);
1904 WRITE_ONCE(next->prev, last);
David S. Miller67fed452008-09-21 22:36:24 -07001905}
1906
1907/**
1908 * skb_queue_splice - join two skb lists, this is designed for stacks
1909 * @list: the new list to add
1910 * @head: the place to add it in the first list
1911 */
1912static inline void skb_queue_splice(const struct sk_buff_head *list,
1913 struct sk_buff_head *head)
1914{
1915 if (!skb_queue_empty(list)) {
1916 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001917 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001918 }
1919}
1920
1921/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001922 * skb_queue_splice_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001923 * @list: the new list to add
1924 * @head: the place to add it in the first list
1925 *
1926 * The list at @list is reinitialised
1927 */
1928static inline void skb_queue_splice_init(struct sk_buff_head *list,
1929 struct sk_buff_head *head)
1930{
1931 if (!skb_queue_empty(list)) {
1932 __skb_queue_splice(list, (struct sk_buff *) head, head->next);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001933 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001934 __skb_queue_head_init(list);
1935 }
1936}
1937
1938/**
1939 * skb_queue_splice_tail - join two skb lists, each list being a queue
1940 * @list: the new list to add
1941 * @head: the place to add it in the first list
1942 */
1943static inline void skb_queue_splice_tail(const struct sk_buff_head *list,
1944 struct sk_buff_head *head)
1945{
1946 if (!skb_queue_empty(list)) {
1947 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
David S. Miller1d4a31d2008-09-22 21:57:21 -07001948 head->qlen += list->qlen;
David S. Miller67fed452008-09-21 22:36:24 -07001949 }
1950}
1951
1952/**
Eric Dumazetd961949662012-04-30 21:29:16 +00001953 * skb_queue_splice_tail_init - join two skb lists and reinitialise the emptied list
David S. Miller67fed452008-09-21 22:36:24 -07001954 * @list: the new list to add
1955 * @head: the place to add it in the first list
1956 *
1957 * Each of the lists is a queue.
1958 * The list at @list is reinitialised
1959 */
1960static inline void skb_queue_splice_tail_init(struct sk_buff_head *list,
1961 struct sk_buff_head *head)
1962{
1963 if (!skb_queue_empty(list)) {
1964 __skb_queue_splice(list, head->prev, (struct sk_buff *) head);
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
Linus Torvalds1da177e2005-04-16 15:20:36 -07001970/**
Stephen Hemminger300ce172005-10-30 13:47:34 -08001971 * __skb_queue_after - queue a buffer at the list head
1972 * @list: list to use
1973 * @prev: place after this buffer
1974 * @newsk: buffer to queue
1975 *
1976 * Queue a buffer int the middle of a list. This function takes no locks
1977 * and you must therefore hold required locks before calling it.
1978 *
1979 * A buffer cannot be placed on two lists at the same time.
1980 */
1981static inline void __skb_queue_after(struct sk_buff_head *list,
1982 struct sk_buff *prev,
1983 struct sk_buff *newsk)
1984{
Gerrit Renkerbf299272008-04-14 00:04:51 -07001985 __skb_insert(newsk, prev, prev->next, list);
Stephen Hemminger300ce172005-10-30 13:47:34 -08001986}
1987
Joe Perches7965bd42013-09-26 14:48:15 -07001988void skb_append(struct sk_buff *old, struct sk_buff *newsk,
1989 struct sk_buff_head *list);
Gerrit Renker7de6c032008-04-14 00:05:09 -07001990
Gerrit Renkerf5572852008-04-14 00:05:28 -07001991static inline void __skb_queue_before(struct sk_buff_head *list,
1992 struct sk_buff *next,
1993 struct sk_buff *newsk)
1994{
1995 __skb_insert(newsk, next->prev, next, list);
1996}
1997
Stephen Hemminger300ce172005-10-30 13:47:34 -08001998/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07001999 * __skb_queue_head - queue a buffer at the list head
2000 * @list: list to use
2001 * @newsk: buffer to queue
2002 *
2003 * Queue a buffer at the start of a list. This function takes no locks
2004 * and you must therefore hold required locks before calling it.
2005 *
2006 * A buffer cannot be placed on two lists at the same time.
2007 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002008static inline void __skb_queue_head(struct sk_buff_head *list,
2009 struct sk_buff *newsk)
2010{
Stephen Hemminger300ce172005-10-30 13:47:34 -08002011 __skb_queue_after(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002012}
Brian Norris4ea7b0c2019-02-11 13:02:25 -08002013void skb_queue_head(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002014
2015/**
2016 * __skb_queue_tail - queue a buffer at the list tail
2017 * @list: list to use
2018 * @newsk: buffer to queue
2019 *
2020 * Queue a buffer at the end of a list. This function takes no locks
2021 * and you must therefore hold required locks before calling it.
2022 *
2023 * A buffer cannot be placed on two lists at the same time.
2024 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002025static inline void __skb_queue_tail(struct sk_buff_head *list,
2026 struct sk_buff *newsk)
2027{
Gerrit Renkerf5572852008-04-14 00:05:28 -07002028 __skb_queue_before(list, (struct sk_buff *)list, newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002029}
Brian Norris4ea7b0c2019-02-11 13:02:25 -08002030void skb_queue_tail(struct sk_buff_head *list, struct sk_buff *newsk);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002031
Linus Torvalds1da177e2005-04-16 15:20:36 -07002032/*
Linus Torvalds1da177e2005-04-16 15:20:36 -07002033 * remove sk_buff from list. _Must_ be called atomically, and with
2034 * the list known..
2035 */
Joe Perches7965bd42013-09-26 14:48:15 -07002036void skb_unlink(struct sk_buff *skb, struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002037static inline void __skb_unlink(struct sk_buff *skb, struct sk_buff_head *list)
2038{
2039 struct sk_buff *next, *prev;
2040
Qian Cai86b18aa2020-02-04 13:40:29 -05002041 WRITE_ONCE(list->qlen, list->qlen - 1);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002042 next = skb->next;
2043 prev = skb->prev;
2044 skb->next = skb->prev = NULL;
Eric Dumazetd7d16a82019-10-23 22:44:48 -07002045 WRITE_ONCE(next->prev, prev);
2046 WRITE_ONCE(prev->next, next);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002047}
2048
Gerrit Renkerf525c062008-04-14 00:04:12 -07002049/**
2050 * __skb_dequeue - remove from the head of the queue
2051 * @list: list to dequeue from
2052 *
2053 * Remove the head of the list. This function does not take any locks
2054 * so must be used with appropriate locks held only. The head item is
2055 * returned or %NULL if the list is empty.
2056 */
Gerrit Renkerf525c062008-04-14 00:04:12 -07002057static inline struct sk_buff *__skb_dequeue(struct sk_buff_head *list)
2058{
2059 struct sk_buff *skb = skb_peek(list);
2060 if (skb)
2061 __skb_unlink(skb, list);
2062 return skb;
2063}
Brian Norris4ea7b0c2019-02-11 13:02:25 -08002064struct sk_buff *skb_dequeue(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002065
2066/**
2067 * __skb_dequeue_tail - remove from the tail of the queue
2068 * @list: list to dequeue from
2069 *
2070 * Remove the tail of the list. This function does not take any locks
2071 * so must be used with appropriate locks held only. The tail item is
2072 * returned or %NULL if the list is empty.
2073 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002074static inline struct sk_buff *__skb_dequeue_tail(struct sk_buff_head *list)
2075{
2076 struct sk_buff *skb = skb_peek_tail(list);
2077 if (skb)
2078 __skb_unlink(skb, list);
2079 return skb;
2080}
Brian Norris4ea7b0c2019-02-11 13:02:25 -08002081struct sk_buff *skb_dequeue_tail(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002082
2083
David S. Millerbdcc0922012-03-07 20:53:36 -05002084static inline bool skb_is_nonlinear(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002085{
2086 return skb->data_len;
2087}
2088
2089static inline unsigned int skb_headlen(const struct sk_buff *skb)
2090{
2091 return skb->len - skb->data_len;
2092}
2093
Willem de Bruijn3ece7822017-08-03 16:29:38 -04002094static inline unsigned int __skb_pagelen(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002095{
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03002096 unsigned int i, len = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002097
Alexey Dobriyanc72d8cd2016-11-19 04:08:08 +03002098 for (i = skb_shinfo(skb)->nr_frags - 1; (int)i >= 0; i--)
Eric Dumazet9e903e02011-10-18 21:00:24 +00002099 len += skb_frag_size(&skb_shinfo(skb)->frags[i]);
Willem de Bruijn3ece7822017-08-03 16:29:38 -04002100 return len;
2101}
2102
2103static inline unsigned int skb_pagelen(const struct sk_buff *skb)
2104{
2105 return skb_headlen(skb) + __skb_pagelen(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002106}
2107
Ian Campbell131ea662011-08-19 06:25:00 +00002108/**
2109 * __skb_fill_page_desc - initialise a paged fragment in an skb
2110 * @skb: buffer containing fragment to be initialised
2111 * @i: paged fragment index to initialise
2112 * @page: the page to use for this fragment
2113 * @off: the offset to the data with @page
2114 * @size: the length of the data
2115 *
2116 * Initialises the @i'th fragment of @skb to point to &size bytes at
2117 * offset @off within @page.
2118 *
2119 * Does not take any additional reference on the fragment.
2120 */
2121static inline void __skb_fill_page_desc(struct sk_buff *skb, int i,
2122 struct page *page, int off, int size)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002123{
2124 skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
2125
Mel Gormanc48a11c2012-07-31 16:44:23 -07002126 /*
Michal Hocko2f064f32015-08-21 14:11:51 -07002127 * Propagate page pfmemalloc to the skb if we can. The problem is
2128 * that not all callers have unique ownership of the page but rely
2129 * on page_is_pfmemalloc doing the right thing(tm).
Mel Gormanc48a11c2012-07-31 16:44:23 -07002130 */
Matthew Wilcox (Oracle)1dfa5bd2019-07-22 20:08:29 -07002131 frag->bv_page = page;
Jonathan Lemon65c84f12019-07-30 07:40:34 -07002132 frag->bv_offset = off;
Eric Dumazet9e903e02011-10-18 21:00:24 +00002133 skb_frag_size_set(frag, size);
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00002134
2135 page = compound_head(page);
Michal Hocko2f064f32015-08-21 14:11:51 -07002136 if (page_is_pfmemalloc(page))
Pavel Emelyanovcca7af32013-03-14 03:29:40 +00002137 skb->pfmemalloc = true;
Ian Campbell131ea662011-08-19 06:25:00 +00002138}
2139
2140/**
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 * As per __skb_fill_page_desc() -- initialises the @i'th fragment of
Mathias Krausebc323832013-11-07 14:18:26 +01002149 * @skb to point to @size bytes at offset @off within @page. In
Ian Campbell131ea662011-08-19 06:25:00 +00002150 * addition updates @skb such that @i is the last fragment.
2151 *
2152 * Does not take any additional reference on the fragment.
2153 */
2154static inline void skb_fill_page_desc(struct sk_buff *skb, int i,
2155 struct page *page, int off, int size)
2156{
2157 __skb_fill_page_desc(skb, i, page, off, size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002158 skb_shinfo(skb)->nr_frags = i + 1;
2159}
2160
Joe Perches7965bd42013-09-26 14:48:15 -07002161void skb_add_rx_frag(struct sk_buff *skb, int i, struct page *page, int off,
2162 int size, unsigned int truesize);
Peter Zijlstra654bed12008-10-07 14:22:33 -07002163
Jason Wangf8e617e2013-11-01 14:07:47 +08002164void skb_coalesce_rx_frag(struct sk_buff *skb, int i, int size,
2165 unsigned int truesize);
2166
Linus Torvalds1da177e2005-04-16 15:20:36 -07002167#define SKB_LINEAR_ASSERT(skb) BUG_ON(skb_is_nonlinear(skb))
2168
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002169#ifdef NET_SKBUFF_DATA_USES_OFFSET
2170static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2171{
2172 return skb->head + skb->tail;
2173}
2174
2175static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2176{
2177 skb->tail = skb->data - skb->head;
2178}
2179
2180static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2181{
2182 skb_reset_tail_pointer(skb);
2183 skb->tail += offset;
2184}
Simon Horman7cc46192013-05-28 20:34:29 +00002185
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002186#else /* NET_SKBUFF_DATA_USES_OFFSET */
2187static inline unsigned char *skb_tail_pointer(const struct sk_buff *skb)
2188{
2189 return skb->tail;
2190}
2191
2192static inline void skb_reset_tail_pointer(struct sk_buff *skb)
2193{
2194 skb->tail = skb->data;
2195}
2196
2197static inline void skb_set_tail_pointer(struct sk_buff *skb, const int offset)
2198{
2199 skb->tail = skb->data + offset;
2200}
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07002201
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002202#endif /* NET_SKBUFF_DATA_USES_OFFSET */
2203
Linus Torvalds1da177e2005-04-16 15:20:36 -07002204/*
2205 * Add data to an sk_buff
2206 */
Johannes Berg4df864c2017-06-16 14:29:21 +02002207void *pskb_put(struct sk_buff *skb, struct sk_buff *tail, int len);
2208void *skb_put(struct sk_buff *skb, unsigned int len);
2209static inline void *__skb_put(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002210{
Johannes Berg4df864c2017-06-16 14:29:21 +02002211 void *tmp = skb_tail_pointer(skb);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002212 SKB_LINEAR_ASSERT(skb);
2213 skb->tail += len;
2214 skb->len += len;
2215 return tmp;
2216}
2217
yuan linyude77b962017-06-18 22:48:17 +08002218static inline void *__skb_put_zero(struct sk_buff *skb, unsigned int len)
2219{
2220 void *tmp = __skb_put(skb, len);
2221
2222 memset(tmp, 0, len);
2223 return tmp;
2224}
2225
2226static inline void *__skb_put_data(struct sk_buff *skb, const void *data,
2227 unsigned int len)
2228{
2229 void *tmp = __skb_put(skb, len);
2230
2231 memcpy(tmp, data, len);
2232 return tmp;
2233}
2234
2235static inline void __skb_put_u8(struct sk_buff *skb, u8 val)
2236{
2237 *(u8 *)__skb_put(skb, 1) = val;
2238}
2239
Johannes Berg83ad3572017-06-14 22:17:20 +02002240static inline void *skb_put_zero(struct sk_buff *skb, unsigned int len)
Johannes Berge45a79d2017-05-24 09:07:47 +02002241{
Johannes Berg83ad3572017-06-14 22:17:20 +02002242 void *tmp = skb_put(skb, len);
Johannes Berge45a79d2017-05-24 09:07:47 +02002243
2244 memset(tmp, 0, len);
2245
2246 return tmp;
2247}
2248
Johannes Berg59ae1d12017-06-16 14:29:20 +02002249static inline void *skb_put_data(struct sk_buff *skb, const void *data,
2250 unsigned int len)
2251{
2252 void *tmp = skb_put(skb, len);
2253
2254 memcpy(tmp, data, len);
2255
2256 return tmp;
2257}
2258
Johannes Berg634fef62017-06-16 14:29:24 +02002259static inline void skb_put_u8(struct sk_buff *skb, u8 val)
2260{
2261 *(u8 *)skb_put(skb, 1) = val;
2262}
2263
Johannes Bergd58ff352017-06-16 14:29:23 +02002264void *skb_push(struct sk_buff *skb, unsigned int len);
2265static inline void *__skb_push(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002266{
2267 skb->data -= len;
2268 skb->len += len;
2269 return skb->data;
2270}
2271
Johannes Bergaf728682017-06-16 14:29:22 +02002272void *skb_pull(struct sk_buff *skb, unsigned int len);
2273static inline void *__skb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002274{
2275 skb->len -= len;
2276 BUG_ON(skb->len < skb->data_len);
2277 return skb->data += len;
2278}
2279
Johannes Bergaf728682017-06-16 14:29:22 +02002280static inline void *skb_pull_inline(struct sk_buff *skb, unsigned int len)
David S. Miller47d29642010-05-02 02:21:44 -07002281{
2282 return unlikely(len > skb->len) ? NULL : __skb_pull(skb, len);
2283}
2284
Johannes Bergaf728682017-06-16 14:29:22 +02002285void *__pskb_pull_tail(struct sk_buff *skb, int delta);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002286
Johannes Bergaf728682017-06-16 14:29:22 +02002287static inline void *__pskb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002288{
2289 if (len > skb_headlen(skb) &&
Gerrit Renker987c4022008-08-11 18:17:17 -07002290 !__pskb_pull_tail(skb, len - skb_headlen(skb)))
Linus Torvalds1da177e2005-04-16 15:20:36 -07002291 return NULL;
2292 skb->len -= len;
2293 return skb->data += len;
2294}
2295
Johannes Bergaf728682017-06-16 14:29:22 +02002296static inline void *pskb_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002297{
2298 return unlikely(len > skb->len) ? NULL : __pskb_pull(skb, len);
2299}
2300
Heiner Kallweitb9df4fd2019-10-06 18:19:54 +02002301static inline bool pskb_may_pull(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002302{
2303 if (likely(len <= skb_headlen(skb)))
Heiner Kallweitb9df4fd2019-10-06 18:19:54 +02002304 return true;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002305 if (unlikely(len > skb->len))
Heiner Kallweitb9df4fd2019-10-06 18:19:54 +02002306 return false;
Gerrit Renker987c4022008-08-11 18:17:17 -07002307 return __pskb_pull_tail(skb, len - skb_headlen(skb)) != NULL;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002308}
2309
Eric Dumazetc8c8b122016-12-07 09:19:33 -08002310void skb_condense(struct sk_buff *skb);
2311
Linus Torvalds1da177e2005-04-16 15:20:36 -07002312/**
2313 * skb_headroom - bytes at buffer head
2314 * @skb: buffer to check
2315 *
2316 * Return the number of bytes of free space at the head of an &sk_buff.
2317 */
Chuck Leverc2636b42007-10-23 21:07:32 -07002318static inline unsigned int skb_headroom(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002319{
2320 return skb->data - skb->head;
2321}
2322
2323/**
2324 * skb_tailroom - bytes at buffer end
2325 * @skb: buffer to check
2326 *
2327 * Return the number of bytes of free space at the tail of an sk_buff
2328 */
2329static inline int skb_tailroom(const struct sk_buff *skb)
2330{
Arnaldo Carvalho de Melo4305b542007-04-19 20:43:29 -07002331 return skb_is_nonlinear(skb) ? 0 : skb->end - skb->tail;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002332}
2333
2334/**
Eric Dumazeta21d4572012-04-10 20:30:48 +00002335 * skb_availroom - bytes at buffer end
2336 * @skb: buffer to check
2337 *
2338 * Return the number of bytes of free space at the tail of an sk_buff
2339 * allocated by sk_stream_alloc()
2340 */
2341static inline int skb_availroom(const struct sk_buff *skb)
2342{
Eric Dumazet16fad692013-03-14 05:40:32 +00002343 if (skb_is_nonlinear(skb))
2344 return 0;
2345
2346 return skb->end - skb->tail - skb->reserved_tailroom;
Eric Dumazeta21d4572012-04-10 20:30:48 +00002347}
2348
2349/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002350 * skb_reserve - adjust headroom
2351 * @skb: buffer to alter
2352 * @len: bytes to move
2353 *
2354 * Increase the headroom of an empty &sk_buff by reducing the tail
2355 * room. This is only allowed for an empty buffer.
2356 */
David S. Miller8243126c2006-01-17 02:54:21 -08002357static inline void skb_reserve(struct sk_buff *skb, int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002358{
2359 skb->data += len;
2360 skb->tail += len;
2361}
2362
Benjamin Poirier1837b2e2016-02-29 15:03:33 -08002363/**
2364 * skb_tailroom_reserve - adjust reserved_tailroom
2365 * @skb: buffer to alter
2366 * @mtu: maximum amount of headlen permitted
2367 * @needed_tailroom: minimum amount of reserved_tailroom
2368 *
2369 * Set reserved_tailroom so that headlen can be as large as possible but
2370 * not larger than mtu and tailroom cannot be smaller than
2371 * needed_tailroom.
2372 * The required headroom should already have been reserved before using
2373 * this function.
2374 */
2375static inline void skb_tailroom_reserve(struct sk_buff *skb, unsigned int mtu,
2376 unsigned int needed_tailroom)
2377{
2378 SKB_LINEAR_ASSERT(skb);
2379 if (mtu < skb_tailroom(skb) - needed_tailroom)
2380 /* use at most mtu */
2381 skb->reserved_tailroom = skb_tailroom(skb) - mtu;
2382 else
2383 /* use up to all available space */
2384 skb->reserved_tailroom = needed_tailroom;
2385}
2386
Tom Herbert8bce6d72014-09-29 20:22:29 -07002387#define ENCAP_TYPE_ETHER 0
2388#define ENCAP_TYPE_IPPROTO 1
2389
2390static inline void skb_set_inner_protocol(struct sk_buff *skb,
2391 __be16 protocol)
2392{
2393 skb->inner_protocol = protocol;
2394 skb->inner_protocol_type = ENCAP_TYPE_ETHER;
2395}
2396
2397static inline void skb_set_inner_ipproto(struct sk_buff *skb,
2398 __u8 ipproto)
2399{
2400 skb->inner_ipproto = ipproto;
2401 skb->inner_protocol_type = ENCAP_TYPE_IPPROTO;
2402}
2403
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002404static inline void skb_reset_inner_headers(struct sk_buff *skb)
2405{
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002406 skb->inner_mac_header = skb->mac_header;
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002407 skb->inner_network_header = skb->network_header;
2408 skb->inner_transport_header = skb->transport_header;
2409}
2410
Jiri Pirko0b5c9db2011-06-10 06:56:58 +00002411static inline void skb_reset_mac_len(struct sk_buff *skb)
2412{
2413 skb->mac_len = skb->network_header - skb->mac_header;
2414}
2415
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002416static inline unsigned char *skb_inner_transport_header(const struct sk_buff
2417 *skb)
2418{
2419 return skb->head + skb->inner_transport_header;
2420}
2421
Tom Herbert55dc5a92015-12-14 11:19:40 -08002422static inline int skb_inner_transport_offset(const struct sk_buff *skb)
2423{
2424 return skb_inner_transport_header(skb) - skb->data;
2425}
2426
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002427static inline void skb_reset_inner_transport_header(struct sk_buff *skb)
2428{
2429 skb->inner_transport_header = skb->data - skb->head;
2430}
2431
2432static inline void skb_set_inner_transport_header(struct sk_buff *skb,
2433 const int offset)
2434{
2435 skb_reset_inner_transport_header(skb);
2436 skb->inner_transport_header += offset;
2437}
2438
2439static inline unsigned char *skb_inner_network_header(const struct sk_buff *skb)
2440{
2441 return skb->head + skb->inner_network_header;
2442}
2443
2444static inline void skb_reset_inner_network_header(struct sk_buff *skb)
2445{
2446 skb->inner_network_header = skb->data - skb->head;
2447}
2448
2449static inline void skb_set_inner_network_header(struct sk_buff *skb,
2450 const int offset)
2451{
2452 skb_reset_inner_network_header(skb);
2453 skb->inner_network_header += offset;
2454}
2455
Pravin B Shelaraefbd2b2013-03-07 13:21:46 +00002456static inline unsigned char *skb_inner_mac_header(const struct sk_buff *skb)
2457{
2458 return skb->head + skb->inner_mac_header;
2459}
2460
2461static inline void skb_reset_inner_mac_header(struct sk_buff *skb)
2462{
2463 skb->inner_mac_header = skb->data - skb->head;
2464}
2465
2466static inline void skb_set_inner_mac_header(struct sk_buff *skb,
2467 const int offset)
2468{
2469 skb_reset_inner_mac_header(skb);
2470 skb->inner_mac_header += offset;
2471}
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002472static inline bool skb_transport_header_was_set(const struct sk_buff *skb)
2473{
Cong Wang35d04612013-05-29 15:16:05 +08002474 return skb->transport_header != (typeof(skb->transport_header))~0U;
Eric Dumazetfda55ec2013-01-07 09:28:21 +00002475}
2476
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002477static inline unsigned char *skb_transport_header(const struct sk_buff *skb)
2478{
2479 return skb->head + skb->transport_header;
2480}
2481
2482static inline void skb_reset_transport_header(struct sk_buff *skb)
2483{
2484 skb->transport_header = skb->data - skb->head;
2485}
2486
2487static inline void skb_set_transport_header(struct sk_buff *skb,
2488 const int offset)
2489{
2490 skb_reset_transport_header(skb);
2491 skb->transport_header += offset;
2492}
2493
2494static inline unsigned char *skb_network_header(const struct sk_buff *skb)
2495{
2496 return skb->head + skb->network_header;
2497}
2498
2499static inline void skb_reset_network_header(struct sk_buff *skb)
2500{
2501 skb->network_header = skb->data - skb->head;
2502}
2503
2504static inline void skb_set_network_header(struct sk_buff *skb, const int offset)
2505{
2506 skb_reset_network_header(skb);
2507 skb->network_header += offset;
2508}
2509
2510static inline unsigned char *skb_mac_header(const struct sk_buff *skb)
2511{
2512 return skb->head + skb->mac_header;
2513}
2514
Amir Vadaiea6da4f2017-02-07 09:56:06 +02002515static inline int skb_mac_offset(const struct sk_buff *skb)
2516{
2517 return skb_mac_header(skb) - skb->data;
2518}
2519
Daniel Borkmann0daf4342017-07-02 02:13:25 +02002520static inline u32 skb_mac_header_len(const struct sk_buff *skb)
2521{
2522 return skb->network_header - skb->mac_header;
2523}
2524
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002525static inline int skb_mac_header_was_set(const struct sk_buff *skb)
2526{
Cong Wang35d04612013-05-29 15:16:05 +08002527 return skb->mac_header != (typeof(skb->mac_header))~0U;
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002528}
2529
2530static inline void skb_reset_mac_header(struct sk_buff *skb)
2531{
2532 skb->mac_header = skb->data - skb->head;
2533}
2534
2535static inline void skb_set_mac_header(struct sk_buff *skb, const int offset)
2536{
2537 skb_reset_mac_header(skb);
2538 skb->mac_header += offset;
2539}
2540
Timo Teräs0e3da5b2013-12-16 11:02:09 +02002541static inline void skb_pop_mac_header(struct sk_buff *skb)
2542{
2543 skb->mac_header = skb->network_header;
2544}
2545
Maxim Mikityanskiyd2aa1252019-02-21 12:39:57 +00002546static inline void skb_probe_transport_header(struct sk_buff *skb)
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002547{
Paolo Abeni72a338b2018-05-04 11:32:59 +02002548 struct flow_keys_basic keys;
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002549
2550 if (skb_transport_header_was_set(skb))
2551 return;
Paolo Abeni72a338b2018-05-04 11:32:59 +02002552
Stanislav Fomichev3cbf4ff2019-04-22 08:55:46 -07002553 if (skb_flow_dissect_flow_keys_basic(NULL, skb, &keys,
2554 NULL, 0, 0, 0, 0))
Tom Herbert42aecaa2015-06-04 09:16:39 -07002555 skb_set_transport_header(skb, keys.control.thoff);
Ying Xuefbbdb8f2013-03-27 16:46:06 +00002556}
2557
Eric Dumazet03606892012-02-23 10:55:02 +00002558static inline void skb_mac_header_rebuild(struct sk_buff *skb)
2559{
2560 if (skb_mac_header_was_set(skb)) {
2561 const unsigned char *old_mac = skb_mac_header(skb);
2562
2563 skb_set_mac_header(skb, -skb->mac_len);
2564 memmove(skb_mac_header(skb), old_mac, skb->mac_len);
2565 }
2566}
2567
Michał Mirosław04fb4512010-12-14 15:24:08 +00002568static inline int skb_checksum_start_offset(const struct sk_buff *skb)
2569{
2570 return skb->csum_start - skb_headroom(skb);
2571}
2572
Alexander Duyck08b64fc2016-02-05 15:27:49 -08002573static inline unsigned char *skb_checksum_start(const struct sk_buff *skb)
2574{
2575 return skb->head + skb->csum_start;
2576}
2577
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002578static inline int skb_transport_offset(const struct sk_buff *skb)
2579{
2580 return skb_transport_header(skb) - skb->data;
2581}
2582
2583static inline u32 skb_network_header_len(const struct sk_buff *skb)
2584{
2585 return skb->transport_header - skb->network_header;
2586}
2587
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002588static inline u32 skb_inner_network_header_len(const struct sk_buff *skb)
2589{
2590 return skb->inner_transport_header - skb->inner_network_header;
2591}
2592
Arnaldo Carvalho de Melo2e07fa92007-04-10 21:22:35 -07002593static inline int skb_network_offset(const struct sk_buff *skb)
2594{
2595 return skb_network_header(skb) - skb->data;
2596}
Arnaldo Carvalho de Melo48d49d0c2007-03-10 12:30:58 -03002597
Joseph Gasparakis6a674e92012-12-07 14:14:14 +00002598static inline int skb_inner_network_offset(const struct sk_buff *skb)
2599{
2600 return skb_inner_network_header(skb) - skb->data;
2601}
2602
Changli Gaof9599ce2010-08-04 04:43:44 +00002603static inline int pskb_network_may_pull(struct sk_buff *skb, unsigned int len)
2604{
2605 return pskb_may_pull(skb, skb_network_offset(skb) + len);
2606}
2607
Linus Torvalds1da177e2005-04-16 15:20:36 -07002608/*
2609 * CPUs often take a performance hit when accessing unaligned memory
2610 * locations. The actual performance hit varies, it can be small if the
2611 * hardware handles it or large if we have to take an exception and fix it
2612 * in software.
2613 *
2614 * Since an ethernet header is 14 bytes network drivers often end up with
2615 * the IP header at an unaligned offset. The IP header can be aligned by
2616 * shifting the start of the packet by 2 bytes. Drivers should do this
2617 * with:
2618 *
Tobias Klauser8660c122009-07-13 22:48:16 +00002619 * skb_reserve(skb, NET_IP_ALIGN);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002620 *
2621 * The downside to this alignment of the IP header is that the DMA is now
2622 * unaligned. On some architectures the cost of an unaligned DMA is high
2623 * and this cost outweighs the gains made by aligning the IP header.
Tobias Klauser8660c122009-07-13 22:48:16 +00002624 *
Linus Torvalds1da177e2005-04-16 15:20:36 -07002625 * Since this trade off varies between architectures, we allow NET_IP_ALIGN
2626 * to be overridden.
2627 */
2628#ifndef NET_IP_ALIGN
2629#define NET_IP_ALIGN 2
2630#endif
2631
Anton Blanchard025be812006-03-31 02:27:06 -08002632/*
2633 * The networking layer reserves some headroom in skb data (via
2634 * dev_alloc_skb). This is used to avoid having to reallocate skb data when
2635 * the header has to grow. In the default case, if the header has to grow
David S. Millerd6301d32009-02-08 19:24:13 -08002636 * 32 bytes or less we avoid the reallocation.
Anton Blanchard025be812006-03-31 02:27:06 -08002637 *
2638 * Unfortunately this headroom changes the DMA alignment of the resulting
2639 * network packet. As for NET_IP_ALIGN, this unaligned DMA is expensive
2640 * on some architectures. An architecture can override this value,
2641 * perhaps setting it to a cacheline in size (since that will maintain
2642 * cacheline alignment of the DMA). It must be a power of 2.
2643 *
David S. Millerd6301d32009-02-08 19:24:13 -08002644 * Various parts of the networking layer expect at least 32 bytes of
Anton Blanchard025be812006-03-31 02:27:06 -08002645 * headroom, you should not reduce this.
Eric Dumazet5933dd22010-06-15 18:16:43 -07002646 *
2647 * Using max(32, L1_CACHE_BYTES) makes sense (especially with RPS)
2648 * to reduce average number of cache lines per packet.
2649 * get_rps_cpus() for example only access one 64 bytes aligned block :
Eric Dumazet18e8c132010-05-06 21:58:51 -07002650 * NET_IP_ALIGN(2) + ethernet_header(14) + IP_header(20/40) + ports(8)
Anton Blanchard025be812006-03-31 02:27:06 -08002651 */
2652#ifndef NET_SKB_PAD
Eric Dumazet5933dd22010-06-15 18:16:43 -07002653#define NET_SKB_PAD max(32, L1_CACHE_BYTES)
Anton Blanchard025be812006-03-31 02:27:06 -08002654#endif
2655
Joe Perches7965bd42013-09-26 14:48:15 -07002656int ___pskb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002657
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002658static inline void __skb_set_length(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07002659{
Yangtao Li5e1abdc2018-11-06 10:45:36 -05002660 if (WARN_ON(skb_is_nonlinear(skb)))
Herbert Xu3cc0e872006-06-09 16:13:38 -07002661 return;
Arnaldo Carvalho de Melo27a884d2007-04-19 20:29:13 -07002662 skb->len = len;
2663 skb_set_tail_pointer(skb, len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002664}
2665
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002666static inline void __skb_trim(struct sk_buff *skb, unsigned int len)
2667{
2668 __skb_set_length(skb, len);
2669}
2670
Joe Perches7965bd42013-09-26 14:48:15 -07002671void skb_trim(struct sk_buff *skb, unsigned int len);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002672
2673static inline int __pskb_trim(struct sk_buff *skb, unsigned int len)
2674{
Herbert Xu3cc0e872006-06-09 16:13:38 -07002675 if (skb->data_len)
2676 return ___pskb_trim(skb, len);
2677 __skb_trim(skb, len);
2678 return 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07002679}
2680
2681static inline int pskb_trim(struct sk_buff *skb, unsigned int len)
2682{
2683 return (len < skb->len) ? __pskb_trim(skb, len) : 0;
2684}
2685
2686/**
Herbert Xue9fa4f72006-08-13 20:12:58 -07002687 * pskb_trim_unique - remove end from a paged unique (not cloned) buffer
2688 * @skb: buffer to alter
2689 * @len: new length
2690 *
2691 * This is identical to pskb_trim except that the caller knows that
2692 * the skb is not cloned so we should never get an error due to out-
2693 * of-memory.
2694 */
2695static inline void pskb_trim_unique(struct sk_buff *skb, unsigned int len)
2696{
2697 int err = pskb_trim(skb, len);
2698 BUG_ON(err);
2699}
2700
Daniel Borkmann5293efe2016-08-18 01:00:39 +02002701static inline int __skb_grow(struct sk_buff *skb, unsigned int len)
2702{
2703 unsigned int diff = len - skb->len;
2704
2705 if (skb_tailroom(skb) < diff) {
2706 int ret = pskb_expand_head(skb, 0, diff - skb_tailroom(skb),
2707 GFP_ATOMIC);
2708 if (ret)
2709 return ret;
2710 }
2711 __skb_set_length(skb, len);
2712 return 0;
2713}
2714
Herbert Xue9fa4f72006-08-13 20:12:58 -07002715/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002716 * skb_orphan - orphan a buffer
2717 * @skb: buffer to orphan
2718 *
2719 * If a buffer currently has an owner then we call the owner's
2720 * destructor function and make the @skb unowned. The buffer continues
2721 * to exist but is no longer charged to its former owner.
2722 */
2723static inline void skb_orphan(struct sk_buff *skb)
2724{
Eric Dumazetc34a7612013-07-30 16:11:15 -07002725 if (skb->destructor) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07002726 skb->destructor(skb);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002727 skb->destructor = NULL;
2728 skb->sk = NULL;
Eric Dumazet376c7312013-08-01 11:43:08 -07002729 } else {
2730 BUG_ON(skb->sk);
Eric Dumazetc34a7612013-07-30 16:11:15 -07002731 }
Linus Torvalds1da177e2005-04-16 15:20:36 -07002732}
2733
2734/**
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002735 * skb_orphan_frags - orphan the frags contained in a buffer
2736 * @skb: buffer to orphan frags from
2737 * @gfp_mask: allocation mask for replacement pages
2738 *
2739 * For each frag in the SKB which needs a destructor (i.e. has an
2740 * owner) create a copy of that frag and release the original
2741 * page by calling the destructor.
2742 */
2743static inline int skb_orphan_frags(struct sk_buff *skb, gfp_t gfp_mask)
2744{
Willem de Bruijn1f8b9772017-08-03 16:29:41 -04002745 if (likely(!skb_zcopy(skb)))
2746 return 0;
Willem de Bruijn185ce5c2019-05-15 13:29:16 -04002747 if (!skb_zcopy_is_nouarg(skb) &&
2748 skb_uarg(skb)->callback == sock_zerocopy_callback)
Willem de Bruijn1f8b9772017-08-03 16:29:41 -04002749 return 0;
2750 return skb_copy_ubufs(skb, gfp_mask);
2751}
2752
2753/* Frags must be orphaned, even if refcounted, if skb might loop to rx path */
2754static inline int skb_orphan_frags_rx(struct sk_buff *skb, gfp_t gfp_mask)
2755{
2756 if (likely(!skb_zcopy(skb)))
Michael S. Tsirkina353e0c2012-07-20 09:23:07 +00002757 return 0;
2758 return skb_copy_ubufs(skb, gfp_mask);
2759}
2760
2761/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07002762 * __skb_queue_purge - empty a list
2763 * @list: list to empty
2764 *
2765 * Delete all buffers on an &sk_buff list. Each buffer is removed from
2766 * the list and one reference dropped. This function does not take the
2767 * list lock and the caller must hold the relevant locks to use it.
2768 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07002769static inline void __skb_queue_purge(struct sk_buff_head *list)
2770{
2771 struct sk_buff *skb;
2772 while ((skb = __skb_dequeue(list)) != NULL)
2773 kfree_skb(skb);
2774}
Brian Norris4ea7b0c2019-02-11 13:02:25 -08002775void skb_queue_purge(struct sk_buff_head *list);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002776
Peter Oskolkov385114d2018-08-02 23:34:38 +00002777unsigned int skb_rbtree_purge(struct rb_root *root);
Yaogong Wang9f5afea2016-09-07 14:49:28 -07002778
Joe Perches7965bd42013-09-26 14:48:15 -07002779void *netdev_alloc_frag(unsigned int fragsz);
Linus Torvalds1da177e2005-04-16 15:20:36 -07002780
Joe Perches7965bd42013-09-26 14:48:15 -07002781struct sk_buff *__netdev_alloc_skb(struct net_device *dev, unsigned int length,
2782 gfp_t gfp_mask);
Christoph Hellwig8af27452006-07-31 22:35:23 -07002783
2784/**
2785 * netdev_alloc_skb - allocate an skbuff for rx on a specific device
2786 * @dev: network device to receive on
2787 * @length: length to allocate
2788 *
2789 * Allocate a new &sk_buff and assign it a usage count of one. The
2790 * buffer has unspecified headroom built in. Users should allocate
2791 * the headroom they think they need without accounting for the
2792 * built in space. The built in space is used for optimisations.
2793 *
2794 * %NULL is returned if there is no free memory. Although this function
2795 * allocates memory it can be called from an interrupt.
2796 */
2797static inline struct sk_buff *netdev_alloc_skb(struct net_device *dev,
Eric Dumazet6f532612012-05-18 05:12:12 +00002798 unsigned int length)
Christoph Hellwig8af27452006-07-31 22:35:23 -07002799{
2800 return __netdev_alloc_skb(dev, length, GFP_ATOMIC);
2801}
2802
Eric Dumazet6f532612012-05-18 05:12:12 +00002803/* legacy helper around __netdev_alloc_skb() */
2804static inline struct sk_buff *__dev_alloc_skb(unsigned int length,
2805 gfp_t gfp_mask)
2806{
2807 return __netdev_alloc_skb(NULL, length, gfp_mask);
2808}
2809
2810/* legacy helper around netdev_alloc_skb() */
2811static inline struct sk_buff *dev_alloc_skb(unsigned int length)
2812{
2813 return netdev_alloc_skb(NULL, length);
2814}
2815
2816
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002817static inline struct sk_buff *__netdev_alloc_skb_ip_align(struct net_device *dev,
2818 unsigned int length, gfp_t gfp)
Eric Dumazet61321bb2009-10-07 17:11:23 +00002819{
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002820 struct sk_buff *skb = __netdev_alloc_skb(dev, length + NET_IP_ALIGN, gfp);
Eric Dumazet61321bb2009-10-07 17:11:23 +00002821
2822 if (NET_IP_ALIGN && skb)
2823 skb_reserve(skb, NET_IP_ALIGN);
2824 return skb;
2825}
2826
Eric Dumazet4915a0d2011-07-11 20:08:34 -07002827static inline struct sk_buff *netdev_alloc_skb_ip_align(struct net_device *dev,
2828 unsigned int length)
2829{
2830 return __netdev_alloc_skb_ip_align(dev, length, GFP_ATOMIC);
2831}
2832
Alexander Duyck181edb22015-05-06 21:12:03 -07002833static inline void skb_free_frag(void *addr)
2834{
Alexander Duyck8c2dd3e2017-01-10 16:58:06 -08002835 page_frag_free(addr);
Alexander Duyck181edb22015-05-06 21:12:03 -07002836}
2837
Alexander Duyckffde7322014-12-09 19:40:42 -08002838void *napi_alloc_frag(unsigned int fragsz);
Alexander Duyckfd11a832014-12-09 19:40:49 -08002839struct sk_buff *__napi_alloc_skb(struct napi_struct *napi,
2840 unsigned int length, gfp_t gfp_mask);
2841static inline struct sk_buff *napi_alloc_skb(struct napi_struct *napi,
2842 unsigned int length)
2843{
2844 return __napi_alloc_skb(napi, length, GFP_ATOMIC);
2845}
Jesper Dangaard Brouer795bb1c2016-02-08 13:14:59 +01002846void napi_consume_skb(struct sk_buff *skb, int budget);
2847
2848void __kfree_skb_flush(void);
Jesper Dangaard Brouer15fad712016-02-08 13:15:04 +01002849void __kfree_skb_defer(struct sk_buff *skb);
Alexander Duyckffde7322014-12-09 19:40:42 -08002850
Florian Fainellibc6fc9f2013-08-30 15:36:14 +01002851/**
Alexander Duyck71dfda52014-11-11 09:26:34 -08002852 * __dev_alloc_pages - allocate page for network Rx
2853 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2854 * @order: size of the allocation
2855 *
2856 * Allocate a new page.
2857 *
2858 * %NULL is returned if there is no free memory.
2859*/
2860static inline struct page *__dev_alloc_pages(gfp_t gfp_mask,
2861 unsigned int order)
2862{
2863 /* This piece of code contains several assumptions.
2864 * 1. This is for device Rx, therefor a cold page is preferred.
2865 * 2. The expectation is the user wants a compound page.
2866 * 3. If requesting a order 0 page it will not be compound
2867 * due to the check to see if order has a value in prep_new_page
2868 * 4. __GFP_MEMALLOC is ignored if __GFP_NOMEMALLOC is set due to
2869 * code in gfp_to_alloc_flags that should be enforcing this.
2870 */
Mel Gorman453f85d2017-11-15 17:38:03 -08002871 gfp_mask |= __GFP_COMP | __GFP_MEMALLOC;
Alexander Duyck71dfda52014-11-11 09:26:34 -08002872
2873 return alloc_pages_node(NUMA_NO_NODE, gfp_mask, order);
2874}
2875
2876static inline struct page *dev_alloc_pages(unsigned int order)
2877{
Neil Horman95829b32016-05-19 11:30:54 -04002878 return __dev_alloc_pages(GFP_ATOMIC | __GFP_NOWARN, order);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002879}
2880
2881/**
2882 * __dev_alloc_page - allocate a page for network Rx
2883 * @gfp_mask: allocation priority. Set __GFP_NOMEMALLOC if not for network Rx
2884 *
2885 * Allocate a new page.
2886 *
2887 * %NULL is returned if there is no free memory.
2888 */
2889static inline struct page *__dev_alloc_page(gfp_t gfp_mask)
2890{
2891 return __dev_alloc_pages(gfp_mask, 0);
2892}
2893
2894static inline struct page *dev_alloc_page(void)
2895{
Neil Horman95829b32016-05-19 11:30:54 -04002896 return dev_alloc_pages(0);
Alexander Duyck71dfda52014-11-11 09:26:34 -08002897}
2898
2899/**
Mel Gorman06140022012-07-31 16:44:24 -07002900 * skb_propagate_pfmemalloc - Propagate pfmemalloc if skb is allocated after RX page
2901 * @page: The page that was allocated from skb_alloc_page
2902 * @skb: The skb that may need pfmemalloc set
2903 */
2904static inline void skb_propagate_pfmemalloc(struct page *page,
2905 struct sk_buff *skb)
2906{
Michal Hocko2f064f32015-08-21 14:11:51 -07002907 if (page_is_pfmemalloc(page))
Mel Gorman06140022012-07-31 16:44:24 -07002908 skb->pfmemalloc = true;
2909}
2910
Eric Dumazet564824b2010-10-11 19:05:25 +00002911/**
Jonathan Lemon7240b602019-07-30 07:40:32 -07002912 * skb_frag_off() - Returns the offset of a skb fragment
2913 * @frag: the paged fragment
2914 */
2915static inline unsigned int skb_frag_off(const skb_frag_t *frag)
2916{
Jonathan Lemon65c84f12019-07-30 07:40:34 -07002917 return frag->bv_offset;
Jonathan Lemon7240b602019-07-30 07:40:32 -07002918}
2919
2920/**
2921 * skb_frag_off_add() - Increments the offset of a skb fragment by @delta
2922 * @frag: skb fragment
2923 * @delta: value to add
2924 */
2925static inline void skb_frag_off_add(skb_frag_t *frag, int delta)
2926{
Jonathan Lemon65c84f12019-07-30 07:40:34 -07002927 frag->bv_offset += delta;
Jonathan Lemon7240b602019-07-30 07:40:32 -07002928}
2929
2930/**
2931 * skb_frag_off_set() - Sets the offset of a skb fragment
2932 * @frag: skb fragment
2933 * @offset: offset of fragment
2934 */
2935static inline void skb_frag_off_set(skb_frag_t *frag, unsigned int offset)
2936{
Jonathan Lemon65c84f12019-07-30 07:40:34 -07002937 frag->bv_offset = offset;
Jonathan Lemon7240b602019-07-30 07:40:32 -07002938}
2939
2940/**
2941 * skb_frag_off_copy() - Sets the offset of a skb fragment from another fragment
2942 * @fragto: skb fragment where offset is set
2943 * @fragfrom: skb fragment offset is copied from
2944 */
2945static inline void skb_frag_off_copy(skb_frag_t *fragto,
2946 const skb_frag_t *fragfrom)
2947{
Jonathan Lemon65c84f12019-07-30 07:40:34 -07002948 fragto->bv_offset = fragfrom->bv_offset;
Jonathan Lemon7240b602019-07-30 07:40:32 -07002949}
2950
2951/**
Masanari Iidae2278672014-02-18 22:54:36 +09002952 * skb_frag_page - retrieve the page referred to by a paged fragment
Ian Campbell131ea662011-08-19 06:25:00 +00002953 * @frag: the paged fragment
2954 *
2955 * Returns the &struct page associated with @frag.
2956 */
2957static inline struct page *skb_frag_page(const skb_frag_t *frag)
2958{
Matthew Wilcox (Oracle)1dfa5bd2019-07-22 20:08:29 -07002959 return frag->bv_page;
Ian Campbell131ea662011-08-19 06:25:00 +00002960}
2961
2962/**
2963 * __skb_frag_ref - take an addition reference on a paged fragment.
2964 * @frag: the paged fragment
2965 *
2966 * Takes an additional reference on the paged fragment @frag.
2967 */
2968static inline void __skb_frag_ref(skb_frag_t *frag)
2969{
2970 get_page(skb_frag_page(frag));
2971}
2972
2973/**
2974 * skb_frag_ref - take an addition reference on a paged fragment of an skb.
2975 * @skb: the buffer
2976 * @f: the fragment offset.
2977 *
2978 * Takes an additional reference on the @f'th paged fragment of @skb.
2979 */
2980static inline void skb_frag_ref(struct sk_buff *skb, int f)
2981{
2982 __skb_frag_ref(&skb_shinfo(skb)->frags[f]);
2983}
2984
2985/**
2986 * __skb_frag_unref - release a reference on a paged fragment.
2987 * @frag: the paged fragment
2988 *
2989 * Releases a reference on the paged fragment @frag.
2990 */
2991static inline void __skb_frag_unref(skb_frag_t *frag)
2992{
2993 put_page(skb_frag_page(frag));
2994}
2995
2996/**
2997 * skb_frag_unref - release a reference on a paged fragment of an skb.
2998 * @skb: the buffer
2999 * @f: the fragment offset
3000 *
3001 * Releases a reference on the @f'th paged fragment of @skb.
3002 */
3003static inline void skb_frag_unref(struct sk_buff *skb, int f)
3004{
3005 __skb_frag_unref(&skb_shinfo(skb)->frags[f]);
3006}
3007
3008/**
3009 * skb_frag_address - gets the address of the data contained in a paged fragment
3010 * @frag: the paged fragment buffer
3011 *
3012 * Returns the address of the data within @frag. The page must already
3013 * be mapped.
3014 */
3015static inline void *skb_frag_address(const skb_frag_t *frag)
3016{
Jonathan Lemon7240b602019-07-30 07:40:32 -07003017 return page_address(skb_frag_page(frag)) + skb_frag_off(frag);
Ian Campbell131ea662011-08-19 06:25:00 +00003018}
3019
3020/**
3021 * skb_frag_address_safe - gets the address of the data contained in a paged fragment
3022 * @frag: the paged fragment buffer
3023 *
3024 * Returns the address of the data within @frag. Checks that the page
3025 * is mapped and returns %NULL otherwise.
3026 */
3027static inline void *skb_frag_address_safe(const skb_frag_t *frag)
3028{
3029 void *ptr = page_address(skb_frag_page(frag));
3030 if (unlikely(!ptr))
3031 return NULL;
3032
Jonathan Lemon7240b602019-07-30 07:40:32 -07003033 return ptr + skb_frag_off(frag);
3034}
3035
3036/**
3037 * skb_frag_page_copy() - sets the page in a fragment from another fragment
3038 * @fragto: skb fragment where page is set
3039 * @fragfrom: skb fragment page is copied from
3040 */
3041static inline void skb_frag_page_copy(skb_frag_t *fragto,
3042 const skb_frag_t *fragfrom)
3043{
3044 fragto->bv_page = fragfrom->bv_page;
Ian Campbell131ea662011-08-19 06:25:00 +00003045}
3046
3047/**
3048 * __skb_frag_set_page - sets the page contained in a paged fragment
3049 * @frag: the paged fragment
3050 * @page: the page to set
3051 *
3052 * Sets the fragment @frag to contain @page.
3053 */
3054static inline void __skb_frag_set_page(skb_frag_t *frag, struct page *page)
3055{
Matthew Wilcox (Oracle)1dfa5bd2019-07-22 20:08:29 -07003056 frag->bv_page = page;
Ian Campbell131ea662011-08-19 06:25:00 +00003057}
3058
3059/**
3060 * skb_frag_set_page - sets the page contained in a paged fragment of an skb
3061 * @skb: the buffer
3062 * @f: the fragment offset
3063 * @page: the page to set
3064 *
3065 * Sets the @f'th fragment of @skb to contain @page.
3066 */
3067static inline void skb_frag_set_page(struct sk_buff *skb, int f,
3068 struct page *page)
3069{
3070 __skb_frag_set_page(&skb_shinfo(skb)->frags[f], page);
3071}
3072
Eric Dumazet400dfd32013-10-17 16:27:07 -07003073bool skb_page_frag_refill(unsigned int sz, struct page_frag *pfrag, gfp_t prio);
3074
Ian Campbell131ea662011-08-19 06:25:00 +00003075/**
3076 * skb_frag_dma_map - maps a paged fragment via the DMA API
Marcos Paulo de Souzaf83347d2011-10-31 15:11:45 +00003077 * @dev: the device to map the fragment to
Ian Campbell131ea662011-08-19 06:25:00 +00003078 * @frag: the paged fragment to map
3079 * @offset: the offset within the fragment (starting at the
3080 * fragment's own offset)
3081 * @size: the number of bytes to map
Mauro Carvalho Chehab771b00a2017-05-12 09:19:29 -03003082 * @dir: the direction of the mapping (``PCI_DMA_*``)
Ian Campbell131ea662011-08-19 06:25:00 +00003083 *
3084 * Maps the page associated with @frag to @device.
3085 */
3086static inline dma_addr_t skb_frag_dma_map(struct device *dev,
3087 const skb_frag_t *frag,
3088 size_t offset, size_t size,
3089 enum dma_data_direction dir)
3090{
3091 return dma_map_page(dev, skb_frag_page(frag),
Jonathan Lemon7240b602019-07-30 07:40:32 -07003092 skb_frag_off(frag) + offset, size, dir);
Ian Campbell131ea662011-08-19 06:25:00 +00003093}
3094
Eric Dumazet117632e2011-12-03 21:39:53 +00003095static inline struct sk_buff *pskb_copy(struct sk_buff *skb,
3096 gfp_t gfp_mask)
3097{
3098 return __pskb_copy(skb, skb_headroom(skb), gfp_mask);
3099}
3100
Octavian Purdilabad93e92014-06-12 01:36:26 +03003101
3102static inline struct sk_buff *pskb_copy_for_clone(struct sk_buff *skb,
3103 gfp_t gfp_mask)
3104{
3105 return __pskb_copy_fclone(skb, skb_headroom(skb), gfp_mask, true);
3106}
3107
3108
Ian Campbell131ea662011-08-19 06:25:00 +00003109/**
Patrick McHardy334a8132007-06-25 04:35:20 -07003110 * skb_clone_writable - is the header of a clone writable
3111 * @skb: buffer to check
3112 * @len: length up to which to write
3113 *
3114 * Returns true if modifying the header part of the cloned buffer
3115 * does not requires the data to be copied.
3116 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04003117static inline int skb_clone_writable(const struct sk_buff *skb, unsigned int len)
Patrick McHardy334a8132007-06-25 04:35:20 -07003118{
3119 return !skb_header_cloned(skb) &&
3120 skb_headroom(skb) + len <= skb->hdr_len;
3121}
3122
Daniel Borkmann36976492016-02-19 23:05:25 +01003123static inline int skb_try_make_writable(struct sk_buff *skb,
3124 unsigned int write_len)
3125{
3126 return skb_cloned(skb) && !skb_clone_writable(skb, write_len) &&
3127 pskb_expand_head(skb, 0, 0, GFP_ATOMIC);
3128}
3129
Herbert Xud9cc2042007-09-16 16:21:16 -07003130static inline int __skb_cow(struct sk_buff *skb, unsigned int headroom,
3131 int cloned)
3132{
3133 int delta = 0;
3134
Herbert Xud9cc2042007-09-16 16:21:16 -07003135 if (headroom > skb_headroom(skb))
3136 delta = headroom - skb_headroom(skb);
3137
3138 if (delta || cloned)
3139 return pskb_expand_head(skb, ALIGN(delta, NET_SKB_PAD), 0,
3140 GFP_ATOMIC);
3141 return 0;
3142}
3143
Patrick McHardy334a8132007-06-25 04:35:20 -07003144/**
Linus Torvalds1da177e2005-04-16 15:20:36 -07003145 * skb_cow - copy header of skb when it is required
3146 * @skb: buffer to cow
3147 * @headroom: needed headroom
3148 *
3149 * If the skb passed lacks sufficient headroom or its data part
3150 * is shared, data is reallocated. If reallocation fails, an error
3151 * is returned and original skb is not changed.
3152 *
3153 * The result is skb with writable area skb->head...skb->tail
3154 * and at least @headroom of space at head.
3155 */
3156static inline int skb_cow(struct sk_buff *skb, unsigned int headroom)
3157{
Herbert Xud9cc2042007-09-16 16:21:16 -07003158 return __skb_cow(skb, headroom, skb_cloned(skb));
3159}
Linus Torvalds1da177e2005-04-16 15:20:36 -07003160
Herbert Xud9cc2042007-09-16 16:21:16 -07003161/**
3162 * skb_cow_head - skb_cow but only making the head writable
3163 * @skb: buffer to cow
3164 * @headroom: needed headroom
3165 *
3166 * This function is identical to skb_cow except that we replace the
3167 * skb_cloned check by skb_header_cloned. It should be used when
3168 * you only need to push on some header and do not need to modify
3169 * the data.
3170 */
3171static inline int skb_cow_head(struct sk_buff *skb, unsigned int headroom)
3172{
3173 return __skb_cow(skb, headroom, skb_header_cloned(skb));
Linus Torvalds1da177e2005-04-16 15:20:36 -07003174}
3175
3176/**
3177 * skb_padto - pad an skbuff up to a minimal size
3178 * @skb: buffer to pad
3179 * @len: minimal length
3180 *
3181 * Pads up a buffer to ensure the trailing bytes exist and are
3182 * blanked. If the buffer already contains sufficient data it
Herbert Xu5b057c62006-06-23 02:06:41 -07003183 * is untouched. Otherwise it is extended. Returns zero on
3184 * success. The skb is freed on error.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003185 */
Herbert Xu5b057c62006-06-23 02:06:41 -07003186static inline int skb_padto(struct sk_buff *skb, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003187{
3188 unsigned int size = skb->len;
3189 if (likely(size >= len))
Herbert Xu5b057c62006-06-23 02:06:41 -07003190 return 0;
Gerrit Renker987c4022008-08-11 18:17:17 -07003191 return skb_pad(skb, len - size);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003192}
3193
Alexander Duyck9c0c1122014-12-03 08:17:33 -08003194/**
Brian Norris4ea7b0c2019-02-11 13:02:25 -08003195 * __skb_put_padto - increase size and pad an skbuff up to a minimal size
Alexander Duyck9c0c1122014-12-03 08:17:33 -08003196 * @skb: buffer to pad
3197 * @len: minimal length
Florian Fainellicd0a1372017-08-22 15:12:14 -07003198 * @free_on_error: free buffer on error
3199 *
3200 * Pads up a buffer to ensure the trailing bytes exist and are
3201 * blanked. If the buffer already contains sufficient data it
3202 * is untouched. Otherwise it is extended. Returns zero on
3203 * success. The skb is freed on error if @free_on_error is true.
3204 */
3205static inline int __skb_put_padto(struct sk_buff *skb, unsigned int len,
3206 bool free_on_error)
3207{
3208 unsigned int size = skb->len;
3209
3210 if (unlikely(size < len)) {
3211 len -= size;
3212 if (__skb_pad(skb, len, free_on_error))
3213 return -ENOMEM;
3214 __skb_put(skb, len);
3215 }
3216 return 0;
3217}
3218
3219/**
3220 * skb_put_padto - increase size and pad an skbuff up to a minimal size
3221 * @skb: buffer to pad
3222 * @len: minimal length
Alexander Duyck9c0c1122014-12-03 08:17:33 -08003223 *
3224 * Pads up a buffer to ensure the trailing bytes exist and are
3225 * blanked. If the buffer already contains sufficient data it
3226 * is untouched. Otherwise it is extended. Returns zero on
3227 * success. The skb is freed on error.
3228 */
3229static inline int skb_put_padto(struct sk_buff *skb, unsigned int len)
3230{
Florian Fainellicd0a1372017-08-22 15:12:14 -07003231 return __skb_put_padto(skb, len, true);
Alexander Duyck9c0c1122014-12-03 08:17:33 -08003232}
3233
Linus Torvalds1da177e2005-04-16 15:20:36 -07003234static inline int skb_add_data(struct sk_buff *skb,
Al Viroaf2b0402014-11-27 21:44:24 -05003235 struct iov_iter *from, int copy)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003236{
3237 const int off = skb->len;
3238
3239 if (skb->ip_summed == CHECKSUM_NONE) {
Al Viroaf2b0402014-11-27 21:44:24 -05003240 __wsum csum = 0;
Al Viro15e6cb42016-11-01 22:42:45 -04003241 if (csum_and_copy_from_iter_full(skb_put(skb, copy), copy,
3242 &csum, from)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -07003243 skb->csum = csum_block_add(skb->csum, csum, off);
3244 return 0;
3245 }
Al Viro15e6cb42016-11-01 22:42:45 -04003246 } else if (copy_from_iter_full(skb_put(skb, copy), copy, from))
Linus Torvalds1da177e2005-04-16 15:20:36 -07003247 return 0;
3248
3249 __skb_trim(skb, off);
3250 return -EFAULT;
3251}
3252
Eric Dumazet38ba0a62012-04-23 17:48:27 +00003253static inline bool skb_can_coalesce(struct sk_buff *skb, int i,
3254 const struct page *page, int off)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003255{
Willem de Bruijn1f8b9772017-08-03 16:29:41 -04003256 if (skb_zcopy(skb))
3257 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003258 if (i) {
Matthew Wilcox (Oracle)d8e18a52019-07-22 20:08:26 -07003259 const skb_frag_t *frag = &skb_shinfo(skb)->frags[i - 1];
Linus Torvalds1da177e2005-04-16 15:20:36 -07003260
Ian Campbellea2ab692011-08-22 23:44:58 +00003261 return page == skb_frag_page(frag) &&
Jonathan Lemon7240b602019-07-30 07:40:32 -07003262 off == skb_frag_off(frag) + skb_frag_size(frag);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003263 }
Eric Dumazet38ba0a62012-04-23 17:48:27 +00003264 return false;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003265}
3266
Herbert Xu364c6ba2006-06-09 16:10:40 -07003267static inline int __skb_linearize(struct sk_buff *skb)
3268{
3269 return __pskb_pull_tail(skb, skb->data_len) ? 0 : -ENOMEM;
3270}
3271
Linus Torvalds1da177e2005-04-16 15:20:36 -07003272/**
3273 * skb_linearize - convert paged skb to linear one
3274 * @skb: buffer to linarize
Linus Torvalds1da177e2005-04-16 15:20:36 -07003275 *
3276 * If there is no free memory -ENOMEM is returned, otherwise zero
3277 * is returned and the old skb data released.
3278 */
Herbert Xu364c6ba2006-06-09 16:10:40 -07003279static inline int skb_linearize(struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003280{
Herbert Xu364c6ba2006-06-09 16:10:40 -07003281 return skb_is_nonlinear(skb) ? __skb_linearize(skb) : 0;
3282}
3283
3284/**
Eric Dumazetcef401d2013-01-25 20:34:37 +00003285 * skb_has_shared_frag - can any frag be overwritten
3286 * @skb: buffer to test
3287 *
3288 * Return true if the skb has at least one frag that might be modified
3289 * by an external entity (as in vmsplice()/sendfile())
3290 */
3291static inline bool skb_has_shared_frag(const struct sk_buff *skb)
3292{
Pravin B Shelarc9af6db2013-02-11 09:27:41 +00003293 return skb_is_nonlinear(skb) &&
3294 skb_shinfo(skb)->tx_flags & SKBTX_SHARED_FRAG;
Eric Dumazetcef401d2013-01-25 20:34:37 +00003295}
3296
3297/**
Herbert Xu364c6ba2006-06-09 16:10:40 -07003298 * skb_linearize_cow - make sure skb is linear and writable
3299 * @skb: buffer to process
3300 *
3301 * If there is no free memory -ENOMEM is returned, otherwise zero
3302 * is returned and the old skb data released.
3303 */
3304static inline int skb_linearize_cow(struct sk_buff *skb)
3305{
3306 return skb_is_nonlinear(skb) || skb_cloned(skb) ?
3307 __skb_linearize(skb) : 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003308}
3309
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003310static __always_inline void
3311__skb_postpull_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3312 unsigned int off)
3313{
3314 if (skb->ip_summed == CHECKSUM_COMPLETE)
3315 skb->csum = csum_block_sub(skb->csum,
3316 csum_partial(start, len, 0), off);
3317 else if (skb->ip_summed == CHECKSUM_PARTIAL &&
3318 skb_checksum_start_offset(skb) < 0)
3319 skb->ip_summed = CHECKSUM_NONE;
3320}
3321
Linus Torvalds1da177e2005-04-16 15:20:36 -07003322/**
3323 * skb_postpull_rcsum - update checksum for received skb after pull
3324 * @skb: buffer to update
3325 * @start: start of data before pull
3326 * @len: length of data pulled
3327 *
3328 * After doing a pull on a received packet, you need to call this to
Patrick McHardy84fa7932006-08-29 16:44:56 -07003329 * update the CHECKSUM_COMPLETE checksum, or set ip_summed to
3330 * CHECKSUM_NONE so that it can be recomputed from scratch.
Linus Torvalds1da177e2005-04-16 15:20:36 -07003331 */
Linus Torvalds1da177e2005-04-16 15:20:36 -07003332static inline void skb_postpull_rcsum(struct sk_buff *skb,
Herbert Xucbb042f2006-03-20 22:43:56 -08003333 const void *start, unsigned int len)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003334{
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003335 __skb_postpull_rcsum(skb, start, len, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003336}
3337
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003338static __always_inline void
3339__skb_postpush_rcsum(struct sk_buff *skb, const void *start, unsigned int len,
3340 unsigned int off)
3341{
3342 if (skb->ip_summed == CHECKSUM_COMPLETE)
3343 skb->csum = csum_block_add(skb->csum,
3344 csum_partial(start, len, 0), off);
3345}
Herbert Xucbb042f2006-03-20 22:43:56 -08003346
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003347/**
3348 * skb_postpush_rcsum - update checksum for received skb after push
3349 * @skb: buffer to update
3350 * @start: start of data after push
3351 * @len: length of data pushed
3352 *
3353 * After doing a push on a received packet, you need to call this to
3354 * update the CHECKSUM_COMPLETE checksum.
3355 */
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01003356static inline void skb_postpush_rcsum(struct sk_buff *skb,
3357 const void *start, unsigned int len)
3358{
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003359 __skb_postpush_rcsum(skb, start, len, 0);
Daniel Borkmannf8ffad62016-01-07 15:50:23 +01003360}
3361
Johannes Bergaf728682017-06-16 14:29:22 +02003362void *skb_pull_rcsum(struct sk_buff *skb, unsigned int len);
Daniel Borkmann479ffccc2016-08-05 00:11:12 +02003363
David S. Miller7ce5a272013-12-02 17:26:05 -05003364/**
WANG Cong82a31b92016-06-30 10:15:22 -07003365 * skb_push_rcsum - push skb and update receive checksum
3366 * @skb: buffer to update
3367 * @len: length of data pulled
3368 *
3369 * This function performs an skb_push on the packet and updates
3370 * the CHECKSUM_COMPLETE checksum. It should be used on
3371 * receive path processing instead of skb_push unless you know
3372 * that the checksum difference is zero (e.g., a valid IP header)
3373 * or you are setting ip_summed to CHECKSUM_NONE.
3374 */
Johannes Bergd58ff352017-06-16 14:29:23 +02003375static inline void *skb_push_rcsum(struct sk_buff *skb, unsigned int len)
WANG Cong82a31b92016-06-30 10:15:22 -07003376{
3377 skb_push(skb, len);
3378 skb_postpush_rcsum(skb, skb->data, len);
3379 return skb->data;
3380}
3381
Eric Dumazet88078d92018-04-18 11:43:15 -07003382int pskb_trim_rcsum_slow(struct sk_buff *skb, unsigned int len);
WANG Cong82a31b92016-06-30 10:15:22 -07003383/**
David S. Miller7ce5a272013-12-02 17:26:05 -05003384 * pskb_trim_rcsum - trim received skb and update checksum
3385 * @skb: buffer to trim
3386 * @len: new length
3387 *
3388 * This is exactly the same as pskb_trim except that it ensures the
3389 * checksum of received packets are still valid after the operation.
Ross Lagerwall6c57f042019-01-17 15:34:38 +00003390 * It can change skb pointers.
David S. Miller7ce5a272013-12-02 17:26:05 -05003391 */
3392
3393static inline int pskb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3394{
3395 if (likely(len >= skb->len))
3396 return 0;
Eric Dumazet88078d92018-04-18 11:43:15 -07003397 return pskb_trim_rcsum_slow(skb, len);
David S. Miller7ce5a272013-12-02 17:26:05 -05003398}
3399
Daniel Borkmann5293efe2016-08-18 01:00:39 +02003400static inline int __skb_trim_rcsum(struct sk_buff *skb, unsigned int len)
3401{
3402 if (skb->ip_summed == CHECKSUM_COMPLETE)
3403 skb->ip_summed = CHECKSUM_NONE;
3404 __skb_trim(skb, len);
3405 return 0;
3406}
3407
3408static inline int __skb_grow_rcsum(struct sk_buff *skb, unsigned int len)
3409{
3410 if (skb->ip_summed == CHECKSUM_COMPLETE)
3411 skb->ip_summed = CHECKSUM_NONE;
3412 return __skb_grow(skb, len);
3413}
3414
Eric Dumazet18a4c0e2017-10-05 22:21:21 -07003415#define rb_to_skb(rb) rb_entry_safe(rb, struct sk_buff, rbnode)
3416#define skb_rb_first(root) rb_to_skb(rb_first(root))
3417#define skb_rb_last(root) rb_to_skb(rb_last(root))
3418#define skb_rb_next(skb) rb_to_skb(rb_next(&(skb)->rbnode))
3419#define skb_rb_prev(skb) rb_to_skb(rb_prev(&(skb)->rbnode))
3420
Linus Torvalds1da177e2005-04-16 15:20:36 -07003421#define skb_queue_walk(queue, skb) \
3422 for (skb = (queue)->next; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003423 skb != (struct sk_buff *)(queue); \
Linus Torvalds1da177e2005-04-16 15:20:36 -07003424 skb = skb->next)
3425
James Chapman46f89142007-04-30 00:07:31 -07003426#define skb_queue_walk_safe(queue, skb, tmp) \
3427 for (skb = (queue)->next, tmp = skb->next; \
3428 skb != (struct sk_buff *)(queue); \
3429 skb = tmp, tmp = skb->next)
3430
David S. Miller1164f522008-09-23 00:49:44 -07003431#define skb_queue_walk_from(queue, skb) \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003432 for (; skb != (struct sk_buff *)(queue); \
David S. Miller1164f522008-09-23 00:49:44 -07003433 skb = skb->next)
3434
Eric Dumazet18a4c0e2017-10-05 22:21:21 -07003435#define skb_rbtree_walk(skb, root) \
3436 for (skb = skb_rb_first(root); skb != NULL; \
3437 skb = skb_rb_next(skb))
3438
3439#define skb_rbtree_walk_from(skb) \
3440 for (; skb != NULL; \
3441 skb = skb_rb_next(skb))
3442
3443#define skb_rbtree_walk_from_safe(skb, tmp) \
3444 for (; tmp = skb ? skb_rb_next(skb) : NULL, (skb != NULL); \
3445 skb = tmp)
3446
David S. Miller1164f522008-09-23 00:49:44 -07003447#define skb_queue_walk_from_safe(queue, skb, tmp) \
3448 for (tmp = skb->next; \
3449 skb != (struct sk_buff *)(queue); \
3450 skb = tmp, tmp = skb->next)
3451
Stephen Hemminger300ce172005-10-30 13:47:34 -08003452#define skb_queue_reverse_walk(queue, skb) \
3453 for (skb = (queue)->prev; \
Linus Torvaldsa1e48912011-05-22 16:51:43 -07003454 skb != (struct sk_buff *)(queue); \
Stephen Hemminger300ce172005-10-30 13:47:34 -08003455 skb = skb->prev)
3456
David S. Miller686a2952011-01-20 22:47:32 -08003457#define skb_queue_reverse_walk_safe(queue, skb, tmp) \
3458 for (skb = (queue)->prev, tmp = skb->prev; \
3459 skb != (struct sk_buff *)(queue); \
3460 skb = tmp, tmp = skb->prev)
3461
3462#define skb_queue_reverse_walk_from_safe(queue, skb, tmp) \
3463 for (tmp = skb->prev; \
3464 skb != (struct sk_buff *)(queue); \
3465 skb = tmp, tmp = skb->prev)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003466
David S. Miller21dc3302010-08-23 00:13:46 -07003467static inline bool skb_has_frag_list(const struct sk_buff *skb)
David S. Milleree039872009-06-09 00:17:13 -07003468{
3469 return skb_shinfo(skb)->frag_list != NULL;
3470}
3471
3472static inline void skb_frag_list_init(struct sk_buff *skb)
3473{
3474 skb_shinfo(skb)->frag_list = NULL;
3475}
3476
David S. Milleree039872009-06-09 00:17:13 -07003477#define skb_walk_frags(skb, iter) \
3478 for (iter = skb_shinfo(skb)->frag_list; iter; iter = iter->next)
3479
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003480
Sabrina Dubrocab50b0582019-11-25 14:48:57 +01003481int __skb_wait_for_more_packets(struct sock *sk, struct sk_buff_head *queue,
3482 int *err, long *timeo_p,
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003483 const struct sk_buff *skb);
Paolo Abeni65101ae2017-05-16 11:20:13 +02003484struct sk_buff *__skb_try_recv_from_queue(struct sock *sk,
3485 struct sk_buff_head *queue,
3486 unsigned int flags,
3487 void (*destructor)(struct sock *sk,
3488 struct sk_buff *skb),
Paolo Abenifd69c392019-04-08 10:15:59 +02003489 int *off, int *err,
Paolo Abeni65101ae2017-05-16 11:20:13 +02003490 struct sk_buff **last);
Sabrina Dubrocab50b0582019-11-25 14:48:57 +01003491struct sk_buff *__skb_try_recv_datagram(struct sock *sk,
3492 struct sk_buff_head *queue,
3493 unsigned int flags,
Paolo Abeni7c13f972016-11-04 11:28:59 +01003494 void (*destructor)(struct sock *sk,
3495 struct sk_buff *skb),
Paolo Abenifd69c392019-04-08 10:15:59 +02003496 int *off, int *err,
Rainer Weikusatea3793e2015-12-06 21:11:34 +00003497 struct sk_buff **last);
Sabrina Dubrocab50b0582019-11-25 14:48:57 +01003498struct sk_buff *__skb_recv_datagram(struct sock *sk,
3499 struct sk_buff_head *sk_queue,
3500 unsigned int flags,
Paolo Abeni7c13f972016-11-04 11:28:59 +01003501 void (*destructor)(struct sock *sk,
3502 struct sk_buff *skb),
Paolo Abenifd69c392019-04-08 10:15:59 +02003503 int *off, int *err);
Joe Perches7965bd42013-09-26 14:48:15 -07003504struct sk_buff *skb_recv_datagram(struct sock *sk, unsigned flags, int noblock,
3505 int *err);
Linus Torvaldsa11e1d42018-06-28 09:43:44 -07003506__poll_t datagram_poll(struct file *file, struct socket *sock,
3507 struct poll_table_struct *wait);
Al Viroc0371da2014-11-24 10:42:55 -05003508int skb_copy_datagram_iter(const struct sk_buff *from, int offset,
3509 struct iov_iter *to, int size);
David S. Miller51f3d022014-11-05 16:46:40 -05003510static inline int skb_copy_datagram_msg(const struct sk_buff *from, int offset,
3511 struct msghdr *msg, int size)
3512{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003513 return skb_copy_datagram_iter(from, offset, &msg->msg_iter, size);
David S. Miller51f3d022014-11-05 16:46:40 -05003514}
Al Viroe5a4b0b2014-11-24 18:17:55 -05003515int skb_copy_and_csum_datagram_msg(struct sk_buff *skb, int hlen,
3516 struct msghdr *msg);
Sagi Grimberg65d69e22018-12-03 17:52:10 -08003517int skb_copy_and_hash_datagram_iter(const struct sk_buff *skb, int offset,
3518 struct iov_iter *to, int len,
3519 struct ahash_request *hash);
Al Viro3a654f92014-06-19 14:15:22 -04003520int skb_copy_datagram_from_iter(struct sk_buff *skb, int offset,
3521 struct iov_iter *from, int len);
Al Viro3a654f92014-06-19 14:15:22 -04003522int zerocopy_sg_from_iter(struct sk_buff *skb, struct iov_iter *frm);
Joe Perches7965bd42013-09-26 14:48:15 -07003523void skb_free_datagram(struct sock *sk, struct sk_buff *skb);
samanthakumar627d2d62016-04-05 12:41:16 -04003524void __skb_free_datagram_locked(struct sock *sk, struct sk_buff *skb, int len);
3525static inline void skb_free_datagram_locked(struct sock *sk,
3526 struct sk_buff *skb)
3527{
3528 __skb_free_datagram_locked(sk, skb, 0);
3529}
Joe Perches7965bd42013-09-26 14:48:15 -07003530int skb_kill_datagram(struct sock *sk, struct sk_buff *skb, unsigned int flags);
Joe Perches7965bd42013-09-26 14:48:15 -07003531int skb_copy_bits(const struct sk_buff *skb, int offset, void *to, int len);
3532int skb_store_bits(struct sk_buff *skb, int offset, const void *from, int len);
3533__wsum skb_copy_and_csum_bits(const struct sk_buff *skb, int offset, u8 *to,
3534 int len, __wsum csum);
Hannes Frederic Sowaa60e3cc2015-05-21 17:00:00 +02003535int skb_splice_bits(struct sk_buff *skb, struct sock *sk, unsigned int offset,
Joe Perches7965bd42013-09-26 14:48:15 -07003536 struct pipe_inode_info *pipe, unsigned int len,
Al Viro25869262016-09-17 21:02:10 -04003537 unsigned int flags);
Tom Herbert20bf50d2017-07-28 16:22:42 -07003538int skb_send_sock_locked(struct sock *sk, struct sk_buff *skb, int offset,
3539 int len);
Joe Perches7965bd42013-09-26 14:48:15 -07003540void skb_copy_and_csum_dev(const struct sk_buff *skb, u8 *to);
Thomas Grafaf2806f2013-12-13 15:22:17 +01003541unsigned int skb_zerocopy_headlen(const struct sk_buff *from);
Zoltan Kiss36d5fe62014-03-26 22:37:45 +00003542int skb_zerocopy(struct sk_buff *to, struct sk_buff *from,
3543 int len, int hlen);
Joe Perches7965bd42013-09-26 14:48:15 -07003544void skb_split(struct sk_buff *skb, struct sk_buff *skb1, const u32 len);
3545int skb_shift(struct sk_buff *tgt, struct sk_buff *skb, int shiftlen);
3546void skb_scrub_packet(struct sk_buff *skb, bool xnet);
Daniel Axtens779b7932018-03-01 17:13:37 +11003547bool skb_gso_validate_network_len(const struct sk_buff *skb, unsigned int mtu);
Daniel Axtens2b16f042018-01-31 14:15:33 +11003548bool skb_gso_validate_mac_len(const struct sk_buff *skb, unsigned int len);
Joe Perches7965bd42013-09-26 14:48:15 -07003549struct sk_buff *skb_segment(struct sk_buff *skb, netdev_features_t features);
Steffen Klassert3a1296a2020-01-25 11:26:44 +01003550struct sk_buff *skb_segment_list(struct sk_buff *skb, netdev_features_t features,
3551 unsigned int offset);
Vlad Yasevich0d5501c2014-08-08 14:42:13 -04003552struct sk_buff *skb_vlan_untag(struct sk_buff *skb);
Jiri Pirkoe2195122014-11-19 14:05:01 +01003553int skb_ensure_writable(struct sk_buff *skb, int write_len);
Shmulik Ladkanibfca4c52016-09-19 19:11:09 +03003554int __skb_vlan_pop(struct sk_buff *skb, u16 *vlan_tci);
Jiri Pirko93515d52014-11-19 14:05:02 +01003555int skb_vlan_pop(struct sk_buff *skb);
3556int skb_vlan_push(struct sk_buff *skb, __be16 vlan_proto, u16 vlan_tci);
Davide Carattifa4e0f82019-10-12 13:55:07 +02003557int skb_mpls_push(struct sk_buff *skb, __be32 mpls_lse, __be16 mpls_proto,
Martin Varghesed04ac222019-12-05 05:57:22 +05303558 int mac_len, bool ethernet);
Martin Varghese040b5cf2019-12-02 10:49:51 +05303559int skb_mpls_pop(struct sk_buff *skb, __be16 next_proto, int mac_len,
3560 bool ethernet);
John Hurleyd27cf5c2019-07-07 15:01:56 +01003561int skb_mpls_update_lse(struct sk_buff *skb, __be32 mpls_lse);
John Hurley2a2ea502019-07-07 15:01:57 +01003562int skb_mpls_dec_ttl(struct sk_buff *skb);
Sowmini Varadhan6fa01cc2016-04-22 18:36:35 -07003563struct sk_buff *pskb_extract(struct sk_buff *skb, int off, int to_copy,
3564 gfp_t gfp);
Arnaldo Carvalho de Melo20380732005-08-16 02:18:02 -03003565
Al Viro6ce8e9c2014-04-06 21:25:44 -04003566static inline int memcpy_from_msg(void *data, struct msghdr *msg, int len)
3567{
Al Viro3073f072017-02-17 23:13:25 -05003568 return copy_from_iter_full(data, len, &msg->msg_iter) ? 0 : -EFAULT;
Al Viro6ce8e9c2014-04-06 21:25:44 -04003569}
3570
Al Viro7eab8d92014-04-06 21:51:23 -04003571static inline int memcpy_to_msg(struct msghdr *msg, void *data, int len)
3572{
Al Viroe5a4b0b2014-11-24 18:17:55 -05003573 return copy_to_iter(data, len, &msg->msg_iter) == len ? 0 : -EFAULT;
Al Viro7eab8d92014-04-06 21:51:23 -04003574}
3575
Daniel Borkmann2817a332013-10-30 11:50:51 +01003576struct skb_checksum_ops {
3577 __wsum (*update)(const void *mem, int len, __wsum wsum);
3578 __wsum (*combine)(__wsum csum, __wsum csum2, int offset, int len);
3579};
3580
Davide Caratti96178132017-05-18 15:44:37 +02003581extern const struct skb_checksum_ops *crc32c_csum_stub __read_mostly;
3582
Daniel Borkmann2817a332013-10-30 11:50:51 +01003583__wsum __skb_checksum(const struct sk_buff *skb, int offset, int len,
3584 __wsum csum, const struct skb_checksum_ops *ops);
3585__wsum skb_checksum(const struct sk_buff *skb, int offset, int len,
3586 __wsum csum);
3587
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003588static inline void * __must_check
3589__skb_header_pointer(const struct sk_buff *skb, int offset,
3590 int len, void *data, int hlen, void *buffer)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003591{
Patrick McHardy55820ee2005-07-05 14:08:10 -07003592 if (hlen - offset >= len)
David S. Miller690e36e2014-08-23 12:13:41 -07003593 return data + offset;
Linus Torvalds1da177e2005-04-16 15:20:36 -07003594
David S. Miller690e36e2014-08-23 12:13:41 -07003595 if (!skb ||
3596 skb_copy_bits(skb, offset, buffer, len) < 0)
Linus Torvalds1da177e2005-04-16 15:20:36 -07003597 return NULL;
3598
3599 return buffer;
3600}
3601
Eric Dumazet1e98a0f2015-06-12 19:31:32 -07003602static inline void * __must_check
3603skb_header_pointer(const struct sk_buff *skb, int offset, int len, void *buffer)
David S. Miller690e36e2014-08-23 12:13:41 -07003604{
3605 return __skb_header_pointer(skb, offset, len, skb->data,
3606 skb_headlen(skb), buffer);
3607}
3608
Daniel Borkmann4262e5c2013-12-06 11:36:16 +01003609/**
3610 * skb_needs_linearize - check if we need to linearize a given skb
3611 * depending on the given device features.
3612 * @skb: socket buffer to check
3613 * @features: net device features
3614 *
3615 * Returns true if either:
3616 * 1. skb has frag_list and the device doesn't support FRAGLIST, or
3617 * 2. skb is fragmented and the device does not support SG.
3618 */
3619static inline bool skb_needs_linearize(struct sk_buff *skb,
3620 netdev_features_t features)
3621{
3622 return skb_is_nonlinear(skb) &&
3623 ((skb_has_frag_list(skb) && !(features & NETIF_F_FRAGLIST)) ||
3624 (skb_shinfo(skb)->nr_frags && !(features & NETIF_F_SG)));
3625}
3626
Arnaldo Carvalho de Melod626f622007-03-27 18:55:52 -03003627static inline void skb_copy_from_linear_data(const struct sk_buff *skb,
3628 void *to,
3629 const unsigned int len)
3630{
3631 memcpy(to, skb->data, len);
3632}
3633
3634static inline void skb_copy_from_linear_data_offset(const struct sk_buff *skb,
3635 const int offset, void *to,
3636 const unsigned int len)
3637{
3638 memcpy(to, skb->data + offset, len);
3639}
3640
Arnaldo Carvalho de Melo27d7ff42007-03-31 11:55:19 -03003641static inline void skb_copy_to_linear_data(struct sk_buff *skb,
3642 const void *from,
3643 const unsigned int len)
3644{
3645 memcpy(skb->data, from, len);
3646}
3647
3648static inline void skb_copy_to_linear_data_offset(struct sk_buff *skb,
3649 const int offset,
3650 const void *from,
3651 const unsigned int len)
3652{
3653 memcpy(skb->data + offset, from, len);
3654}
3655
Joe Perches7965bd42013-09-26 14:48:15 -07003656void skb_init(void);
Linus Torvalds1da177e2005-04-16 15:20:36 -07003657
Patrick Ohlyac45f602009-02-12 05:03:37 +00003658static inline ktime_t skb_get_ktime(const struct sk_buff *skb)
3659{
3660 return skb->tstamp;
3661}
3662
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003663/**
3664 * skb_get_timestamp - get timestamp from a skb
3665 * @skb: skb to get stamp from
Deepa Dinamani13c6ee22019-02-02 07:34:48 -08003666 * @stamp: pointer to struct __kernel_old_timeval to store stamp in
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003667 *
3668 * Timestamps are stored in the skb as offsets to a base timestamp.
3669 * This function converts the offset back to a struct timeval and stores
3670 * it in stamp.
3671 */
Patrick Ohlyac45f602009-02-12 05:03:37 +00003672static inline void skb_get_timestamp(const struct sk_buff *skb,
Deepa Dinamani13c6ee22019-02-02 07:34:48 -08003673 struct __kernel_old_timeval *stamp)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003674{
Deepa Dinamani13c6ee22019-02-02 07:34:48 -08003675 *stamp = ns_to_kernel_old_timeval(skb->tstamp);
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003676}
3677
Deepa Dinamani887feae2019-02-02 07:34:50 -08003678static inline void skb_get_new_timestamp(const struct sk_buff *skb,
3679 struct __kernel_sock_timeval *stamp)
3680{
3681 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3682
3683 stamp->tv_sec = ts.tv_sec;
3684 stamp->tv_usec = ts.tv_nsec / 1000;
3685}
3686
Patrick Ohlyac45f602009-02-12 05:03:37 +00003687static inline void skb_get_timestampns(const struct sk_buff *skb,
Arnd Bergmanndf1b4ba2019-10-25 22:04:46 +02003688 struct __kernel_old_timespec *stamp)
Patrick Ohlyac45f602009-02-12 05:03:37 +00003689{
Arnd Bergmanndf1b4ba2019-10-25 22:04:46 +02003690 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3691
3692 stamp->tv_sec = ts.tv_sec;
3693 stamp->tv_nsec = ts.tv_nsec;
Patrick Ohlyac45f602009-02-12 05:03:37 +00003694}
3695
Deepa Dinamani887feae2019-02-02 07:34:50 -08003696static inline void skb_get_new_timestampns(const struct sk_buff *skb,
3697 struct __kernel_timespec *stamp)
3698{
3699 struct timespec64 ts = ktime_to_timespec64(skb->tstamp);
3700
3701 stamp->tv_sec = ts.tv_sec;
3702 stamp->tv_nsec = ts.tv_nsec;
3703}
3704
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003705static inline void __net_timestamp(struct sk_buff *skb)
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003706{
Eric Dumazetb7aa0bf2007-04-19 16:16:32 -07003707 skb->tstamp = ktime_get_real();
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003708}
3709
Stephen Hemminger164891a2007-04-23 22:26:16 -07003710static inline ktime_t net_timedelta(ktime_t t)
3711{
3712 return ktime_sub(ktime_get_real(), t);
3713}
3714
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003715static inline ktime_t net_invalid_timestamp(void)
3716{
Thomas Gleixner8b0e1952016-12-25 12:30:41 +01003717 return 0;
Ilpo Järvinenb9ce2042007-06-15 15:08:43 -07003718}
Patrick McHardya61bbcf2005-08-14 17:24:31 -07003719
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003720static inline u8 skb_metadata_len(const struct sk_buff *skb)
3721{
3722 return skb_shinfo(skb)->meta_len;
3723}
3724
3725static inline void *skb_metadata_end(const struct sk_buff *skb)
3726{
3727 return skb_mac_header(skb);
3728}
3729
3730static inline bool __skb_metadata_differs(const struct sk_buff *skb_a,
3731 const struct sk_buff *skb_b,
3732 u8 meta_len)
3733{
3734 const void *a = skb_metadata_end(skb_a);
3735 const void *b = skb_metadata_end(skb_b);
3736 /* Using more efficient varaiant than plain call to memcmp(). */
3737#if defined(CONFIG_HAVE_EFFICIENT_UNALIGNED_ACCESS) && BITS_PER_LONG == 64
3738 u64 diffs = 0;
3739
3740 switch (meta_len) {
3741#define __it(x, op) (x -= sizeof(u##op))
3742#define __it_diff(a, b, op) (*(u##op *)__it(a, op)) ^ (*(u##op *)__it(b, op))
3743 case 32: diffs |= __it_diff(a, b, 64);
Gustavo A. R. Silva82385b02018-10-17 15:01:37 +02003744 /* fall through */
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003745 case 24: diffs |= __it_diff(a, b, 64);
Gustavo A. R. Silva82385b02018-10-17 15:01:37 +02003746 /* fall through */
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003747 case 16: diffs |= __it_diff(a, b, 64);
Gustavo A. R. Silva82385b02018-10-17 15:01:37 +02003748 /* fall through */
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003749 case 8: diffs |= __it_diff(a, b, 64);
3750 break;
3751 case 28: diffs |= __it_diff(a, b, 64);
Gustavo A. R. Silva82385b02018-10-17 15:01:37 +02003752 /* fall through */
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003753 case 20: diffs |= __it_diff(a, b, 64);
Gustavo A. R. Silva82385b02018-10-17 15:01:37 +02003754 /* fall through */
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003755 case 12: diffs |= __it_diff(a, b, 64);
Gustavo A. R. Silva82385b02018-10-17 15:01:37 +02003756 /* fall through */
Daniel Borkmannde8f3a82017-09-25 02:25:51 +02003757 case 4: diffs |= __it_diff(a, b, 32);
3758 break;
3759 }
3760 return diffs;
3761#else
3762 return memcmp(a - meta_len, b - meta_len, meta_len);
3763#endif
3764}
3765
3766static inline bool skb_metadata_differs(const struct sk_buff *skb_a,
3767 const struct sk_buff *skb_b)
3768{
3769 u8 len_a = skb_metadata_len(skb_a);
3770 u8 len_b = skb_metadata_len(skb_b);
3771
3772 if (!(len_a | len_b))
3773 return false;
3774
3775 return len_a != len_b ?
3776 true : __skb_metadata_differs(skb_a, skb_b, len_a);
3777}
3778
3779static inline void skb_metadata_set(struct sk_buff *skb, u8 meta_len)
3780{
3781 skb_shinfo(skb)->meta_len = meta_len;
3782}
3783
3784static inline void skb_metadata_clear(struct sk_buff *skb)
3785{
3786 skb_metadata_set(skb, 0);
3787}
3788
Alexander Duyck62bccb82014-09-04 13:31:35 -04003789struct sk_buff *skb_clone_sk(struct sk_buff *skb);
3790
Richard Cochranc1f19b52010-07-17 08:49:36 +00003791#ifdef CONFIG_NETWORK_PHY_TIMESTAMPING
3792
Joe Perches7965bd42013-09-26 14:48:15 -07003793void skb_clone_tx_timestamp(struct sk_buff *skb);
3794bool skb_defer_rx_timestamp(struct sk_buff *skb);
Richard Cochranc1f19b52010-07-17 08:49:36 +00003795
3796#else /* CONFIG_NETWORK_PHY_TIMESTAMPING */
3797
3798static inline void skb_clone_tx_timestamp(struct sk_buff *skb)
3799{
3800}
3801
3802static inline bool skb_defer_rx_timestamp(struct sk_buff *skb)
3803{
3804 return false;
3805}
3806
3807#endif /* !CONFIG_NETWORK_PHY_TIMESTAMPING */
3808
3809/**
3810 * skb_complete_tx_timestamp() - deliver cloned skb with tx timestamps
3811 *
Richard Cochranda92b192011-10-21 00:49:15 +00003812 * PHY drivers may accept clones of transmitted packets for
3813 * timestamping via their phy_driver.txtstamp method. These drivers
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003814 * must call this function to return the skb back to the stack with a
3815 * timestamp.
Richard Cochranda92b192011-10-21 00:49:15 +00003816 *
Richard Cochranc1f19b52010-07-17 08:49:36 +00003817 * @skb: clone of the the original outgoing packet
Benjamin Poirier7a76a022015-08-07 09:32:21 -07003818 * @hwtstamps: hardware time stamps
Richard Cochranc1f19b52010-07-17 08:49:36 +00003819 *
3820 */
3821void skb_complete_tx_timestamp(struct sk_buff *skb,
3822 struct skb_shared_hwtstamps *hwtstamps);
3823
Willem de Bruijne7fd2882014-08-04 22:11:48 -04003824void __skb_tstamp_tx(struct sk_buff *orig_skb,
3825 struct skb_shared_hwtstamps *hwtstamps,
3826 struct sock *sk, int tstype);
3827
Patrick Ohlyac45f602009-02-12 05:03:37 +00003828/**
3829 * skb_tstamp_tx - queue clone of skb with send time stamps
3830 * @orig_skb: the original outgoing packet
3831 * @hwtstamps: hardware time stamps, may be NULL if not available
3832 *
3833 * If the skb has a socket associated, then this function clones the
3834 * skb (thus sharing the actual data and optional structures), stores
3835 * the optional hardware time stamping information (if non NULL) or
3836 * generates a software time stamp (otherwise), then queues the clone
3837 * to the error queue of the socket. Errors are silently ignored.
3838 */
Joe Perches7965bd42013-09-26 14:48:15 -07003839void skb_tstamp_tx(struct sk_buff *orig_skb,
3840 struct skb_shared_hwtstamps *hwtstamps);
Patrick Ohlyac45f602009-02-12 05:03:37 +00003841
Richard Cochran4507a712010-07-17 08:48:28 +00003842/**
3843 * skb_tx_timestamp() - Driver hook for transmit timestamping
3844 *
3845 * Ethernet MAC Drivers should call this function in their hard_xmit()
Richard Cochran4ff75b72011-06-19 03:31:39 +00003846 * function immediately before giving the sk_buff to the MAC hardware.
Richard Cochran4507a712010-07-17 08:48:28 +00003847 *
David S. Miller73409f32013-12-27 13:04:33 -05003848 * Specifically, one should make absolutely sure that this function is
3849 * called before TX completion of this packet can trigger. Otherwise
3850 * the packet could potentially already be freed.
3851 *
Richard Cochran4507a712010-07-17 08:48:28 +00003852 * @skb: A socket buffer.
3853 */
3854static inline void skb_tx_timestamp(struct sk_buff *skb)
3855{
Richard Cochranc1f19b52010-07-17 08:49:36 +00003856 skb_clone_tx_timestamp(skb);
Miroslav Lichvarb50a5c72017-05-19 17:52:40 +02003857 if (skb_shinfo(skb)->tx_flags & SKBTX_SW_TSTAMP)
3858 skb_tstamp_tx(skb, NULL);
Richard Cochran4507a712010-07-17 08:48:28 +00003859}
3860
Johannes Berg6e3e9392011-11-09 10:15:42 +01003861/**
3862 * skb_complete_wifi_ack - deliver skb with wifi status
3863 *
3864 * @skb: the original outgoing packet
3865 * @acked: ack status
3866 *
3867 */
3868void skb_complete_wifi_ack(struct sk_buff *skb, bool acked);
3869
Joe Perches7965bd42013-09-26 14:48:15 -07003870__sum16 __skb_checksum_complete_head(struct sk_buff *skb, int len);
3871__sum16 __skb_checksum_complete(struct sk_buff *skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003872
Herbert Xu60476372007-04-09 11:59:39 -07003873static inline int skb_csum_unnecessary(const struct sk_buff *skb)
3874{
Tom Herbert6edec0e2015-02-10 16:30:28 -08003875 return ((skb->ip_summed == CHECKSUM_UNNECESSARY) ||
3876 skb->csum_valid ||
3877 (skb->ip_summed == CHECKSUM_PARTIAL &&
3878 skb_checksum_start_offset(skb) >= 0));
Herbert Xu60476372007-04-09 11:59:39 -07003879}
3880
Herbert Xufb286bb2005-11-10 13:01:24 -08003881/**
3882 * skb_checksum_complete - Calculate checksum of an entire packet
3883 * @skb: packet to process
3884 *
3885 * This function calculates the checksum over the entire packet plus
3886 * the value of skb->csum. The latter can be used to supply the
3887 * checksum of a pseudo header as used by TCP/UDP. It returns the
3888 * checksum.
3889 *
3890 * For protocols that contain complete checksums such as ICMP/TCP/UDP,
3891 * this function can be used to verify that checksum on received
3892 * packets. In that case the function should return zero if the
3893 * checksum is correct. In particular, this function will return zero
3894 * if skb->ip_summed is CHECKSUM_UNNECESSARY which indicates that the
3895 * hardware has already verified the correctness of the checksum.
3896 */
Al Viro4381ca32007-07-15 21:00:11 +01003897static inline __sum16 skb_checksum_complete(struct sk_buff *skb)
Herbert Xufb286bb2005-11-10 13:01:24 -08003898{
Herbert Xu60476372007-04-09 11:59:39 -07003899 return skb_csum_unnecessary(skb) ?
3900 0 : __skb_checksum_complete(skb);
Herbert Xufb286bb2005-11-10 13:01:24 -08003901}
3902
Tom Herbert77cffe22014-08-27 21:26:46 -07003903static inline void __skb_decr_checksum_unnecessary(struct sk_buff *skb)
3904{
3905 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3906 if (skb->csum_level == 0)
3907 skb->ip_summed = CHECKSUM_NONE;
3908 else
3909 skb->csum_level--;
3910 }
3911}
3912
3913static inline void __skb_incr_checksum_unnecessary(struct sk_buff *skb)
3914{
3915 if (skb->ip_summed == CHECKSUM_UNNECESSARY) {
3916 if (skb->csum_level < SKB_MAX_CSUM_LEVEL)
3917 skb->csum_level++;
3918 } else if (skb->ip_summed == CHECKSUM_NONE) {
3919 skb->ip_summed = CHECKSUM_UNNECESSARY;
3920 skb->csum_level = 0;
3921 }
3922}
3923
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003924/* Check if we need to perform checksum complete validation.
3925 *
3926 * Returns true if checksum complete is needed, false otherwise
3927 * (either checksum is unnecessary or zero checksum is allowed).
3928 */
3929static inline bool __skb_checksum_validate_needed(struct sk_buff *skb,
3930 bool zero_okay,
3931 __sum16 check)
3932{
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003933 if (skb_csum_unnecessary(skb) || (zero_okay && !check)) {
3934 skb->csum_valid = 1;
Tom Herbert77cffe22014-08-27 21:26:46 -07003935 __skb_decr_checksum_unnecessary(skb);
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003936 return false;
3937 }
3938
3939 return true;
3940}
3941
David S. Millerda279882018-02-16 15:52:42 -05003942/* For small packets <= CHECKSUM_BREAK perform checksum complete directly
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003943 * in checksum_init.
3944 */
3945#define CHECKSUM_BREAK 76
3946
Tom Herbert4e18b9a2015-04-20 14:10:04 -07003947/* Unset checksum-complete
3948 *
3949 * Unset checksum complete can be done when packet is being modified
3950 * (uncompressed for instance) and checksum-complete value is
3951 * invalidated.
3952 */
3953static inline void skb_checksum_complete_unset(struct sk_buff *skb)
3954{
3955 if (skb->ip_summed == CHECKSUM_COMPLETE)
3956 skb->ip_summed = CHECKSUM_NONE;
3957}
3958
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003959/* Validate (init) checksum based on checksum complete.
3960 *
3961 * Return values:
3962 * 0: checksum is validated or try to in skb_checksum_complete. In the latter
3963 * case the ip_summed will not be CHECKSUM_UNNECESSARY and the pseudo
3964 * checksum is stored in skb->csum for use in __skb_checksum_complete
3965 * non-zero: value of invalid checksum
3966 *
3967 */
3968static inline __sum16 __skb_checksum_validate_complete(struct sk_buff *skb,
3969 bool complete,
3970 __wsum psum)
3971{
3972 if (skb->ip_summed == CHECKSUM_COMPLETE) {
3973 if (!csum_fold(csum_add(psum, skb->csum))) {
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003974 skb->csum_valid = 1;
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003975 return 0;
3976 }
3977 }
3978
3979 skb->csum = psum;
3980
Tom Herbert5d0c2b92014-06-10 18:54:13 -07003981 if (complete || skb->len <= CHECKSUM_BREAK) {
3982 __sum16 csum;
3983
3984 csum = __skb_checksum_complete(skb);
3985 skb->csum_valid = !csum;
3986 return csum;
3987 }
Tom Herbert76ba0aa2014-05-02 16:29:18 -07003988
3989 return 0;
3990}
3991
3992static inline __wsum null_compute_pseudo(struct sk_buff *skb, int proto)
3993{
3994 return 0;
3995}
3996
3997/* Perform checksum validate (init). Note that this is a macro since we only
3998 * want to calculate the pseudo header which is an input function if necessary.
3999 * First we try to validate without any computation (checksum unnecessary) and
4000 * then calculate based on checksum complete calling the function to compute
4001 * pseudo header.
4002 *
4003 * Return values:
4004 * 0: checksum is validated or try to in skb_checksum_complete
4005 * non-zero: value of invalid checksum
4006 */
4007#define __skb_checksum_validate(skb, proto, complete, \
4008 zero_okay, check, compute_pseudo) \
4009({ \
4010 __sum16 __ret = 0; \
Tom Herbert5d0c2b92014-06-10 18:54:13 -07004011 skb->csum_valid = 0; \
Tom Herbert76ba0aa2014-05-02 16:29:18 -07004012 if (__skb_checksum_validate_needed(skb, zero_okay, check)) \
4013 __ret = __skb_checksum_validate_complete(skb, \
4014 complete, compute_pseudo(skb, proto)); \
4015 __ret; \
4016})
4017
4018#define skb_checksum_init(skb, proto, compute_pseudo) \
4019 __skb_checksum_validate(skb, proto, false, false, 0, compute_pseudo)
4020
4021#define skb_checksum_init_zero_check(skb, proto, check, compute_pseudo) \
4022 __skb_checksum_validate(skb, proto, false, true, check, compute_pseudo)
4023
4024#define skb_checksum_validate(skb, proto, compute_pseudo) \
4025 __skb_checksum_validate(skb, proto, true, false, 0, compute_pseudo)
4026
4027#define skb_checksum_validate_zero_check(skb, proto, check, \
4028 compute_pseudo) \
Sabrina Dubroca096a4cf2015-02-06 18:54:19 +01004029 __skb_checksum_validate(skb, proto, true, true, check, compute_pseudo)
Tom Herbert76ba0aa2014-05-02 16:29:18 -07004030
4031#define skb_checksum_simple_validate(skb) \
4032 __skb_checksum_validate(skb, 0, true, false, 0, null_compute_pseudo)
4033
Tom Herbertd96535a2014-08-31 15:12:42 -07004034static inline bool __skb_checksum_convert_check(struct sk_buff *skb)
4035{
Davide Caratti219f1d792017-05-18 15:44:39 +02004036 return (skb->ip_summed == CHECKSUM_NONE && skb->csum_valid);
Tom Herbertd96535a2014-08-31 15:12:42 -07004037}
4038
Li RongQinge4aa33a2019-07-04 17:03:26 +08004039static inline void __skb_checksum_convert(struct sk_buff *skb, __wsum pseudo)
Tom Herbertd96535a2014-08-31 15:12:42 -07004040{
4041 skb->csum = ~pseudo;
4042 skb->ip_summed = CHECKSUM_COMPLETE;
4043}
4044
Li RongQinge4aa33a2019-07-04 17:03:26 +08004045#define skb_checksum_try_convert(skb, proto, compute_pseudo) \
Tom Herbertd96535a2014-08-31 15:12:42 -07004046do { \
4047 if (__skb_checksum_convert_check(skb)) \
Li RongQinge4aa33a2019-07-04 17:03:26 +08004048 __skb_checksum_convert(skb, compute_pseudo(skb, proto)); \
Tom Herbertd96535a2014-08-31 15:12:42 -07004049} while (0)
4050
Tom Herbert15e23962015-02-10 16:30:31 -08004051static inline void skb_remcsum_adjust_partial(struct sk_buff *skb, void *ptr,
4052 u16 start, u16 offset)
4053{
4054 skb->ip_summed = CHECKSUM_PARTIAL;
4055 skb->csum_start = ((unsigned char *)ptr + start) - skb->head;
4056 skb->csum_offset = offset - start;
4057}
4058
Tom Herbertdcdc8992015-02-02 16:07:34 -08004059/* Update skbuf and packet to reflect the remote checksum offload operation.
4060 * When called, ptr indicates the starting point for skb->csum when
4061 * ip_summed is CHECKSUM_COMPLETE. If we need create checksum complete
4062 * here, skb_postpull_rcsum is done so skb->csum start is ptr.
4063 */
4064static inline void skb_remcsum_process(struct sk_buff *skb, void *ptr,
Tom Herbert15e23962015-02-10 16:30:31 -08004065 int start, int offset, bool nopartial)
Tom Herbertdcdc8992015-02-02 16:07:34 -08004066{
4067 __wsum delta;
4068
Tom Herbert15e23962015-02-10 16:30:31 -08004069 if (!nopartial) {
4070 skb_remcsum_adjust_partial(skb, ptr, start, offset);
4071 return;
4072 }
4073
Tom Herbertdcdc8992015-02-02 16:07:34 -08004074 if (unlikely(skb->ip_summed != CHECKSUM_COMPLETE)) {
4075 __skb_checksum_complete(skb);
4076 skb_postpull_rcsum(skb, skb->data, ptr - (void *)skb->data);
4077 }
4078
4079 delta = remcsum_adjust(ptr, skb->csum, start, offset);
4080
4081 /* Adjust skb->csum since we changed the packet */
4082 skb->csum = csum_add(skb->csum, delta);
4083}
4084
Florian Westphalcb9c6832017-01-23 18:21:56 +01004085static inline struct nf_conntrack *skb_nfct(const struct sk_buff *skb)
4086{
4087#if IS_ENABLED(CONFIG_NF_CONNTRACK)
Jeremy Sowden261db6c2019-09-13 09:13:14 +01004088 return (void *)(skb->_nfct & NFCT_PTRMASK);
Florian Westphalcb9c6832017-01-23 18:21:56 +01004089#else
4090 return NULL;
4091#endif
4092}
4093
Jeremy Sowden261db6c2019-09-13 09:13:14 +01004094static inline unsigned long skb_get_nfct(const struct sk_buff *skb)
Linus Torvalds1da177e2005-04-16 15:20:36 -07004095{
Jeremy Sowden261db6c2019-09-13 09:13:14 +01004096#if IS_ENABLED(CONFIG_NF_CONNTRACK)
4097 return skb->_nfct;
4098#else
4099 return 0UL;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01004100#endif
Jeremy Sowden261db6c2019-09-13 09:13:14 +01004101}
4102
4103static inline void skb_set_nfct(struct sk_buff *skb, unsigned long nfct)
4104{
4105#if IS_ENABLED(CONFIG_NF_CONNTRACK)
4106 skb->_nfct = nfct;
4107#endif
4108}
Florian Westphaldf5042f2018-12-18 17:15:16 +01004109
4110#ifdef CONFIG_SKB_EXTENSIONS
4111enum skb_ext_id {
4112#if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
4113 SKB_EXT_BRIDGE_NF,
4114#endif
Florian Westphal41650792018-12-18 17:15:27 +01004115#ifdef CONFIG_XFRM
4116 SKB_EXT_SEC_PATH,
4117#endif
Paul Blakey95a72332019-09-04 16:56:37 +03004118#if IS_ENABLED(CONFIG_NET_TC_SKB_EXT)
4119 TC_SKB_EXT,
4120#endif
Mat Martineau3ee17bc2020-01-09 07:59:19 -08004121#if IS_ENABLED(CONFIG_MPTCP)
4122 SKB_EXT_MPTCP,
4123#endif
Florian Westphaldf5042f2018-12-18 17:15:16 +01004124 SKB_EXT_NUM, /* must be last */
4125};
4126
4127/**
4128 * struct skb_ext - sk_buff extensions
4129 * @refcnt: 1 on allocation, deallocated on 0
4130 * @offset: offset to add to @data to obtain extension address
4131 * @chunks: size currently allocated, stored in SKB_EXT_ALIGN_SHIFT units
4132 * @data: start of extension data, variable sized
4133 *
4134 * Note: offsets/lengths are stored in chunks of 8 bytes, this allows
4135 * to use 'u8' types while allowing up to 2kb worth of extension data.
4136 */
4137struct skb_ext {
4138 refcount_t refcnt;
4139 u8 offset[SKB_EXT_NUM]; /* in chunks of 8 bytes */
4140 u8 chunks; /* same */
4141 char data[0] __aligned(8);
4142};
4143
Paolo Abeni8b69a802020-01-09 07:59:24 -08004144struct skb_ext *__skb_ext_alloc(void);
4145void *__skb_ext_set(struct sk_buff *skb, enum skb_ext_id id,
4146 struct skb_ext *ext);
Florian Westphaldf5042f2018-12-18 17:15:16 +01004147void *skb_ext_add(struct sk_buff *skb, enum skb_ext_id id);
4148void __skb_ext_del(struct sk_buff *skb, enum skb_ext_id id);
4149void __skb_ext_put(struct skb_ext *ext);
4150
4151static inline void skb_ext_put(struct sk_buff *skb)
4152{
4153 if (skb->active_extensions)
4154 __skb_ext_put(skb->extensions);
4155}
4156
Florian Westphaldf5042f2018-12-18 17:15:16 +01004157static inline void __skb_ext_copy(struct sk_buff *dst,
4158 const struct sk_buff *src)
4159{
4160 dst->active_extensions = src->active_extensions;
4161
4162 if (src->active_extensions) {
4163 struct skb_ext *ext = src->extensions;
4164
4165 refcount_inc(&ext->refcnt);
4166 dst->extensions = ext;
4167 }
4168}
4169
4170static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *src)
4171{
4172 skb_ext_put(dst);
4173 __skb_ext_copy(dst, src);
4174}
4175
4176static inline bool __skb_ext_exist(const struct skb_ext *ext, enum skb_ext_id i)
4177{
4178 return !!ext->offset[i];
4179}
4180
4181static inline bool skb_ext_exist(const struct sk_buff *skb, enum skb_ext_id id)
4182{
4183 return skb->active_extensions & (1 << id);
4184}
4185
4186static inline void skb_ext_del(struct sk_buff *skb, enum skb_ext_id id)
4187{
4188 if (skb_ext_exist(skb, id))
4189 __skb_ext_del(skb, id);
4190}
4191
4192static inline void *skb_ext_find(const struct sk_buff *skb, enum skb_ext_id id)
4193{
4194 if (skb_ext_exist(skb, id)) {
4195 struct skb_ext *ext = skb->extensions;
4196
4197 return (void *)ext + (ext->offset[id] << 3);
4198 }
4199
4200 return NULL;
4201}
Florian Westphal174e2382019-09-26 20:37:05 +02004202
4203static inline void skb_ext_reset(struct sk_buff *skb)
4204{
4205 if (unlikely(skb->active_extensions)) {
4206 __skb_ext_put(skb->extensions);
4207 skb->active_extensions = 0;
4208 }
4209}
Florian Westphal677bf082019-11-21 06:56:23 +01004210
4211static inline bool skb_has_extensions(struct sk_buff *skb)
4212{
4213 return unlikely(skb->active_extensions);
4214}
Florian Westphaldf5042f2018-12-18 17:15:16 +01004215#else
4216static inline void skb_ext_put(struct sk_buff *skb) {}
Florian Westphal174e2382019-09-26 20:37:05 +02004217static inline void skb_ext_reset(struct sk_buff *skb) {}
Florian Westphaldf5042f2018-12-18 17:15:16 +01004218static inline void skb_ext_del(struct sk_buff *skb, int unused) {}
4219static inline void __skb_ext_copy(struct sk_buff *d, const struct sk_buff *s) {}
4220static inline void skb_ext_copy(struct sk_buff *dst, const struct sk_buff *s) {}
Florian Westphal677bf082019-11-21 06:56:23 +01004221static inline bool skb_has_extensions(struct sk_buff *skb) { return false; }
Florian Westphaldf5042f2018-12-18 17:15:16 +01004222#endif /* CONFIG_SKB_EXTENSIONS */
4223
Florian Westphal895b5c92019-09-29 20:54:03 +02004224static inline void nf_reset_ct(struct sk_buff *skb)
Patrick McHardya193a4a2006-03-20 19:23:05 -08004225{
Yasuyuki Kozakai5f79e0f2007-03-23 11:17:07 -07004226#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01004227 nf_conntrack_put(skb_nfct(skb));
4228 skb->_nfct = 0;
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01004229#endif
Patrick McHardya193a4a2006-03-20 19:23:05 -08004230}
4231
Patrick McHardy124dff02013-04-05 20:42:05 +02004232static inline void nf_reset_trace(struct sk_buff *skb)
4233{
Florian Westphal478b3602014-02-15 23:48:45 +01004234#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Gao feng130549fe2013-03-21 19:48:41 +00004235 skb->nf_trace = 0;
4236#endif
Linus Torvalds1da177e2005-04-16 15:20:36 -07004237}
4238
Ye Yin2b5ec1a2017-10-26 16:57:05 +08004239static inline void ipvs_reset(struct sk_buff *skb)
4240{
4241#if IS_ENABLED(CONFIG_IP_VS)
4242 skb->ipvs_property = 0;
4243#endif
4244}
4245
Florian Westphalde8bda12018-12-18 17:15:17 +01004246/* Note: This doesn't put any conntrack info in dst. */
Eric Dumazetb1937222014-09-28 22:18:47 -07004247static inline void __nf_copy(struct sk_buff *dst, const struct sk_buff *src,
4248 bool copy)
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07004249{
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 dst->_nfct = src->_nfct;
4252 nf_conntrack_get(skb_nfct(src));
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01004253#endif
Florian Westphal478b3602014-02-15 23:48:45 +01004254#if IS_ENABLED(CONFIG_NETFILTER_XT_TARGET_TRACE) || defined(CONFIG_NF_TABLES)
Eric Dumazetb1937222014-09-28 22:18:47 -07004255 if (copy)
4256 dst->nf_trace = src->nf_trace;
Florian Westphal478b3602014-02-15 23:48:45 +01004257#endif
Yasuyuki Kozakaiedda5532007-03-14 16:43:37 -07004258}
4259
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07004260static inline void nf_copy(struct sk_buff *dst, const struct sk_buff *src)
4261{
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07004262#if defined(CONFIG_NF_CONNTRACK) || defined(CONFIG_NF_CONNTRACK_MODULE)
Florian Westphala9e419d2017-01-23 18:21:59 +01004263 nf_conntrack_put(skb_nfct(dst));
KOVACS Krisztian2fc72c72011-01-12 20:25:08 +01004264#endif
Eric Dumazetb1937222014-09-28 22:18:47 -07004265 __nf_copy(dst, src, true);
Yasuyuki Kozakaie7ac05f2007-03-14 16:44:01 -07004266}
4267
James Morris984bc162006-06-09 00:29:17 -07004268#ifdef CONFIG_NETWORK_SECMARK
4269static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4270{
4271 to->secmark = from->secmark;
4272}
4273
4274static inline void skb_init_secmark(struct sk_buff *skb)
4275{
4276 skb->secmark = 0;
4277}
4278#else
4279static inline void skb_copy_secmark(struct sk_buff *to, const struct sk_buff *from)
4280{ }
4281
4282static inline void skb_init_secmark(struct sk_buff *skb)
4283{ }
4284#endif
4285
Florian Westphal7af8f4c2018-12-18 17:15:19 +01004286static inline int secpath_exists(const struct sk_buff *skb)
4287{
4288#ifdef CONFIG_XFRM
Florian Westphal41650792018-12-18 17:15:27 +01004289 return skb_ext_exist(skb, SKB_EXT_SEC_PATH);
Florian Westphal7af8f4c2018-12-18 17:15:19 +01004290#else
4291 return 0;
4292#endif
4293}
4294
Eric W. Biederman574f7192014-04-01 12:20:24 -07004295static inline bool skb_irq_freeable(const struct sk_buff *skb)
4296{
4297 return !skb->destructor &&
Florian Westphal7af8f4c2018-12-18 17:15:19 +01004298 !secpath_exists(skb) &&
Florian Westphalcb9c6832017-01-23 18:21:56 +01004299 !skb_nfct(skb) &&
Eric W. Biederman574f7192014-04-01 12:20:24 -07004300 !skb->_skb_refdst &&
4301 !skb_has_frag_list(skb);
4302}
4303
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07004304static inline void skb_set_queue_mapping(struct sk_buff *skb, u16 queue_mapping)
4305{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07004306 skb->queue_mapping = queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07004307}
4308
Stephen Hemminger92477442009-03-21 13:39:26 -07004309static inline u16 skb_get_queue_mapping(const struct sk_buff *skb)
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07004310{
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07004311 return skb->queue_mapping;
Pavel Emelyanov4e3ab472007-10-21 17:01:29 -07004312}
4313
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07004314static inline void skb_copy_queue_mapping(struct sk_buff *to, const struct sk_buff *from)
4315{
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07004316 to->queue_mapping = from->queue_mapping;
Peter P Waskiewicz Jrf25f4e42007-07-06 13:36:20 -07004317}
4318
David S. Millerd5a9e242009-01-27 16:22:11 -08004319static inline void skb_record_rx_queue(struct sk_buff *skb, u16 rx_queue)
4320{
4321 skb->queue_mapping = rx_queue + 1;
4322}
4323
Stephen Hemminger92477442009-03-21 13:39:26 -07004324static inline u16 skb_get_rx_queue(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08004325{
4326 return skb->queue_mapping - 1;
4327}
4328
Stephen Hemminger92477442009-03-21 13:39:26 -07004329static inline bool skb_rx_queue_recorded(const struct sk_buff *skb)
David S. Millerd5a9e242009-01-27 16:22:11 -08004330{
Eric Dumazeta02cec22010-09-22 20:43:57 +00004331 return skb->queue_mapping != 0;
David S. Millerd5a9e242009-01-27 16:22:11 -08004332}
4333
Julian Anastasov4ff06202017-02-06 23:14:12 +02004334static inline void skb_set_dst_pending_confirm(struct sk_buff *skb, u32 val)
4335{
4336 skb->dst_pending_confirm = val;
4337}
4338
4339static inline bool skb_get_dst_pending_confirm(const struct sk_buff *skb)
4340{
4341 return skb->dst_pending_confirm != 0;
4342}
4343
Florian Westphal2294be0f2018-12-18 17:15:20 +01004344static inline struct sec_path *skb_sec_path(const struct sk_buff *skb)
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04004345{
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07004346#ifdef CONFIG_XFRM
Florian Westphal41650792018-12-18 17:15:27 +01004347 return skb_ext_find(skb, SKB_EXT_SEC_PATH);
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07004348#else
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07004349 return NULL;
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07004350#endif
Denis Kirjanov0b3d8e02013-10-02 05:58:32 +04004351}
Alexey Dobriyandef8b4f2008-10-28 13:24:06 -07004352
Pravin B Shelar68c33162013-02-14 14:02:41 +00004353/* Keeps track of mac header offset relative to skb->head.
4354 * It is useful for TSO of Tunneling protocol. e.g. GRE.
4355 * For non-tunnel skb it points to skb_mac_header() and for
Eric Dumazet3347c962013-10-19 11:42:56 -07004356 * tunnel skb it points to outer mac header.
4357 * Keeps track of level of encapsulation of network headers.
4358 */
Pravin B Shelar68c33162013-02-14 14:02:41 +00004359struct skb_gso_cb {
Alexander Duyck802ab552016-04-10 21:45:03 -04004360 union {
4361 int mac_offset;
4362 int data_offset;
4363 };
Eric Dumazet3347c962013-10-19 11:42:56 -07004364 int encap_level;
Alexander Duyck76443452016-02-05 15:27:37 -08004365 __wsum csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07004366 __u16 csum_start;
Pravin B Shelar68c33162013-02-14 14:02:41 +00004367};
Konstantin Khlebnikov9207f9d2016-01-08 15:21:46 +03004368#define SKB_SGO_CB_OFFSET 32
4369#define SKB_GSO_CB(skb) ((struct skb_gso_cb *)((skb)->cb + SKB_SGO_CB_OFFSET))
Pravin B Shelar68c33162013-02-14 14:02:41 +00004370
4371static inline int skb_tnl_header_len(const struct sk_buff *inner_skb)
4372{
4373 return (skb_mac_header(inner_skb) - inner_skb->head) -
4374 SKB_GSO_CB(inner_skb)->mac_offset;
4375}
4376
Pravin B Shelar1e2bd512013-05-30 06:45:27 +00004377static inline int gso_pskb_expand_head(struct sk_buff *skb, int extra)
4378{
4379 int new_headroom, headroom;
4380 int ret;
4381
4382 headroom = skb_headroom(skb);
4383 ret = pskb_expand_head(skb, extra, 0, GFP_ATOMIC);
4384 if (ret)
4385 return ret;
4386
4387 new_headroom = skb_headroom(skb);
4388 SKB_GSO_CB(skb)->mac_offset += (new_headroom - headroom);
4389 return 0;
4390}
4391
Alexander Duyck08b64fc2016-02-05 15:27:49 -08004392static inline void gso_reset_checksum(struct sk_buff *skb, __wsum res)
4393{
4394 /* Do not update partial checksums if remote checksum is enabled. */
4395 if (skb->remcsum_offload)
4396 return;
4397
4398 SKB_GSO_CB(skb)->csum = res;
4399 SKB_GSO_CB(skb)->csum_start = skb_checksum_start(skb) - skb->head;
4400}
4401
Tom Herbert7e2b10c2014-06-04 17:20:02 -07004402/* Compute the checksum for a gso segment. First compute the checksum value
4403 * from the start of transport header to SKB_GSO_CB(skb)->csum_start, and
4404 * then add in skb->csum (checksum from csum_start to end of packet).
4405 * skb->csum and csum_start are then updated to reflect the checksum of the
4406 * resultant packet starting from the transport header-- the resultant checksum
4407 * is in the res argument (i.e. normally zero or ~ of checksum of a pseudo
4408 * header.
4409 */
4410static inline __sum16 gso_make_checksum(struct sk_buff *skb, __wsum res)
4411{
Alexander Duyck76443452016-02-05 15:27:37 -08004412 unsigned char *csum_start = skb_transport_header(skb);
4413 int plen = (skb->head + SKB_GSO_CB(skb)->csum_start) - csum_start;
4414 __wsum partial = SKB_GSO_CB(skb)->csum;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07004415
Alexander Duyck76443452016-02-05 15:27:37 -08004416 SKB_GSO_CB(skb)->csum = res;
4417 SKB_GSO_CB(skb)->csum_start = csum_start - skb->head;
Tom Herbert7e2b10c2014-06-04 17:20:02 -07004418
Alexander Duyck76443452016-02-05 15:27:37 -08004419 return csum_fold(csum_partial(csum_start, plen, partial));
Tom Herbert7e2b10c2014-06-04 17:20:02 -07004420}
4421
David S. Millerbdcc0922012-03-07 20:53:36 -05004422static inline bool skb_is_gso(const struct sk_buff *skb)
Herbert Xu89114af2006-07-08 13:34:32 -07004423{
4424 return skb_shinfo(skb)->gso_size;
4425}
4426
Eric Dumazet36a8f392013-09-29 01:21:32 -07004427/* Note: Should be called only if skb_is_gso(skb) is true */
David S. Millerbdcc0922012-03-07 20:53:36 -05004428static inline bool skb_is_gso_v6(const struct sk_buff *skb)
Brice Goglineabd7e32007-10-13 12:33:32 +02004429{
4430 return skb_shinfo(skb)->gso_type & SKB_GSO_TCPV6;
4431}
4432
Daniel Axtensd02f51c2018-03-03 03:03:46 +01004433/* Note: Should be called only if skb_is_gso(skb) is true */
4434static inline bool skb_is_gso_sctp(const struct sk_buff *skb)
4435{
4436 return skb_shinfo(skb)->gso_type & SKB_GSO_SCTP;
4437}
4438
Willem de Bruijn4c3024d2019-03-06 14:35:15 -05004439/* Note: Should be called only if skb_is_gso(skb) is true */
Willem de Bruijnb90efd22019-02-07 14:54:16 -05004440static inline bool skb_is_gso_tcp(const struct sk_buff *skb)
4441{
Willem de Bruijn4c3024d2019-03-06 14:35:15 -05004442 return skb_shinfo(skb)->gso_type & (SKB_GSO_TCPV4 | SKB_GSO_TCPV6);
Willem de Bruijnb90efd22019-02-07 14:54:16 -05004443}
4444
Daniel Borkmann5293efe2016-08-18 01:00:39 +02004445static inline void skb_gso_reset(struct sk_buff *skb)
4446{
4447 skb_shinfo(skb)->gso_size = 0;
4448 skb_shinfo(skb)->gso_segs = 0;
4449 skb_shinfo(skb)->gso_type = 0;
4450}
4451
Daniel Axtensd02f51c2018-03-03 03:03:46 +01004452static inline void skb_increase_gso_size(struct skb_shared_info *shinfo,
4453 u16 increment)
4454{
4455 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4456 return;
4457 shinfo->gso_size += increment;
4458}
4459
4460static inline void skb_decrease_gso_size(struct skb_shared_info *shinfo,
4461 u16 decrement)
4462{
4463 if (WARN_ON_ONCE(shinfo->gso_size == GSO_BY_FRAGS))
4464 return;
4465 shinfo->gso_size -= decrement;
4466}
4467
Joe Perches7965bd42013-09-26 14:48:15 -07004468void __skb_warn_lro_forwarding(const struct sk_buff *skb);
Ben Hutchings4497b072008-06-19 16:22:28 -07004469
4470static inline bool skb_warn_if_lro(const struct sk_buff *skb)
4471{
4472 /* LRO sets gso_size but not gso_type, whereas if GSO is really
4473 * wanted then gso_type will be set. */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04004474 const struct skb_shared_info *shinfo = skb_shinfo(skb);
4475
Alexander Duyckb78462e2010-06-02 12:24:37 +00004476 if (skb_is_nonlinear(skb) && shinfo->gso_size != 0 &&
4477 unlikely(shinfo->gso_type == 0)) {
Ben Hutchings4497b072008-06-19 16:22:28 -07004478 __skb_warn_lro_forwarding(skb);
4479 return true;
4480 }
4481 return false;
4482}
4483
Herbert Xu35fc92a2007-03-26 23:22:20 -07004484static inline void skb_forward_csum(struct sk_buff *skb)
4485{
4486 /* Unfortunately we don't support this one. Any brave souls? */
4487 if (skb->ip_summed == CHECKSUM_COMPLETE)
4488 skb->ip_summed = CHECKSUM_NONE;
4489}
4490
Eric Dumazetbc8acf22010-09-02 13:07:41 -07004491/**
4492 * skb_checksum_none_assert - make sure skb ip_summed is CHECKSUM_NONE
4493 * @skb: skb to check
4494 *
4495 * fresh skbs have their ip_summed set to CHECKSUM_NONE.
4496 * Instead of forcing ip_summed to CHECKSUM_NONE, we can
4497 * use this helper, to document places where we make this assertion.
4498 */
Eric Dumazet05bdd2f2011-10-20 17:45:43 -04004499static inline void skb_checksum_none_assert(const struct sk_buff *skb)
Eric Dumazetbc8acf22010-09-02 13:07:41 -07004500{
4501#ifdef DEBUG
4502 BUG_ON(skb->ip_summed != CHECKSUM_NONE);
4503#endif
4504}
4505
Rusty Russellf35d9d82008-02-04 23:49:54 -05004506bool skb_partial_csum_set(struct sk_buff *skb, u16 start, u16 off);
Shirley Maa6686f22011-07-06 12:22:12 +00004507
Paul Durranted1f50c2014-01-09 10:02:46 +00004508int skb_checksum_setup(struct sk_buff *skb, bool recalculate);
Linus Lüssing9afd85c2015-05-02 14:01:07 +02004509struct sk_buff *skb_checksum_trimmed(struct sk_buff *skb,
4510 unsigned int transport_len,
4511 __sum16(*skb_chkf)(struct sk_buff *skb));
Paul Durranted1f50c2014-01-09 10:02:46 +00004512
Alexander Duyck3a7c1ee42012-05-03 01:09:42 +00004513/**
4514 * skb_head_is_locked - Determine if the skb->head is locked down
4515 * @skb: skb to check
4516 *
4517 * The head on skbs build around a head frag can be removed if they are
4518 * not cloned. This function returns true if the skb head is locked down
4519 * due to either being allocated via kmalloc, or by being a clone with
4520 * multiple references to the head.
4521 */
4522static inline bool skb_head_is_locked(const struct sk_buff *skb)
4523{
4524 return !skb->head_frag || skb_cloned(skb);
4525}
Florian Westphalfe6cc552014-02-13 23:09:12 +01004526
Edward Cree179bc672016-02-11 20:48:04 +00004527/* Local Checksum Offload.
4528 * Compute outer checksum based on the assumption that the
4529 * inner checksum will be offloaded later.
Otto Sabartd0dcde62019-01-06 00:29:15 +01004530 * See Documentation/networking/checksum-offloads.rst for
Edward Creee8ae7b02016-02-11 21:03:37 +00004531 * explanation of how this works.
Edward Cree179bc672016-02-11 20:48:04 +00004532 * Fill in outer checksum adjustment (e.g. with sum of outer
4533 * pseudo-header) before calling.
4534 * Also ensure that inner checksum is in linear data area.
4535 */
4536static inline __wsum lco_csum(struct sk_buff *skb)
4537{
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004538 unsigned char *csum_start = skb_checksum_start(skb);
4539 unsigned char *l4_hdr = skb_transport_header(skb);
4540 __wsum partial;
Edward Cree179bc672016-02-11 20:48:04 +00004541
4542 /* Start with complement of inner checksum adjustment */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004543 partial = ~csum_unfold(*(__force __sum16 *)(csum_start +
4544 skb->csum_offset));
4545
Edward Cree179bc672016-02-11 20:48:04 +00004546 /* Add in checksum of our headers (incl. outer checksum
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004547 * adjustment filled in by caller) and return result.
Edward Cree179bc672016-02-11 20:48:04 +00004548 */
Alexander Duyck9e74a6d2016-02-17 11:23:55 -08004549 return csum_partial(l4_hdr, csum_start - l4_hdr, partial);
Edward Cree179bc672016-02-11 20:48:04 +00004550}
4551
Linus Torvalds1da177e2005-04-16 15:20:36 -07004552#endif /* __KERNEL__ */
4553#endif /* _LINUX_SKBUFF_H */